MIND, CODE AND

CONTEXT

Essays in Pragmatics

T. Givon
University of Oregon

LAWRENCEERLBAUM ASSOCIATES, PUBLISHERS

1989 Hillsdale, New Jersey London
Copyright © 1989 by Lawrence Erlbaum Associates, Inc.
All rights reserved. No part of this book may be reproduced in
any form, by photostat, microform, retrieval system, or by any other
means, without the prior written permission of the publisher.

Lawrence Erlbaum Associates, Inc., Publishers

365 Broadway
Hillsdale, New Jersey 07642

Library of Congress Cataloging in Publication Data

Givon, T., 1936-

Mind, code, and context : essays in pragmatics / T. Givon
p. cm.
Bibliography: p.
Includes index.
ISBN 0-89859-607-6. ISBN 0-8058-0482-X(pbk.)
1. Pragmatics. I. Title.
B831.5.G58 1989
121’.68—dcl9 88-21298
CIP

PRINTED IN THE UNITED STATES OF AMERICA

10 987654321
To my parents, Victoria and Alexander Givon, who, by stranding me
early in life between their abundant love and unbounded expectations,
have inadvertently taught me the rudiments of pragmatics.
Contents

Preface xvii

1. INTRODUCTION 1

1.1. Instead of definition 1

1.1.1. Context, frame and point of view 1
1.1.2. Systems and meta-levels: The two predicaments 2
1.2. The scope of pragmatics: Recurrent themes 5
1.2.1. Gradation, continuum and non-discreteness 5
1.2.2. Relevance and importance 6
1.2.3. Similarity and analogy 6
1.2.4. Kind vs. degree: The arbitrariness of taxonomy 6
1.2.5. Abductive inference and analogic reasoning 7
1.2.6. The semeiotic relation 8
1.2.7. Purpose, function explanation and understanding 8
1.2.8. Norm, ground, saliency, frequency and markedness 8
1.3. Early roots 9
1.3.1. The dialecticians 9
1.3.2. Socrates 10
1.3.3. Aristotle 10
1.3.3.1. Non-discrete gradual change 11
1.3.3.2. The mean’ 11
1.3.3.3. Aristotle’s functionalism 12
1.3.3.4. Abductive inference 13
1.4. Kant 14
1.4.1. Extreme reductionism in Western Epistemology 14
1.4.2. Kant’s middle-ground alternative 16
1.5. Peirce and Pragmatism 20
1.5.1. Abductive inference 20
1.5.2. The interpretant as a third term in semeiotics 21
1.5.3. Non-Platonic categorization 21
1.5.3.1. The multi-dimensionality of semantic space 21
1.5.3.2. The non-discreteness of semantic dimensions 22
1.6. Ludwig Wittgenstein 23
1.6.1. The rejection of deductive logic as a means for transacting
new information 24
1.6.2. The attack on Platonic categorization 25

v
VI COfsTTENTS

1.7. Other strands of pragmatics 25

1.7.1. The language-and-culture tradition 25
1.7.2. Early functionalism: The communicative context 26
1.7.3. The Speech-act tradition: Purposive context 27
1.7.4. The presupposition tradition: Shared knowledge 28
1.7.5. Modal logic and possible worlds: The indexing of context 29
1.7.6. Ethnography of Speech: Social interaction as context 30
1.7.7. Developmental pragmatics 30
1.7.8. Pragmatics and machines: Plans, Goals, Scripts 31
1.7.9. Gestalt, prototypes and metaphors 31
1.7.10. Biology and evolution 31
1.8. Projections: Toward an integrated, pragmatics of organism,
mind and behavior 32

2. CATEGORIES AND PROTOTYPES: BETWEEN PLATO

AND WITTGENSTEIN 35

2.1. Discrete categories, fuzzy gradations and prototypes 35

2.1.1. Preamble 35
2.1.2. Extreme Platonic categorization 36
2.1.3. Extreme Wittgensteinean categorization 37
2.1.4. The prototype approach to categorization 38
2.1.5. Categorization and frequency distribution 40
2.2. Ludwig Wittgenstein and categorization 43
2.2.1. Preamble 43
2.2.2. The context-dependence of meaning 44
2.2.3. The non-discreteness of meaning 48
2.2.4. Exemplars, norms and prototypes 51
2.3. Prototypes and metaphoric extension 54
2.3.1. Preamble 54
2.3.2. Metaphoric induction into a lexical prototype 54
2.3.3. Metaphoric extension of a lexical prototype 56
2.3.4. Prototypes, metaphoric extension and grammaticalization 57
2.4. Metaphoric extension in the grammar of transitivity 59
2.5. The functional motivation for a hybrid system of categories 64
2.6. Closure 67
 CONTENTS vii

3. THE LINGUISTIC CODE AND THE ICONICITY OF

GRAMMAR 69

3.1. Preamble 69
3.2. The interpretant as context 70
3.2.1. Background 70
3.2.2. Steps to an ecology of contexts 73
3.3. ’Code’ and ’designatum’ 76
3.3.1. The designatum: Some background 76
3.3.2. Language as a complex coding system 81
3.3.3. Functional realms coded by language 82
3.3.4. In search of the designatum 84
3.3.4.1. The lexical code 84
3.3.4.2. The propositional code 86
3.3.4.3. The discourse-pragmatic code 91
3.4. Iconicity and isomorphism in grammar 94
3.4.1. Background 94
3.4.2. Preliminaries 96
3.4.2.1. Iconicity and isomorphism 96
3.4.2.2. Motivation and explanation 97
3.4.2.3. Degree of conscious access
to iconic code relations 98
3.4.3. Isomorphism, abstraction and prototypicality 99
3.4.4. Iconicity in syntax: Case studies 104
3.4.4.1. Propositions and temporal order 104
3.4.4.2. Quantity scales in the grammar of
referential identification 105
3.4.4.3. The use of word-order in the grammar of referential
identification 107
3.4.4.4. The grammatical coding of complement clauses 108
3.4.4.5. Passivization and the coding of transitivity 112
3.4.4.6. Antipassivization and the coding of transitivity 115
3.4.4.7. Stereotypicality, predictability
and coding saliency 117
3.4.4.8. Linear ordering, proximity and scope 118
3.4.5. Some meta-iconicity considerations 120
3.4.5.1. Arbitrary grammatical norms vs.
natural iconic principles 120
3.4.5.2. Conflicts between iconicity principles 121
 CONTENTS

3.5. Some bio-adaptive considerations 122

3.5.1. The difference of Man 122
3.5.2. The biological basis of isomorphic coding 123
3.5.2.1. Function, teleology and organisms 123
3.5.2.2.Ontogeny, phylogeny, language change and the
iconicity of language 124
3.5.2.3. From icon to symbol 125

4. PROPOSITIONAL MODALITIES: TRUTH, CERTAINTY,

INTENT, AND INFORMATION 127

4.1. Propositions, sentences and information 127

4.2. Epistemic modalities: The logico-semantic tradition 128
4.3. Epistemic modalities and the communicative contract 129
4.3.1. Preamble: Epistemic vs. other modalities 129
4.3.2. The epistemic space and evidentiality 130
4.3.2.1. The communicative contract 130
4.3.2.2. Evidentiality and epistemic modes 131
4.3.2.3. Subjective certainty 133
4.3.2.4. Source of information and defensibility 135
4.3.2.5. Rules of evidence 137
4.3.2.6. Certainty, challenge and response 139
4.4. The continuum space of propositional modalities 141
4.4.1. Preliminaries 141
4.4.2. The non-discreteness of presupposition 142
4.4.2.1. General considerations 142
4.4.2.2. Truth, knowledge, belief or familiarity? 144
4.4.2.3. Presupposition vs. backgroundedness 146
4.4.3. The non-discreteness of assertions 147
4.4.4. Epistemic modalities and manipulative speech-acts 148
4.4.4.1. Preamble 148
4.4.4.2. From irrealis to intent to power 148
4.4.4.3. Subjunctives: From certainty to manipulation 149
4.4.5. The speech act continuum 151
4.4.5.1. Preamble 151
4.4.5.2. From imperative to interrogative 153
4.4.5.3. From imperative to declarative 154
4.4.5.4. From declarative to interrogative 155
4.5. The pragmatics of NEG-assertion 156
4.5.1. Preamble 156
4.5.2. Negation as a speech act 156
4.5.3. Explicit and implicit normative expectations 158
 CONTENTS IX

4.5.4. The ontology of negative states and events 159

4.5.4.1. The pragmatic status of negative events 159
4.5.4.2. The pragmatics of negative states 161
4.5.5. Negation and irrealis 162
4.6. Certainty, power and deference 164
4.6.1. Preamble 164
4.6.2. Certainty and authority 164
4.6.3. Authority, negation and politeness 165
4.6.4. Certainty, modesty and politeness 166
4.6.5. Knowledge, certainty, responsibility and blame 166
4.7. The hybrid nature of propositions: Between Tautology and
contradiction 167
4.8. The negotiation of modality 170

5. THE PRAGMATICS OF REFERENCE: EXISTENCE,

REFERENTIAL INTENT AND THEMATIC IMPORT 173

5.1. Reference and existence 173

5.1.1. The Real World vs. the Universe of Discourse 173
5.1.2. Reference vs. referential intent 174
5.2. The semantics of indefinite reference 175
5.2.1. Referential opacity 175
5.2.2. Reference and propositional modalities 177
5.2.3. The grammatical marking of indefinite reference 179
5.3. The pragmatics of reference: Denotation
vs. referential import 181
5.3.1. The numeral’one’in Hebrew 181
5.3.2. Methodological preliminaries: The empirical measurement
of the importance of referents in discourse 183
5.3.3. The numeral’one’ in Krio 183
5.3.4. The numeral one in Mandarin Chinese 186
5.3.5. The unstressed indefinite’this’in spoken English 189
5.4. Normative action and referential interpretation: The context-
sensitivity of referential intent 192
5.5. The non-discreteness of reference and definiteness 193
5.5.1. Preamble 193
5.5.2. Non-discreteness of reference: Vague speaker’s intent 194
5.5.3. Gradation of referential intent 196
5.5.4. Scales of definiteness: Degree of specification 197
X CONTENTS

5.6. Coreference, topicality and evocation: The pragmatic grey

margins of Russell’s Theory of Types 198
5.6.1. Preamble 198
5.6.2. The pragmatics of coreference 198
5.6.3. Coreference and a non-Platonic theory of types 202

6. THE PRAGMATICS OF ANAPHORIC REFERENCE:

DEFINITENESS AND TOPICALITY 205

6.1. Introduction: Definiteness, presupposition and knowing

other minds 205
6.2. Sources of definiteness: Grounds
for knowing other minds 206
6.3. Definiteness and topicality 208
6.3.1. Topic and theme 208
6.3.2. In search of the elusive topic 209
6.3.3. Interim summary: Searching for the cognitive base of
topicality 213
6.4. The measurement of topicality: Predictability, importance
and attention 214
6.4.1. Preliminaries: Anaphoric predictability and cataphoric im¬
portance of topics in discourse 214
6.4.2. Predictability and the code-quantity scale 216
6.4.3. Code quantity and referential confusion 218
6.4.4. Code quantity and thematic predictability 219
6.4.5. Subject, object and transitivity: Referential predictability
and referential importance 220
6.4.5.1. Preamble: Referential importance 220
6.4.5.2. Subject and object 220
6.4.5.3. Direct and indirect object 221
6.4.5.4. Grammatical marking of important definites 222
6.4.6. Word-order and topicality 222
6.4.6.1. The topic-comment ordering principle 222
6.4.6.2. Word-order flexibility and referential
predictability 223
6.4.6.3. Dislocated order and anaphoric predictability 225
6.4.7. Word order and topic importance 226
6.4.7.1. Preamble 226
6.4.7.2. Definiteness and word-order in Papago 226
6.4.7.3. Word-order and topic persistence in Klamath 227
6.4.7.4. Direct and indirect object 228
6.4.7.5. Contrastive NPs and pre-posed order 228
 CONTENTS xi

6.4.7.6. Predictability, importance and pre-posed subject

pronouns 230
6.4.7.7. Pre-posed indefinite subjects in verb-first lan¬
guages: Where predictability and importance coin¬
cide 231
6.4.7.8. Cleft-focus constructions: Counter-expectancy and
word-order 232
6.4.7.9. Word order and thematic predictability 233
6.5. Discussion 234
6.5.1. The pragmatic nature of definite reference 234
6.5.2. Task urgency, code quantity, linear order and attention 235

7. MODES OF KNOWLEDGE AND MODES OF

PROCESSING: THE ROUTINIZATION OF BEHAVIOR
AND INFORMATION 237

7.1. Introduction 237

7.2. The philosophical tradition: Modes of inference 238
7.2.1. Preamble 238
7.2.2. The ’early’ Wittgenstein on induction and deduction 238
7.2.3. Peirce and abduction 242
7.2.3.1. Abduction and hypothesis 242
7.2.3.2. Abduction, causation and teleology 244
7.2.3.3. Abduction, similarity and relevance 244
7.2.4. Interim summary 245
7.3. The linguistic tradition: Grammar and modes of information
processing 246
7.3.1. Grammatical vs. pre-grammatical speech 246
7.3.2. Pre-grammatical rules 247
7.3.3. Grammar and automated processing 248
7.3.4. Routinized processing and context 249
7.3.5. Interim summary 249
7.4. The psychological tradition: Attended vs.
automated processing 250
7.4.1. Preamble 250
7.4.2. Cognitive psychology: Perception,
memory and attention 251
7.4.2.1. Consciousness vs. automation 251
7.4.2.2. Levels of consciousness and automaticity 252
7.4.3. Kinesiology and motor skills 253
7.4.4. Neurological aspects of automated processing 254
7.4.4.1. Routinization learning 254
7.4.4.2. Genetically determined routinization 255
xii CONTENTS

7.5. The rise and function of automated processing 256

7.5.1. Properties of automated processing: Interim summary 256
7.5.2. The functional motivation for automated processing 257
7.5.2.1. The adaptive tradeoff 257
7.5.2.2. Task urgency, predictability and economy 258
7.5.2.3. The automaticity continuum 258
7.5.3. Evolutionary aspects of automated processing 259
7.5.3.1. Rise of consciousness vs. rise of automation 259
7.5.3.2. Pre-conditions for routinization: Taxonomy
and frequency of experience 260
7.5.3.3. Bottom-up routinization 260
7.5.3.4. The biological unity of automation 261
7.5.3.5. Linguistic parallelisms 261
7.6. Some experimental evidence of grammar as an automated lan¬
guage processing mode 262

8. FACT, LOGIC AND METHOD: THE PRAGMATICS OF

SCIENCE 269

8.1. Introduction 269

8.1.1. Epistemology and organized science 269
8.1.2. The legacy of reductionism 270
8.1.3. What is an apt metaphor for behavioral and cognitive
science? 270
8.2. Reductionism in the philosophy of science 271
8.2.1. Theory vs. practice in science 271
8.2.2. Inductivism 273
8.2.2.1. Extreme inductivism in philosophy 273
8.2.2.2. Extreme inductivism in the behavioral sciences 277
8.2.3. Deductivism 279
8.2.3.1. Popper and the Hypothetico-Deductive (HD)
method 279
8.2.3.2. Pragmatic elements in Popper’s method 281
8.2.3.3. Deductivism in a social science 283
8.3. Pragmatic accounts of the scientific method 286
8.3.1. Preamble 286
8.3.2. The fact-driven nature of hypothesis formation 286
8.3.3. The status of ’observable’ facts 289
8.3.4. The relevance of data 291
8.3.5. The impetus for an empirical cycle: Puzzling facts 293
8.3.6. Intermezzo: A case-study of an empirical cycle 293
 CONTENTS xiii

8.4. The pragmatics of explanation 300

8.4.1. Preamble 300
8.4.2. Deductivism: Generalization as explanation 301
8.4.3. Causal explanation: Co-occurrence, dependency and com¬
plex patterns 304
8.4.3.1. Preamble 304
8.4.3.2. The context-dependent, theory-laden nature of
’cause’ 304
8.4.3.3. The multi-variable empirical environment 305
8.4.3.4. The bio-ontology of causal reasoning 306
8.5. Functional explanation 308
8.5.1. Preamble 308
8.5.2. Structure and function in biology 310
8.5.3. Some features of bio-behavior 311
8.5.3.1. Adaptive tasks and survival imperatives 311
8.5.3.2. The law-governed behavior of bio-organisms: Unifor¬
mity and adaptive choice 311
8.5.3.3. Bio-organisms as interactive systems: Mutual de¬
pendencies and cross-constraints 312
8.5.3.4. Higher-level interaction governing bio-behavior 313
8.5.3.5. Some examples of law-governed behavior
in bio-culture 314
8.5.4. The reductionist attack on functional explanation 317
8.5.4.1. Motive as’cause’in socio-behavior 317
8.5.4.2. The denial of bio-teleology 319
8.6. Closure: Explanatory domains, ranges of facts and the
legacy of structuralism 320

9. LANGUAGE, CULTURE AND TRANSLATION 323

9.1. Introduction 323

9.2. Modern extreme reductionists 325
9.2.1. Preamble 325
9.2.2. The anti-universalism of modern empiricists 325
9.2.3. The anti-universalism of modern mentalists 327
9.2.4. The universalism of modern Rationalists 329
9.3. Grammar and translation 330
9.3.1. Serial verb constructions 331
9.3.2. Pre-posed adverbial clauses in strict
verb-final language 333
9.3.3. Verb complement structure 334
9.3.4. Co-lexicalized complement-taking verbs 335
9.3.5. Non-embedded background clauses 336
XIV CONTENTS

9.3.6. The grammatical coding of evidentiality 337

9.3.7. Manipulative speech acts in Korean 338
9.4. Meaning, culture and translation 339
9.4.1. Preamble: Meaning, understanding and context 339
9.4.2. Translating across lexical gaps: Case studies 340
9.4.2.1. The red herring of technological vocabulary 340
9.4.2.2. Cosmology and the physical universe 342
9.4.2.3. Spatial orientation in Guugu Yimidhirr 345
9.4.2.4. Science and popular culture: Righteous food 349
9.4.2.5. Sin, taboo and Christianity 350
9.4.2.6. Ute leadership 351
9.4.2.7. You ’owe’, I ’pay’: Obligation and exchange in
Melanesia 352
9.4.2.8. Culture and grammar: Deference 354
9.5. The kinship controversy: An old battlefield revisited 355
9.5.1. Preamble 355
9.5.2. Kroeber’s incipient structuralism 355
9.5.3. Malinowski and Leach: Naive (’biological’) vs. complex
(’socio-cultural’) functionalism 357
9.5.4. Structural universalism: The kinship semantics of Scheffler
and Lounsbury 359
9.5.5. The crux of the matter: Kinship terms and translation 361
9.6. The methodological conundrum: Navigating between un¬
tenable extremes 362
9.7. A pragmatic approach to translation 364
9.7.1. The mirage of reductionism 364
9.7.2. The illusion of perfect communication 364
9.7.3. The negotiation of ’shared’ context 365
9.7.4. Relative overlap of contexts 365

10. ADAPTIVE BEHAVIOR, GROUP VARIABILITY AND

THE GENETIC CODE: THE PRAGMATICS OF
BIO-EVOLUTION 369

10.1. Prospectus 369

10.2. Pre-evolutionary biology 371
10.3. The rise of evolutionary thinking: Biological diversity,
embryology and the fossil record 373
10.3.1. Preamble 373
10.3.2. Biological diversity 373
10.3.3. Embryonic development 374
10.3.4. The fossil record 375
 CONTENTS xv

10.4. Rear-guard defense of Platonism 376

10.4.1. Preamble 376
10.4.2. Naturphilosophie 377
10.4.3. Lamarck’s adaptive functionalism 378
10.4.4. Darwin 379
10.5. The siren-call of Lamarckism 380
10.6. The transfer mechanism: Adaptive behavior, population
dynamics and species evolution 384
10.6.1. Preamble 384
10.6.2. The software of adaptive behavior: Skilled user
strategies 384
10.6.3. The adaptive advantage of skilled learning and
automated processing 386
10.7. Socio-cultural, transmitted learning and evolutionary
continuity 389
10.7.1. Background 389
10.7.2. The interaction between individual learning and group
transmission in evolution 391
10.7.2.1. Role of the individual’s routinized behavior in
gaining adaptive-selectional advantage 391
10.7.2.2. The role of socio-cultural transmission in
accelerating the spread of favorable
genetic traits 392
10.7.3. Summary 392
10.8. The difference of mind and the difference it makes 393
10.8.1. The difference of Man 393
10.8.2. Darwin, Wallace and evolutionary contiguity 394
10.8.3. More rear guard skirmishes of Platonism: The abduction
of Gould’s punctuated equilibrium 396
10.8.4. Modularity and the Platonization of mental faculties 399
10.9. Closure: The language parallels 400
10.9.1. Variation and change 400
10.9.2. History, diachrony and recapitulation 401
10.9.3. The evolution of communicative context 403
XVI CONTENTS

11. THE MYSTIC AS PRAGMATIST: LAO TSE AND

TAOISM 405

11.1. Introduction 405

11.2. Taoist metaphysics: Unity in diversity 406
11.2.1. The One 406
11.2.2. The Two 408
11.2.3. The circle: Tao as drift 412
11.3. Taoist epistemology 413
11.3.1. Two realities, two modes of knowledge 413
11.3.2. Yin and Yang: The context-dependence of categories
11.4. Taoist ethics 416
11.4.1. The ethics of entropy 416
11.4.2. Wu-Wei and Tao: The water metaphor 417
11.4.3. Wu-Wei as a utilitarian creed 419
11.4.4. The Code: Li vs. natural kinship 420
11.4.5. The doctrine of Straw Dogs 421
11.5. Closure 423

BIBLIOGRAPHY 425

INDEX 453
Preface

Pragmatics is an approach to description, to information processing, thus to

the construction, interpretation and communication of experience. At its core
lies the notion of context, and the axiom that reality and/or experience are not
absolute fixed entities, but rather frame-dependent, contingent upon the ob¬
server’s perspective.
Pragmatics traces its illustrious ancestry to the pre-Socratic Greek dialec¬
ticians, then via Aristotle to Locke, Kant and Peirce, eventually to 19th Century
phenomenologists, and-last but not least-to Ludwig Wittgenstein. In cogni¬
tive psychology, pragmatics underlies figure-ground perception, primed
storage and maleable recall, attended (’context-scanning’) information
processing, and flexible (’prototype’) categorization. In linguistics, prag¬
matics animates the study of contextual meaning and metaphoric extension,
frame semantics and the semeiotics of grammar-in-discourse, the sociology
of language, and the acquisition of communicative competence. In anthropol¬
ogy, pragmatics is reflected in the exploration of cultural relativity, ethno-
methodology and cross-cultural cognition.
In spite of such exalted lineage and wide applicability, the academic study
of pragmatics remains narrow, insular and fractious. On the one hand,
various formal schools have undertaken to keep pragmatics firmly attached
to the very discipline which it purported to overthrow-formal deductive
logic. On the other hand, a plethora of informal schools have taken the
intoxicating freedom of contextual relativity as license for extreme
methodological nihilism, unfettered intuitionism, and an anything goes rejec¬
tion of sensible empirical constraints. What unites these extreme interpreta¬
tions is, paradoxically, an antipragmatic faith in the Platonic excluded middle:
The lack of total order means a total lack of order; the lack of total understanding is a
total lack of understanding. In this way, the very essence of pragmatics is
subverted by its most impassioned proponents.
Pragmatics, at its somewhat unadorned middle-ground best, closely reflects
the evolutionary compromise practiced by biological organisms. In adapting
to life in a less-than-ideal environment, bio-organisms have invariably opted
for the proposition that half a loaf is infinitely better than none; that life is
precariously suspended mid-way between absolute order and unmitigated
chaos; that while full determinism is a dangerous evolutionary trap, un¬
bounded freedom is an unrealistic evolutionary mirage. In their humble
travail to adapt and survive, bio-organisms have recognized what contentious
xvii
xviii PREFACE

academics all too often ignore—that GoedePs observed limits on systems—

neither fully consistent, nor ever complete-sum up rather well the pragmatic
predicament of life and mind in a real environment.

The writing of this book was supported by a number of generous sources. I

would like to take this opportunity and acknowledge their help. In order: the
Office of Naval Research, research grant "Mechanisms of Cognitive Perfor¬
mance" (with S. Keele and M. Posner) (1983-1985), the John Simon Guggenheim
Memorial Foundation, through a fellowship , "The pragmatics of human lan¬
guage" (1986); The Fulbright-Hays Exchange Program, through a visiting lec¬
tureship "The American Indian: Past and Present" (Auckland, 1986); The
National Endowment for the Humanities, through research grants "Ute tradition¬
al Narratives and the pragmatics of word-order change" (Ignacio, 1980-1985);
"Serial verbs and the mental reality of’event’: An empirical approach to the
Sapir-Whorf Hypothesis" (Papua-New Guinea, 1985-1987; renewed 1988-
1990); and the DeutscheForschungGemeinschaft, for a visiting fellowship "Lec¬
tures on pragmatics" (Koln, 1987). To make the picture complete, I would also
like to acknowledge the consistent non-support of the National Science Foun¬
dation.

T. Givon
Eugene, Oregon
CHAPTER 1

INTRODUCTION

1.1. Instead of definition *

1.1.1. Context, frame and point of view

Pragmatics may be likened to a vast terrain whose boundaries are so

distant that we perceive them only dimly, given our less-than-exalted
vantage point. This is somewhat embarrassing for a discipline so intent¬
ly focused upon the study of vantage points; although there is perhaps a
certain measure of poetic justice involved in the embarrassment. As a
serious empirical discipline, pragmatics is still in its infancy, clumsily at¬
tempting to grasp for its own meaning. It would thus be presumptive,
and perhaps even alien to the very spirit of pragmatics, to saddle it
prematurely with a rigid definition. Still, if there is a unifying theme to
the entire enterprise, it must have at its very core the notion of context,
or frame, or point of view.
Pragmatics as a method may be first likened to the way one goes about
constructing a description. The reason why I've chosen 'description' as
my first metaphor for pragmatics may trace back to dimly recalled times
in military reconnaissance. When one was sent to draw a panoramic view
of some Godforsaken hill, the resulting sketch-cum-commentary had to
always specify the map coordinates of one's vantage point; that is
where one stood when drawing the picture. Your description -- pictorial-
cum-verbal, you were told -- was useless without those coordinates. The
first metaphor for pragmatics as a method may thus be given as:

(1) Description and point of view:

"The description of an entity is incomplete, indeed un¬

interpretable, unless it specifies the point of view from whence
the description was undertaken".

*1 am indebted to T.K. Bikson, Hartmut Haberland, Fred Kroon, Dennis Robinson and Martin
Tweedale for many helpful suggestions concerning the early history of pragmatics. In revis¬
ing this chapter, I benefitted greatly from having had the opportunity to present an earlier ver¬
sion at the Philosophy Department Colloquium, Auckland University.
2 Mind, Code and Context

The very same idea may be re-phrased in terms of a 'picture and frame
metaphor:

(2) Picture and frame:

"A picture is not fully specified unless its frame is also specified".

Pragmatics as a method may also be rendered in terms of the relation

between meaning and context:

(3) Meaning and context:

"The meaning of an expression cannot be fully understood

without understanding the context in which the expression is
used".

As revealing as the three metaphors above may be, there is still a more
general way of approaching the definition of pragmatics. Somewhat
surprising, one may trace it back to the work of an eminent logician,
Bertrand Russell.

1.1.2. Systems and meta-levels: The two predicaments

"...There was only one catch and that was Catch-22, which
specified that a concern for one's own safety in the face of dangers
that were real and immediate was the process of a rational mind.
Orr was crazy and could be grounded. All he had to do was ask;
and as soon as he did, he would no longer be crazy and would
have to fly more missions. Orr would be crazy to fly more mis¬
sions and sane if he didn't, but if he was sane he had to fly them.
If he flew them he was crazy and didn't have to; but if he didn't
want to he was sane and had to..."

J. Heller Catch-22 (1962, p. 54)

The three core metaphors for the pragmatic method given above — point
of view, frame, context — may be further generalized via the notions of
systems and meta-levels. Let a system be, at its most general, a hierarchic
arrangement of parts and sub-parts. When one undertakes to specify
('describe') a system, it is desirable, from a purely practical point of view.
 Introduction 3

to impose some limit on the description, otherwise the descriptive task

may be infinite. This requirement is the one we call closure.
The system, as a hierarchic entity, is made out of a progression of
levels, each one acting as meta-level to the sub-level(s) embedded within
it. Each meta level is thus the context for the sub-levels embedded within
it. For purely practical reasons, if the system is to remain finite (i.e.
describable within finite time,space and means), the last -- highest —
meta-level must remain context-less; it lacks its own meta-level. In terms
of our picture metaphor, the last meta-level is the frame, yet itself remains
un-framed, therefore not fully specified. And here lies our first predica¬
ment of pragmatics, that of completeness:

(4) The predicament of completeness:

"So long as the system is fully specified, i.e. closed, it must remain
in principle incomplete".

Bertrand Russell in his Introduction to Mathematical Philosophy (1919), dealt

with the second predicament, that of consistency. In his celebrated Theory
of Types, he observed that the classical logical paradoxes, such as the liar's
paradox (T never tell the truth') are all instances of a more general
phenomenon, that of self inclusion. That is, within the same description,
one level acts as both the meta-level the sub-level. In other words, it
'includes itself'.1 When such self-inclusion - or 'crossing of meta-levels'
-- is allowed, the system becomes inconsistent. And here lies the second
predicament of pragmatics:

(5) The predicament of consistence:

"So long as one is allowed to switch meta-levels -- or points of

view - in the middle of a description, the description is logically
inconsistent".

To bypass the predicament of inconsistency, Russell, in his Theory of Types,

resorted to legislation. Logical descriptions, he insisted, must remain
within the scope of one specified meta-level. In other words:

1 Russell formulated the general case of this paradox as: "The set of all the sets that don't in¬
clude themselves; does it or doesn't it then include itself?"
4 Mind, Code and Context

(6) Russell's Constraint on systems:

"A self-consistent (though in an obvious sense incomplete) logi¬

cal description can only operate within a fixed point of view, con¬
text, meta-level".

In imposing his constraint, Russell, with one wave of his magic wand,
exorcised the specter of pragmatics out of deductive logic. This exorcism
yielded two results, the first intended, the second perhaps not altogether
obvious to the exorcist himself at the time:

(a) Deductive logic was rescued as a closed, internally-consistent, co¬

herent system.

(b) The instrument of deductive logic was removed, once and for all, as
serious contender for modeling, describing or explaining human lan¬
guage -- or mind.

Put another way, Russell indeed saved the instrument, by giving up its
original -- historic — purpose.
Deductive logic for the moment aside, the pragmatically engendered
predicaments of closure and consistence continue to haunt any attempt
to describe language and mind. Neither language nor mind abides by the
requirement of closure, except perhaps temporarily, for limited tasks.
Both mind and language are necessarily open systems that constantly ex¬
pand, add meta-levels, learn and modify themselves. Equally, both lan¬
guage and mind are notoriously promiscuous in violating Russell's
constraint on self-inclusion and reflexivity. Consciousness is indeed
forever adjusting its frame, shifting meta-levels; it keeps re-framing and
reflexively framing itself. This propensity of consciousness is neither an
aberration nor an accident. Rather, it is a necessary, adaptively motivated
capacity; it stands at the very core of our perceptual and cognitive
processing mechanisms. It is a precondition for the mind's ability to
select, evaluate, file, contextualize and respond appropriately to moun¬
tains of information.
The key notions here are 'select', 'evaluate', 'contextualize' and
'appropriately'. In the immense, Herculean task of natural -- biological -
- information processing, the bulk of the input is in fact blocked, i.e.
deemed — in the appropriate context — to be either irrelevant or not ur¬
gent. Only small morsels of the input, judged to be either relevant or ur¬
gent in context, are let through for further processing. The selective
exercise of the mind's contextual judgement -- the readjustment of the
frame for the particular occasion and task -- is the sine qua non of natural,
biological information processing, which is undertaken under severe
limits: Finite time, finite storage capacity, finite means.
 Introduction 5

As we shall see throughout, it is the mind's pragmatic flexibility and

open-endedness (i.e. 'incompleteness' and 'inconsistence'), its capacity to
re-frame and re-contextualize, that enables it to perform - sometimes
concurrently, often selectively — the multitude of its complex processing
tasks.

1.2. The scope of pragmatics: Recurrent themes

As noted above, at the core of pragmatics lies the notion of context.

Pragmatics is a context-dependent approach to analysis -- of behavior, of
tasks, of systems, of meaning. A number of recurrent themes have been
traditionally associated with pragmatics. All of them are mediated by --
or founded upon -- the core notion of context. In this section I will brief¬
ly survey pragmatics' most common leitmotifs.

1.2.1. Gradation, continuum and non-discreteness

As we shall see in Chapter 2, below, non-pragmatic approaches to

description, thus to sub-levels within a hierarchic system, have always
assumed that categories are discrete. In other words, membership in a
category is governed by the strict rule of the excluded middle. A major fea¬
ture of pragmatics has been, ever since its inception, that categories are
not fully discrete, but may display shades and gradations. Not all ex¬
ponents of non-discreteness have explicitly related it to the central notion
of context. I would like to argue here that context is indeed the crucial
mediator that makes non-discreteness both possible and necessary. The ar¬
gument runs roughly as follows:

(a) The point of view -- being itself outside the picture -- cannot be con¬
strained by the frame-internal system of discrete categories. Outside
the upper meta-level of the system the context -- frame -- is undifferen¬
tiated.

(b) In principle, therefore, any adjustment in the ultimate point-of-view

is bound to be non-discrete, it may be made gradually, without sharp
categorial breaks.

(c) In principle, then, the system inside the frame will display the conse¬
quences of non-discrete adjustments of the frame.
6 Mind, Code and Context

This necessary connection between context and non-discreteness can

only be broken by discretizing the notion of context. This has been at¬
tempted repeatedly in formal logic in the last four decades (see discus¬
sion in sections 1.7.4. and 1.7.5., below). However, the minute such gambit
is accomplished, the erstwhile context ceases to be context. The erstwhile
frame merely joins the system/picture on the inside. And on the outside
-- mutatis mutandis — there remains context, the last meta-level, open-
ended, non-discrete, and as disdainful as ever of logic's slights of hand.

1.2.2. Relevance and importance

The two partly related notions of importance and relevance surface
wherever one traffics in pragmatics. Both are a matter of degree, rather
than of discrete choice. Both involve contextual judgement that can be
captured neither by deductive nor inductive logic. Rather, the judgement
involved in both is abductive;2 reference must be made to the context
relative to which something is judged to be either important or relevant.

1.2.3. Similarity and analogy

Much like importance and relevance, the notions of similarity and
analogy are in principle impervious to deductive or inductive reasoning.
Equally, they are non-discrete, a matter of degree. In principle, anything
can be similar to anything else, and anything may be viewed by analogy
to anything else -- provided the appropriate context, frame, or point of
view is construed. And construing the appropriateness of context is, in
principle, a purely abductive enterprise.

1.2.4. Kind vs. degree: The arbitrariness of taxonomy

Somewhat related to the question of non-discreteness of categories is
the issue of supra-categories, meta-levels and thus -- in the universe of
individual tokens — the hierarchy of types. Typically, in a non-pragmatic
('Platonic') approach to the taxonomy of types and sub-types, one always
assumes -- often implicitly 3 -- that it is somehow possible to tell a 'major'
from a 'minor' property of individual tokens within a population by some
algorithmic means. Individuals that differ from each other by only
'minor' properties are then said to differ only by degree; they therefore
belong to the same type. In contrast, individuals that differ from each

2 See discussion directly below.

3 See for example Russell (1919).
 Introduction 7

other by a 'major' property are said to be different in kind; they belong

to different types. We may illustrate this diagrammatically as follows:

(7) Difference in kind and difference in degree

The fundamental distinction between 'degree' and 'kind' is the basis for
all taxonomies — pragmatic and non-pragmatic alike. The difference be¬
tween the two approaches lies in their attitude toward the absoluteness
of such a distinction. To the non-pragmatist, it is somehow a rigid, prin¬
cipled distinction that yields absolute categorial boundaries. The prag¬
matist merely points out the obvious, namely that the distinction between
'major' and 'minor' properties -- and thus between 'difference in kind'
and 'difference in degree' -- is in principle arbitrary; it cannot be made
on either deductive or inductive grounds. It is typically a matter of ab-
ductive judgement about the context -- often the purpose -- of the
taxonomy. Only relative to that context can some properties or distinc¬
tions be deemed important, and others less so.

1.2.5. Abductive inference and analogic reasoning

The non-pragmatic tradition speaks of two modes of knowledge -- or
modes of inference -- deductive and inductive. The first proceeds from
the general rule to its specific instances. The second presumably proceeds
from specific instances to the general rule. Pragmatically-based abduc¬
tive inference4 -- concerning appropriateness of context, importance,
relevance, similarity or explanation -- is in principle a different kind of
reasoning. It proceeds by hypothesis, guesswork or intuition, often by
analogy. It is thus, in principle, unconstrained.

4 See discussion of Peirce, further below, as well as in Chapter 7, below.

8 Mind, Code and Context

1.2.6. The semeiotic relation

The discussion of the relation between a sign and its designatum al¬
most always drags in pragmatics by either the front or the back door. This
is due to the presence of the third term in the semeiotic equation, the silent
partner — the interpretant. The interpretant, as will be argued below,
turns out to involve the perceived context, within which the semeiotic
relation holds for its perceiver. The semeiotic relation, of even the most
abstract symbol, is thus in principle a context dependent entity, i.e. a mat¬
ter of pragmatics.

1.2.7. Purpose, function explanation and understanding

As will be suggested further below,5 6 intent is already a crypto-prag¬

matic notion, involving the mind of the intender as context for a proposi¬
tion. In addition, the teleological notions of purpose and function — in
the philosophy of science as well as in the behavior of organisms -- are in
principle pragmatic entities that cannot be captured by deductive or in¬
ductive means. They involve an abductive, analogical, context-depend¬
ent judgement. Further, these teleological entities often form the bulk of
the context within which cognition, behavior and communication must
be understood or explained. And 'explanation' is itself a pragmatic no¬
tion, since it involves relating an entity to its wider frame, its context.

1.2.8. Norm, ground, saliency, frequency and markedness

In the study of perception, cognition and communication, the notion

of norm ('ground'), vis-a-vis which information (the 'figure') is salient, is
a core notion. But 'norm' is itself a prime pragmatic construct: The norm
in information processing is merely the context vis-a-vis which informa¬
tion is salient. As we shall see at a number of junctures below (in par¬
ticular Chapters 2, 4 and 6), the notion of norm is inherently
distributional; it is based on a certain skewing in the relative frequency
of the 'figure' and the 'ground': The ground ('norm') is relatively fre¬
quent, while the figure ('counter-norm') is relatively rare. As we shall see
later on, the substantive sense of markedness in natural language, per¬
ception and cognition is grounded in the pragmatics of norm vs. counter¬
norm and their skewed relative frequency.

5 See Chapter 3, below.

6 See Chapter 5, below.
 Introduction 9

1.3. Early roots

"...Any philosophical doctrine that should be completely new

could hardly fail to prove completely false; but the rivulets at the
head of the river of pragmatism are easily tracked back to almost
any desired antiquity. Socrates bathed in these waters. Aristotle
rejoiced when he could find them. They run, where one would
least suspect them, beneath the dry rubbish-heaps of Spinoza.
Those clean definitions that strew the pages of Essay Concerning
Humane Understanding (I refuse to reform the spelling) had been
washed out in these same pure springs. It was this medium, and
not tar-water, that gave health and strength to Berkeley's earlier
work... From it the general views of Kant derive such clearness as
they have..."

C. S. Peirce, "Pragmatism in Retrospect:

The Last Formulation" [1940, p. 269]

1.3.1. The dialecticians

In these somewhat catty yet generous words, the founder of modem
Pragmatism pays homage to his antecedents. These antecedents in fact
go further back, to the pre-Socratic Dialecticians. Allen (1966) charac¬
terizes Anaximander's dialectics as follows:

"...The world, like a pendulum, maintains equilibrium through the al¬

ternation of its extremes..." [1966, p. 3]

And Diogenes Laertius ascribes to Heraclitus the following observation:

"...All things come into being by conflict of opposites, and the sum of
things flows like a stream..." [1925, vol. II, p. 415]

The classical dictum also attributed to Heraclitus: "...You never step twice
into the same river..." is also essentially pragmatic, presumably pointing
out to the ever-changing context -- time, the river, the self. Similarly, in
the anonymous manuscript Divided Logi, ascribed to the Protagoran tradi¬
tion, we find the following observation:

"...If we sat in a row and said 'I am an initiate', we would all say the
same thing, but only I would speak the truth, since I really am one..."

7 Cited from Haberland (1985) after Diels (1969, vol. II).

10 Mind, Code and Context

A much fuller dialectic-pragmatic approach, apparently ignored by

Peirce, is of course found in mystic Oriental philosophers, most succint-
ly in Lao Tse's Tao Teh Ching.8 The mystic's route indeed remained an ever¬
present alternative to the two non-pragmatic reductionist schools --
empiricism and rationalism -- in the history of Western philosophy. But
the mystics were easy to ignore in Western philosophy; they were in¬
herently non-analytic; they also rejected the major tenet of Western epis¬
temology — the separation of Mind from World.

1.3.2. Socrates
The inclusion of Socrates in the roots of Pragmatism is at first surpris¬
ing, given our largely Platonic access to his philosophy. However, in Hip-
pias Major Socrates (or was it Plato?) adopts a Pragmatic — indeed
Wittgensteinean—approach in discussing the context relativity of the
meaning of adjectives such as 'fine' and 'foul'. Indeed, Socrates invokes
Heraclitus for that purpose:

"...Don't you know that what Heraclitus said holds good: The finest of
monkeys is foul when put together with another class, and the finest of
pots is foul when put together with the class of girls...If you put the class
of girls together with the class of gods, won't the same thing happen as
happened when the class of pots was put together with that of girls?
Won't the finest girl be seen to be foul? And didn't Heraclitus (whom you
bring in) say the same thing too, that 'the wisest of men is seen to be a
monkey compared to god in wisdom and fineness and everything
else'?..." [Woodruff, 1982, pp. 10-11]

1.3.3. Aristotle

Also somewhat surprising is Peirce's inclusion of Aristotle in the Prag¬

matist family tree. Surprising first because of Aristotle's extreme em¬
piricist view of the source of mental categories, or 'forms'. According to
Aristotle, 'forms' contain the entire categorial array (i.e. 'meaning') of
perceived objects -- minus their existence. Aristotle's 'forms' -- though
objective-external in origin, are just as fixed, absolute and universal as
Plato's innate categories.9 Like Plato, Aristotle views categories as dis¬
crete, requiring necessary-and-sufficient membership criteria.10 Further,
contrary to Morris's (1938) attribution, Aristotle's semeiotics shows no

8 Dealt with in some detail in Chapter 11, below.

9 See discussion of Aristotle's 'forms' as given in his Metaphysics in Tweedale (1986).
10 See Aristotle's Categories, in Ackrill (tr.& ed.,1963).
 Introduction 11

trace of Peirce s interpretant'. Nonetheless, in four distinct areas Aris¬

totle may be indeed shown to be an early exponent of pragmatics. We will
take them up in order.

1.3.3.1. Non-discrete gradual change

This topic recurs in many of Aristotle's work, but is perhaps most clear¬
ly developed in the Metaphysics. The problem that necessitated a prag¬
matic departure for Aristotle arose from the inability of his absolute
'forms' to account for gradual change, such as 'coming-into-being' or
'passing away'. Aristotle rushed to the rescue of his 'forms' with the con¬
cept of synolon. As Tweedale (1986) observes:

"...The synolon is Aristotle's concession to the Heraclitean flux. It

manages an existence of sorts even while it is coming-to-be or passing
away. This is because it contains matter, the sine qua non for anything that
undergoes a process of change..."[1986, p. 5]12

1.3.3.2. ’The mean’

Aristotle's 'mean' (or 'the intermediate') in the Ethics may be inter¬
preted as the pragmatist's Golden Mean, whereby the extreme, pure cases
are to be eschewed in favor of the delicately balanced, hard to grasp thin
line in-between.13 Thus consider the following passage from the Ethics
(McKeon, ed., 1941):

"...In everything that is continuous and divisible it is possible to take

more, less or an equal amount, and that either in terms of the thing itself
or relative to us; and the equal is an intermediate between excess and
defect. By the intermediate in the object I mean that which is equidistant
from each of the extremes, which is one and the same for all men; by in¬
termediate relative to us, that which is neither too much nor too little --
and this is not one, nor the same for all. For instance, if ten is many and
two is few, six is the intermediate, taken in terms of the object... But the
intermediate relative to us is not to be taken so; if ten pounds is too much

11 See discussion in Chapter 3, below.

12 While forms are presumably immutable, absolute and undergo no flux. I owe my frail un¬
derstanding of these issues to Martin Tweedale (in personal communication); see also
Tweedale (1986).
13 I am again indebted to Martin Tweedale (in personal communication) for suggesting this
area of Aristotle's pragmatism to me.
12 Mind, Code and Context

for a particular person to eat and two too little, it does not follow that the
trainer will order six pounds..." [1941, pp. 957-958]

Aristotle is confronting here the problem of graded continuum, and also

implicitly the problem of context, especially in making the intermediate
'relative to us' a contingent notion. This is even clearer in the next pas¬
sage:

"...both fear and confidence and appetite and anger and pity and in
general pleasure and pain may be felt both too much and too little, and
in both cases not well; but to feel them at the right time, with reference to
the right objects, toward the right people, with the right motive, and in
the right way, is what is both intermediate and best..." [McKeon, ed., 1941,
p. 958]

Aristotle thus enumerates the various elements of context upon which

'the intermediate for us' is contingent: time, object, participants, motive,
manner. And while this is pragmatics in the realm of ethics, it is classi¬
cal pragmatics nonetheless.

1.3.3.3. Aristotle’s functionalism

Aristotle's approach to biology is suffused with functionalist teleol¬
ogy. His arguments are constructed first against an earlier structuralist
tradition, manifest in the works of pre-Socratic materialists such as Em¬
pedocles. These materialist sought to describe and explain the nature of
organisms in the same way as they described the nature of inorganic
physical objects — by reference to their component elements, or in the
case of Democritus to their component structures. Aristotle begins his ar¬
gument in De Partibus Animalium by advocating a move from
Empedocles' elemental description to structural description, i.e. descrip¬
tion in terms of the histology and anatomy of organs:

"...But if men and animals are natural phenomena, then natural

philosophers must take into consideration not merely the ultimate sub¬
stances of which they are made, but also flesh, bone, blood and all the
other homogeneous parts; not only these but also the heterogenous parts,
such as face, hand, foot..." [McKeon, ed., 1941, p. 647].

He then proceeds to argue that even such structuralist approach, such as

that of Democritus, won't do:

"...Does, then, configuration and color constitute the essence of the

various animals and their several parts? For if so, what Democritus says
 Introduction 13

will be strictly correct... no hand of bronze or wood or constituted in any

but the appropriate way can possibly be a hand in more than name. For
like a physician in a painting, or like a flute in a sculpture, it will be un¬
able to do the office which that name implies..." (ibid, p. 647; emphases are
mine; TG)

Next, a teleological approach to biology is outlined, using the analogy of

usable artifacts:
"...What, however, I would ask, are the forces by which the hand or the
body was fashioned into its shape? The woodcarver will perhaps say, by
the axe or the auger; the physiologist, by air and by earth. Of these two
answers the artificer's is the better, but it is nevertheless insufficient. For
it is not enough for him to say that by the stroke of his tool this part was
formed into a concavity, that into a flat surface; but he must state the
reasons why he struck his blow in such a way as to affect this, and what
his final object was..." [ibid, pp. 647-648; emphases are mine; TG]
The need for form-function correlation is now driven home, again by
using the analogy of tools:
"...if a piece of wood is to be split with an axe, the axe must of neces¬
sity be hard; and, if hard, must of necessity be made of bronze or iron.
Now exactly in the same way the body, which like the axe is an instru¬
ment -- for both the body as a whole and its several parts individually
have definite operations for which they are made; just in the same way,
I say, the body if it is to do its work, must of necessity be of such and such
character..." [ibid, p. 650]
In outlining his functionalist argument, Aristotle rejects two extremist
views of biology: The structuralist's anti-teleology, and Plato's abstract
super-teleology (whereby all purposive phenomena -- in the inorganic
and organic universe alike -- are attributed to some higher purposive in¬
telligence). While not addressing himself specifically to language,
Aristotle's functionalism in biology indeed presages an important in¬
gredient of modem pragmatics.

1.3.3.4. Abductive inference

Flanson (1958) points out that Peirce's abductive inference, the 'third
mode' of knowledge, was first described by Aristotle:

"...Aristotle lists the types of inferences. These are deductive, induc¬

tive and one other called apagoge. This is translated as reduction . Peirce
translates it as 'abduction' or 'retroduction'. What distinguishes this kind
of argument for Aristotle is that the relation of the middle to the last term
is uncertain, though equally or more probable than the conclusion; or
14 Mind, Code and Context

again an argument in which the terms intermediate between the last term
and the middle are few. For in any of these cases it turns out that we ap¬
proach more nearly to knowledge... since we have taken a new term..."
(1958, p. 85; cited from Aristotle's Prior Analytic, tr. by Jenkinson, ed. by
Ross, Oxford, vol. II, p. 25)
In another reference, in the Posterior Analytic, Aristotle characterizes the
mode of abduction as follows:
"...The particular facts are not merely brought together but there is a
new element added to the combination by the very act of thought by
which they are combined... The pearls are there, but they will not hang
together till someone provides the string..." (Posterior Analytic, vol. II, p.
19; cf. Whewell, Organum Renovatum, pp. 72-73).
We will return to discuss the modes of inference, and abduction in par¬
ticular, in Chapters 7 and 8, below.

1.4. Kant

1.4.1. Extreme reductionism in Western Epistemology

The nature of the categories of human mind, their origin, and what ex¬
actly they stand for, are persistent questions that have been with us since
the dawn of Western thought. What is the source of all those entities that
inhabit our mind? How did they get there? What do they stand for? Do
we experience the world in a certain way because our mind is so pre-cast,
chock-full of concepts that have been placed there by the Deity (as Des¬
cartes would have it)? Are those innate ideas there because of that latter-
day deity. Genetics or 'the Bio-Program' (as Chomsky, 1966, 1968 or
Bickerton, 1981, would)? Or do we, alternatively, owe our mental
categories to experience? Has some real or perceived outside world been
imprinted on our passive mind via sensory perception?
Ever since post-Socratic Greek philosophy began to contemplate the
puzzle of mind vs. world (or 'in vs.out', or 'subject vs. object'), and until
the latter part of the 18th Century, only two extreme answers to this eter¬
nal puzzle had been seriously contemplated in Western philosophy.
Rationalists, from Plato down, had opted for the primacy of innate
categories of the mind. Empiricists, from Aristotle onward, had argued
for the primacy of experience, whether attributed to some putative Real
World (the 'objectivist' option) or to the vagaries of our senses (the
'subjectivist' option).
It is perhaps not altogether an accident that the two extreme reduc¬
tionist schools of Western epistemology. Rationalism and Empiricism,
 Introduction 15

turn out nonetheless to share two of their most fundamental assumptions

about the nature of mind:

(a) Separateness of mind from world: The mind and the world
— internal and external -are distinct and separate entities, al¬
lowing no direct access from one to the other.
(b) Absoluteness of mental categories: Categories of the mind
— be they Plato's innate ideas or Aristotle's perceived forms
— are discrete, universal and absolute, with membership ad¬
judicated by rigid — necessary and/or sufficient — criteria,
which members either do or do not possess.
With the first assumption (a). Western analytic philosophy detached it¬
self from its old antecedence in the mystic tradition.14 The second as¬
sumption (b) characterizes equally well rationalists from Plato to St.
Augustine to Descartes to Chomsky, as it does empiricists from Aristotle
to St. Thomas Aquinas to Hume to Skinner or Bloomfield. Its roots are
presumably found in the logical doctrine of the excluded middle.
The motivation for assumption (b) is both legitimate and, I believe, of
some interest. Kant has alluded to it in his discussion of the potential in¬
finity of the 'manifold multiplicity of intuitions' (i.e. 'sensations' of the
outside world; see further below). Both Plato and Aristotle had to reckon
with the very same problem. If reality is in principle unconstrained, non¬
discrete and multifarious, and if the senses do not rigidly pre-select and
constrain that multiplicity, then the understanding ('cognition') cannot
match that unconstrained multifarious continuum without itself being
infinite and unconstrained. Kant's solution to this dilemma was, I believe,
the right one: Disengage 'the thing for itself' from 'the thing for us', by
invoking the — potentially discretizing — transcendental schema. That
schema is pre-condition to both perception and cognition. Plato, on the
other hand, chose to ascribe the discretizing role to the innate categories
of the mind.
Aristotle's position remains somewhat ambiguous. At first, glance he
seems to have refused to acknowledge the problem. His mental categories
— the perceived 'forms' — were said to correspond to actual objects in all
detail - except for actual existence. Presumably, if the world (or at least
the world in process of change) is graduated and unconstrained, the
'forms' should be likewise. Thus, consider the following from De Anima
(McKeon, ed., 1941):

14 See Chapter 11, below, for an illustration of that tradition, though from a non-Western
source.
15 See Haberland (1985, p. 379) for a discussion of the history of the law of the excluded mid¬
dle.
16 Mind, Code and Context

"...Actual knowledge is identical with its object... the mind which is ac¬
tively thinking is the object which it thinks..." [pp: 593; 595]

If Aristotle had identified the dilemma, only two elegant ways out would
have remained for him:

(a) Allow the mind an infinite capacity; or

(b) Construe the world itself as constrained, categorial, discrete.

Aristotle's way of scotching the problem was, as we have seen above, by

postulating the synolon, which would account for the mind's ability to
reflect the world's unconstrained flux. In terms of elegance, this is admit¬
tedly somewhat of a fudge.

1.4.2. Kant’s middle-ground alternative

There are three main reasons, I believe, for considering Kant a pioneer
of modern pragmatics. At issue are three major aspects of his epistemol¬
ogy. I will survey them in order.

1.4.2.1. The source of mental categories

Kant was clearly the first modern analytic philosopher to elucidate the
middle ground, pragmatic alternative to the two extreme, reductionist
schools of Western epistemology. I quote here from Kemp (1968), a
rigorous interpreter:

"...Human knowledge arises through the joint functioning of intuition

(the product of sensibility)16 and concept (the product of understanding).
Sensibility is a passive receptivity, the power of receiving representations
of the objects by which understanding is an active spontaneity, the power
of exercising thought over objects given us in sensible intuition. Neither
by itself gives us knowledge: 'Thoughts without contents are empty', says

16 Kemp notes that Kant's Anchaaung means 'immediate apprehension' and is often trans¬
lated as 'perception'. However, it subsumes not only 'sensory perception', but also 'pure
intuition'. For this reason, Kemp prefers to render Anchaaung as 'intuition'. I personally
feel that such a rendition is bound to be even more confusing, in that it suggests the ex¬
clusion of sensory perception from the meaning of the term.
 Introduction 17

Kant in a famous phrase (KRV A51 B75), 'intuitions without concepts are
blind ..."[1968, p. 16]17 F

Kant's middle-ground epistemology stresses the interactive, mutually-

dependent ( dialectic') relation between percepts and concepts. Concepts
and percepts thus form the respective context for each other, a context
within which each receives its respective interpretation. None is by itself
a viable prime. Kant's contribution thus lies in his attempt to do away
with the artificial mind-vs.-world dilemma, via his constructivist inter¬
pretation of 'reality'. To quote Kemp again:

"...We do not find them [our sense impressions of the worlds; TG] al¬
ready organized... but rather organize them ourselves..." [1968, p. 23]

And further:

"...'the order and regularity in the appearance, which we entitle na¬

ture, we introduce ourselves. We could never find them in appearance,
had not we ourselves, or the nature of our mind, originally set them there'
(A125)..." [1968, p. 32]

And finally:

"...Our empirical knowledge has two sources, sensibility and under¬

standing. If our mind consisted of nothing but the passive ability to
receive sensations, we should have a manifold multiplicity of intuitions,
but no knowledge of what the manifold contained..." [1968, p. 23]19

17 1 owe T.K. Bikson (in personal communication) an alternative -- and to my mind preferable
— rendition of Kant's dictum: "Concepts without percepts are empty percepts without con¬
cepts are blind".
18 In much the same way, neither the chicken nor the egg, neither nature nor nurture, neither
environment nor behavior, are independent primes. The Western intellectual tradition is
replete with reductionist pseudo-puzzles.
19 Kant makes matters more complicated yet by observing that there must exist a mediating
schema for matching categories of the understanding with sensible appearance, given that
there is no guarantee that the two would be otherwise 'homogenous' — i.e. isomorphic.
This mediating function is presumably part of Kant's apriori synthetic, involving our
space-time grid. Kant's transcendental schema is thus a precondition for matching con¬
cepts with percepts. To quote Kemp once again:
"...Now since there is no homogeneity between category and appearance, and since never¬
theless the appearance has in the end to be subsumed under the category, there must, Kant
says, be some third thing which is homogenous both with the category and with the ap¬
pearance, and which mediates between the two; this third thing, which must clearly be in
one respect intelligible and in another sensible, consists in what Kant calls the transcen¬
dental schema. The key to this is the notion of time..." [1968, p. 30],
For a somewhat compatible view, see Givon (1979a, Ch. 8).
18 Mind, Code and Context

1.4.2.2. Kant’s analytic and apriori-synthetic

First, Kant introduces the concept of analytic knowledge; that is,
knowledge that comes to us by knowing the definitions of terms, the rules
of logic, or the rules of games. Such knowledge is thus in principle
tautological.20 Analytic knowledge is in some sense thus presupposed,
being true by convention.
Second, Kant introduces the notion of apriori-synthetic knowledge;
that is, knowledge of facts that are always true of this world -- but not by
analytic necessity. The latter is Kant's transcendental schema of space
and time. That schema underlies our perceptual/conceptual apparatus
itself as pre-condition to all possible experience of this world. Possible
experience presupposes such a schema, rather than comprises it. Thus,
consider the following passage from the Critique of Pure Reason of Pure
Reason (Smith, 1929):

"...the representation of space must be presupposed. The representation

of space cannot, therefore, be empirically obtained from the relations of
outer appearance. On the contrary, this outer experience is itself possible
at all only through that representation..." [p. 68]

It is possible to argue — as I shall indeed argue later on 21 — that Kant's

analytic and apriori-synthetic are both early precursors of the pragmatic
notion of presupposition, or shared background.

1.4.2.3. The transcendental schema as a relative point-of-view

In his discussion of the apriori-synthetic transcendental schema of
space, Kant observes (Smith, 1929):

"....It is, therefore, only from the human standpoint that we can speak
of space, of extended things etc. If we depart from the subjective condition
under which we alone can have outer intuition...the representation of
space stands for nothing whatsoever..." [p. 71; emphases are mine; TG]

Our point of view vis-a-vis the outside world is thus relative. Again, con¬
sider (Smith, 1929):

20 See discussion of Wittgenstein's Tractatus further below.

21 See further discussion in Chapter 4, below.
 Introduction 19

"...we cannot judge in regard to the intuitions of other thinking beings,

whether they are bound by the same conditions as those which limit our
intuition and which for us are universally valid. If we add to the concept
of the subject of a judgement the limitation under which the judgement is
made, the judgement is then unconditionally valid..." [p. 72]

In other words, a judgement -- knowledge, description -- is only uncon¬

ditionally valid ('fully characterized') if the subject making it and his
point of view -- or context — are taken into account. This is certainly a
clear articulation of a pragmatic, context-dependent theory of descrip¬
tion.

1.4.2.4. The world for us vs. the world for itself

Given the discussion above, it is thus possible to view Kant's caution
about the low likelihood of our ever coming to know 'the thing for itself'
(as distinct from 'the thing for us') not as a mere species of healthy skep¬
ticism, but rather as an early precursor of the pragmatics of gestalt and
point of view. Kant is willing to concede that there must be 'something
out there' to stimulate our sensory perception. For a constructionist, that
is a fairly safe abduction from the mere existence of experience, fine
details notwithstanding. 'The thing for itself', however, remains unavail¬
able to us, given our limited, relative point of view. Thus, in Kemp's dis¬
cussion of Kant's Transcendental Analytic we find:

"...'the most the understanding can achieve a priori is to anticipate the

form of a possible experience in general' (A246 B303). It does not give us
knowledge of things as they are in themselves, but only of things as they
appear to us..." [1968, p. 38]

I think we are entitled to read Kant as suggesting that it is our limited

point of view — the frame -- that determines the form of reality as it ap¬
pears to us. Therefore, to discuss the nature of reality outside that frame
-- or outside any frame (i.e. 'the thing for itself') -- really makes little sense.
Kant thus anticipated the central theme of modern pragmatics. L

22 I am indebted to T. K. Bikson and Fred Kroon (in personal communication; but see also
Kroon, 1981) for pointing out to me how Kant's 'transcendental schema', 'apriori synthetic'
and 'the world for us' indeed presage a pragmatic approach to mind and meaning.
20 Mind, Code and Context

1.4.2.5. Kant the Platonist

There is no evidence in Kant's work to suggest a pragmatic approach
to the formal nature of categories of the mind, i.e. a rejection of
Platonic/Aristotelian discreteness and rigid criteria. For this, one must
await Peirce or, more explicitly, Wittgenstein.

1.5. Peirce and Pragmatism

We turn now to consider, not for the last time, the acknowledged God¬
father of modern pragmatics, Charles Sanders Peirce. His multiple in¬
sights range over major landmarks of the pragmatic agenda, often
without being acknowledged by latter-day practitioners. At this point we
will survey, rather briefly, the main areas of his contribution to prag¬
matics.23 The first two -- abductive inference and semeiotics -- will be
covered in greater detail in Chapters 3 and 7 below, respectively.

1.5.1. Abductive inference

The dichotomy between the two extreme schools of Western epistemol¬
ogy, empiricist and rationalist, was later echoed in the Philosophy of
Science (see Chapter 8, below), in the discussion of modes of knowledge
('modes of inference'). Empiricists traditionally tended to emphasize in¬
duction as the prime mode of knowledge, while rationalists, quite pre¬
dictably, have tended to emphasize deduction. Echoing Kant's rejection
of both reductionist extremes, Peirce resuscitated Aristotle's third mode,
which he rechristened abduction. This mode of hypothesis often invol¬
ves analogical reasoning, and thus the pragmatic, context-dependent no¬
tions of similarity and relevance. As suggested earlier above, neither
analogy nor similarity nor relevance may be characterized by inductive
or dedutive means. All three also involve shades and gradations, again a
characteristic feature of pragmatics. 24

23 Such a short survey is both inadequate and presumptuous, given the enormous range of
subjects Peirce had covered, his voluminous written output, and last but not least the con¬
siderable difficulty one experiences in penetrating his prose. I will thus not be surprised if
in some dark corner of Peirce's writings further invaluable gems of pragmatic insight still
lurk, waiting to be discovered. My own access to Peirce owes much to personal contact
with a number of bona fide Peirceans, most particularly T. K. Bikson, Raimo Anttila, John
Haiman and Michael Shapiro. The latter's book The Sense of Grammar (1983) was particular¬
ly helpful to me in my attempt to understand Peirce's semeiotics.
24 Deductive, inductive and abductive inference are further discussed in Chapters 7 and 8,
below.
 Introduction 21

1.5.2. The interpretant as a third term in semeiotics

Peirce is responsible for introducing the notion of interpretant as the

third term in the semeiotic relation, the term that binds the sign to its
designatum in the minds of the interpreter. As will be suggested in Chap¬
ter 3, below, the interpretant is merely a stand-in for perceived context.
Peirce thus took the non-pragmatic semeiotics that had prevailed since
Aristotle and pointed out to its necessary pragmatic dimension.

1.5.3. Non-Platonic categorization

Peirce's well documented penchant for discrete trichotomies leaves off

a distinct Platonic flavor. Nonetheless, in his explicit treatment of a num¬
ber of sub-areas of pragmatics, he comes down -- in the final analysis —
on the side of non-discrete, context-dependent categorization.
Consider first Peirce's discussion of iconic signs (pictures, diagrams,
metaphors), which hinges upon the concept of 'similarity'. In order for
objects to resemble each other, the categories underlying them must be -
- in principle -- non-Platonic. This is so because of two inherent features
associated with 'resemblance':

(a) It may be infinitely and non-discretely graded; and

(b) it must be contextually mediated.

Implicitly, then, Peirce's semeiotics is founded upon non-Platonic

categorization.
In two other respects, Peirce's discussion of categories goes con¬
siderably beyond Kant's Platonism, often presaging Wittgenstein. We
will discuss the two in order.

1.5.3.1. The multi-dimensionality of semantic space

Peirce considered most natural (thus also linguistic) signs to be a mix¬

ture of three types of signs -- symbol, index, and icon. The three pure
types are merely extreme facets of complex, multi-dimensional signs. Im¬
plicitly, Peirce thus pioneered the notion of 'impure categories',
categories that are distributed along a multi-dimensional space.
22 Mind, Code and Context

1.5.3.2. The non-discreteness of semantic dimensions

Peirce's three types of icons -- picture, diagram, and metaphor -- are
clearly three points along a continuum. Whether we are fully entitled
to interpret Peirce as an explicit non-Platonic categorizer remains to some
degree a matter of conjecture. Buchler's interpretation seems to suggest
contextually-mediate meaning:

"...A term has meaning, in other words, if it is definable by other terms

describing sensible properties. The way in which, generally speaking,
these other terms, which may not be in turn thus defined, acquire mean¬
ing, is by being associated or correlated in their usage, according to empiri¬
cal conventions..." [1939, p. 113; emphases are mine; TG]

The reference to conventions and usage is decidedly Wittgensteinean.

Further:

"...Peirce's emphasis on 'percept' and 'operation' is in effect an em¬

phasis that meaning is something public. We properly explain the mean¬
ing of a term to someone not by eloquently attempting to evoke familiar
images in his mind, but by prescribing how he can gain perceptual ac¬
quaintance with the word denoted. The appeal is not primarily to im¬
agination but to tests that can be undertaken by everybody..." [1939, p.
115].

With sufficient leeway, one may read here a rendition of post-Wit¬

tgenstein Philosophy of Ordinary Language.
Further suggestions that Peirce may have considered categorial space
to be scalar flow naturally from two key components upon which, he
believed, meaning was founded — circumstances and habit. Peirce clear¬
ly considers both to be graded and non-discrete. Thus:

"...Mill says that it means that if in all circumstances attending two

phenomena are the same, they will be alike. But taken strictly this means
absolutely nothing, since no two phenomena ever can happen in cir¬
cumstances precisely alike, nor are two phenomena precisely alike..."
[Peirce, 1940, p. 221]

And further:

"...Habits have grades of strength varying for complete dissociation

to inseparable association. These grades are mixtures of promptitudes of

25 See further discussion in Chapter 3, below.

 Introduction 23

action, say excitability and other ingredients... Habit change oftenconsists

of raising or lowering the strength of habit...." [Peirce, 1940, pp. 277-278]

Consider next Peirce's discussion of vagueness:

"...Logicians have too much neglected the study of vagueness, not

suspecting the important part it plays in mathematical thought...
Wherever degree or any other possibility of continuous variation sub¬
sists, absolute precision is impossible. Much else must be vague, because
no man's interpretation of words is based on exactly the same experience
as any other man's..." [Peirce, 1940, pp. 294-295]

If one reads 'experience' to mean 'context', then a clearly Wittgensteinean

theory of meaning indeed emerges.
Finally, in his discussion of natural categories (such as biological clas¬
sification), Peirce anticipates rather prophetically the current discussion
of prototype semantics (see Chapter 2, below). The crucial ingredient is
his observation on the role of frequency distribution and the clustering
of populations around some normative mean:

"...It may be quite impossible to draw a sharp line of demarkation be¬

tween two classes, although they are real and natural classes in strict
truth. Namely, this will happen when the form about which the in¬
dividuals of one class cluster is not so unlike the form about which the
individuals of another class cluster, but the variations from each middling
form may not precisely agree. In such a case, we may know in regard to
any intermediate form what proportion of the objects of that form has one
purpose and what proportion the other; but unless we have some sup¬
plementary information we cannot tell which one had one purpose and
which one the other..." [Peirce, 1931, pp. 87-88; emphases are mine; TG]

1.6. Ludwig Wittgenstein

Wittgenstein's pragmatic approach to semantic categories will be dis¬

cussed in considerable detail in Chapter 2, below. His flexible, non-dis¬
crete, context-dependent approach to meaning has had an enormous
impact on the cognitive sciences - in particular psychology, anthropol¬
ogy and linguistics.26 In the space below I will merely summarize the
two major points that constitute the core of Wittgenstein's contribution
to the modern pragmatic agenda. One of those has been consistently ig-

26 Paradoxically, after two short decades of intense excitement, the interest in Wittgenstein
within Philosophy seems to have degenerated into, largely, textual exegesis and cultish in¬
sularity.
24 Mind, Code and Context

nored, presumably because of its association with the 'early' Wit¬

tgenstein.

1.6.1. The rejection of deductive logic as a means for transacting

new information
In his Tractatus (1918), a book often -- and to my mind mistakenly --
lumped together with Logical Positivism, Wittgenstein observes that the
propositions of deductive logic can be all reduced to either tautologies or
contradictions. Neither, he points out, could possibly serve to transact
new knowledge. In rejecting deductive logic as possible model for com¬
munication, Wittgenstein—probably unintentionally— came tantalizingly
close to defining two major parameters of the pragmatics of communica¬
tion, both anathemas to formal logic:

(a) The use of Language in communication is not a matter of truth

or falsity of atomic propositions, but rather a process of trans¬
ferring information from one mind to another.

(b) In communication, totally old (thus tautological to the

receiver) information and totally new (thus incoherent to the
receiver) information are equally useless. This is so because:

(i) Information that is totally old is totally redundant, it offers

no motivation for the receiver to participate in the com¬
municative transaction; and

(ii) Information that is totally new is totally incoherent, it of¬

fers no overlap with the receiver's existing knowledge, it
thus cannot be integrated with it.

In order for information to be processed by some receiver — i.e. integrated

coherently into the receiver's pre-existing knowledge — it must lie some¬
where between these two extreme poles. Implicitly thus, the crucial role
of the receiver in information processing, and thus of the receiver's back¬
ground knowledge, which forms an important part of the transmitter's
discourse context, is anticipated in the 'early' Wittgenstein. We will
return to discuss this subject in Chapter 4, below.
One may argue, I think legitimately, that deductive logic, in setting up
the notions of tautology and contradiction, has already made an in¬
cipient, covert move toward the pragmatics of context. This is so because
tautology and contradiction are not properties of an atomic proposition,
but rather of a proposition vis-a-vis another proposition. That other
proposition may be considered the context. In the embryonic discourse
 Introduction 25

called 'logical proof', the premise(s) is equivalent to our pragmatic no¬

tion of context. Deductive logic has thus been dealing with context im¬
plicitly -- surreptitiously -- all along. But the 'logical' context has always
remained rigidly constrained, or closed. In the pragmatics of natural lan¬
guage, and thus of mind, context is, forever open ended.

1.6.2. The attack on Platonic categorization

Wittgenstein's contribution to pragmatics is of course much more in¬
timately associated with his Philosophical Investigations (1953). Here one
may observe three interlocking themes:

(a) The context-dependence of meaning;

(b) The flexibility, non-discreteness and scalarity of semantic

dimensions; and

(c) The 'family resemblance' -- thus non-criterial - nature of

semantic categories.

We will return to discuss these three themes in considerable detail in

Chapter 2, below.

1.7. Other strands of pragmatics

The pragmatic agenda in the second half of the 20th Century reminds
one, in retrospect, of a dam burst. Many of the streams feeding into the
swollen river, to borrow Peirce's metaphor, had been meandering, slug¬
gishly and unobtrusively, for a long time. They had accumulated, slow¬
ly, leisurely, behind the dam of deductive logic. Somewhere along the
way, the flow began to quicken, the tempo picked up. And all of a sud¬
den the entire dam is threatened. In the space below I will briefly recount
what seem to me to be some of the more obvious components of this prag¬
matic dam burst.

1.7.1. The language-and-culture tradition

With illustrious antecedents in philosophy, such as Kant, Peirce and
Wittgenstein, it is easy to forget that other cognitive disciplines - in their
own way and within their own peculiar methodologies and preoccupa¬
tions — have also participated in the growth of modern pragmatics. At
the intersection of linguistics and anthropology stands one such contribu¬
tion, arising from the comparative study of cultures. The shared
26 Mind, Code and Context

('generic') cultural world view indeed constitutes the bulk of the back¬
ground knowledge -- context — vis-a-vis which inter-personal behavior
and communication take place.
The balance between universality and specificity of human mind and
human language easily translates into the question of how universal -- or
how group-specific -- human culture is. Aristotle had already occupied
himself with this issue, setting up the agenda for later work on univer-
sals.27 , .

But the bulk of substantive contributions came from pioneer cross-cul¬

tural investigators, such as Whitney, von Humboldt, Boaz, Kroeber,
Malinowski, Sapir and Whorf. The profound functionalism that per¬
meated the work of these illustrious scholars is already, by definition, a
brand of pragmatics (see section 1.7.2., below). Their consciousness of the
intertwined nature of the language-and-culture complex enriched our
understanding of culture as the ever-present context. Finally, their inves¬
tigation into the balance between linguistic-cultural universality and
specificity defined the agenda for a yet ongoing debate. We will return to
discuss these issues in Chapter 9, below.

1.7.2. Early functionalism: The communicative context

The description of any structured instrument can be approached from

two distinct points of view. It may be described purely in terms of its
structure. In biology, this would amount to describing the anatomy of the
various organs. Alternatively, the mere structural description of an in¬
strument may be supplemented by a description of its correlated func¬
tion. In biology, this would amount to describing the physiology of the
organs. In biology, purely structural description, without seeking corre¬
lated function(s), has not been seriously entertained since Aristotle. In¬
deed, what would be the point? Likewise in the early history of
linguistics, an implicit functionalism prevailed, beginning with the
semantically motivated categories of Aristotle and the medieval Modis-
tae, continuing with the implicit cognitive assumptions of the Cartesians,

27 In his De Interpretatione, Aristotle makes the following brief observation concerning the
non-universal nature of the relation between language and mind, as compared with the
universal relation between mind and world:
"...Now spoken sounds are symbols of affectations of the soul, and written marks sym¬
bols of spoken sounds. And just as written marks are not the same for all men, neither are
spoken sounds. But what these are in the first place signs of — affectations of the soul — are
the same for all; and what these affectations are likenesses of - actual things - are also the
same..." [Ackrill, tr. & ed., 1963, p. 43],
 Introduction 27

and following with the mentalism of late 19th Century German Roman¬
tics such as von Humboldt and Paul. Even the more structure-conscious
linguistic traditions, such as the 19th Century's Neo-Grammarians, or the
so-called 'Traditional Grammarians' such as Jespersen,28 were profound¬
ly functionalist in their interpretation of linguistic structure. Only in the
early 20th century, under the impact of behaviorism — that extreme brand
of empiricism — did explicit, unvarnished structuralism emerge upon the
linguistic scene, in the forceful person of Leonard Bloomfield.29
Bloomfield's extreme position, closely bound to his naive, dogmatic
empiricism, proclaimed that the instrument -- language — may be
described without reference to its function -- meaning or communicative
use. It is indeed curious that the other modern school of linguistic struc¬
turalism, Generative Grammar, somehow contrives to trace its
philosophical roots to the philosophical rationalism of Descartes.
The work of all 20th Century functionalist schools, from Jespersen and
Sapir onward,31 falls squarely within the scope of pragmatics, insofar as
they have all insisted that the structure of language can only be under¬
stood and explained (rather than merely described) in the context of its
communicative use, i.e. its function. We will return to discuss some of
these issues in Chapters 7 and 10, below.

1.7.3. The Speech-act tradition: Purposive context

The Tate' Wittgenstein (1953) left behind as legacy the distinct British
school of Ordinary Language Philosophy. A sub-tradition of that school,
clustering around 'performative', 'speech-act' or 'illocutionary force'
analysis, has contributed massively to the recent pragmatic agenda,
beginning with Austin (1962) and continuing with Searle (1970), Grice
(1968/1975) and a veritable avalanche that came in their wake. This sub-
tradition occupies itself primarily with one important aspect of context -
- the goal or purpose of verbal acts. Still, many works within this tradition
branched out into the study of other important themes in pragmatics.
We will return to this topic in more detail in Chapter 4, below.

28 See Jespersen's Philosophy of Grammar (1924).

29 See Bloomfield's Language (1933), as well as further discussion in Chapter 8, below.
30 This tracing back of antecedence was accomplished somewhat post-hoc in Chomsky (1966).
Chomsky trained in linguistics under Z. Harris, a leading Bloomfieldian structurahst. His
training in formal logic was also within a purely structuralist tradition, which his early
book. Syntactic Structures (1957), clearly reflects.
31 One may list here at least the Prague School, Kenneth Pike, Dwight Bolinger and Michael
Halliday.
32 See in particular Cole (ed., 1981).
28 Mind, Code and Context

1.7.4. The presupposition tradition: Shared knowledge

Formal logic, from its very inception a non-pragmatic, atomistic, con¬

text-free mode of semantic analysis, has been forced to gradually accom¬
modate the pragmatic facts of life, as it wrestled with the complexities of
natural language interpretation. The concept of presupposition, itself at
least as old as Kant, was pressed into service to account for a range of
context-related phenomena. The need, for this arose from what was seen
initially as a -- slight but disturbing — 'distortion' in the interpretation of
propositions of natural language. Some propositions -- or sometimes por¬
tions within them — seemed to not be 'asserted' as true or false. Rather,
they seemed to be 'presupposed', as preconditions for the meaningful¬
ness (in logic truth or falsity) of other propositions.
The intuitive connection between 'logical' presupposition and the
pragmatics of discourse context, communicative intent and speech act is
of course obvious. Formal logicians from Strawson (1950) onward,
however, have fought valiantly to constrain presupposition as a proper¬
ty of atomic propositions. They thus attempted to gloss over the obvious
ontology of presupposition in the speaker's communicative intent, and
in his/her context-bound assumptions about the hearer's communica¬
tive intent. This purging of speaker and hearer from the description of
meaning is only a natural consequence of Russell's (1919) injunctions. It
was transferred into linguistics beginning with Keenan (1969), thereby
precipitating a litany of presuppositional literature in the 1970s and
1980's.33 The main thrust of that literature has been, to this day, an at¬
tempt to sub-divide context on its multifarious shades and varieties into
neat, discrete categories. The entire package is then to be grafted into a
truth-conditional logic.34In this way, from having begun initially as an
important opening toward pragmatics and the study of language in con¬
text, presuppositional analysis soon degenerated into a rear-guard action,
purporting to reinforce the claim of a non-pragmatic instrument — deduc¬
tive logic -- on the semantic analysis of natural language. We will return
to discuss these issues in Chapters 4 and 6, below.

33 See e.g. Fillmore (1971), Horn (1972), Karttunen (1974), inter alia. The seminal paper which
triggered the linguistic world's interest in presupposition was actually rather detached
from logic, Kiparski and Kiparski (1968).
34 See e.g. Keenan (1971), Gazdar (1979), Oh and Dinneen (eds, 1979), or Sperber and Wilson
(1985), inter alia. The fatal attraction of the siren song of deductive logic to otherwise sober
linguists, in the face of recurrent frustration and a veritable mountain of recalcitrant data,
remains a great mystery.
 Introduction 29

1.7.5. Modal logic and possible worlds: The indexing of context

Modal logic began as a rather modest attempt to deal formally with
non-fact propositional modalities such as 'possible' and 'future', with
'intended states' and 'tense', and most particularly with the problem of
reference under the scope of such modalities.35 Soon, however, possible
worlds logic was pressed into service to deal wholesale with the problem
of context. Consider, for example, the following lines from Montague
(1970):

"...To interpret a pragmatic language L we must specify several things.

In the first place, we must determine the set of all possible contexts of use
— or rather, of all complexes of relevant aspects of possible contexts of use;
we may call such complexes indices, or to borrow Dana Scott's term, points
of reference..." [p. 144; first two emphases are mine; TG]

As possible candidates for 'indices' only the indexical 'Y is explicitly men¬
tioned, although in outlining the problem of context earlier, Montague
refers to indexical expressions in general (i.e. person, time, place). The in¬
tractability of this list-approach, and the open endedness of 'possible
contexts' and 'relevant aspects', somehow never seemed to daunt the late
Montague. In his wake, others began to move in the obvious direction
and enlarge the list of indices. Lewis (1972), for example listed eight types
of context; so that a fully specified 'index' is now:

"...any octuple of which the first coordinate is a possible world, the

second coordinate is a moment of time, the third coordinate is a place, the
fourth coordinate is a set of persons (or other creatures capable of being
a speaker), the fifth coordinate is a set of person (or other creatures capable
of being an audience), the sixth coordinate is a set (possibly empty) of con¬
crete things capable of being pointed at, the seventh coordinate is a segment
of discourse, and the eighth coordinate an infinite sequence of things..."
[1972, p. 176; emphases are mine; TG]

The considerable futility of such an enterprise -- the instability and open-

endedness of each 'indexed' element of the context -- was glossed over at
the time. In a later work, Lewis (1979) acknowledges the problem more
explicitly, with detailed illustrations from a more realistic sample ap-

35 See e.g. Kripke (1963, 1972), Cocchiarella (1965), Hinttika (1967), Purtill (1968) or Scott
(1970), inter alia.
30 Mind, Code and Context

proximating natural conversation. Similar misgivings were expressed

earlier by Creswell (1972):

"...Writers who, like David Lewis[ll], do try to give a bit more body to
these notions, talk about time, place, speakers, hearers,...etc. and then go
through agonies of conscience in trying to decide whether they have
taken account of enough..." [p. 8] 36
» • »

The logicians, again in Russell's well-motivated tradition, thus attempt

to sanitize, constrain -- and sweep under the neat rug of a closed deduc¬
tive system -- the intractable problem of context.

1.7.6. Ethnography of Speech: Social interaction as context

An important strand in modern pragmatics comes to us from the in¬

terface of sociology, anthropology and linguistics, via the works of Labov,
Gumperz, Goffman, Sachs and Schegloff, Geerz, Garfinkel, Brown and
Levinson, and many others, under various labels.37 These works probe
into the social, interactional and inter-personal context of communication
and the face-to-face speech situation. Notions such as power, status
gradients, politeness and deference, and control of the floor during com¬
munication (i.e. 'turn-taking') have been studied and elaborated. Many
of those intersect rather naturally with the earlier pragmatic tradition of
speech-act analysis. We will return to discuss a number of these topics in
Chapter 4, below.

1.7.7. Developmental pragmatics

An important recent addition to the pragmatic agenda is the com¬

municatively oriented study of child language acquisition.38 An earlier
rationalist-structuralist dogma39 had held that the human language
capacity is largely innate, and that therefore the role of linguistic input
-- and thus of the communicative interaction and context -- in the acquisi-

36 Cresswell (1972) goes on to suggest — tongue in cheek -- that in interpreting the bar-room
poem line 'Just fetch your Jim another quart' one might perhaps need to index "...a 'previous
drinks' context...”.
37 Sub-strands of this tradition have been called at one time or another 'Ethnolinguistics'
(Gumperz, 1977,1982), 'Conversational Analysis' (Sacks, Schegloff and Jefferson, 1974;
Brown and Levinson, 1978, 1979), 'Ethnomethodology' (Goffman, 1974, 1976; Garfinkel,
1972; Geertz, 1972), or 'Sociolinguistics' (Labov, 1972a, 1972b; Ochs, 1979a).
38 See for example Scollon (1974), Bates (1976), Keenan and Schieffelin (1976), Ochs (1979b),
Ochs and Schieffelin (eds, 1979), inter alia.
39 As in, e.g., Chomsky (1968) and the long list of structuralist acquisition studies inspired
thereby.
 Introduction 31

tion of language is minimal. Developmental pragmatics has evolved as

an empirically-motivated reaction to that early dogma, endeavoring to
study the acquisition of language in its natural, functional, communica¬
tive context.

1.7.8. Pragmatics and machines: Plans, Goals, Scripts

Within Artificial Intelligence (AI) there emerged a small but persistent

pragmatic tradition, beginning with Winograd's (1970) attempt to incor¬
porate speech-act notions into the algorithm, and continuing more recent¬
ly with the scripts-plans-and-goals approach of Schank and Abelson
(1977) or Levy (1979), inter alia. Since in principle pragmatics cannot be
constrained within a closed algorithm without ceasing to be pragmatics,
the various pragmatic moves within the AI community represent at least
a tacit admission of the limits of the deductive algorithm as model for
natural language, communication and behavior.

1.7.9. Gestalt, prototypes and metaphors

One must also note that pragmatics has had a vigorous presence
within post-Behaviorist cognitive psychology, in the areas of gestalt and
cognitive priming (see discussion in Chapter 3, below), and similarly in
the psychology of perception (Attneave, 1959). More recently, a conver¬
gence has taken place between several disciplines looking at the nature
of human categorization and metaphoric meaning, spanning Goguen's
(1969) work on fuzzy sets logic, Rosch's (1973a, 1973b) work on natural
categories in cognitive psychology, Ortonyi's (ed., 1979) work in the
psychology of communication, Lakoff and Johnson's (1980) linguistic
work on metaphor, and Johnson's (ed., 1981) work on metaphor in
philosophy. This inter-disciplinary convergence, ascribed by some to
'late' Wittgensteinean influence, will be discussed in considerable detail
in Chapter 2, below.

1.7.10. Biology and evolution

In the history of the study of the diversity of life forms, a certain paral¬
lel can be seen with the history of Western epistemology, with the balance
32 Mind, Code and Context

shifting periodically between non-pragmatic and pragmatic schools of

thought. Thus, creationist pre-evolutionary schools considered all extant
biological species to be rigidly fixed -- much like Platonic/Aristotelian
categories. In contrast, a dynamic, evolutionary, flux-like view of species
is by itself already crypto-pragmatic.
Within the evolutionary camp itself, innatist non-pragmatic schools
first predominated, expounding upon the wonders of pre-ordained
speciation. The agent of such pre-ordination was the Deity itself, which
mandated life to evolve toward some spiritual apex, and for the glory of
God. (In the pinch, more recently, the secular deity of molecular genetics
can also be pressed into the service of pre-determinism). An early version
of this approach can be seen in Charles Bonnet's preformation,40 which
proposed to view all development as already present in the embryo (for
ontogenesis), or in the earlier species (for phylogenesis). In somewhat the
same vein, the late 18th Century German school of Naturphilosophie en¬
visioned a universal, transcendental order governing evolution toward
aesthetically and ethically higher peaks.41
A true, potentially pragmatic transformation eventually took place,
first with Lamarck underlying interactive functionalism, later with
Darwin's recognition of the role of the environment -- i.e., context — in
shaping the evolution of organisms. Here once again, one soon observes
a familiar philosophical split. On the one hand there is what Gould
(1980) characterizes as the extreme Neo-Darwinian approach, according
to which the role of environmental selection is fully deterministic, with
the organism's purposive behavior having no role. This school is rather
reminiscent of extreme philosophical empiricism. At the other extreme
one finds naive Lamarckians,42 according to whom the organism's pur¬
posive behavior is paramount in directing its own evolution. This school
may be likened to extreme philosophical rationalism. As Gould (1980)
points out, neither extreme is likely to be tenable. A more realistic ap¬
proach is likely to be, once again, an interactionist -- thus pragmatic —
middle-ground. We will return to explore these issue in Chapter 10,
below.

40 See discussion of Bonnet (1764) in Gould (1977, Ch. 2).

41 With august literary figures such as Schelling and Goethe, but also eminent biologists such
as Kielmeyer and Meckel; for discussion again see Gould (1977, ch. 2) as well as Chapter
10, below.
42 Gould (1980) mentions in particular Koestler (1972).
 Introduction 33

1.8. Projections: Toward an integrated, pragmatics of

organism, mind and behavior

In this book I will attempt to emulate pragmatics itself in one crucial

respect: I will let the book remain unabashedly open-ended and incom¬
plete. This flows rather naturally from my conviction that many of
pragmatics' most profound ramifications have yet to be discovered, un¬
derstood and articulated. It is, in addition, my suspicion that pragmatics
-- as a method -- may hold the key to an integrated understanding of be¬
havior and cognition, thus ultimately to an understanding of the evolu¬
tion of biological organisms. Within this general context, pragmatics also
hold the key to our understanding of language and mind. To the extent
that so-called Cognitive Science is ever to become more than a political
catch phrase, a passing fad, I suspect that pragmatics also holds the key
to this yet-to-be unified field. Lastly, to the extent that the humanities
have a realistic hope of becoming amenable to systematic investigation
by some science-like method, pragmatics may hold the key to that as
well.43

43 Feyerabend's (1975) wholesale rejection of all method, essentially on the grounds that no
method by itself is 100% effective, is an unnecessarily extremist construction of science, in
that it implicitly concedes the reductionists' passion for all-or-nothing solutions. In con¬
trast, a pragmatic middle-ground of methodological pluralism has been taken for granted
in the more 'mess/ sciences such as biology and cognitive psychology (and, if Hanson
(1958) is to be believed, also in physics). Indeed, earlier on, Feyerabend (1970) himself had
advocated such pluralism. By 1975, Feyerabend had apparently opted for throwing the
baby out with the bathwater, presumably on the theory that if one cannot have complete
order one might as well embrace total chaos. If cognizing and behaving organisms were
to adopt the same methodological purism, their path to swift evolutionary extinction
would be assured. These topics will be discussed at considerable length in chapters 7 and
8, below.
 '

'

> ..
CHAPTER 2

CATEGORIES AND PROTOTYPES:

BETWEEN PLATO AND
WITTGENSTEIN

2.1. Discrete categories, fuzzy gradations and prototypes*

"There is a mysterious wisdom by which phenomena among

themselves disparate can be called analogous names, just as
divine things can be designated by terrestrial terms, and through
equivocal symbols God can be called lion or leopard; and death
can be called sword; joy, flame; flame, death; death, abyss; abyss,
perdition; perdition, raving; and raving, passion...The more it is a
dissimilar similitude and not literal, the more a metaphor reveals
its truth..."

Umberto Eco,
The Name of the Rose (1980, p. 248)

2.1.1. Preamble
In the preceding chapter we surveyed the philosophical background
upon which our discussion of mental categories can now proceed. From
the dawn of Western epistemology, we noted, three traditional themes
pervaded the discussion of mental categories:

(a) The source of categories; that is, whether they are innate
('conceptual', as rationalists would have it), or external
('perceptual', as empiricists would have it).

*1 am indebted to Fr. Garth Hallett SJ, Michael Posner and John Verhaar for helpful comments
and suggestions. An early precursor of this chapter was presented at the Symposium on
Categorization and Noun Classification (Eugene, Oregon, September 1983), and benefited
greatly from the discussion there, in particular from comments by George Lakoff and Pete
Becker.
36 Mind, Code and Context

(b) The designatum of mental categories; that is, what they stand for in
the semeiotic relation.
(c) The formal nature of mental categories; i.e. whether they are discrete
and absolute, or graded and context dependent.

Question (a) will not concern us here. Question (b) will be discussed in
considerable detail in Chapter 3, below. In this chapter we will concern
ourselves primarily with question (c). >

2.1.2. Extreme Platonic categorization

As noted earlier,1 both rationalists and empiricists, from Plato and

Aristotle down, have subscribed to the view that mental categories are
discrete and absolute. As we have also noted, both Plato (posing as
Socrates) and Aristotle had their difficulties, or moments of pragmatist's
doubt, concerning the absoluteness of categories. For brevity's sake, I will
refer to the absolutist approach to categories from now on as the Platonic
tradition. According to this tradition, membership in a category depends
on the unambiguous possession or non-possession of some criterial
properties. Such properties must be both necessary and sufficient in
order to determine membership in a category. The possession or non-pos¬
session of a property is not a matter of degree or equivocation. The law
of the excluded middle is strictly adhered to. Vagueness and indecision
may invade the mental system due to insufficient information, but in
principle, gradation and ambiguity have no theoretical status in this
system. Such a system may be illustrated diagrammatically as:

(1) Diagram of extreme Platonic categorization

1 See Chapter 1, section 1.4.1.

 Categories and Prototypes 37

In (1) above, A is the criterial property determining membership in the

category. Individuals in sub-category B possess that property, and are
thus members of A. Individuals in sub-category C do not possess that
criterial property and are not members of A.

2.1.3. Extreme Wittgensteinean categorization

A second approach to categories, diametrically opposed to the

Platonic, will be tentatively labeled as the extreme Wittgensteinean, after
the 'late' Wittgenstein (1953). The question of whether Wittgenstein was
indeed an extremist proponent of this approach will be discussed further
below. According to this approach, categories are neither absolute nor
discrete. To begin with, they are relative, or contingent upon context (use,
purpose, point of view, general schema, etc.). Second, they are fuzzy
edged, and the space defining them is a continuum. Category member¬
ship maybe a matter of degree. Members of a category may relate to each
other through family resemblance -- a non-logical, pragmatic notion.
Member a may thus resemble member b, b may resemble c, c may
resemble d, etc. But a and z may not resemble each other at all -- and still
belong to the same category. This second extreme type of categorization
may be illustrated diagrammatically as:

(2) Diagram of extreme Wittgensteinean categorization

In (2) above, sub-categories a,b,c,d, are equally members of a single

meta-category, even though the intersection of shared properties (or
'similarities') between each adjacent pair is not the same as for the other
pairs.
38 Mind, Code and Context

2.1.4. The prototype approach to categorization

It is important to recognize that both extreme approaches to
categorization, represent important aspects of category formation, in cog¬
nition, language and behavior. There is indeed a certain measure of
categorial discreteness in our perceptual, conceptual and linguistic or¬
ganization. At the sign ('code') level, lexical items, morphemes, discrete
syntactic structures and even some of the rules governing behavior and
communication, all constitute -- up to a point -- a huge reservoir of prima
facie evidence for the existence of such discreteness. At the same time, the
very same phenomena in cognition, language and behavior, when
studied closely and with sufficient care, furnish a large body of evidence
in support of non-discreteness, fuzzy edgedness, scalarity and context
relativity of categories and of the rules that govern their application. If
one is then to be empirically responsible and philosophically honest, one
could not subscribe to either extreme position in attempting to under¬
stand mind, language and behavior. Rather, one must recognize a middle
ground, a hybrid solution, the pragmatic alternative.
The story of how the pragmatic alternative to the extremist approaches
to categorization -- prototype theory -- came into being, is a happy case
of convergence in Cognitive Science. As is often the case in empirical
work, the philosophical lineage of this convergence may have been
adopted largely post hoc, and it is not clear whether this adoptive lineage
indeed played a significant role initially. The inception point of prototype
categorization may be traced to experimental psychology, via Posner and
Keele's (1968) work on the genesis of abstract ideas. Posner (1986) further
credits, in retrospect, the works of Attneave (1959) in the psychology of
perception, as well as Bransford and Franks (1971).2 The thread was then
picked up by Rosch (1973a, 1973b, 1975),3 who was also responsible for
suggesting the aptness of Wittgenstein's 'family resemblance' metaphor.
Quite independently, Berlin and Kay's (1969) study of folk taxonomies
and the cross-language distribution of color terms 4 brought Anthropol¬
ogy into the arena. And linguists soon joined the fray with Ross's work
on squishes (1972,1973,1974) and Lakoff's on fuzzy concepts and hedges
(1973, 1977,1982).5

2 Posner (in personal communication) also points out that prototypes came into psychology
from the notion of 'schema', via Kant, Head and Bartlett; then through Galton's idea of
composite photographs representing different but related people. The notion of 'family
resemblance' was discussed by Price in his Thinking and Experience, probably independently
of Wittgenstein.
3 See also Rosch and Lloyd (eds, 1978).
4 See also Kay and McDaniel (1978) and Coleman and Kay (1981), inter alia.
5 See also Lakoff and Johnson (1980), inter alia.
 Categories and Prototypes 39

Like the extreme Wittgensteinean approach, prototype semantics

recognizes non-discrete categorial space both within and between
categories.6 Like Wittgenstein, again, it observes that natural mental
categories are seldom defined in terms of one — or even few — necessary
and sufficient criterial properties. Rather, natural categories are formed
at the intersection of several 'characteristic', 'typical', 'normative' fea¬
tures. Such intersections, where significant clustering of individual mem¬
bers of the category's population is to be found, tend to be relatively stable
and replicable, but not absolutely so. We may represent this third type of
categorization, prototype clustering, by the following intersection
diagram:

(3) Diagram of prototype-clustering categorization

The shaded area in (3) represents the portion of the categorial space where
individual members display the largest number of characteristic fea¬
tures. Individuals occupying that portion of the categorial space are the
most typical members, the normative exemplars of the category, its
prototype. Now, in other areas of the categorial space, fewer — though

6 The expression 'within as well as between categories' can be, of course, translated as 'at
various meta-levels of the categorial hierarchy'.
40 Mind, Code and Context

still 'enough' - characteristic features intersect, say 3 out of 4. Such areas

are still populated by many fairly typical individuals, ones that are fairly
close to the category's prototype core. Certainly, individuals occupying
those areas have stronger claim to categorial membership than those who
occupy further outlying zones (say, at the intersection of only two proper¬
ties, or one, or none). Membership is thus a matter of degree.
The characteristic gradation found in natural, prototype-like
categories may be ascribed to two distinct sources:7

(a) Members may be ranked as to how many characteristic features they

possess.
(b) The features themselves may be ranked according to their relative
importance for defining the prototype.

Neither (a) nor (b) above can be determined by deductive or inductive

means. Both involve context-dependent abductive judgement. By com¬
bining the gradation arising from both sources, one may derive an overall
measure of degree of proto typicality, or distance from the prototype core,
for each individual member of the category.

2.1.5. Categorization and frequency distribution

The way we have described proto type-like categorization thus far does
not seem to diverge sufficiently, in some crucial respects, from our
description of extreme Platonic categorization. Thus, in our diagram (3)
above, individual members of the class either do or do not possess four,
three, two, one or none of the features that define the class's prototype.
What is missing is of course a crucial element of natural categorization,
the element of frequency distribution. Put another way, nothing in the
Venn diagram (3) guarantees that a substantial maj ority of the category's
population will be prototypical, i.e. will display the largest number of
the most important features. The notions of norm and counter-norm, so
crucial in the phenomenology of natural categories, are completely miss¬
ing. Indeed, Venn diagrams are, in principle, a deductive instrument. On

7 It is easy to show, of course, that the two factors are not really distinct from each other, but
rather represent the application of the very same principle at two different meta-levels.
This is so because in order to create gradation of types (b), in the rank-ordering of features
according to their importance, one must resort to classifying features by their sub-features,
how many of those each feature possesses, and of what importance. We have already noted
(see Ch. 1) that in principle it is impossible to distinguish between a difference in kind and
a difference in degree. Rather, it is the pragmatic context that makes possible such a
distinction, whereby some features are considered 'more important' for the purpose at
hand.
 Categories and Prototypes 41

the other hand, natural categorization, at least at some level, requires an

added inductive element.
In order to demonstrate how the prototype approach is indeed a real
alternative to both the Platonic and Wittgensteinean extremes, one must
note that a natural category, be it biological, behavioral, cognitive or
linguistic, is characterized by the degree of clustering of its membership
around the categorial mean, or prototype core. It requires charac¬
terization in terms of the distribution of its population. To illustrate this,
let us return to the Platonic extreme; there, 100% of a category's popula¬
tion must distribute at exactly the same point of the categorial space, as
in:

(4) Frequency distribution of a population exhibiting

extreme Platonic categorization

% of members
within sub-
segments of the
category space

members of members of
category A category B

In contrast, the extreme Wittgensteinean approach to categorization

would predict a uniform distribution of the population along the
categorial space. This remains true regardless of whether individual
members exhibit the possessed property to an equal or unequal degree.
This type of categorization is illustrated in:
42 Mind, Code and Context

(5) Frequency distribution of a population exhibiting

extreme Wittgensteinean categorization

% of members
within sub-
segments of the
category space

categorial space

One may now contrast both extremes with prototype-like categoriza¬

tion, whereby a certain majority of the population clusters somewhere
near the categorial mean. Both 'majority' and 'near' are relative, context-
dependent, terms. Their actual absolute values depend on the saliency
relation of the category vis-a-vis its purposive pragmatic context. Such
absolute values cannot be assigned deductively or inductively. This type
of categorization may be illustrated as:
(6) Frequency distribution of a population exhibiting
prototype-like categorization

% of members
within sub- B
segments of the ✓ %
category space \ / \
\ / \
\ /
\
\ /
\
\ /
\
\ /
\
_L-\_
A \
categorial u
prototype
mean for ^
space V prototype
mean for
category A category B

A number of details concerning this clustering distribution of natural

populations around the categorial mean - or prototype core -- must remain
open. Abell-shaped distribution, whereby 2/3 of the population distributes
within one standard deviation from the mean, is only one possible distribu¬
tion. There are some grounds for suspecting that many - perhaps most -
biological sub-systems, in neurology, perception, cognition, language or
culture, require a sharper distribution curve, with a larger portion of the
 Categories and Prototypes 43

categorial population clustering closer to the mean. Such tighter clustering

produces more salient categorial peaks. If this is indeed the case, it must
surely be motivated by functional-adaptive considerations relative to each
biological sub-domain. But whatever the functional context, the categorial
peaks characteristic of prototype-like distribution must be distinct enough
for the organism to associate a substantial majority of the category's popula¬
tion with its prototype.8 Again, 'distinct enough' and 'substantial majority'
cannot, in principle, be determined by deductive means.

2.2. Ludwig Wittgenstein and categorization

2.2.1. Preamble

As noted above, Rosch (1973a) adopted Wittgenstein's family

resemblance metaphor in her work on natural categories and prototypes.
However post-hoc this adoption may have been, it was far from accidental.
Indeed, Wittgenstein's attack on Platonic categorization in his Philosophical
Investigations (1953) had an exceptional trail-blazing impact on the modern
pragmatic approach to categorization. This is particularly true because
Peirce's earlier work on the subject has, until recently, gone almost entirely
unnoticed, and is still ignored by the philosophical main stream.9

8 For an illuminating discussion of these issues, correlating the neuro-physiology of color

perception and the semantic distribution of color terms in human languages, see Kay and
McDaniel (1978). For an extensive review of the role of context, purpose and saliency in
the construction and retention of classificatory schemata, see Medin and Schaffer (1978).
The 'hybrid' prototype approach to categorization is defined there in terms that accom¬
modate both 'analytic' and 'analogical' learning, i.e. both the Platonic and Wittgensteinean
modes, respectively.
9 There is no evidence in the Investigations that Wittgenstein was at all aware of Peirce's work
(Garth Hallett, in personal communication; but see also Hallett, 1977). Wittgenstein's
attitude to antecedence is of course well documented: "...I do not wish to judge how far my
efforts coincide with those of other philosophers. Indeed, what I have written here makes
no claim to novelty in detail, and the reason why I give no sources is that it is a matter of
indifference to me whether the thoughts that I have had have been anticipated by someone
else..." (Tractatus Logico-Philosophicus, 1918, p. 3).
44 Mind, Code and Context

Whether one could also read a full-fledged prototype theory into

Wittgenstein's work remains to some extent a moot question, given
Wittgenstein's undeniable impact on cognitive psychology and linguis¬
tics. One must note, however, that it was in cognitive psychology, with
its traditional -- largely methodological -- preoccupation with popula¬
tions and frequency distribution, that an explicit prototype theory of
categories arose. Wittgenstein's impact on Anthropology has been, on the
other hand, far from unambiguous. His vigorous denial of Platonism has
been often used in Anthropology to mean that if generalization cannot be
absolute, no real generalizations exist.10
Like the other two Godfathers of modern pragmatics, Kant and Peirce,
Wittgenstein's prose style is not always the epitome of clarity. As in the
case of Kant and Peirce, his richly metaphoric style has over the years
given rise to a veritable cottage industry of textual interpretation. In the
following three sections I will discuss what seem to be Wittgenstein's
three major themes in his attack on Platonic categorization.

2.2.2. The context-dependence of meaning

Technically speaking, one could easily conceive of an interpretation of

context-dependent semantics that would be -- to all intent and purpose
-- fully compatible with Platonic categorization. This would indeed be
the case if one depicted the contexts themselves as absolute, discrete and
sharply bounded. We could thus Platonically categorize all possible
contexts; and as a consequence, if the frame is Platonically stable and
discrete, the picture framed in it would be equally stable and discrete. We
have already noted earlier why this solution to the problem of context is
impractical, given the central predicaments of pragmatics. One may note
that this solution has also been rejected historically. Both Plato and
Aristotle, as noted earlier, refused to contend with the embarrassing
messiness and open-endedness of real world context. Because of this

10 One must happily acknowledge, however, that at least some anthropologists who were
profoundly touched by Wittgenstein's relativistic, context-dependent theory of meaning,
nonetheless recognized the need to retain some measure of categoriality, both in our
description of the subject matter and in our methodology. In referring to methodology, for
example, C. Geertz writes: "...The besetting sin of interpretative approaches to anything --
literature, dreams, symptoms, culture — is that they tend to resist, or to be permitted to
resist, conceptual articulation and thus to escape systematic modes of assessment. You
either grasp an interpretation or you do not, see the point of it or you do not, accept it or
you do not. Imprisoned in the immediacy of its own detail, it is presented as self-validat¬
ing... Any attempt to cast what it says in terms other than its own is regarded as a travesty
— as the anthropologist's severest term of moral abuse, ethnocentric..." (The Interpretation
of Cultures, 1973, p. 24). And in referring to the subject matter: "...To abandon hope of finding
the "logic" of cultural organization in some Pythagorean "realm of meaning" is not to
abandon the hope of finding it at all. It is to turn our attention toward that which gives
symbols their life: their use..." (ibid., p. 405).
 Categories and Prototypes 45

messiness and open-endedness of real world context. Because of this

messiness, they idealized their categories (in Aristotle's case, the 'forms')
and made them absolute and context free.
In modern linguistics, one may note striking parallels to the way
Aristotle dealt with the predicament of pragmatics, i.e. by postulating a
double system — the forms and the synolon. One finds this, for example,
in Saussure's distinction between langue, the idealized, Platonic, under-
lying system of the language, and parole, the messy universe of language
use and language behavior.
Chomsky's distinction between linguistic 'competence' and linguistic
'performance' is a similar attempt to rescue Platonism by confining the
discussion of linguistic categories to an idealized realm of the mind. And
earlier, Bloomfield's rejection of the relevance of meaning to linguistic
analysis was motivated, in no small measure, by a wish to avoid the
pragmatic mess of the real world.11
History aside, there are good reasons why the theoretical possibility
of Platonizing context is doomed from the very start. First, the number
of potential distinct contexts in our experiential universe is in principle
unbounded, thus potentially infinite. However large our storage capacity
is, and however fast our retrieval system, they would never be large
enough or fast enough, respectively, to store and retrieve all possible
contexts. Second, contexts do not exist objectively, and thus cannot be
objectively specified. Rather they are construed — or abduced -- by the
interpreter's mind, given specific goals, tasks, functions. This is, of
course, the main reason why their number is potentially infinite. But it is
also a good reason why they cannot be specified and stored in advance of
experience, but rather must be construed by the experiencer di novo, for
the occasion. Lastly, the process by which one construes the 'right' con¬
text, or ponders 'how right' it appears, is unspecifiable by deductive
means. It involves abduction, which in turn hinges upon judgements of
relevance, similarity, or analogy.
The best known and most often cited metaphor used by Wittgenstein
(1953) to illustrate the context-dependent, nature of meaning is that of
games 12:

"...A move in chess doesn't consist simply in moving a piece in such-

and-such way on the board -- nor yet in one's thoughts and feelings as
one makes the move; but in the circumstances that we call "playing a
game of chess", "solving a chess problem", and so on..." [33; p. 17].

11 This was so because Bloomfield, as a born-again empiricist, believed that meaning resided
in the objective fabric of the Real World. The investigation of meaning was thus the proper
domain of the natural sciences rather than of linguistics (see Bloomfield, 1933, pp. 139-140).
12 All quotations below are from the Philosophical Investigations (1953), and are identified by
paragraph and page numbers, in that order.
46 Mind, Code and Context

"...When one shows someone the king in chess and says: "This is the
king", this does not tell him the use of the piece — unless he already knows
the rules of the game up to this last point: the shape of the king... The
shape of the chessmen corresponds here to the sound or shape of a
word..." [31; p. 15].

Summarizing this point explicitly, Wittgenstein adds:

"...For a large class of cases -- though not for all -- in which we employ
the word "meaning" it can be defined as thus: the meaning of a word is
its use in language..." [43; p. 20].

Driving at the same point from another angle, Wittgenstein discusses

the contrast between referring to a composite whole and referring to it
through component parts:

"...Asking "Is this object a composite?" outside a particular language-

game is like what a boy did once, who had to say whether the verbs in a
certain sentence were in the active or the passive voice, and who racked
his brain over the question whether the verb "to sleep" meant something
active or passive. We use the word "composite" (and therefore the word
"simple") in an enormous number of different and differently-related
ways..." [47; p. 22].
The thrust of the argument is now given:

"...Suppose that instead of saying "Bring me the broom”, you said "Bring
me the broomstick and the brush which is fitted on it". -- Isn't the answer:
"Do you want the broom? Why do you put it so oddly?"... Imagine a
language-game in which someone is ordered to bring certain objects
which are composed of several parts... and two ways of playing it: in one
(a) the composite objects (brooms, chairs, tables etc.) have names... in the
other (b) only the parts are given names and the wholes are described by
means of them. In what sense is an order in the second game an analyzed
form of an order in the first?..." [60; p. 30].

And in summary:

"...You may say: "The point of the two orders is the same". I should say
so too. — But it is not everywhere clear what should be called 'the point'
of an order..." [62; p. 30].

Here Wittgenstein's 'point' is clearly the pragmatic context, or more

specifically, the relevant portion of the potentially infinite number of
 Categories and Prototypes 47

different contexts that may be logically construed for each item, situation,
event.
Another of Wittgenstein's ingenious metaphors for context should
strike a familiar chord in the heart of travellers and mountaineers:

"...Language is a labyrinth of paths. You approach from one side and

know your way about; you approach the same place from another side
and no longer know your way about..." [203; p. 82].

Another equation of use with context is seen in a passing reference to

ostensive definitions, where Wittgenstein -- presumably unawares --
echoes Peirce:

"...Whether the word "number" is necessarily in the ostensive definition

depends on whether without it the other person takes the definition
otherwise than I wish. And that will depend on the circumstances under
which it is given, and on the person I give it to. And how he 'takes' the
definition is seen in the use he makes of the word defined..." [29; p. 14].

Another observation yet equates context with purpose when one

constructs a classificatory schema:

"...But how we group words into kinds will depend on the aim of the
classification..." [17; p. 8].

Wittgenstein alludes to the public aspect of context in a way that is

again reminiscent of Peirce:

"...Here the term language-game is meant to bring into prominence the

fact that the speaking of language is part of an activity, or a form of life..."
[23; p. 11].

Finally, Wittgenstein makes a number of memorable observations

concerning the difficulty we seem to experience in understanding the
'real' meaning of words, and how deep we are mired in the mud of our
old Platonic habit. While his language here is richly poetic, he nonetheless
alludes to some of science's most ancient predicaments:

"...One thinks that one is tracing the outline of the thing's nature over
and over again, and one is merely tracing round the frame through which
we look at it..." [114; p. 48].

"...A main source of our failure to understand is that we do not command

clear view of the use of words. -- Our grammar is lacking in this sort of
48 Mind, Code and Context

perspicuity. A perspicuous representation produces just that under¬

standing which consists in 'seeing connexions'. Hence the importance of
finding and inventing intermediate cases..." [122; p. 49].

And:

"...(For the crystalline purity of logic was, of course, not the result of
investigation: it was a requirement). The conflict becomes intolerable; the
requirement is now in danger of becoming empty. -- We have got on the
slippery ice where there is no friction and so in a certain sense the
conditions are ideal, but also, just because of that, we are unable to walk.
We want to walk: so we need friction. Back to the rough ground..." [107;
p. 46].

2.2.3. The non-discreteness of meaning

The idea that context is the prime source of non-discreteness in our

mental categories had, of course, been expressed by Peirce, as in his
criticism of Mill:

"...since no two phenomena ever can happen in circumstances precisely

alike, nor are two phenomena precisely alike..." [1940, p. 221].

As noted above, the non-discreteness of context is not logically necessary,

but is in fact the case. Wittgenstein's depiction of the gradual, open-ended
nature of meaning is indeed brilliant. But like all frontal attacks against
a dominant dogma, it is often couched in equally extreme terms. This
clearly serves the useful purpose of highlighting one's differences with
the prevailing orthodoxy. Perhaps the reason why Wittgenstein may be
interpreted as an extremist on the issue of categories may be traced to the
historical context within which he had to elaborate his heretic views —
the preminence of Logical Positivism in philosophy.13
Many of Wittgenstein's metaphors deal with the open endedness of
linguistic meaning, implicitly contrasting it with logic, which is closed
and internally-consistent:

"...ask yourself whether our language is complete; -- whether it was so

before the symbolism of chemistry and the notation of infinitesimal
calculus were incorporated into it; for these are, so to speak, the suburbs

13 For the Logical Positivists - Frege, Russell, Carnap, Tarski, Montague — Peirce apparently
never existed. The same was apparently true for Wittgenstein. For all intent and purpose,
then, Wittgenstein labored, within his own immediately-perceived intellectual environ¬
ment, in the splendid isolation of a heretical pioneer.
 Categories and Prototypes 49

of our language. (And how many houses or streets does it take before a
town begins to be a town...)..." [18; p. 18].

...But how many kinds of sentences are there? Say assertion, question,
command? -- There are countless kinds: countless different kinds of use of
what we call "symbols", "words", "sentences". And this multiplicity is not
something fixed, given once and for all; but new types of language, new
language-games, as we may say, come into existence, and others become
obsolete and get forgotten..." [23; p. 11].

"...We call many different things "names"; the word "name" is used to
characterize many different kind of usage of a word, related to one
another in many different ways..." [38; p. 18-19].

Wittgenstein next makes the classical observation that the boundaries

between criterial ('necessary', 'sufficient') and non-criterial properties
for determining membership in a meaning category are not as sharp as
the Platonic tradition would have them:

"...The essential thing is that this is a lamp, that it serves to give light; --
that it is an ornament to the room, fills an empty space, etc. is not essential.
But there is not always a sharp distinction between essential and inessen¬
tial..." [62; p. 30].

The discussion eventually leads to gradualism, then to the central

metaphor of family resemblance:

"...[You may wish to challenge me by saying:] You talk about all sorts of
language-games, but have nowhere said what the essence of a language-
game, and hence of language, is: what is common to all these activities...
And this is true. — Instead of producing something common to all that
we call language, I am saying that these phenomena have no one thing
in common which makes us use one word for all, -- but they are related to
one another in many different ways..." [65; p. 31].

"...(we) can see how similarities crop up and disappear. And the result
of this examination is this: we see a complicated network of similarities
overlapping and criss-crossing, sometimes overall similarities, some¬
times similarities of detail. I can think of no better expression to charac¬
terize these similarities than "family resemblances"; for the various
resemblances between members of a family: build, features, colour of
eyes, gait, temperament etc. etc. overlap and criss-cross in the same way.
-- And I shall say: 'games' form a family..." [66,67; p. 32].
50 Mind, Code and Context

Another of Wittgenstein's metaphors for the similarity bond between

members of a category is the woven thread:

"...But someone wishes to say: "There is something in common to all the

constructions -- namely the disjunction of their common properties" — I
should reply: Now you are only playing with words. One might as well
say: "Something runs through the whole thread, namely the overlapping
of those fibers..." [67; p. 32].
\ • »

What Wittgenstein is presumably after is the idea that a pattern -- the

whole — cannot be defined by the sum of its constituent parts.
The fuzziness of inter-categorial boundaries is another major theme in
Wittgenstein's argument:

"...How should we explain to someone what a game is? I imagine that

we should describe games to him, we might add: "This and similar things
are called 'game'". And do we know any more about it ourselves? Is it
only other people whom we cannot tell exactly what a game is? -- But this
is not ignorance. We do not know the boundaries because none have been
drawn..." [69; p. 33].

"...One might say that the concept 'game' is a concept with blurred
edges. — "But is a blurred concept a concept at all? Is it even always an
advantage to replace an indistinct picture by a sharp one? Isn't the
indistinct one often exactly what we need?..." [71; p. 34].

"...Frege compares a concept to an area and says that an area with vague
boundaries cannot be called an area at all. This presumably means that
we cannot do anything with it. -- But is it senseless to say: "Stand roughly
here"?..." [71; p. 34].

In his gentle admonition of Frege, Wittgenstein reminds us that the

reason why fuzzy-edged concepts are not totally useless is the real-world
context where they are employed. This is true when a concept's fuzziness
does not impinge upon the core function it serves, upon the domain
where it actually applies. Now, one may wish to argue that this is, at least
implicitly, an articulation of prototype semantics. Indeed, in articulating
a theory of prototypes, one must specify the areas of meaning that can
tolerate fuzziness, and then contrast those with areas that require a higher
degree of categorial stability; one must also explain why this is so,
presumably in terms of the functional context. Wittgenstein's tantalizing
references, suggestive though they may be, are never developed further.
They are reiterated, suggestively, obliquely, through other metaphors:
 Categories and Prototypes 51

"...[if you ask yourself] How did we learn the meaning of this word
("good" for instance)? From what sort of examples? In what language-
game? then it will be easier for you to see that the word must have a family
of meanings..." [77; p. 36].

"...I use the name "N" without a fixed meaning. (But that detracts little
from its usefulness, as it detracts from that of a table that it stands on four
legs instead of three and so sometimes wobbles..." [79; p. 37].

"...I said that the application of a word is not everywhere bounded by

rules... A rule stands there like a sign-post. -- Does the sign-post leave no
doubt open about the way I have to go? Does it show which direction I
have to take when I've passed it; whether along the road or the footpath
or cross-country?..." [84,85; p. 39].

"...We understand what it means to set a pocket watch to the exact time
or to regulate it to be exact. But... how nearly does it approach the ideal?...
"Inexact" is really a reproach, and "exact" is praise. And that is to say that
what is inexact attains its goal less perfectly than what is more exact. Thus
the point here is what we call "goal". Am I inexact when I give our distance
from the sun to the nearest foot, or tell a joiner the width of a table to the
nearest thousandth of an inch?..." [88; pp. 44-45].

2.2.4. Exemplars, norms and prototypes

It is a bit harder to read into Wittgenstein's text an explicit elaboration

of the prototype alternative to Platonic categories. In the space below, I
have culled from the Investigations the strongest passages that suggest, at
least under a sympathetic reading, something like a prototype approach.
Let us begin with:

"...in certain cases, especially when one points 'to the shape' or 'to the
number', there are characteristic experiences and ways of pointing — char¬
acteristic because they recur often (not always) when shape or number
are meant..." [35; p. 17; emphasis is mine; TG].

Taken at face value, 'characteristic experiences' that 'recur often' may be

construed to mean norms based on frequency. In fact, however, Wit¬
tgenstein refers here to characteristic mental experiences the speaker may
be having while 'pointing to a shape, number etc.'.
Consider next:
52 Mind, Code and Context

"...But then the use of the word is unregulated, the 'game' we play with
it is unregulated". -- It is not everywhere circumscribed by rules; but no
more are there any rules for how high one throws the ball in tennis, or
how hard; yet tennis is a game for all that and has rules too..." [68; p. 33].

One can interpret this passage to mean that while some areas of meaning
remain less regulated, others abide by stricter rules; and presumably then,
the latter would require a higher measure of categoriality, perhaps
prototype-like categories. But such an interpretation remains only im¬
plicit.
Consider next:

"...So I am shown various different leaves and told "This is 'a leaf'", I get
an idea of the shape of a leaf, a picture of it in my mind. -- But what does
the picture of a leaf look like when it does not show us any particular
shape, but what is common to all shapes of leaf'? Which shade is the
'sample in my mind' of the colour green -- the sample of what is common
to all shades of green? "But might there be no such 'general' sample? Say
a schematic leaf, or a sample of pure green?" -- Certainly there might. But
for such a schema to be understood as a schema, and not as the shape of
a particular leaf, and for a slip of pure green to be understood as a sample
of all that is greenish and not as a sample of pure green -- this in turn
resides in the way the samples are used..." [73; p. 35].

This passage comes tantalizingly close to articulating a psychological

notion of 'prototype' or 'typical exemplar' as a necessary ingredient in
the formation of natural categories. But again, the interpretation remains
implicit.
Consider next:

"...Here also belongs the idea that if you see this leaf as a sample of 'leaf
shape in general' you see it differently from someone who regards it as,
say, a sample of this particular shape..." [74; p. 35].

Wittgenstein argues two separate points here. First, he is recapitulating

his position concerning the context-dependent nature of meaning.
Second, he continues to attack the Platonic variant of 'idealized example'.
A more explicit formulation of an alternative to Platonic categorization
does not follow.
Consider next:

"...I want to say: we have here a normal case and abnormal cases. It is
only in normal cases that the use of a word is clearly prescribed; we know,
we are in no doubt, what to say in this or that case. The more abnormal
 Categories and Prototypes 53

a case, the more doubtful it becomes what we are to say... if there were
for instance no characteristic expressions of pain, of fear, of joy; if rule
became exception and exception became rule; or if both became
phenomena of roughly equal frequency - this would make our normal
language-games lose their point..." [141,142; p. 56].

This is, as far as I know, the most explicit elucidation in the Investigations
of one fundamental prerequisite for a prototype theory - as distinct from
mere contextual gradualism: The correlation between typicality and
frequency of an experience, and the attendant pragmatically-based no¬
tions of norm and counter-norm.14
Finally consider:

"...[a suggestive example that] is not meant to apply to anything but the
examples given is different from that which 'points beyond' them. "But
then doesn't our understanding reach beyond all the examples?" - A very
queer expression, and a quite natural one! -- But is that all? Isn't there a
deeper explanation; or mustn't at least the understanding of the explana¬
tion be deeper?... "But do you really explain to the other person what you
yourself understand? Don't you get him to guess the essential thing? You
give him examples, — but he has to guess their drift, to guess your
intention..." [208, 209, 210; pp. 83-84].

This passage echoes Peirce's notion of abduction or pragmatic intuition

(Wittgenstein's 'guessing'), whereby the central features of similarity,
import and relevance are -- in principle -- logically undefinable. One could
also read into the text a prototype-semantic interpretation: General
categories are inferred from typical exemplars, from normative cases. But,
such a reading is only implicit in the text.
In one clear sense, Wittgenstein's contribution to the development of
prototype semantics has already been acknowledged by Rosch and
others, who considered him their inspiration in developing an empirical¬
ly viable pragmatic alternative to Platonism.15 But, to do justice to the
spirit of a great master, one must also add that this interpretation of
Wittgenstein serves to illustrate a central theme of Pragmatics, namely
that for all practical purpose, meaning is the product of how you press it
into use.

14 See discussion in Givon (1979a, Ch. 3).

15 Most of us cite Wittgenstein in ignorance of Peirce's work.
54 Mind, Code and Context

2.3. Prototypes and metaphoric extension

2.3.1. Preamble
The most prototypical member of a population ('category'), the one
displaying the largest number of the most important characteristic fea¬
tures, is presumably the one that best approximates our mental concept
of a prototype-like category. It is worth noting that Wittgenstein (1953)
recognized this possibility, but did not explicitly resolve its seeming
internal contradictions:

"...But for such a schema to be understood as a schema, and not as the

shape of a particular leaf, and for a slip of pure green to be understood
as a sample of all that is greenish but not as a sample of pure green — this
in turn resides in the way the samples are used..." [73; p. 35].

Indeed, the most striking feature of Rosch's work on natural categories

is her demonstrating how our notion of 'class' -- the prototype -- evolves
from repeated experience of typical exemplars. Members of a category
can thus be ranked according to their distance from (or 'degree of
resemblance to') the most typical exemplar -- the prototype core. And the
core may be considered as something akin to the population's — or the
category's — mean.16
The notion of resemblance or similarity is indeed the crux of how we
form our mental categories. It is also crucial for understanding how we
modify and extend existing prototype-like categories. One may call this
process of extension analogy or metaphor.17 There is nothing logically
necessary, or Platonically discrete and absolute, about resemblance or
'being like'. Anything can, in principle, be said to 'be like' anything else,
given the appropriate context. Analogy and metaphor are thus -- in
principle — pragmatic, open ended, context-dependent, abduction-
driven notions.18

2.3.2. Metaphoric induction into a lexical prototype

The extension of category membership to new members is one of the
most ubiquitous processes in human categorization, be it at the sensory,

16 See extensive discussion in Medin and Schaffer (1978).

17 These two names for the same phenomenon come to linguistics via two distinct traditions.
'Metaphor' comes to us via of literary analysis, while 'analogy7 comes more directly from
the philosophical tradition, most recently via Peirce. For a general discussion see Anttila
(1977) and Lakoff and Johnson (1980).
18 See further discussion in Chapter 7, below, as well as in Givon (1982a).
 Categories and Prototypes 55

cognitive or linguistic level. It is precisely the fuzzy-edged, prototype-like

nature of mental categories that makes such extension possible. Through
this process, less typical individuals may nonetheless join a category, given
an appropriate (though 'less typical' or 'less frequent') context. Since the
literature is rich with examples of such metaphoric extension of lexical
meaning, one illustration here will suffice.19
Suppose one said:

(7) George built a wall around himself

For (7) to be interpreted metaphorically, there must have been -- from the
speaker's point of view — something in George's behavior which
resembled the literal meaning of build a wall around oneself. In principle,
this could have involved the brick or mortar used, the characteristic
wall-shape, wall-height, wall-color etc. But in this particular case, the
analogy is likely to have involved the following two pragmatic inferen¬
ces ('abductions') concerning the normative goals and consequences of
literal wall-building:

(8) a. Goals: Walls are often built for the purpose of defending
oneself from outside threat;
b. Consequences: Wall-building often results in undue isolation
from the outside.

The metaphoric extension of the membership of the category 'wall¬

building' thus allowed defensive and isolating behavior to join the cluster of
features characterizing it. These two features will remain, at least initially,
less characteristic (less prototypical). They may at first define -- i.e. admit
into the category -- marginal, less-typical, more exceptional members,
ones judged to be included metaphorically rather than literally.
There is nothing logically necessary in selecting the two pragmatic
inferences (8a) and (8b) over all others as the relevant or important
inferences in this case. They are not endowed with such privileged status
in principle. It is the purpose or context of this particular metaphoric
extension of 'wall-building' that makes these two abductions relevant. It
is of course true that some contexts are more frequent, normative or
stable, given the general layout of the culture ~ or of human cognition.
In this particular case, 'defensiveness' and 'isolation' are rather ob¬
vious metaphoric extensions, given the cluster of normal purposes of
wall-building. But suppose the same physical walls were normatively
built for an altogether different purpose, say laying siege to a citadel or
reaching higher up. In the first instance, 'defensiveness' would change to
'offensiveness'; and 'isolation' will change its valence -- from reflexive

19 For a book-size collection, see Lakoff and Johnson (1980).

56 Mind, Code and Context

('isolating oneself') to transitive ('isolating someone else'). In the second,

nothing related to either 'defensiveness' or 'isolation' would be relevant
any more. Rather, one could construe another set of apt metaphors, such
as reaching up, ambition, self-improvement, intellectual growth etc.

2.3.3. Metaphoric extension of a lexical prototype

The flexibility, open-endedness and' context-sensitivity of prototype¬
like categories also provide for the eventual change in the definition of
the prototype itself. Such change is the common end product of the
process we have just discussed, i.e. metaphoric induction of new mem¬
bers into a category. The cumulative impact of this influx of new and
less-prototypical members inevitably leads to either the re-definition of
the prototype core features, or to changes in their relative ranking.
Metaphoric extension is indeed the major source of diachronic change in
lexicon, phonology, morphology and grammar. We will discuss only one
case from lexical semantics as illustration.
The English verbs know and can may be traced back to the same
Indo-European root, with the meaning 'know' being presumably the
older one.20 The extension from know to can is due to one less-central sense
of know — 'know how (to do)'. And presumably this less-central sense
extended itself via the following pragmatic inference:

(9) 'know how to do' 'be better able to do'

Now, not all uses of know involve the sense of 'know how'; two other
senses are presumably more central to the prototype definition of know:

(10) a. 'know that P is true/false'

b. 'be familiar with N'21

Next, can as we know it now involves not only the sense 'know how', but
also another — more central -- sense, 'have power to act'. When can
completed its historical divergence from know, along the line taking
advantage of a fortuitous phonological variation, the newer sense of
'have power to act' became more central to its prototype definition; while
the sense of 'know how' -- which came from know as the historical 'bridge'
in the original metaphoric extension, became less central to the newly-as¬
sembled prototype definition of can.

20 I owe the details of this example to Raimo Anttila (in personal communication).
21 Wherestands for 'proposition' and 'N' for 'individual entity'.
 Categories and Prototypes 57

To take the example a bit further, can soon developed a rather predict¬
able deontic sense, that of 'be permitted to act'. This extension occurred,
presumably, through the following chain of pragmatic inferences:

(11) a. 'have the power to act' => 'not be restrained from acting'
b. 'not be restrained from acting => 'be permitted to act'

There is nothing deductively necessary about inferences (lla,b). The

consequent in (11a) is only one of many possible reasons why one may
'have power to act'. And the consequent in (lib) is only one possible
reason why one may be 'not restrained from acting'. It is the specific
context which motivated the choice of these particular abductions as
relevant in this case.
Finally, can quite commonly also develops an epistemic sense, that of
probability, presumably via the pragmatic inference:

(12) 'ability to act' 3 'higher probability of acting'

And although this epistemic sense may not be central for the current
prototype definition of can, it is clearly recognized by speakers and may
even be on the ascendance.22 In this way, each new sense of know, and of
can after it split from know, re-defines the old prototype to some degree,
by extending membership to new items that had been excluded from the
category before. The process of inducting new members into a category
is thus the very same process that eventually modifies the category's
prototype core.

2.3.4. Prototypes, metaphoric extension and grammaticalization

Metaphoric extension of prototype-like categories is also central to the
process known as grammaticalization, through which lexical words
become grammatical morphemes (and eventually inflectional morphol¬
ogy). I will illustrate this process here with one widely-attested change,
where the verb go and come give rise, respectively, to the modalities
'future' and 'perfective'.23 This change involves a metaphoric extension
of the sense of spatial motion of go and come to temporal motion, a rather
common type of semantic bleaching.24

22 See discussion, and some experimental results, in Givon (1979a, ch. 3).
23 For grammaticalization in general see Givon (1971,1975a; 1979a, ch. 6).
24 See discussion in Givon (1979a, ch. 8). The connection between grammaticalization and
metaphoric extension is also noted in Qaudi and Heine (1984) and Heine and Reh (1984).
The original discussion of the widespread transformation of 'go' and 'come' into tense-
aspect markers can be found in Givon (1973a). A related discussion may be also found in
Traugott (1974).
58 Mind, Code and Context

One may construe this change as a metaphoric switch from spatial to

temporal deixis. As is well known, go and come differ, in addition to their
semantic features of motion from some source or toward some goal, also
by their pragmatic feature of deixis, whereby go is used when the motion
is away from the position of the speaker, come is used when the mostion
is toward the speaker's position. The metaphoric extension that occurs in
grammaticalization may thus hinge upon the following pragmatic in¬
ference:

(13) (a) GO = 'move away from this place' 3

FUTURE = 'move away from this time'
(b) COME = 'move toward this place' 3
PERFECTIVE = 'move toward this time'

Two semantic elements of the verbs go and come have been preserved in
the metaphoric transformation above: (a) the sense of motion to/from;
and (b) the deixis. Both have been transformed from space to time. But
in addition, another transformation also occurs. Spatial motion is not
uni-directional; one could go or come either to or from a place. But
temporal motion is unidirectional. So that the erstwhile go as a 'future'
modality could only mean going away from the present. While the
erstwhile come as a 'perfective' modality could only mean coming toward
the present. And the features 'away from' and 'toward' were inherited
from the deixis feature of the erstwhile verbs.
The inferences in (13) are typically abductive. Many other semantic
features of go and come were deemed irrelevant for this particular
metaphoric extension. Motion in time is indeed an inseperable ingredient
of motion in space, but may not be the core of the prototype go or come.
In this metaphoric extension, however, it was relevant to to the purpose
of the extension, and became a core feature of the prototypes of 'future'
and 'perfective'.
To drive the latter point home, consider the equally common gram¬
maticalization of the serial verbs go and come into the locational case-
markers 'to' and 'from', respectively.25 This process may he summarized as:

(14) a. GO = 'He walked going (to the) market' >

TO = 'He walked to (the) market'
b. COME = 'He walked coming (from the) market' >
FROM = 'He walked from (the) market'

Unlike the metaphoric extension in (13), in (14) it is not the deixis feature
of go and come that have been preserved, but rather the semantic features
of 'motion toward a goal' and 'motion away from a source', respectively.

25 For the grammaticalization of serial verbs into case markers, see Givon (1975a).
 Categories and Prototypes 59

Further, the directionality of go and come as case-markers is exactly the

opposite of their directionality as tense-aspect markers. In other words,
for the purpose of the latter metaphoric extensions, other features of the
verbs were judged more relevant, and those are the ones that became the
prototype core of the new grammatical morphemes.

2.4. Prototypes and metaphoric extension in the grammar of

transitivity

Prototype-like categorization and metaphoric extension are just as

pervasive in grammar as they are in the lexicon and inflectional morphol¬
ogy. In the space below I will illustrate this by citing a number of
metaphoric extensions in the grammar of transitivity.
In grammar -- and grammars -- the notion of transitivity is central to
our understanding of events and actions, and how these two differ from
mere states. The prototype definition of 'event' involves three core fea¬
tures:26

(15) a. Agency: An event is a change involving a highly

visible/salient, volitional, controlling, acting agent/cause;

b. Affectedness: An event is a change involving a highly

visible/ salient, non-volitional, non-acting affected patient/ef-
fect;

c. Rate of change: An event is a change that is highly

visible/salient in terms of rate of change over time.

Note first that the three conditions in (15) can each be couched in terms
of degree; indeed, many intermediates may be found on these three
scales. Further gradation may also arise when an event scores high on one
or two of these scales but not on the other(s). Finally, gradation - and
fuzzy-edgedness -- may also arise from the fact that these three features
are only the most important features in defining the transitive prototype;
other -- less central -- features may still exist.
In a broad sense, features (15a,b,c) refer to the prototypical agent,
patient and verb, respectively, involved in the transitive event. Some
examples of prototypical transitive events are:

■it

26 For further discussion see Hopper and Thompson (1980; 1982 eds) or Givon (1984, chapters
4,5,8,11).
60 Mind, Code and Context

(16) a. Mary cut the meat

b. John destroyed the house
c. They killed the goat
d. She broke the chair

Many events involve objects that are less-than-prototypical patients.

That is, in terms of our prototype feature (15b) they themselves are either
less concrete, less visible, less salient; or the effect of the event on them
may be less salient; or the entire change may be relatively slow (feature
(15c)) and thus again less salient.27 Some of these less-prototypical objects
are not really patients by the strictest definition. They do not --
'objectively speaking' -- undergo any visible change during the event.
Nonetheless, grammars often offer the option of construing such patients
-- metaphorically -- as patient-like. They are then marked grammatically
as direct objects, the most common way of coding prototypical patients.
This option offered in grammars is not trivial; it reflects subtle differences
in the way an event may be viewed, construed, contextualized. To
illustrate this, consider first the following paired expressions, each in¬
volving the same lexical verb. Each expression renders the -- 'objectively
same' -- event as either intransitive (on the left) or transitive (on the right).
And in a clear sense, the transitive version is a metaphoric extension of
the intransitive:

(17) literal sense metaphoric sense

(intransitive) (transitive)

a. He rode on the horse He rode the horse

b. He swam across the channel He swam the channel
c. They entered into the citadel They entered the citadel
d. She took the money from him She robbed him of his money
e. She walked with her dog She walked her dog

The horse in (17a) on the left is construed as a location; it is less affected,

less controlled. On the right it is construed as an affected, manipulated,
controlled patient. The channel in (17b) on the left is mere geographic space
to be crossed. On the right it is an adversary to be conquered, subdued,
tamed. The citadel in (17c) on the left is a location to be entered. On the
right it is construed as a target, to be breached and taken over. In (17d) on
the left 'he' is construed as the source from which money was removed.
On the right 'he' is the affected, aggrieved victim. Finally, the dog in (17e)
on the left is a companion, a self-controlling partner. But on the right the
dog is a controlled, manipulated patient. Thus, regardless of the near

27 Saliency or affectedness of the patient (condition (17b)) and rate of change (condition (17c))
are not totally independent of each other.
 Categories and Prototypes 61

identity of the 'objective' physical event, one can construe these non¬
patients metaphorically as 'patient-like', then code them syntactically as
one would prototypical patients. Consequently, the entire clause is coded
as a prototypical transitive clause.
In the grammar of transitivity, the exact opposite metaphoric extension
can also take place. That is, one can demote an object that is 'objectively7
a prototypical patient, by construing it as if it were 'not quite a patient'.
It is then stripped of direct object status, and the entire clause is coded as
intransitive. As an illustration of this process, consider the following
paired expressions. In each pair, the 'original' transitive appears on the
left and the metaphoric ('derived') intransitive on the right:

(18) original transitive derived intransitive

with prototype patient with 'demoted' patient

a. She ate the fish She ate (regularly)

b. They hunted the deer They hunt (each fall)
c. She hunted the deer She went deer-hunting
d. He drank his beer He drinks (excessively)
e. He fought Joe Lewis He fights (for a living)
f. She shot only one bullet She started shooting

The 'logical' object in the expressions in (18) is expressed on the right as

either zero or an 'incorporated' part of the verb. Both are syntactic devices
used to demote the object and render the clause intransitive. The reason
for this demotion is that the object is non-specific ('non-referential'); it's
type is predictable, it is highly stereotyped. Thus, one stereo typically eats
'food', drinks 'liquor', hunts 'game animals' or 'deer', fights 'prize
fighters' or shoots 'bullets'. The stereotypicality and non-specificity of the
object make it cognitively less salient, thus lowering the transitivity of
the clause (recall condition (15b) above).28
It is also quite common in language to metaphorically construe events
in which — strictly speaking -- the subject is not a prototypical agent (and
thus the event does not abide by condition (15a) above) as if they were
transitive events. As illustrations consider:

(19) a. He saw the mountain

b. She knew the answer
c. He heard the song
d She feared him
e. He wanted a sandwich
f. She remembered the conversation

28 In many languages, the transitivity of the clause is coded morphologically on both the
subject and the verb. In such languages, traditionally referred to as 'ergative', the expres¬
sions on the right hand side in (18) will exhibit intransitive morphology.
62 Mind, Code and Context

Examples (19a-f) violate all three prototype transitivity conditions in (15).

First, the subject is neither volitional nor active nor a causer; second, the
object is not visibly affected and thus not a prototype patient; and third,
the change is neither fast nor salient. In fact these expressions code,
strictly speaking, states rather than events. In many languages, especially
those that display an ergative or active-stative morphology,29 the expres¬
sions in (19) would be coded syntactically as intransitive clauses. English
allows coding them as if they were transitive, i.e. with subject and direct
object. In construing them as transitives, English seems to be relaxing an
important part of the definition of 'agent' in (15a), moving from the
stricter condition of 'having volition, control and decision-to-act' to a
weaker condition of 'having conscious mental activity'. Such extension
is again an instance of generalization or semantic bleaching, based on the
fact that both volition and decision-to-act are themselves mental ac¬
tivities. The pragmatic inferences involved in this generalization may be
given as:

(20) a. 'subject capable of mental activity' r>

'subject capable of volition and decision'

b. 'subject capable of volition and decision' =>

'subject capable of acting as agent'

Consider next the converse case, where an event is semantically quite

transitive -- with prototypical agent (and patient); but the agent is unim¬
portant, unknown, stereotyped/predictable, non-referential or communicative¬
ly irrelevant. Under such conditions the agent may be metaphorically
downgraded, and the clause is then coded as passive or impersonal, as in:30

(21) a.The window was broken yesterday (by John)

b. They found him dead on the beach
c. You buy newspapers from a vending machine
d. English is spoken here
e. One doesn't do this sort of a thing any more
f. He got killed in a traffic accident
g. Everybody knows him well around here

29 In ergative languages, the morpho-syntax is sensitive to all three prototype conditions in

(15). In active-stative languages, clausal morpho-syntax is sensitive primarily to condition
(15a). For further discussion see Givon (1984a, Ch. 5).
30 This downgrading closely parallels the situation of objects-patients in (18), above. Con¬
structions with stereotyped, downgraded patients, as in (18), are often referred to as
'antipassive'.
 Categories and Prototypes 63

These English expressions are grammatically quite diverse, and the

'irrelevant' agent in them is suppressed by a variety of means. The agent
is demoted via the use of oblique case-marking in (21a). It is left unex¬
pressed altogether in (21d,f). Or it is expressed as a non-referential,
impersonal pronoun in (21b,c,e,g). Such passive-impersonal construc¬
tions are often used to convey a state resulting from an event, rather than
the event itself (which is more commonly coded as an transitive active
clause). Impersonal-passive clauses as in (21) may be viewed as
'metaphorically intransitive'.31
The last example involves what linguists somewhat implausibly call
'raising', and are fond of discussing in purely syntactic terms. Consider
the following two pairs of English sentences, each with alternating verbs
that are semantically rather close:

(22) a. He wished that she would visit

b. He wanted her to visit

c. She agreed that he could leave

d. She let him leave

Verbs of volition or permission as in (22) above are not prototypically

transitive. They involve mental rather than physical processes, no dis¬
cernible action by an agent, and little if any visible physical effect on the
object. In fact, strictly speaking the verbs in (22) are intransitives, they
'have no object'. In the variants (22b,d), however, a syntactic object -- the
so-called 'raised' one — appears. It is, strictly speaking, the subject of the
complement verb. But there is a subtle difference in meaning between the
transitive (object-taking) and the intransitive (objectless) member of each
pair. In (22a) a purely internal 'wishing' took place, with no hint of contact
or communication between the two participants. The subject of visit —
'she' -- could have been miles away, and thus not likely to have been
manipulated by the subject of wish.'She' is coded as subject, thus em¬
phasizing her possible independence as a decision-making agent. In
(22b), on the other hand, there is a higher likelihood that the wish was
actually communicated, that there was direct contact between subject and
object -- and thus an opportunity for the subject to manipulate the object.
Syntactically this is expressed by the 'raising'. That is, the logical subject
of the complement verb is coded as if it were a manipulated object of want.
It is thus treated metaphorically like a prototype patient.
In the same vein, the subject of leave in (22c) may not even be there and
may not be aware of what went on in the mind of the subject of agree.
There is thus no implication that he actually left. In contrast, in (22d) there
is a stronger implication of communicative -- or even physical -- contact

31 See further discussion in Givon (1981a).

64 Mind, Code and Context

between the two participants, and also an implication that the object of
let in fact left. The subject of leave is thus construed, metaphorically, as
more like a patient, and coded syntactically as an object. The pragmatic
inferences involved in these metaphoric extensions may be given as:32

(23) a. If the subject of 'wish' (or agree) establishes physical contact with
a person whose action they desire (or agree to), the probability
of communicating their wish (br agreement) to that person in¬
creases;

b. If you communicate to another person your desire (or agreement)

that they take an action, the probability of their taking action is
higher than if you did not;

c. The more you adjust your actions to the motives of another

person, the less you act as an independent agent; and the more
you resemble a manipulated patient.

None of these inferences are deductively necessary. They are prob¬

abilistic, abductive, pragmatic.

2.5. The functional motivation for a hybrid system of

categories

"...The solution to great arguments is usually close

to the golden mean..."

S.J. Gould,
Ontogeny and Phytogeny, (1977, p. 18 )

It is reasonably clear that the philosophical poles of Platonic rigidity

and Wittgensteinean flexibility represent two extreme aspects of the cog¬
nitive and neurological tool-box possessed by living organisms. Under
specific task conditions, blends of the two ~ at various ratios - are both
possible and necessary. In summarizing the history of recent experimen¬
tal research into these issues, my colleague Michael Posner characterized
the pragmatic, hybrid nature of human categorization - and the function-

32 For more discussion of the semantics and pragmatics of manipulative verbs in relation to
the syntax of their complements, see Givon (1975b, 1980a).
 Categories and Prototypes 65

"...The major lesson of this research seems to me to lie in its confirmation

of the general information processing tenet of bounded rationality
(March and Simon, 1958). The constraints imposed upon information
processing by the nature of the human brain lead to some heuristics that
provide means for allowing us to operate efficiently in the world to which
we are adapted..." (1986, p. 9 of ms)

Linguistic or mental categories, or rules of grammar, are seldom totally

discrete and exceptionless. Still, linguistic and mental categories — and
rules of grammar -- do exist. And among them, those that display high
frequency — and thus presumably high functional load, i.e. those that
are more predictable, turn out to also be more stable and more discrete,
given their functional context.
As will be noted in Chapter 7, the same imperfect but manageable
blend between stronger and weaker categoriality also characterizes infor¬
mation processing in general, in humans as well as in other biological
organisms. In the functional domains of neurology, motor-control, atten¬
tion and memory, the interplay between automated (more category-de¬
pendent, less context-scanning) and attended (less category-dependent,
more context-scanning) processing is well documented. Within this
biological hybrid system, the relative flexibility and open-endedness of
categories is just as important as their relative stability, distinctness and
saliency. The more Wittgensteinean aspects of the system, at whatever
level, allow the organism to perform context-sensitive, feedback-de¬
pendent tasks. Such tasks require finer discriminations of shades and
gradations in the scanned context; they also demand the exercise of subtle
and contingent judgement in interpreting the phenomenological con¬
tinuum. This type of information processing tends to depend on using
more narrow, sequential, analytic, time consuming, conscious, attended
processing channels. Neither the use of human language, nor its analogy-
driven adaptive change over ontogeny, phylogeny and history, could
ever proceed without this flexible component of our categorization sys¬
tem.
It may be argued that the flexibility required by the organism in
processing information at whatever level may be theoretically achieved
through assigning discrete categorial status to all imaginable contexts.
But, as we have already noted above, such Platonization of context is both
logically and psychologically a mirage, and for a number of reasons:

(i) Memory storage capacity:

The number of theoretically possible contexts is in principle both un-

denumerable and unspecifiable. Anything at all can be the context for
anything else, given some judgement of relevance. And as already noted,
66 Mind, Code and Context

relevance -- or the similarity decisions that underlie judgements of

relevance -- are both open-ended entities.

(ii) Processing time:

Our lexicon, for all its Wittgensteinean open-endedness, is still finite and
largely discrete at any particular point in time. This allows the organism
to embark upon search-and-retrieval operations that can be concluded,
at least in principle, within finite time. But to search through, and retrieve
from, a potentially infinite pool of discrete contexts would be a futile
procedure. It is precisely because contexts can be construed upon contin¬
gency, or abduced for the occasion, that meaning is processable at all.

(iii) Expansion and modification:

The acquisition and integration of new information into an existing

knowledge-base is in principle impossible within a full discrete, deduc¬
tive system. Any non-trivial change in a knowledge-base must rely on
analogy, abduction, metaphor and judgements of relevance. None of
these could operate -- or even be defined -- within a system of discretized
context.33

While a processing system must incorporate a considerable measure

of Wittgensteinean flexibility, equally well some decisions made by or¬
ganisms — or tasks performed at various levels -- must proceed at high
speed, categorially, unattended, in parallel, and with little regard to finer
shades and gradations. Under task conditions that demand such
categoriality, the context itself must be construed in starker, discrete,
Platonic terms. Task that are highly repetitious, thus predictable, are
carried out that way, as well as tasks that require, high speed response.
Thus, if a deer must decide — within a fraction of a second — whether a
potential predator is too small to worry about (say a bobcat), or large

33 For an extensive discussion of the relevant literature in the psychology of learning, see
Medin and Schaffer (1978). Hyman and Frost (1975) also suggested that during earlier
stages in the acquisition of a new category, actual instances — exemplars — are stored as
mental representation of the category (see also Hintzman and Ludlam, 1980; Hintzman,
1983). But later on such exemplars are converted into prototypes; and later still, prototype¬
like categories are converted into more discrete Platonic categories. The philosophical
literature, from Peirce to (even the early) Wittgenstein (1918), had of course made similar
observations, although on less firm empirical grounds.
 Categories and Prototypes 67

enough to run away from (say a cougar), the categorization supporting

such a decision is likely to be rather discrete and Platonic. Such
categorization would tend to compress all potential intermediates on the
scale into either one extreme or the other. And it would tend to pay much
less attention to fine shades of context.
A hybrid, prototype-like system of categories is thus, most likely, a
necessary optimal solution to the real-world task environment within
which organisms process information. It allows them to make gross
discriminations and snap decisions whenever necessary. Still, the system
retains an ingrained potential for flexibility, context-sensitivity and ex-
tendability, to be used when such features are called for.34

2.6. Closure

The hybrid solution to the problem of categorization — the theory of

prototypes -- in a way recapitulates Kant's anti-reductionist approach to
epistemology: Neither concepts nor percepts are by themselves viable as
sole source of our mental categories. Only together, interactively, can
they give rise to our mental representation — construction — of ex¬
perience. Posner's 'bounded rationality' -- rational enough to allow some
categorial stability, yet incomplete and flexible -- is a similar kind of
pragmatic compromise. Within such a framework, both extreme features
of the system — the Platonic and the Wittgensteinean -- are equally
necessary, in order for the organism to perform adaptive tasks.
I would like to close this chapter with Gould's comment on another
futile argument between two reductionist extremes, this one in the his¬
tory of biology:

"...I doubt that such a controversy could have arisen

unless both positions were valid (though incomplete)..."

S.J. Gould
Ontogeny and Phytogeny (1977, p. 59)

34 As Posner and Warren (1972) note, more rigid and discrete categories are a crucial
prerequisite for automated, unattended processing. Conversely, more flexible prototype¬
like categories must be involved in analytic, attended processing. In the discussion of the
three modes of inference ~ induction, deduction and abduction (Chapters 7 and 8) - we
will return to this subject in some detail. In discussing the nature of human categorization,
we thus introduce a number of leitmotifs that will remain with us throughout much of this
book.
 *

. ..
 Chapter 3

THE LINGUISTIC CODE AND

THE ICONICITY OF GRAMMAR

"...A sign is something by knowing which

we know something more..."

C.S. Peirce

3.1. Preamble*

In this chapter we return to discuss the justification for treating

human language as a system of signs, or a code. I will use these two
terms interchangeably. Our discussion of this rather complex subject
will be divided into three headings:

(a) The role of the interpretant -- or context -- in binding the

sign and the designatum together in the semeiotic relation;

(b) The identification of the linguistic designator ('sign'), the

designatum of language, and the notion of 'code';

(c) The Iconicity of the linguistic code, its non-arbitrary nature,

particularly in so far as grammar ('syntax') is concerned.

An attempt will be made to support the latter with a certain body of

evidence illustrating the pervasive iconicity found in the grammar of
natural language. In a summary section, some broader bio-adaptive im¬
plications of the iconicity of natural language will be outlined.

*1 am indebted to T.K. Bikson, John Haiman, Michael Shapiro, Martin Tweedale and John
Verhaar for many helpful comments and suggestions. This chapter also benefitted from the
general discussion at the Symposium on Iconicity, Stanford University, May 1983; particularly
from comments by Henning Andersen, Joseph Greenberg and Dan Slobin.
70 Mind, Code and Context

3.2. The interpretant as context

3.2.1. Background
As noted in Chapter 1, above, the semeiotic relation between a sign
and its designatum must be taken as a trichotomy , demanding a third
term ~ the interpretant. The notion that the coding relation is triadic
rather than (as is superficially more pleasing) diadic, is attributed by
Morris (1938) to Aristotle, but the relevant passage in De Interpretation
does not support this attribution. 'Interpretant' has always been — and
still remains -- somewhat of a murky concept. Morris (1938), for ex¬
ample, suggests that the interpreter -- the mind that perceives the sig¬
nificance of the semeiotic relation - is a fourth term in the semeiotic rela¬
tion:

"...These three components in the semeiosis may be called, respective¬

ly, the sign vehicle , the designatum , and the interpretant; the interpreter
may be included as a fourth factor..." [1938, p. 3].

However, in another passage Morris seems to identify 'interpretant'

with the mind of the interpreter:

"...The interpretant of the sign is the mind; the interpretant is a

thought or concept..." [1938, p. 30]

In one of his more desperate attempts to give a coherent definition of

'interpretant', Peirce identifies it as an effect upon the interpreter:

"...I define a Sign as anything which is so determined by something

else, called its Object [i.e. the designatum; TG], and so determines an ef¬
fect upon a person, which effect I call its Interpretant... My insertion of
'upon a person' is a sop to Cerebrus, because I despair of making my
own broader conception understood (H 80-1)..." [Shapiro, 1983, p. 47].

Citing Aristotle again, Morris also seems to define 'interpretant' as

the generic-lexical knowledge of all entities and their properties, held
in common by members of the speech community (i.e. 'culture'):

"...The interpretant is a thought or concept; these thoughts or concepts

are common to all men and arise from the apprehension by the mind of
objects and their properties..." [1938, p. 30].

And citing Peirce, Morris also defines the 'interpretant' as a set of

shared habits and rules of usage:
 The Linguistic Code: The Iconicity of Grammar 71

"...Charles S. Peirce, whose work is second to none in the history of

semeiotic, came to the conclusion that in the end the interpretant of a
symbol must reside in a habit and not in the immediate physiological
reaction which the sign vehicle evoked or in the attendant image or
emotion -- a doctrine which prepared the way for the contemporary em¬
phasis on rules of usage ..." [1938, p. 31; emphases are mine; TG]

Morris's somewhat narrow definition of 'interpretant' is perhaps un¬

derstandable, given that his (and Aristotle's) discussion is confined, to
the most arbitrary type of sign, the symbol. Within human language,
this type of sign is confined, almost exclusively, to the lexicon. Morris
makes this rather clear in his discussion:

"...the interpretant is a thought or concept; these thoughts or concepts

are common to all men and arise from the apprehension by the mind of
objects and their properties; uttered words are then given by the mind
the function of directly representing these concepts, and indirectly the
corresponding things; the sounds chosen for this purpose are arbitrary
and vary from social group to social group; the relations between the
sounds are not arbitrary but correspond to the relations of concepts and
so of things..." [1938, p. 30].

Morris's oblique reference to 'relations between sounds', which may

probably be translated as 'relations between words within the pro¬
position', is but a faint echo of Peirce's own discussion of iconicity in
grammar.
We are, I believe, again indebted to Peirce for providing the critical
mass in identifying 'interpretant' with context. This is so because, un¬
like Aristotle and Morris, Peirce discusses not only symbols but also in¬
dices and icons. For each of these three types of sign, he identifies a
wide range of possible types of interpretant. Further, Peirce's recognized
complex signs, especially in our grammatical code. And within those
complex signs, a mixture of symbolic (i.e. 'more arbitrary'), indexical
and iconic sign-elements can be observed. Each sign-element, in turn,
may motivate a different interpretant-element in the complex sign. The
precise internal structure of this multiplicity, and its potential Wit-
tgensteinean scalarity, is often obscured by rather Platonic proclivities
of Peirce's expository style, which is indeed replete with discrete
trichotomies and the trichotomies of trichotomies. (See Peirce, 1940, pp.
115-118, as well as Shapiro, 1983, pp. 63-65). But it is perhaps possible to
72 Mind, Code and Context

interpret Peirce's discrete trichotomies as standing for a scalar con¬

tinuum.1
As for the function of the Peirce's 'interpretant', Shapiro (1983) offers
the following observation:

"...It is only the genuine triadic relation obtaining between sign, object
and interpretant that enables the sign to offer assurance to the inter¬
pretant that the object for which it stands is simultaneously identical
with the object of the sign the interpretant interprets..." (1983, p. 61).

The 'interpretant' here seems to be equated with the interpreter.

'Assurance of interpretation' is thus guaranteed by having the ap¬
propriate interpretant (or, in my terminology, the relevant context) as¬
signed to the sign:object diad as its third member. Peirce himself
proceeds to give here one more trichotomy, of the types of 'assurances
of interpretation'.2
Perhaps the most suggestive rendition of Peirce's 'interpretant' as
'context' is to be found in the following passage (Shapiro, 1983):

"...Each sign has one immediate interpretant, but any finite number of
dynamic interpretants, beginning with zero. Peirce's theory of inter-
pretants argues that semeiosis is teleological, that interpretation is
shaped by a normative goal..." (1983, p. 55; emphases are mine; TG).

An equally suggestive rendition is given by Short (1981) in his discus¬

sion of 'immediate', 'dynamic' and 'final' interpretants:

"...First, the idea of immediate interpretant presupposes that of a ground

of interpretation: the significance of a sign is its immediate interpretant
and this is its grounded interpretability, which the sign possesses regard¬
less of whether it is actually interpreted. Second, the idea of a final inter¬
pretant presupposes that of a goal of interpretation. For, apart from such
a goal, consideration of a sign would not lead the interpreter to any
'destined' conclusion... any conclusion should remain just as good..."
(1981, p. 214; emphases are mine; TG).

1 I owe this suggestion to T.K. Bikson (in personal communication).

2 Assurance by instinct, by experience and by form ; see Shapiro (1983, p. 60). If I am not
mistaken these three correspond to the Kantian propositional modalities, 'possible truth',
'factual truth' and 'analytic truth', and thus presumably also to Peirce's 'abduction',
'induction' and 'deduction, respectively. For further discussion of those, see Chapter 4,
below.
 The Linguistic Code: The Iconicity of Grammar 73

Rather than follow the elaborate sub-classification of signs and their

interpretants offered by Peirce and later interpreters, I will suggest in
the following section that Peirce's notion of 'interpretant' be equated
with a Wittgensteinean notion of 'context'.

3.2.2. Steps to an ecology of contexts3

In this section I propose to integrate Wittgenstein's notion of context-

dependent meaning with Peirce's notion of 'interpretant'. I will do so
initially by fiat, declaring them to be one and the same. This will be done
through an axiom concerning the application of 'context' (or 'inter¬
pretant') to a cognitively-based4 * * theory of communication ('information
processing'):

(1) Axiom concerning the pragmatics of 'relevant context':

(a) The relevant context for the communication of knowledge

('information', 'belief') from one mind to another is not
some objective context; nor is it a set of propositions about such
objective context.

(b) Rather, the relevant context for the communication of

knowledge ('information', 'belief') is itself some knowledge,
information or belief held by some interpreter, i.e. by some
participant ('speaker', 'hearer') in the communication.

As one can see, axiom (1) makes it easy to understand why Peirce's
'interpretant' has received such a wide range of potentially conflicting
renditions. Once again we have come face to face with the three prover¬
bial blind men describing the 'same' elephant. One may, on the one
hand, choose to dwell on the fact that the 'relevant interpretant' is not
independent of the mind of the interpreter. Given such a choice, it
would be easy to equate the two. But this merely defers enumerating all
the highly specific types of context that figure as the most normative
('likely7, 'important') interpretants. In doing so, one downplays the
'more objective' aspect of 'interpretant'.
One may, on the other hand, choose to dwell on the contexts themsel¬
ves, for the moment ignoring the fact that those contexts can only play

3 With obvious indebtedness for the title of this section to Gregory Bateson's Steps to an
Ecology of Mind (1972).
4 By 'cognitive-based' I mean a theory of information processing that emphasizes the fact
that information is transferred from the mind of one organism to the mind of another. Such
a theory must thus take into account the cognitive and perceptual mechanisms that
constrain the way the mind/brain processes information.
74 Mind, Code and Context

their 'interpretant role' inside the mind of the interpreter. In doing so,
one tends to underplay the role of the interpreter's mind -- i.e. the 'more
subjective' aspect of 'interpretant'.
The obvious resolution to this artificial dilemma must again be Kan¬
tian in nature. It must involve the recognition that 'interpretant' -- the
relevant context for communication -- is an interactive product, to which
both mind and environment contribute in liberal measures.5
Let us turn now to consider the major sub-divisions of context
relevant for the use of language in communication. One must recognize
that this list is to some extent open ended, and rightly so. Nonetheless,
both the specific categories and the three major foci under which they
are grouped are highly stable and well attested in the traditional lin¬
guistic literature.6

(a) The generic focus: Shared world and culture

(i) Knowledge of (and beliefs concerning) the so-called real

world (including society and culture), assumed by the
speaker to be held by hearer, as member of the same speech
community ('culture'), and manifest first and foremost in the
commonly-held lexicon.7 This tacitly subsumes whatever
universal cognitive constraints that underlie the human
mind, including any universal capacities for logical in¬
ference.8 It also subsumes whatever perceptual universals
that underlie the human sensory apparatus.9 At least in prin¬
ciple, it thus also subsumes Kant's Transcendental Schema
of time-and-space.10

5 It is reasonably dear that Peirce occupied himself, on different occasions and in different
contexts, with different aspects of 'interpretant', a fact that perhaps explains the great
variety of definitions of this elusive entity he seems to have produced.
6 See for example Cole and Morgan (eds, 1975), Cole (ed. 1978, ed. 1981) or Levinson (1983),
inter alia.
7 The lexicon — dictionary -- is the most conspicuous repository of the shared generic
knowledge of the (relatively stable) properties of entities, institutions, customs, beliefs etc.
Much of Aristotle's 'interpretant' in his De Interpretation boils down to generic, lexically-
coded knowledge.
8 This corresponds, broadly, to Kant's analytic knowledge. See Johnson-Laird (1983) for an
extensive survey of such universal capacities from an experimental psychologist's point of
view. I suspect Johnson-Laird has over-estimated by a considerable margin the scope and
universality of these components of our shared generic background. But some capacities
to do deduction and induction (and obviously also abduction, on which Johnson-Laird is
silent) are surely cognitively based and human-universal.
9 See for example the discussion of color perception and its linguistic significance in Kay and
McDaniel (1978).
10 I.e. Kant's apriori synthetic.
 The Linguis tic Code: The lconicity of Grammar 75

(b) The deictic focus: Shared speech situation

(ii) Deixis:
The knowledge, shared by the speaker and hearer in a par¬
ticular speech act and by virtue of being together on the
same scene at the same time, of the immediate ('deictic')
speech situation. This includes, among others, the shared
reference for T and 'you', 'now' and 'then', 'this' and 'that',
or 'here' and 'there'.

(iii) Socio-personal relations:

Knowledge, shared by the speaker and hearer, in their respec¬
tive roles , of their respective socio-personal relation. This in¬
cludes respective power, status, long-term social goals,
obligations, entitlements, needs and expectations, most
specifically as they are relevant to the communicative trans¬
action at hand .

(iv) Speech-act Teleology:

The shifting goals of the communicative transaction, clause
by clause, primarily from the speaker's perspective. The
more localized , linguistically coded speech act designation of
clauses (declarative, interrogative, manipulative, etc.). This
may or may not also include the speaker's information¬
processing goals, such as foregrounding vs. backgrounding,
focus, emphasis or contrast, and the designation of impor¬
tant topics.11

(c) The discourse focus: Shared prior text

(v) Overt and covert propositions:

Knowledge, shared by the speaker and hearer in a particular
communication transaction, of the specific discourse that
has been transacted, in particular the immediately preced¬
ing discourse.12 This includes the specific propositions com¬
prising the uttered text, but also whatever other entailed

11 See discussion in Levy (1979) as well as further below. Whether one wishes to include here
detailed information-processing goals remains a thorny issue. Such goals are hierarchically
layered and tend to open the door for infinite meta-regress. Further, speakers are increas¬
ingly less conscious of harboring these increasingly more abstract processing goals. One
must recognize here the potential for another Wittgensteinean continuum. For further
discussion of 'degree of consciousness', see Chapter 7, below.
12 Pete Becker (in personal communication) refers to this as 'prior text'.
76 Mind, Code and Context

propositions the speaker assumes that the hearer can derive

from the text by whatever means. It also includes such en¬
tailed propositions the speaker assumes that the hearer can
derive from the thematic structure of the text, again by
whatever means.

(vii) Meta-propositional modalities:

Knowledge, held by the speaker and hearer and shared to
various degrees (and not always symmetrically), of the
strength of their respective belief, certainty, evidential
support or valuative preferences, all pertaining to the
propositions comprising the specific discourse (or entailed
from it). This also includes some probabilistic assessment of
the strength of each other's beliefs and preferences.13

One must, finally, own up to the existence of an irreducible residue, a

recalcitrant escape clause concerning the open-endedness of 'context'.
This residue can never be fully captured, however exhaustive and
refined one's taxonomy may be. I refer here to the provision, seemingly
unprincipled yet sine qua non of any empirical pragmatics, which in
principle admits any information at all as legitimate context (or
'interpretant') in natural communication, provided the speaker deems it
relevant to the transaction -- on whatever grounds -- and available to
the hearer, by whatever means.14

3.3. ’Code’ and ’designatum’

We now turn to deal with the other two components of the semeiotic
relation, sign and designatum. We will begin by discussing a question
that has traditionally been rather controversial, namely, what precisely
is the designatum of the linguistic sign.

3.3.1. The designatum: Some background

"...Now spoken sounds are symbols of affections in the soul, and

written marks are symbol of spoken sounds. And just as written
marks are not the same for all men, neither are spoken sounds.

13 See discussion in Chapter 4, below.

14 One's private mental life. Divine revelation, hunches, intuitions, inspiration and the like
are all perfectly legitimate grounds for assuming the relevance of some context. Whether
— and to what degree — such sources insure successful communication is a separate,
empirical, issue. But there is no principled grounds on which this open-endedness of
context can be ruled out, nor is it particularly desirable to attempt such strict delimitation.
 The Linguistic Code: The Iconicity of Grammar 77

But what these are in the first place signs of -- affections of the soul
-- are the same for all; and what these affections are likenesses of
-- actual things — are also the same..."

Aristotle, De lnterpretatione
(J.L. Ackrill, tr., 1963)

In this terse passage Aristotle opened, for all eternity, the agenda of
Western semeiotics. The arbitrariness of the linguistic symbol -- the
sound code -- is declared boldly. The two-step coding relation is out¬
lined, between 'sounds', 'affections of soul' and 'actual things'. And an
iconic relation -- likeness — is posited between mental entities and the
external reality they stand for. (This 'likeness' was expressed through
Aristotle's 'forms'; see Chapter 1, above). What is more, the relation be¬
tween mind and reality is said to be universal.
Aristotle's two-step coding relation has survived to this day in the
discussion of language by both philosophers and linguists. It is com¬
monly given as:

(2) LANGUAGE—THOUGHT—REALITY

Consider, for example, the following rendition from Haiman (1980):

"...Since the transformational revolution, it has been claimed that the

structure of language reflects the structure of thought, and that its study
provides a window on the mind... I contend that the structure of
thought in its turn reflects the structure of reality to an extent greater
than is now fashionable to recognize... (1980, p. 537)

In this passage Haiman argues specifically against Chomsky's dismissal

of any systematic coding relationship between 'mind, and 'reality'.
Chomsky's position is to some extent predictable from his Cartesian
rationalism. Implicitly, Chomsky construes the linguistic code (see in
particular Chomsky, 1968) as involving a bi-partite coding relationship
between 'language' and 'thought/mind', without worrying about
whether thoughts may code some 'reality':

"...our interpretation of the world is based in part on representational

systems that derive from the structure of the mind itself and do not mir¬
ror in any direction the form of things in the external world..."
(Chomsky, 1981, p. 3)
78 Mind, Code and Context

At another point, Haiman seems to be throwing in a skeptical -- per¬

haps a Kantian or Lockean - hedge, suggesting that perhaps 'external
reality' is not an issue here, but rather perceived reality :

"...many linguistic universals reflect, in a rather obvious way, our

common perceptions about our world..." (to appear, p. 3)

However, this passage may be also read with 'our common perceptions'
standing for 'thought', thus allowing at least tacitly for the second
coding relationship, the one between 'thought' and 'external reality'.
A different, more expanded and more explicit argument for a bi-partite
Aristotelian coding relation has been advanced by Verhaar (1983). Ver-
haar considers first a triadic code, with two coding relations.

(a) 'Internal': The relation of sound to meaning; and

(b) 'External': The relation of meaning to some X.

I will henceforth refer to these two relations as, respectively:

(a) The sign-relation ('S-relation' for short); and

(b) The experience-relation ('X-relation' for short).

Verhaar points out, I think correctly, that — at least for the lexical code
— the S-relation is rather trivial. The more interesting and difficult issue
concerning the lexicon is the precise nature of the X-relation, more
specifically, what exactly is the designatum 'X':

"...The "target" of consciousness is some "X"..." (1983, ms, p. 2)

Verhaar next proceeds to discuss a number of cases where a triadic

relation is not feasible, and where the linguistic code seems to boil down
to a diadic relation. His first case involves the well-known self-inclusion
paradox, or 'consciousness of consciousness':

"...Now, it is by all means possible to be conscious of one's being con¬

scious of X, but in such a state of consciousness, the original X has dis¬
appeared "out of the picture"...the original consciousness, of which one
is now conscious, has lost its intentionality and therefore also its X..."
(1983, ms, p. 2).

There are two issues that Verhaar neglects to consider in his argument.
First, while the X of self-reflection may not be the same X ('external/per¬
ceived reality'), a definite new X has come into being via self-reflection,
namely 'the "being conscious" of which one is now conscious'. The X --
 The Linguistic Code: The Iconicity of Grammar 79

and with it the X-relation - is not lost but only shifted. Second, the dif¬
ficulty in identifying the X here is not specific to self-reflection, but is
rather a feature of all predications reporting mental events, as in:

(3) a. Mary knew that John went to Seattle.

b. John felt a great pain.
c. John contemplated the apple.

The designatum ('X') of (3a) is not the external event of John's going to
Seattle, but rather the internal mental event of Mary's knowing it.
Similarly, the 'X' of (3b) is not the pain itself, but rather John's feeling it.
And the 'X' of (3c) is not the apple, but rather John's contemplating it.
As we shall see further below, the fact that the 'X' in an X-relation (or
even one of the terms in the S-relation) is a 'subjective' mental entity in
no way disqualifies it from being a legitimate designatum, nor does it
impinge upon the more-than-diadic nature of the linguistic coding rela¬
tion.
The second argument used by Verhaar, in his attempt to challenge tri¬
adic Aristotelian semeiotics, involves the grammatical coding of the
pragmatic function of 'topic:

"...There is, in syntax, a kind of genuine X-relatedness that is not

"iconic": That of topicalization. It is not "iconic" because topicalization is
not "triadic", but dyadic: it is like "pointing", and comprises therefore
only (of) that pointing and that which is being pointed at..." (1983,
ms, pp. 4-5)

As we shall see further below, Verhaar's perceptive comments on

topicalization in fact pertain to the entire functional realm of discourse-
pragmatics, as coded by morpho-syntax. When we discuss this issue
more fully, I will attempt to sketch out the reasons why the absence of
an obvious — even conscious -- mental designatum, should not in itself
be taken to mean the total lack of such a designatum.
The third argument raised by Verhaar against Aristotle's triadic
semeiotics revolves around the issue of language and culture . In this con¬
nection, while acknowledging the Sapir-Whorf Hypothesis, Verhaar
quotes Cassierer's (1933) account of the way children acquire their first
language:

"...language does not just come into a world of objective perceptions

which are already there... [it] is itself a mediator in the formation of ob¬
jects; it is, to a certain extent, the mediator par excellence , the most im¬
portant and precious instrument for the conquest and construction of a
true world of objects..." (Cassierer, 1933, p. 33)
so Mind, Code and Context

This argument is indeed Verhaar's most general, and thus presumably

the most challenging. It is framed primarily in terms of the lexical code.
It opens the door, potentially, for a would-be Wittgensteinean rejection
of either the distinctness of language from thought (if one goes by the
Investigations ), or of the distinctness of thought and/or language from
'reality' (if one goes by the Tractatus ). In the former, Wittgenstein char¬
acterizes Augustine's Platonic approach to the linguistic code as fol¬
lows: t ■ »

"...Every word has a meaning. This meaning is correlated with the

word. It is the object for which the word stands..." (1953; 1; p. 2)

This approach seems to dispense with Verhaar's X-relation. (Much like

Chomsky, Augustine's faith in a systematic relation between 'reality'
and 'mind' was indeed weak). The bulk of the Investigations argues
against such Platonism; but it is apparently possible to construe the ar¬
gument not only as a rejection of absolute categories, but also of all
categories. And if so, one must also dispense with Verhaar's S-relation.
To compound matters, an extreme would-be Tate' Wittgensteinean
could also argue for giving up Verhaar's X-relation, roughly along the
following lines:

(a) Meaning does not map extensionally to some objective

reality; rather, it maps onto a contextually-mediated -- thus
highly subjective -- reality.
(b) Since the external 'X' cannot be defined rigorously, the X-
relation between meaning and 'X' lacks a proper desig-
natum; therefore such a relation does not exist.

I myself feel that such an extreme interpretation is unsupported by 'late'

Wittgenstein's text, which does not argue for a total contextualization of
meaning, nor for total subjectivization of all external designata. In this
connection one may recall the following passage from the Investigations :

"...But then the use of the word is unregulated, the 'game' we play
with it is unregulated". -- It is not everywhere circumscribed by rules; but
no more are there rules for how high one throws the ball in tennis... yet
tennis is a game for all that and has rules too..." (1953; 68; p. 33; em¬
phasis is mine; TG)

One may perhaps infer a rejection of the 'X' ~ and thus of the X-rela¬
tion — from the following passage of the Tractatus:
 The Linguistic Code: The Iconicity of Grammar 81

"...the limits of my language means the limits of my world..." (1918, p.

115)

arguing that 'since language and world are one and the same' there is
no sense in talking about a relation between 'X' and itself, unless it was
a rather strange reflexive relation. Again, I find this an unnecessarily ex¬
treme reading of the 'early' Wittgenstein. Let us turn now to outline a
more realistic approach to language as a communicative code.

3.3.2. Language as a complex coding system

In the discussion thus far I have deliberately followed a tradition that

must now be transcended. That tradition, in discussing language as
semeiotics ('coding system'), has been for a long time beclouded by al¬
most total disregard for some rather fundamental facts concerning
human language. Chief among those is the fact that language is a com¬
plex, multi-level code, rather than a relatively simple matching of sound
sequences with lexical meanings. The reason why the complexity of the
linguistic code is so crucial for our discussion of the designatum — and
the nature of both the S-relation and X-relation -- involves all three
aspects of the semeiotics of language:

(a) The sign level:

Each of the three major functional realms of language -- lexicon, proposi¬

tional semantics and discourse pragmatics — displays its own highly
specific coding properties (or 'coding processes'). It is thus incumbent
upon us to define the S-relation in language three separate times, in
terms that are highly specific to each functional realm.

(b) The mind level and the S-relation:

(i) The 'internal', mental entity associated with the linguistic

sign in the S-relation is highly specific to each of the three
functional realms. Our discussion of this mental entity -- and
of its S-relation -- must be thus highly specific to the func¬
tional realm under study.

(ii) The S-relation between the linguistic sign at one end and the
mental entity at the other is not a single coding step. Rather,
it involves several successive steps of coding and re-coding.
And this complex process of successive coding is highly
specific — cognitively and neurologically -- to each of the
three functional realms.
82 Mind, Code and Context

(c) The level of experienced reality:

(i) The external 'X', paired with the mental entity through the
X-relation, is itself highly specific to each of the three func¬
tional realms. Our discussion of this 'X' and its X-relation
must be thus highly specific to the functional realm under
discussion.
t . »

(ii) Finally, the X-relation between the external 'X' and whatever
mental entity it may code, may also be a multi-step coding
process. It may involve highly specific perceptual,
neurological and cognitive steps.

The complex, multi-coding nature of both our S-relation and X-rela¬

tion, while well documented in the neurology and psychology of both
cognition and perception, is largely outside the immediate scope of our
discussion here. We will thus proceed to discuss the three functional
realms coded in human language.

3.3.3. Functional realms coded by language

"...When one shows someone the king in chess and says: "This
is the king", this does not tell him the use of the piece -- unless
he already knows the rules of the game up to this last point:
the shape of the king... The shape of the chessmen
corresponds here to the sound or shape of a word..."

L. Wittgenstein,
Philosophical Investigations
(1953; 31; p. 15)

In the above passage Wittgenstein coined, perhaps unintentionally,15 a

lovely metaphor for both the reality and complexity of the linguistic
code. The complex, multi-level meaning system underlying the game of
chess, and the codes used to make that system manipulable by the

15 Wittgenstein - unlike Peirce -- deals almost always with the most obvious level of
semeiology, the sound-coding of lexical meaning. However, Wittgenstein's reference to
'rules' is clearly not limited to lexical-semantic regularities (as in the definition of the
chessmen's "roles"). Rather, it also pertains to the rules governing possible board positions,
moves, and coherent move sequences ('Sicilian Defense', 'King's Pawn Opening'). Extend¬
ing Wittgenstein's discussion to the other two functional realms coded by language —
propositional semantics and discourse pragmatics - is thus natural.
 The Linguistic Code: The Iconicity of Grammar 83

players, indeed mirror, in a simplified microcosm, the multi-level func¬

tional system of language and its S-related codes:

(i) Lexical meaning:

Taken broadly, the chess pieces are all lexical words; so are the defini¬
tion of the various locations ('squares') and their properties ('colors').
Complex spatial configurations of squares, where pieces may make well-
defined clusters , are then complex lexical items ('compounds', 'derived
vocabulary'). And this again mirrors an important property of the
human lexicon.
The shape of the king is not, per se, an expression of its lexical mean¬
ing (i.e. its role in the game). Rather, it is the code-expression ('sign') of
that role, a highly concrete visual code standing in this metaphor for our
sound code. Unlike the predominantly symbolic-arbitrary sound-code
of the human lexicon ('king', 'Konig', 'shah', etc.), the visual shape of
the king is to some extent -- residually -- iconic; although many chess-
sets use a much more abstract -- increasingly arbitrary -- lexical code for
the chessmen.

(ii) Propositional information:

The spatial positions of the individual chess pieces are states, and
their descriptions are state propositions. The moves made by indivi¬
dual pieces from one spatial positions to another are events, and their
descriptions are event propositions.
In the human mind, then, information about specific states or events,
concerning specific individuals at specific times and places, is coded by
propositions. And human language, in turn, codes those mental
propositions by syntax; syntax specifies how words are combined,
together with grammatical morphemes to make sentences; and senten¬
ces are the signs that stand in an S-relation to mental propositions.

(iii) Discourse-pragmatic function:

The concatenations of individual moves into well-defined move-se¬

quences, the hierarchic meta-structure of those combinations from the
lowest local level, through increasingly global ('tactical' or 'strategic')
levels, is governed by discourse pragmatics. Through discourse-prag¬
matic conventions or rules, propositions expressing individual states
('positions') and events ('moves') receive their thematic coherence; they
become meaningful within the multi-propositional structure of the dis¬
course -- the game.
84 Mind, Code and Context

The functional realm of discourse-pragmatics is jointly coded in

human language, together with propositional-semantics, by sentence-
level syntax. One may, to quite an extent, identify within sentential syn¬
tax the code elements that bear specific responsibility for coding
propositional-semantic information, and those that are more specifical¬
ly responsible for coding discourse-pragmatic function.
If we are to identify the designatum ('X') associated with human lan¬
guage, and at the same time understand the complex, multi-level coding
relation of 'X', 'language' and 'mind', we must pursue the investigation
separately for the three main functional realms, (i) meaning, (ii) infor¬
mation, and (iii) function.

3.3.4. In search of the designatum

"...Enormous simplifications were possibly necessary to

carry a deeper truth on the surface of a mass of unsorted
detail. That was, after all, what happened when history was
written; many, if not most, of the true facts discarded..."

A. Powell Temporary Kings

(1973, p. 27)

In this section we will consider further the questions raised by Ver-

haar (1983) concerning the S-relation and X-relation that bind together
the linguistic code, its mental designatum and the latter's experiential
designatum. We will treat each functional realm separately.

3.3.4.1. The lexical code

a) The S-relation

The nature of the sound-code via which the mental lexicon of con¬
cepts is coded is relatively well understood empirically. It is also per¬
haps slightly less controversial philosophically. Cross-cultural diversity
in lexical-semantic organization indicates that the mental designatum
associated with the S-relation (and thus the experiential designatum
that stands in a X-relation to it) is in part culturally mediated. Culture as
a context for meaning is indeed one sub-category of context (see above);
we will discuss it more fully in Chapter 9, below.
Whatever one may choose to say about the existence and nature of the
experiential designatum, the meaning of a word — whatever it may be —
 The Linguistic Code: The Iconicity of Grammar 85

must itself be a mental entity. Each lexical mental entity ('concept') thus
has some sound-code label attached to it.16
For the speaker, the lexical code-label serves in the retrieval and trans¬
mission of intended lexical meanings. For the hearer, the code-label
evokes lexical meanings intended by the speaker, together with,
presumably, other meanings that were not necessarily intended by the
speaker. Presumably, the code-label also serves as a major mnemonic
device for the storage (and retrieval) of conceptual knowledge in the
mind/brain. (The other major mnemonic device is of course meaning it¬
self, which presumably is represented in the mind not as an unordered,
and thus unretrievable list, but rather as a system of semantic interrela¬
tions).
Finally, an enormous body of experimental psycho-linguistic litera¬
ture demonstrates, through specific behavioral consequences, the men¬
tal reality of both the sound code and mental lexical meaning that par¬
take in the lexical S-relation.17

(b) The X-relation

While the reality of the S-relation associated with the lexicon is rela¬
tively well established, Verhaar's X-relation -- and thus the question of
what is the experiential designatum of mental lexical item -- remains
controversial. The cleavage lines in this controversy follow the predict¬
able philosophical division:

16 The sound-label can be substituted by other 'superficial coding' labels. It can be changed
into an alphabetic code that represents the sounds more abstractly, with a wide latitude as
to how faithful the written code is to the spoken one. Further, lexical meaning can be coded
pictographically or - as often evolves from pictographic codes ~ ideographically . Finally,
lexical meaning (as well as the entire language) may be also coded gesturally, as in ASL.
Thus, if the same person is a fluent speaker, a literate writer and also a user of ASL, that
person is capable of controlling potentially many codes pertaining to the very same mental
entities.
17 For an exhaustive up-to-date review of the literature on what psychologists call lexical
access', in terms of the mental reality - and experimental- behavioral consequences - of
both the code-labels and lexical meanings, see Carr (1985). For details concerning the
storage and retrieval of words and mental categories, see Rosch and Lloyd (eds,1978). Smith
and Medin (1981), Anderson and Bower (1973), Neely (1976,1977), Nissen (1976) or Kintsch
(1974), inter alia . For many details concerning the neurological reality of the lexicon and its
sound-code, see Patterson and Marcel (1977), Pirozzolo and Rayner (1977), Saffran and
Marin (1977), Searleman (1977), Shallice & Warrington (1975), Whittaker (1983), inter aim .
For details of the mental reality and behavioral consequences of the sound-code itself, see
Adams (1979), Cole and Rudnicky (1983), Ganong (1980), Klatt (1979), Luce and Pisoni
(1984), Nooteboom (1981), Cohen and Nooteboom (eds, 1975), Pollatsek and Carr (1979),
inter alia.
86 Mind, Code and Context

(i) For extreme rationalists, the X-relation is trivial, presumably

because of the primacy of innate ideas (the mental designa-
tum), and also because of the alleged lack of systematic iso¬
morphism between mind and reality. Thus, language may in¬
deed be 'the mirror of mind', but mind is not 'the mirror of
reality'.

(ii) For extreme empiricists, the X-relation between external (or

'perceived') reality and mental entities is taken for granted,
given the primacy of world (or 'percept') over concept. At its
extreme logical-positivist, experiential designata are referen¬
tial, real-world entities standing in an extensional relation to
mental concepts.

(iii) For Kantian constructivists, the X-relation is presumably

highly systematic, complex, bi-directional and interactive. It
is — to add the Peircean-Wittgensteinean element — also a
context-mediated relation.

But what is then the experiential designatum of our mental lexicon?

The bulk of our generically-shared mental-lexical concepts could not
possibly stand in an X-relation to referential designata, but only to clas¬
ses of potential experiential designata. Still, mental concepts are
formed, contextually, interactively, through the experience of individual
referential instances.18 It is thus primarily when placed in the context of
specific proposition, about specific individuals partaking in specific
states or events, anchored at specific times and places, that lexical items
stand in an X-relation to experiential designata. Therefore, in order to
determine the experiential designata of lexical items ('concepts'), one
must study the experiential designata of propositional information.

3.3.4.2. The propositional code

(a) The S-relation

As noted above, propositional-semantic information is coded in

human language by syntactic structure. 'Structure' is a much more
abstract and complex coding instrument than the sound-code. At its
most concrete, syntactic structure is made out of the following three
coding devices:

18 See discussion in Ch. 2, above, concerning the role of typical referential exemplars in
constructing prototype-like generic categories.
 The Linguistic Code: The Iconicity of Grammar 87

(i) grammatical morphemes

(ii) intonation contours
(iii) linear order

Grammatical morphemes are phonologically coded much like the lexicon

(from whence they evolve over time). Their 'meaning' however, is not
lexical but much more general and abstract.19 Further, most grammatical
morphemes — much like the bulk of word-order and intonational devices
that make up syntax -- are not involved in the coding of propositional in¬
formation, but rather in the coding of discourse-pragmatic function.
The propositional information coded by syntax pertains primarily to
three major aspects of the state or event:

(4) (a) Designating the roles of the participants in the state or

event (agent, patient, recipient, instrument, associate,
beneficiary, manner, possessor, predicate, time, place
etc.);
(b) Designating the predication type ('verbal' vs. 'adj¬
ectival' vs. 'nominal' predicate); and
(c) Identifying the transitivity value of the proposition
('transitive' vs. 'intransitive'; 'state' vs. 'process' vs.
'action').

When combined together, aspects (4a,b,c) above make up the proposi¬

tional frame, sometimes referred to by philosophers as 'logical syntax'
or 'logical structure' (see e.g. Carnap, 1959). When further combined
with specific lexical items, propositional frames are transformed into
full-fledged propositions, bearing information concerning the state or
event (i.e. who did what to whom, where, when, with what, for whom,
with whom, how, why or under what circumstances, etc.).
Sentences (or 'clauses'), which are syntactically-coded propositions,
are thus the signs that stand in an S-relation to propositions. And
propositions are, in turn, the mental entities coded by sentences. Since
both linguistics and psychology have established the independent
reality of both sentences and propositions beyond any reasonable
doubt, we will forego belaboring the point here.20

(b) The X-relation

At first glance, an X-relation between mental propositions and some

experiential designatum ('states', 'events') seems a safe bet. This seems

19 For an extensive discussion of the relation between lexicon and grammatical morphology,
see Givon (1984a, Ch. 3).
20 For extensive surveys of the linguistic evidence, see Chafe (1970,1979,1980, ed.), Langacker
(1984) or Givon (1984a), inter alia . For an equally extensive surveys of the experimental
psychological literature, see Kintsch (1974,1977), Kintsch and Keenan (1973), Kintsch and
van Dijk (1978), van Dijk and Kintsch (1983) or Johnson-Laird (1983), inter alia .
88 Mind, Code and Context

true even in spite of the somewhat murky status of the triadic relation
between language, thought and reality. In the space below I will attempt
to illustrate how the experiential designatum of propositions, much like
that of lexical items, is a context-dependent, pragmatically-mediated en¬
tity.
Consider the following, eminently transparent, proposition describ¬
ing a rather concrete event:

(5) A man was beating a dog

Suppose you reported event (5) to me; suppose the mental proposition
expressed in (5) referred ('stood in an X-relation') to some real event;21
suppose, further, that you meant to report that event, which you had in
fact witnessed, truthfully and accurately. The mental proposition coded
by (5) refers to what ('beating') some agent ('a man') did to some patient
('a dog'). There seems to be, at first glance, a simple isomorphic X-rela-
tion between the proposition in (5) and the event it purports to code.
But this seeming isomorphism dissolves rather quickly under the glare
of even the most cursory inspection.
Suppose you witnessed that event from about 100 yards away, with
the following details registering on your retinas rather sharply:

(6) (a) The man was beating the dog with a stick, which he
held in his left hand;
(b) The dog was tied to a tree, with a yellow leash;
(c) The man kept raising and lowering the stick at certain
intervals (say every 3 seconds);
(d) The dog kept yelping, rather pitifully, producinga thin,
high-pitched, rather unnerving wail;
(e) The man was tall, dark-haired, athletic, dressed in a
green jogging outfit;
(f) The dog was a small, black-and-white off-spaniel mutt;
(g) It was a bright, cool, breezy day, the 29th of October,
around 10:27 am;
(h) The whole scene was located on a green lawn, in the
park above the river;
(i) You were standing 107 yards away, to the south-east, at
the bottom of the fir-covered butte.
(j) There was another man watching the scene from about
70 yards on the opposite side; a woman holding the

21 See further discussion in Chapter 5, below, concerning the domain of reference of linguis¬
tically-coded information, the so-called 'Real World', and its relation to the 'Universe of
Discourse'. Verhaar's first exception to the straight-forward X-relation between 'thought'
and 'reality' indeed involves thought as the object ('designatum') of thought.
 The Linguistic Code: The Iconicity of Grammar 89

hand of a small child was walking diagonally across the

lawn, coming up to about 40 yards from the man-and-
dog scene, then slowly withdrawing toward the public
bathrooms; other people -- both adults and children --
were playing all over the park;
(k) On the freeway in the background, about half a mile
away, cars were zooming along at a fast clip.

Your reporting (5) to me was prompted by my inquiring: "What hap¬

pened at the park?". All you told me, upon a certain reflection, was (5).
Yet, it turns out that you also saw, at the very same time and place and
from the same vantage point, all of the details in (6a-k) above. I know
this because when I indeed persisted in asking, you readily volunteered
all the added details. And you could have, had I persisted further,
added even more. In fact, the sky is the limit* to what you could have
added. Still, to begin with, you contended yourself with reporting only
(5). Why?
The point of the exercise is by now transparent:

(i) There is, in principle, no limit to the detail that could have
been included in your description of that rather concrete, per¬
ceptually transparent, seemingly unambiguous 'single
event' that you saw. We quit at (6k) for lack of time and
space, but could have gone on forever.

(ii) In principle, the choice of what to include and what to ex¬

clude in descriptions depends on one's pragmatic framing,
i.e. on one's judgements of saliency, relevance, importance.
None of these judgements are 'objective'; nor can they be ar¬
rived at deductively or inductively; they are all are a matter
of point of view; of context; or the relevant interpretant.

There is a vast chasm, in principle, between the mental propositional

representation of states and events and the 'real' (or 'experienced')
states or events they purport to code. There is an equally vast chasm be¬
tween mental pictorial representation of states and events and what ac¬
tually registers on the eye's retina.22 The mind/brain is highly selective
about what it sees, retains or reports. And this selectivity is pragmatical¬
ly controlled, except when processing has become fully automated.23

22 Viz Pylyshyn's (1973) "What the mind's eye tells the mind's brain". The problem we have
here with propositionally-coded information about 'experience' is in no way unique. The
Interpretant is ubiquitously here, as it is at other levels of information processing.
23 See Chapter 7, below, for extensive discussion of the automation ('routinization) of
information processing at various levels.
90 Mind, Code and Context

What we noted above should in no way impel us to beat a hasty

retreat for the illusory safety of Platonic (or Chomskyan) idealism. The
lack of total iconic systematicity in the X-relation between mind and ex¬
perience in no way means a total lack of such systematicity. One's rejec¬
tion of naive, empiricist objectivism concerning 'reality' does not auto¬
matically commit one to embracing an equally extreme Wittgensteinean
mush. In this connection, one may recall Geertz:
t • »

"...To abandon the hope of finding the "logic" of cultural organization

in some Pythagorean "realm of meaning" is not to abandon the hope of
finding it at all. It is to turn our attention toward that which gives sym¬
bols their life: Their use..." ( The Interpretation of Cultures, 1973, p. 405)

Once again, a pragmatic middle-ground is the only empirically viable

alternative to both extremes. The mind and reality interact in the con¬
struction of a mental analog -- or isomorph -- of some 'experienced
reality'. That analog is only similar, but never identical, to whatever it
purports to code. And similarity, the cornerstone of iconic-isomorphic
representation, is a pragmatic, context-mediated relation. The context
or 'interpretant', whether invoked or uninvoked,24 is thus the ubiqui¬
tous silent partner in any coding relation.
Note, finally, that much empirical evidence from experimental
psychology demonstrates how selective people are in constructing
mental representations of states and events. This pertains to directly-
witnessed states and events, as well as to those derived from reading or
oral presentation. The mental representation of states and events is
shown to be highly context-dependent, sensitive to many types of con¬
textual priming. Finally, mental representation of states and events may
remain, under some task conditions, visual rather than verbal-proposi¬
tional.25

24 An apocryphal account has it that Xenophon, seeking a favorable omen from the Delphic
oracle concerning his projected Persian Expedition, tried to outwit the God by not invoking
its name directly. The Pythia came back with the sobering observation: "Invoked or
uninvoked, the God will be there". The interpretant — invoked or univoked -- is just as
ubiquitous.
25 For the details of verbal accounts of witnessed events, see - most conspicuously - Loftus
(1980), as well as many details in Paivio (1971), Potter et al (1977), Tversky (1969,1975), or
Pylyshyn (1973), inter alia . For various aspects of the context-dependent nature of the
storage and retrieval of propositionally-coded experience, see Mandler and Johnson (1977),
Mandler (1978), Kintsch (1974), Kintsch and Green (1978), Rumelhart (1977), Rumelhart
and Ortony (1977), Anderson and Paulson (1978), Anderson and Pichert (1978), Anderson,
Garrod and Sanford (1983), Perfetti and Goldman (1974) or Schank and Abelson {1977),
inter alia. For similar suggestions reported by linguists, see Labov and Waletzky (1967) or
Chafe (1980), inter alia.
 The Linguis tic Code: The Iconicity of Grammar 91

3.3.4.3. The discourse-pragmatic code

3 3.4.3.1. Preamble

We return now to deal with the full impact of Verhaar's (1983) obser¬
vations concerning 'topicalization'. Traditionally, philosophers and lin¬
guists have tended to take for granted their direct, intuitive access to
lexical meaning and propositional information. Both were taken to be
mental entities of which one is conscious. The same has never been true
for discourse pragmatic functions. Both their mental designata and their
experiential designata remain unexplored. Linguists have only recently
begun to understand, just barely, how syntax codes various discourse-
pragmatic functions; how the coding of such functions is distinct from
the coding of propositional information; even what discourse-pragmatic
functions are all about. This is of course natural. Discourse-pragmatics
is a set of highly abstract meta-phenomenological operations and
processing instructions. The processor is often unconscious of using
them in processing the text. What is presented below is, therefore, only
an incomplete sketch.

3.3.4.3.2. Discourse-pragmatic functions

In this section I list the major discourse-pragmatic functions that tend

to be morpho-syntactically coded in natural language. The list is not ex¬
haustive, nor is it the final formulation, given the rather tentative
present state of our understanding.

(7) Major discourse-pragmatic functions

(a) Topicality and reference functions:

(i) The assignment of (relative) topic importance to clausal par¬

ticipants ('subjects', 'objects'), presumably in a way that is
sensitive to their thematic importance in the discourse; this
often intersects with the marking of semantic referentiality
(see Chapter 5, below).

(ii) The referential tracking of clausal participants ('referents')

as they are introduced into the discourse for the first time
('indefinite') or re-introduced ('definite'); the marking of
anaphoric predictability (or 'accessibility) of clausal par¬
ticipants; the assignment of coreference relations in the dis¬
course (see Chapter 6, below).
92 Mind, Code and Context

(iii) The assignment of contrastive or emphatic status to clausal

participants.

This cluster of function is performed by a large number of grammatical

devices, such as case-marking of subject and object, the syntax of tran¬
sitivity, passivization and anti-passivization, articles, demonstratives,
pronouns and agreement, word-order and intonation.

(b) Propositional modality functions:

(iv) The assignment of temporal, aspectual, epistemic, eviden¬

tial, evaluative or speech-act modality to propositions, all in¬
volving various aspects of the speaker's attitude toward the
proposition (see Chapter 4, below).

Syntactically, these functions are coded primarily by verb morphology,

occasionally also with some participation of word-order and intonation.

(c) Thematic coherence functions:

(v) Coding coherence relations between propositions, and be¬

tween larger units, in discourse. This may be done first,
anaphorically, in terms of thematic predictability and
thematic coherence of propositions vis-a-vis the preceding
discourse. Second, it may done cataphorically, in terms of
thematic importance and thematic predictability of
propositions vis-a-vis the following discourse.

(vi) Coding foreground-background status to propositions, i.e.

which ones constitute the gist of the message and which are
peripheral, respectively.

(vii) Coding more specific local inter-clausal relations, such as

sequentiality or counter-sequentiality, temporality, con¬
ditionality, causality, contrast etc.

Syntactically, these functions are coded by either verb morphology or

subordinating or coordinating conjunctions, as well as by word-order
and intonation.

(d) Socio-personal functions:

(viii) Coding the socio-affective relations between speaker and

hearer, in terms of affect, status, obligations, power, needs,
entitlements etc.
 The Linguistic Code: The Iconicity of Grammar 93

Syntactically, these functions may be coded together with speech-acts

(see (iv) above). It is more common, however, for the bulk of these func¬
tions to be coded primarily by intonation, gesture and facial expression.

As should be obvious, discourse-pragmatic functions represent --

without exception -- various facets of the speakers' attitude, perspec¬
tive, or point of view. They may 'be about' the reported propositional
information; they may 'be about' various aspects of the context -- cul¬
tural, situational or textual -- that are relevant to how these propositions
are to be integrated into coherent communication; they may even 'be
about' some discourse-pragmatic functions themselves. Discourse-prag¬
matic functions are mental meta-operations, meta comments, or even
processing instructions.26 Such mental operations can be -- and often
are -- highly abstract; they are clearly 'mental'; they are clearly coded by
syntax; they are clearly, therefore, mental designata standing in an S-
relation to some portion of the linguistic code. Their S-relation, as we
will note below, is also highly isomorphic-iconic. Still, their S-relation is
complex and abstract, it is much less accessible to conscious inspection
than the S-relation holding between mental propositions and their syn¬
tactic code.27

3.3.4.3.3. The experiential designatum of discourse-pragmatic

function: In search of a missing X-relation

As noted earlier, Verhaar (1983) suggests that the semeiotic relation of

discourse-pragmatic function (his 'topicalization') is not tri-partite but
rather bi-partite . That is, there is no experiential designatum ('X') in
these mental operations, and thus no X-relation. On the face of it, this
observation seems rather sound. Unlike for lexical concepts and mental
propositions, one is hard-pressed to identify some real-world or ex¬
periential 'X' that might correspond to discourse-pragmatic functions.
While I do not propose to resolve this problem for eternity at this point,
one word of caution is in order. In the realm of propositional informa¬
tion too, the experiential 'X' need not have real-world extension. It
could have, and often does have, a discourse-universe extension (see
Chapter 5, below). Further, this designatum may also have a purely in¬
ternal mental extension, as in the case of propositions involving verbs
of thinking, knowing, wanting or feeling (see discussion of Verhaar's

26 In this sense, they are also akin to Verhaar's (1983, p. 2) self-reflection paradox, although
perhaps in a sense that Verhaar himself did not anticipate; and certainly not in a sense that
depletes meaning, as 'knowing that one knows...' indeed seems to do.
27 The syntactic structure that codes these operations is also more abstract. In this sense, a
clear meta-iconic relation seems to hold between the degree of complexity/abstractness of
the designatum and the degree of complexity/abstraction of the code.
94 Mind, Code and Context

reflexivity argument and example (3), above). It is, I believe, premature

to rule out the possibility that discourse-pragmatic mental operations
stand in some X-relation to some -- likewise internal, mental — desig-
nata.

3.4. Iconicity and isomorphism in grammar

t . *

3.4.1. Background
"...every animal communication system that is known...either
consists of a fixed, finite number of signals, each associated
with a specific range of behavior...or it makes use of a fixed,
finite number of linguistic dimensions, each of which is
associated with a particular non-linguistic dimension in such
a way that selection of a point along the linguistic dimension
determines and signals a certain point along the associated
non-linguistic dimension... When I make an arbitrary state¬
ment in a human language... I am not selecting a point along
some linguistic dimension that signals a corresponding point
along an associated non-linguistic dimension..."

N. Chomsky, Language and Mind

(1968, pp. 69-70)

In surveying the history of our preoccupation with language as

semeiology, one is struck by the persistent clustering of certain
philosophical propensities, a clustering that seems less-than-accidental
and yet not quite logically necessary. These propensities may be given
as the following pre-empirical postulates:

(i) There is an evolutionary dis-continuity between pre-human

and human communication (cf. Chomsky, 1968);

(ii) Language is a separate module of the mind/brain, not part

of 'general cognition' (cf. Chomsky, 1968; see also Chom¬
sky's various contributions in Piattelli-Palmarini ed., 1979);

(iii) Structuralism in the analysis of language; that is, language

structure can be analyzed independently of its communica¬
tive function (cf. Saussure, 1915; Bloomfield, 1933; Chomsky,
1957, inter alia);
 The Linguistic Code: The Iconicity of Grammar 95

(iv) Some abstract, idealized entity — be it langue or competence —

is the 'object' of linguistic analysis (cf. Saussure, 1915;
Chomsky, 1964);

(v) The synchronic facts of language must be studied in strict

separation from diachronic facts (Saussure, 1915);

(vi) The sign-relation between the linguistic code and its mental
designatum is arbitrary, unlike the obvious iconicity seen in
pre-human communication (Saussure, 1915; Chomsky, 1968).

The Uniqueness of Man doctrine expressed in (i) and (vi) was

espoused by Descartes on religious grounds, and is closely connected
with his belief in innate ideas and their divine source.28 Innate ideas,
those Platonic absolutes underlying the human mind, are presumably
also at the bottom of the idealization of the object of linguistic inves¬
tigation (iv). And that idealization of the data-base (iv) is reflected in the
further idealization in (v), where 'steady state' is abstracted from the
ever-present process of change. The so-called modular approach to lan¬
guage and cognition (ii) is an unprecedented non-Cartesian elaboration
by latter-day structuralists, closely linked to the view of language struc¬
ture as autonomous from language function (iii).
We come at last to the presumed arbitrariness of the human linguistic
code. It seems to be intimately bound with all the other components of
the cluster:

(a) The arbitrariness doctrine is surely implicit in the anti-evolu¬

tionary and anti-developmental approach of the structur¬
alists (i), as well as in the strict separation of synchronic state
from diachronic change (v). (It is only through the three
developmental processes — phylogeny, ontogeny and dia¬
chrony — that the isomorphic pairing of language code and
language function is mediated).

(b) The arbitrariness doctrine is consistent with the modular

separation of language and cognition (ii), since cognition is a
major functional parameter underlying language.

(c) The arbitrariness doctrine is virtually a logical consequence

of the structuralist's insistence on separating the study of

28 Chomsky (1968) and Bickerton (1980) merely substitute DNA for the Divine. See discussion
of the history of evolutionary thinking in Chapter 10, below.
96 Mind, Code and Context

structure and function (iii); functional considerations — i.e.

the needs for coherent storage and quick retrieval -- furnish
the most obvious motivation for an iconic relation between
code and meaning.

(d) Finally, the arbitrariness doctrine is also an implicit conse¬

quence of the idealized data base (iv, v). This is so because
only in studying structures and functions in their com¬
municative context -- i.e. in language use -- can we begin to
appreciate the nature and extent of their isomorphic-iconic
S-relation.

3.4.2. Preliminaries

3.4.2.1. Iconicity and isomorphism

'Iconicity' presupposes 'isomorphism'. An iconic code is necessarily
one that displays an isomorphism to its designatum. As Peirce has
noted, an icon is a sign that bears some similarity to its designatum.
'Iconicity' and 'similarity' are thus closely linked. Here is what Aristotle
had to say on this matter:

"...Things are "alike" [ homoios ] when, though they are not absolutely
the same and though they are different individuals, they nevertheless
have the same form ..." (Metaphysics , p. 206)

And our dictionary defines 'isomorphism' as "...similarity in form..."

(Morris, ed., 1969, pp. 694-695). And form, in turn, involves two com¬
ponents:

(a) A matching number of corresponding points; and

(b) A matching number of corresponding relations.

Isomorphism is thus neither an explanation of nor motivation for iconic

coding, but simply part of its definition.29

29 In this I am diverging from ITaiman (to appear), who asserts that 'isomorphic' repre¬
sentation requires only the matching of points, while 'diagrammatic' representation re¬
quires the matching of both points and relations. I find this distinction arbitrary; it is
supported by neither precedent nor argument. The same goes for Haiman's (1983) attempt
to view iconicity as a 'motivation' ('explanation') for isomorphism.
 The Linguis tic Code: The Iconicity of Grammar 97

3.4.2.2. Motivation and explanation

The motivation for the existence, or for the rise, of iconic code-rela¬
tions in any sub-system within a biological organism falls within the
same domain as all other explanations in biology. It must be first and
foremost functionally explained.30 This is true either for our ultimate un¬
derstanding of the synchronic structure of a system through its bio-cog¬
nitive context, or for our understanding of the developmental mecha¬
nisms by which the system has evolved.31 In biology, this functionalist
approach to explanation has been taken for granted ever since Aristotle.
Thus, consider the following from a standard textbook of anatomy:

"...anatomy is the science that deals with the structure of the

body... physiology is defined as the science of function.
Anatomy and physiology have more meaning when studied
together..." (Crouch, 1978, pp. 9-10; emphases are mine; TG)

The first layer of explanation of iconicity in language is not strictly

empirical; rather it is a pre-empirical, common-sensical postulate, a
meta-assumption concerning the use of human language as a com¬
municative tool. In one guise or another, it has been with linguistics
since its very beginning.

(8) The iconic imperative:

"All other things being equal, a coded experience is easier to

store, retrieve and communicate if the code is maximally
isomorphic to the experience".

Wherever language and its mental representation may have their ex¬
periential designatum, postulate (8) pertains to both Verhaar's 'S-
relation' and 'X-relation'. Otherwise, it always pertains to the S-relation
between the linguistic code and its mental designatum.32
Haiman (to appear) points out that principle (8) is a version of the
traditional one form one meaning or one-to-one correlation between form
and meaning principle. Bolinger renders this principle as:

30 In this connection, note Posner's (1985) discussion of the functional-adaptive motivation

for the hybrid prototype categorization (Chapter 2, section 2.6., above). See also discussion
in Givon (1979, Ch. 1; 1984a, Ch. 2), as well as Chapter 8, below.
31 In language, these developmental mechanisms include evolution, ontogeny and diachrony.
32 Since it is fairly clear that many more perceptual, neurological and cognitive coding steps
(or 'mapping relations') occur in the production and perception of language (and in mental
representation of experience in general), one must assume that principle (8) also pertains
to those other mapping relations.
98 Mind, Code and Context

"...The natural condition of language is to preserve one form for one

meaning, and one meaning for one form..." (1977, p. x)

As suggested earlier, language processing probably involves multiple

coding steps. Within each step, principle (8) manifests itself in ways that
are specific to the processing level involved. Further, the three function¬
al realms coded by language may exhibit the motivating power of prin¬
ciple (8) in ways that are specific to their processing mode (cf. sections
3.3.3, 3.3.4, above). Thus, whenever economy is invoked as a motivating
principle, it must be considered relative to the level and mode of real¬
time physiological and mental processing associated with each coding
level.33

3.4.2.3. Degree of conscious access to iconic code relations

Andersen (1983), in a Peircean vein, points out that the notion of
'similarity' requires the consciousness of similarity in the mind of some
interpreter. In our re-formulated terms (see axiom (1), section 3.2.2.,
above), this follows naturally from the observation that the relevant in-
terpretant -- or context -- for similarity relations is always a mental con¬
struct in the mind of some interpreter. However, 'being in the mind'
may involve different degrees of consciousness. One's degree (and
mode) of consciousness vis-a-vis the three coded realms of language --
lexicon, propositional meaning and discourse-pragmatics function --
may be radically different. One is, no doubt, quite conscious of the S-
relation cum X-relation of at least some lexical concepts. One may be
much less conscious of the S-relation and X-relation of propositional in¬
formation. Although it is still possible to point out those relations, at
least in more concrete cases, to the naive speaker. Finally, one is, on the
whole, quite unaware of either the existence of any mental or experien¬
tial designata of the syntactic structure that codes discourse-pragmatic
function, or of the nature of either its S-relation or X-relation. The
reasons for this differential awareness are transparent. They may be
summarized here as follows:34

(a) The experiential designata of lexical items are more concrete

and obvious than those of propositions; and those of pro-

33 Haiman (1983, to appear) attempts to separate iconic from economic motivation. As I have
argued elsewhere (Givon, 1984b), I think this is a confusing dichotomy. Rather, economy
of mental, neurological or sensory processing ultimately motivates all cases of isomorphic
coding. Postulate (8) is thus as pertinent to cognition, perception, neurology and biology
as it is to language.
34 These three facts concerning the code, the coded and their relation are themselves meta-
iconically matched.
 The Linguistic Code: The Iconicity of Grammar 99

positions are in turn more concrete and obvious than those

of discourse-pragmatic function;

(b) Whatever one-to-one coding relation existing between code

and designatum is thus most transparent for lexical items,
less so for propositions, and least so for discourse-pragmatic
functions;

(c) The code level itself is most concrete for the lexicon (strings
of sounds), less concrete for propositional meaning, and
most abstract for discourse-pragmatic function.

What this gradation implies for the Peircean 'interpretant' remains for
the moment somewhat open. Both Andersen (1983) and Haiman (1983)
seem to reject the abstract iconic relations matching syntax with dis¬
course-pragmatic function as legitimate extensions of Peirce's concept
of icon. They do so in part because of the unclear nature of
'consciousness' and 'interpretant' associated with this isomorphic code.
My own inclination, derived at least in part from recent neuro¬
psychological studies of levels of attention,35 is to resist such apriori
delimitation of the subject matter. It may yet turn out that unwarranted
rigidity will lead us to disregard the most interesting if least obvious
iconic code associated with human language. Many other mental
coding operations are equally unavailable to conscious perusal. The fact
that the discourse-pragmatic code is more complex, abstract, and less
given to conscious scrutiny is no reason for ignoring its existence, its
specific properties and its broader cognitive and biological implications.

3.4 3. Isomorphism, abstraction and prototypicality

"...Hypoicons may be roughly divided according to the mode of
Firstness of which they partake. Those which partake of simple
qualities, or First Firstness, are images ; those which represent the
relations, mainly diadic, or so regarded, of the parts of one thing
by analogous relations in their own parts, are diagrams; those
which represent the representative character of a representamen
by representing a parallelism in something else, are metaphors ...."

C.S. Peirce (1940, p. 105)

In this section I propose to recapitulate the obvious, namely that the

Peircean icon and symbol are merely two extreme poles on a con-

35 See Posner (1985), as well as Chapter 7, below.

ioo Mind, Code and Context

tinuum, one that represents the degree of abstraction in coding rela¬

tions. The increase of abstraction along this scale means, perforce, an in¬
crease in arbitrariness of the code. There is nothing particularly
earthshaking about these observations; they are akin to Bateson's (1979)
note concerning 'patterns and patterns of patterns'.36 Haiman also
recognizes a similar gradation between image and diagram:

"...Although Peirce himself did not emphasize this point, it should be

clear that the distinction between an icon that is an image (like a
photograph), and one that is a diagram (like a stick figure), is also mainly
a matter of degree..." (to appear; p. 16 of ms)

To illustrate with a simple example the continuum between iconic

and symbolic representation, consider the gradual evolution of the let¬
ter A in our alphabet. To judge from its Semitic etymology,37 A is derived
historically from a pictograph representing 'bull' or 'cattle' (Hebrew7!/).
The first step in the abstraction process had already occurred when the
head alone -- rather than the entire body -- was used to represent the en¬
tire animal:

(9)

36 Bateson (1979) views this issue, I think correctly, as a matter of increasingly more general
meta-levels of description, deriving his formulation from Russell's (1919) Theory of Types .
37 The earliest attestation of the script is usually ascribed to Phoenician. However, the name
of the pictograph ?alef has the requisite meaning in Hebrew rather than in Phoenician.
 The Linguistic Code: The Iconicity of Grammar 101

Presumably, the head was considered the most relevant, important,

prototypical part of the body. And this, in turn, expresses a clear hierar¬
chic ('taxonomic') judgement at two distinct levels:38

(a) A functional-biological judgement concerning the crucial,

governing role the head plays vis-a-vis the body; and
(b) A perceptual judgement concerning what are the most
salient visual features of a bull, especially in terms of the
need to distinguish it from other domestic quadrupeds of a
roughly similar shape.

In addition, the representation of the animal's head in (9) already invol¬

ves considerable abstraction, dispensing with more minor structural
features such as hair, skin folds or exact detail of the eye, ear, nose etc.
Bateson considers such abstraction or simplification, inherent in the
modeling ('representation') of any phenomenon, "...in principle distor-
tive of the phenomena to be mapped..." (1979, p. 53). But such
'distortion' is inherent in all processes of generalization, abstraction,
and concept formation; or, for that matter, in all coding and repre¬
sentation.
The taxonomic-hierarchic decision referred to above is just as obvious
in the next step of abstraction, where again more minor details are dis¬
pensed with:

(10)

38 While the two considerations — functional and structural — are formally distinct, it is most
likely that the mind does not keep them apart, leastwise operationally. And this is probably
another manifestation of our postulate (8).
102 Mind, Code and Context

Curving lines are gradually regularized and straightened, to conform to

their prototypical extension. The head now displays the more abstract
iconism of a triangle, carrying the original difference between the wider
top and narrower bottom to its ultimate prototypical conclusion. From
here the path is indeed short to a recognizable version of the letter ?alef.

(11)

What keeps version (11) residually iconic, i.e. isomorphic to the original
bull's head, is its vertical orientation vis-a-vis the scanning eye, an
orientation that preserves the spatial arrangement of the bulls head in
its prototypical posture. But the number of isomorphic points, thus also
of their inter -relations, has shrunk to a bare minimum (horns, ears,
crown, snout). The isomorphism in pattern is still there. The next, his¬
torically attested, development in the Phoenician/Hebrew script effec¬
tively pulled the rug from under this residual iconicity, by turning the
?alef on its side:

(12)

And once the inherent iconism of the prototypical spatial orientation

has been removed, the transition to the Greek alpha — via further inver¬
sion of the capital letter -- is relatively trivial:

The magic of iconic-isomorphic representation is lost altogether, and we

have a unitary symbol standing for a unitary sound or concept.
The gradual process of abstraction seen above may be summarized as
follows:
 The Linguis tic Code: The Iconicity of Grammar 103

(i) Number of isomorphic points:

The abstraction of signs involves a gradual reduction in the
number of points in the sign that map onto points in the
designatum, all the way to the absolute minimum that still
allows an S-relation — one. Presumably, this depletion
proceeds up the hierarchy of generality . That is, more specific
('minor') points are removed first; more generic ('major')
points, those found at higher meta-levels of the classificatory
hierarchy, survive longer.

(ii) Pattern of relations between points:

When the number of points is reduced, the number of inter¬
point relations is also reduced. The pattern is thus gradually
abstracted or generalized. But however abstract and elusive,
an isomorphic pattern of relation must remain there, since
the points alone, even when matching the designatum point
by point, do not guarantee isomorphism.

The end-point in the transition of a sign from icon to symbol displays

two related features. First, only a single point now stands in an S-rela¬
tion—trivially isomorphic—with the designatum. Second, the symbol
displays no obvious pattern of inter-point relation isomorphism with the
designatum. One may thus concede that symbolic representation, as an
extreme pole on the continuum of degree of iconicity, is indeed striking¬
ly different from the imagic extreme. But if we now insist on viewing this
as a difference in kind rather than in degree, we are bound to ignore two
fundamental points, one methodological, the other ontological:

(a) Methodological:
There is no principled way by which one can decide at what point, on
the graded continuum between image and symbol, one has traversed
the boundary between iconic and symbolic representation.

(b) Ontological:
In some fundamental way, a one-point symbolic sign is just as isomor¬
phic to its designatum as an n-point diagram or image. It is isomorphic,
however, at the appropriate meta-level of abstraction. That is, it stands
for the designatum as a whole .

One may of course wish to dismiss point (b) as sheer sophistry. How¬
ever, our lexicon and its phonological representation constitute a large
body of prima facie evidence in support of precisely such single-point
iconic representation (see discussion further below). In sum, the degree
of abstractness of the isomorphic relation between linguistic signs and
104 Mind, Code and Context

their mental or experiential designata, should not prejudice our study of

the complex, abstract iconicity of the grammatical code.

3.4.4. Iconicity in syntax: Case studies

3.4.4.1. Propositions and temporal order

% • %

Two relatively concrete, transparent types of iconicity in syntax will

be mentioned here without much detail. The first involves the rather su¬
perficial isomorphism between components of the proposition and com¬
ponents of a state or event. Thus, in the sentence

(14) John cut the meat for Mary with a knife

each noun phrase corresponds to a participant in the 'event' (or rather,

in the mental proposition). The same line of reasoning would,
presumably, also suggest that the verb itself corresponds to one element
in the proposition, the 'action'. As one can see, however, both the S-rela-
tion and X-relation of the verb are much more abstract and constructed
than those of the nominal arguments.39
Another relatively transparent case of iconicity in grammar is dis¬
cussed by Haiman (1985; to appear). It involves the fact that events tend
to be coded, by propositions, in the same temporal order in which they
occurred in real time. Thus, in (15) below, (15a) is a more preferred order
of reporting. And (15b), while possible, is a more marked order:40 Pre¬
posed clauses are preferred over post-posed ones regardless of temporal
order in the coded experience. The principle controlling such preference
is discourse-pragmatic, having to do with the strategies by which earlier
discourse is recalled back at a relevant point later in the discourse.

(15) a. Having finished his dinner, John left,

b. John left, having finished his dinner.

A more trivial case of such iconicity involves the simple concatena¬

tion of propositions, as in:

(16) a. John got up and left,

b. ?John left and got up.

39 See discussion in section 3.3.4., above, concerning the designatum of propositions.

40 For some text-based quantified studies of the frequency of pre-posed and post-posed
adverbial clauses, see Thompson (1985) and Ramsay (1985). The problem is not as simple
as it first appears.
 The Linguis tic Code: The Iconicity of Grammar 105

3.4.4 2. Quantity scales in the grammar of referential identification

In a series of cross-linguistic studies (Givon, ed.,1983; 1984c; see also

Chapter 6, below) it was noted that the most common topic-coding
anaphoric devices used in languages may be ranked along a scale of
referential predictability ('accessibility', 'continuit/). The points on the
scale that are most stable cross-linguistically are:

(17) most predictable/continuous

a. zero anaphora
b. unstressed pronoun
c. stressed pronoun
d. definite noun
e. restrictively modified definite noun
least predictable/continuous

Let us illustrate these anaphoric devices with simple examples from

English. Respectively:

(18) a. John came in, [ o ] looked around and [o ] gasped.

b. John talked to Bill; then he left. (= John left)
c. John talked to Bill; then he left. (=Bill left).
d. John came in and paused. The woman got up.
e. The tall woman remained in the room. Later on, the man
who had been hiding under the couch emerged.

The grounds on which a referent may be more predictable ('accessible')

may vary, with the most obvious ones being:

(a) Gap of absence from the previous appearance in discourse

(b) Referential complexity of the preceding discourse
(c) Available 'redundant' propositional-semantic information
(d) Available 'redundant' discourse-thematic information

And at least two of these, (a) and (b) above, can be measured quantita¬
tively in text-based studies or in psycho-linguistic experiments.41

41 For the text-based studies, see Giv6n (ed., 1983) as well as Chapter 6, below. For the growing
psycho-linguistic literature see Fletcher (1982), Dahl and Gundel (1982), Tomlin (1985),
Givon et al (1985) and Gernsbacher and Hargreaves (1986), inter alia .
106 Mind, Code and Context

The coding principle underlying the scale in (17) is rather transparent,

and may be summarized first as:

(19) Code-quantity and informational predictability:

"The less predictable the information is, the more coding
material is used coding it"

Since psycho-linguistic evidence 42 suggests a clear correlation between

the predictability of information and the amount of mental effort used
in processing it, one may re-phrase (19) in a more general way, as a cor¬
relation between code quantity and mental effort:

(20) Code-quantity and mental effort:

"The more mental effort is used in processing information,
the more coding material will be used in representing the in¬
formation in language".

Note first that the designatum of the pragmatic function 'degree of

referential predictability', i.e. the mental dimension underlying scale
(17), is itself neither a mental concept of some real-world entity, nor a
mental proposition concerning some experienced state or event. Rather,
the designatum is itself an abstract mental process, having to do with
the speaker's assessment as to how accessible the referent is to the
hearer. The linguist may reflect upon the use of the anaphoric devices in
(17/18), and may indeed conclude that they are ranked along a scale of
mental accessibility or retieval effort.43 But it is reasonably safe to as¬
sume that speakers are largely unaware of the rather simple iconic rela¬
tion existing there.
Further, the scale in (17/18) is merely one of many instances where
principle (19/20) manifests itself in language. For example, lexical mor¬
phemes bear more (and more specific) information than grammatical
morphemes, and are on the whole phonologically larger. In the prag¬
matic use of intonation to designate degree of informational novelty
('contrast', 'surprise'), perceptually more prominent intonation (higher
pitch, louder amplitude, longer duration) always codes less predictable
information (cf. Bolinger, 1978, 1984).44
Note also that the code-quantity scale in (17) and its clear iconic rela-

42 Dahl and Gundel (1982), Chang (1980), Corbett and Chang (1983), Garrod and Sanford
(1977), Anderson, Garrod and Sanford (1983), Dell, McKoon and Ratcliff (1983), Ratcliff
and McKoon (1980), McKoon & Ratcliff (1980), Givon et al (1985), inter alia .
43 As, for example, in Givon (1979, ch. 2).
44 In our scale (17) we have a partial reflection of this fact, in that in many languages - English
included — the difference between unstressed pronouns ('clitic pronouns', 'verb
agreement') and stressed ('independent', 'contrastive') pronouns is coded primarily by
stressing the latter.
 The Linguistic Code: The Iconicity of Grammar 107

tion with some mental dimension is clearly at odds with Chomsky's

characterization of human grammatical structure as non-iconic, non¬
scalar, non-matching:

"...When I make some arbitrary statement in human language,...I am

not selecting a point along some linguistic dimension that signals a cor¬
responding point along an associated non-linguistic dimension..." (1968,
p. 70)

Bolinger (1984) has pointed out that the use of intonation in human
language is a natural extension -- in his word, part and parcel -- of the
gestural inventory used in human communication. This is the very sub¬
system that Chomsky himself concedes resembles animal communica¬
tion:
"...The examples of animal communication that have been examined
to date do share many of the properties of human gestural systems, and
it might be reasonable to explore the possibility of direct [i.e. evolution¬
ary; TG] connection there..." (1968, p. 70)

Principle (18/19), underlying the grammar of referential tracking in

human language, thus seems a natural extension of the phylogenetically
older gestural coding system found in pre-human communication.

3.4.4 3. The use of word-order in the grammar of referential

identification

In a number of recent studies (see Chapter 6, below),45 it was shown

that the pragmatic use of flexible word-order in language, with par¬
ticular reference to the position of the subject and object vis-a-vis the
verb, is highly sensitive to both the relative predictability ('continuity7,
'accessibility') and relative importance of those nominal referents

45 See summary in Givon (ed., 1983a). The cross-linguistic evidence on the discourse-prag¬
matic conditioning of word-order flexibility now includes text-based studies of U te (Givon,
ed., 1983a), Spanish (Bentivoglio, 1983), Biblical Hebrew (Fox, 1983; Givon, 1977), Yagua
(Payne, 1983), Papago (Payne, 1984), Tagalog (Fox, 1985), Iroquois (Mithun, 1985), Man¬
darin Chinese (Sun and Givon, 1985), Indonesian (Verhaar, 1984; Rafferty, 1985), Chamorro
(Cooreman, 1985), Nez Perce (Rude, 1985), Hixkaryana (Derbyshire, 1985) and Pidgin
English (Givon, 1984b).
108 Mind, Code and Context

('topics') in discourse. The correlation between topic predictability,

topic importance and word-order can be summarized initially as:46

(21) a. "More unpredictable/inaccessible referents tend to be

fronted"

b. "More important referents tend to be fronted"

> • »

It is possible, further, to integrate (21a,b) into a single formulation, by

noting their common denominator -- task urgency:47

(22) "When a referent is either less accessible or more important

in the discourse, the communicative task of coding it well
('strongly7, 'distinctly7) is more urgent".

And once (22) is granted, it is possible to express (21) as the unitary cog¬
nitive principle (see also Chapter 6, below):

(23) "Attend first to the most urgent task"

Much like principle (19/20), which motivates the S-relation between

the syntactic code and some internal operation of mental effort, prin¬
ciple (23) suggests an isomorphic S-relation between the syntactic code
and some mental dimension of task urgency. Both of these internal
designata, which may ultimately reflect the very same mental process,
are highly abstract. They involve the inner workings of the processor it¬
self, rather than any coded experience per se . Both are relatively inacces¬
sible to conscious reflection (compared to propositional information or
lexical concepts).

3.4.4.4. The grammatical coding of complement clauses

There exists a systematic correlation between the semantic properties

of complement-taking verbs, most specifically the type of semantic bond

46 While English is a rigid SVO language, it still reflects this principle in dislocated construc¬
tions. Thus compare: a. L-dislocation John, I saw him yesterday; b. Neutral order; I was John
yesterday; c. R-dislocation I saw him yesterday, John. In quantified text sudies (Givon, ed.,
1983a) it was shown that (a) is characteristic of large referential gaps, (b) of intermediate
ones and (c) of smaller ones.
47 For the initial suggestion see Givon (ed. 1983a, 1984b). See also further discussion in
Chapters 5 and 6, below.
 The Linguistic Code: The Iconicity of Grammar 109

they display vis-a-vis their complement clause, and the syntactic struc¬
ture of that complement clause. The semantic properties involved, sub¬
sumed originally under the label of binding,48 can be expressed in terms
of three major scalar features:

(24) a. Control: "The higher the main verb is on the binding

scale, the stronger is the control exerted by its agent
over the agent of the complement clause".

b. Success: "The higher the main verb is on the binding

scale, the more likely it is that the intended manipula¬
tion was successful, and the more likely it is that the
event/state coded in the event coded in the comple¬
ment indeed occurred".

c. Independence: "The higher the main verb is on the

binding scale, the less likely it is that the agent of the
complement clause can act independently".

The binding scale thus ranks manipulative over non-manipulative

verbs, implicative ('successful') over non-implicative manipulations,
intended over unintended manipulations, and direct over mediated
manipulations.49
The syntactic structure of complement clauses involves four distinct
elements that together code the degree of binding exerted by the main
verb over its complement clause:

(25) a. Subject/agent case-marking: "The higher the main verb

is on the binding scale, the less likely is the sub¬
ject/agent of the complement to display the case-mark¬
ing characteristic of subjects/agents of main clauses".

b. Verb modalities: "The higher the main verb is on the

binding scale, the less likely is the complement verb to
display the tense-aspect-modality markings charac¬
teristic of main-clause verbs".

48 Givon (1980a), with some of the semantic details due to an earlier work (Givon, 1975b).
Since that account was written, Chomsky has come out with his own 'binding', which has
nothing to do with the topic discussed here. Unlike Peirce, I do not feel compelled to change
the name of my brain-child in order to save it from kidnappers.
49 That is, 'order' >'wish', 'causeVorder', 'makeVcause', 'make'>'have', respectively.
no Mind, Code and Context

c. Fusion or co-lexicalization: "The higher the main verb

is on the binding scale, the more likely is the comple¬
ment verb to co-lexicalize with the main verb".

d. Separation: "The higher the main verb is on the binding

scale, the less likely it is that a subordinating morpheme
would separate the complement clause from the main
clause". > • »

The coding correlates (25a,b,c,d) are rather transparent iconic-isomor¬

phic expressions of three major transitivity properties of the comple¬
ment clause:

(25a) codes the agent of the complement as either being or

not being an independent, willful, controlling initiator
of the event;

(25b) codes the action/verb of the complement as either being

or not being independent in time, perfective in aspect
and realis in mode;

(25c,d) jointly code the independent status, or lack thereof, of

the event expressed in the complement clause.

Note next that the isomorphism between the syntactic devices in (25)
and the transitivity properties of complement clause derives its power
from two radically different sources. Both coding principles (25a) and
(25b) derive their power — at least in part — from some system-internal,
abstract norms, or rules of grammar . Only because these system-internal
norms for case-marking and tense-aspect marking have already been es¬
tablished for the grammar of main clauses, can coding correlations (25a,b)
assume their iconic power. The main components of these system-internal
norms are highly universal, and may be summarized as follows:

(26) a. Universal norms for subject-marking in transitive

clauses:

"Given the case-role hierarchy:50

AGENT > DATIVE > PATIENT > OTHERS

50 For a detailed discussion of the various case-role hierarchies, see Givon (1984a, Chapters
4,5).
 The Linguistic Code: The Iconicity of Grammar 111

the more the subject case-marking resembles the one nor¬

mally displayed by the highest-ranking term of this hierar¬
chy (i.e. the agent ), the more is the event likely to be an
independent transitive event".

b. Universal norms for verb-marking in transitive

clauses:

"Given the verb-modality hierarchies:51

PERFECTIVE > IMPERFECTIVE
SEQUENTIAL > PERFECT/ANTERIOR
COMPACT > DURATIVE
REALIS > IRREALIS

the closer the verb morphology of the clause is to the higher¬

ranking term in these hierarchies, the more likely is the event
to be an independent transitive event".

While the universal hierarchies in (26a,b) are widely attested cross-lin-

guistically, and while they are no doubt themselves founded upon general
cognitive, communicative and socio-cultural principles, the actual gram¬
mar of case-marking and tense-aspect-modality is highly language-specific.
It involves partially arbitrary language-specific considerations of rigid
word-order, of morphology, of diachronic attrition, etc. In contrast, the
iconic power of coding correlates (25c,d) derives rather transparently
from a general cognitive principle, one that may be given as:

(27) The proximity principle:

"The closer two mental entities are to each other semantical¬
ly or functionally, the more likely they are to be placed
together, in linear proximity, in the linguistic code".

Haiman (to appear) renders principle (27) as:

"...linguistic fusion signals conceptual fusion; linguistic independence

signals conceptual independence..." [p. 209, ms]

The overt linguistic code is substantially linear. What principle (27) does
is establish an iconic relation between linear proximity at the code level
and conceptual proximity (of a great variety of types) at the mental
designatum level. Both the co-lexicalization of verbs (i.e. (25c), also

51 For a detailed discussion of these hierarchies, see Hopper & Thompson (1980) and Givon
(1982b, 1984a ch. 8).
112 Mind, Code and Context

called 'predicate raising') and the absence of a subordinating morpheme

(25d), are thus coding manifestations of greater conceptual proximity
('affinity', 'unity') of the two propositions (main and complement). The
use of these devices signals that the speaker construes the two proposi¬
tions as a single event. The first device, (25c) is manifest at the lexical
level. The second, (25d), is manifest at a wider-scoped syntactic level.
Note, in passing, that principle (27) also predicts the co-lexicalization
of derivational affixes with their relevant stems, the co-lexicalization or
adjacency of grammatical operators with their operand lexical stems,
the proximity of operator words to their operand words, and the cor¬
relation between linear proximity and conceptual scope.
The combined syntactic coding scale that corresponds, in a particular
language, to the conceptual scale of 'degree of binding' between main
and complement clauses, is indeed complex. It involves the case-mark¬
ing system, the tense-aspect-modality system, morphemic subordinator
and the degree of co-lexicalization.52 To illustrate this briefly, consider the
following examples from English:

(28) a. He let-go of her (pred-raising)

b. He let her go (bare-stern comp.)
c. He wanted her to go (infinitive comp.)
d. He wished that she would go (subjunctive comp.,
nominative subject,
subordinator)
e. He thought that she had gone (finite comp.,
nominative subject,
subordinator)

This syntactic scale is totally integrated , with the highly abstract, gram¬
mar-dependent code elements such as case-marking and verb modality
(25a,b) interspersed throughout with the more concrete, cognitively-
transparent coding devices (25c,d). This suggests, at least tentatively,
that in spite of the disparate cluster of motivations that give rise to the
iconicity of the syntactic code, at some level this code may assume a
unified mental reality.

3.4.4 5. Passivization and the coding of transitivity

Passivization is a complex phenomenon, ranging over at least three

major functional domains:53

52 While (25c) is formulated here in categorial (either/or) terms, there is good cross-language
evidence in support of a scalar interpretation of this feature as well. For details see Givon
(1980a).
53 For the original formulation of this cluster approach to passivization, see Givon (1981a).
 The Linguistic Code: The Iconicity of Grammar 113

(29) Functional domains of passivization:

(a) Topic assignment:

"Marking some non-agent -- most commonly an object
— as the primary topic of the clause";

(b) Impersonalization:
"Downgrading the importance or topicality of the sub¬
ject/ agent of the event";

(c) Stativization:
"Construing the event as the resulting state of an ac¬
tion".

As one can see, as a pragmatic operation of re-orienting the speaker's

point of view vis-a-vis an event, passivization tampers with all three
major components of transitivity.54 It suppresses the saliency of the
agent-cause, upgrades the saliency of the patient-effect, and focuses on
the resulting state of the event, rather than on the change ('process') it¬
self.
Each of the three functional domains of passivization (29a,b,c) is as¬
sociated with a distinct grammatical coding-feature that stands in a
transparently iconic S-relation to it. The three coding features of pas¬
sivization are respectively:

(30) Coding features of passivization:

a. Non-agent promotion:
"Some non-agent (most commonly 'object') argument of
the transitive event receives in the passive clause the
type of case-marking (in terms of word-order, morphol¬
ogy and grammatical agreement)55 that in the active
clause characterizes the agent of the event. That is,
given the case hierarchy:
AGT > DAT > PAT > OTHER,
the promoted non-agent is 'moved', syntactically, up to
the top of the hierarchy".

b. Agent/subject demotion:
"The agent/subject of the active clause receives in the
passive clause the type of case-marking characteristic of

54 See also discussion of transitivity in Chapter 2, section 2.5.1..

55 See discussion in Keenan (1975).
114 Mind, Code and Context

a lower positions on the case hierarchy. Most common¬

ly, this involves either oblique or unmarked case, or
total deletion".

c. Verb stativization:
"The verb in the passive clause is likely to display syn¬
tactic characteristics more commonly associated with
stative or adjectival predicates, in terms of the auxiliary
'be', adjectival or gerund morphology, and perfect or
imperfective aspectual morphology".

To illustrate some of these grammatical features briefly, consider the fol¬

lowing examples from English:

(31) a. Mary found John on the beach (active)

b. John was found by Mary on the beach (passive with
oblique AGT)
c. John was found on the beach (passive with
deleted AGT)

Each syntactic feature of the passive, as given in (30), corresponds to

one functional feature in (29). Most of the detailed features in (30a,b,c)
depend for their iconic power on abstract, system-internal structural
norms that are highly specific in the grammar of particular languages.
For example, feature (30a) isomorphically matches non-agent topical-
ization only because of the norm that establishes the prototype syntactic
status of agents/subjects as occupying the top of the case hierarchy. It is
only vis-a-vis this abstract norm that the promotion of the non-agent to
subject-of-passive is truly iconic. In the same vein, feature (30c) receives
its iconic power from specific grammatical norms for marking stative
predicates. In contrast, coding feature (30b) -- isomorphic to agent sup¬
pression (29b) -- derives its iconic power from two distinct sources.
First, the assignment of oblique case-marking again depends on lan¬
guage-specific grammatical norms. But second, total agent deletion (by
far the most common means of agent-demotion in passive clauses cross
languages)56 is iconic in a much more concrete and cognitively trans¬
parent sense, abiding by the following coding principle:

56 In most human languages, the agent of the event cannot appear overtly in the passive
clause. For discussion and many details, see Givon (1979, Ch. 2; 1981a).
 The Linguistic Code: The Iconicity of Grammar 115

(32) The relevance or importance principle:

"A piece of information that is either unimportant or ir¬

relevant receives less over coding".

Note now that the participation of principle (32) in the coding of the
passive is integrated into a scale of degree of agent demotion, where total
deletion appears as the most extreme coding device:57

(33) Scale of agent demotion in passivization:

a. object-like linear position

b. object-like case-marking morphology
c. total deletion

And it is again possible that scale (33) is treated as a unified mental en¬
tity, isomorphic to the functional scale of agent suppression, regardless of
the divergent source of its iconic power.

3.4.4.6. Antipassivization and the coding of transitivity

We have discussed at a number of points above the three major com¬

ponents of semantic transitivity, involving the prototypical agent,
patient, and verb modality.58 As we have already seen in the discussion
of passives, the grammatical coding of these three main components of
transitivity is highly iconic. When the functional text-distribution of
semantically-transitive clauses is studied, other facets of iconicity in the
grammatical code are revealed, ones that involve more clearly dis¬
course-pragmatic function.
In a series of quantified text studies covering a number of languages,59
the degree of topicality ('predictability' and 'importance') of agents and
patients of semantically-transitive clauses was measured, using quan¬
titative text-based methods. In particular, the ergative-transitive clause
in these languages was compared with both the passive and antipas¬
sive clause-type. The relative topicality, and semantic referentiality, of
the agent and patient arguments in these clauses may be summarized as
follows:

57 The scalarity observation here is due to Keenan (1975).

58 See discussion in the preceding section, as well as in Ch. 2, above. More extensive
treatments can be found in Hopper and Thompson (1980) and Givon (1984a, Chapters 4,5).
59 See Cooreman (1982a, 1982b) for Chamorro; Cooreman, Fox and Givon (1983) for Tagalog;
Rude (1983) for Nez Perce; Cooreman (1985) for Dyirbal. For more details of the measure¬
ments, see Chapter 6, below.
116 Mind, Code and Context

(34) Topicality of clause types:

clause type topicality referentiality

TRANS-ergative AGT>PAT both AGT & PAT highly REF

INTRANS-passive PAT»AGT AGT mostly NON-REF
INTRANS-antipass AGT»PAT PAT mostly NON-REF

In the transitive-ergative clause both agent and patient tend to be high¬

ly referential, and the agent is more topical ('continuous', 'important')
in the discourse than the patient. In contrast, the agent in the passive
clause tends to be highly non-referential and non-topical; and the
patient in the antipassive clause tends to be highly non-referential and
non-topical. The use of the passive and antipassive constructions to
code discourse-pragmatic function is thus determined by the relative
topicality, and referentiality, of the agent and patient of the clause.
In the ergative languages that were studied, the morphology of agent,
patient and verb is sensitive to the degree of transitivity of the clause.
In such languages, the syntactic coding of passivization and antipas-
sivization, in particular as it pertains to the marking of agent and
patient, may be summarized as follows (see also section 3.4.4.5., above
for passivization):

(35) (a) Passive:

The agent either assumes oblique-case morphology or,
more commonly, is left unexpressed. In addition, if the
verb had any transitive morphology, it is lost in pas¬
sivization.

(b) Antipassive:
The patient loses its direct-object morphology. It may
then assume oblique-case morphology; or it may be in¬
corporated into the verb ('co-lexicalized'); or, most com¬
monly, it may be left unexpressed. Again, if the verb
had any transitive morphology, it is lost in anti-pas-
sivization.

The iconic-isomorphic S-relation between the semantic-pragmatic

properties of agents and patients (as given in (34)) and their syntactic
coding properties (as given in (35)) is rather transparent. We have al¬
ready discussed the passive earlier above. For the antipassive, we find a
reminiscent scale, involving the degree of demotion of the patient when
it is low in topicality and/or referentiality:
 The Linguistic Code: The Iconicity of Grammar 117

(36) least demoted

a. Direct-object marking
b. Loss of direct-object case-marking; sometimes assump¬
tion of oblique case-marking
c. Co-lexicalization with the verb ('incorporation')
d. Total deletion ('zero coding')
most demoted

Of the four coding points on scale (36), two -- (36a,b) -- receive their
iconic power via more abstract grammar-dependent norms. The other
two, however -- (36c,d) -- are cognitively transparent, bringing to mind
immediately coding principle (32), the one that relates the strength
('saliency') of the code with the importance or relevance of the coded
meaning/function. Principle (32) may be recast in slightly more general
terms as (37):

(37) The saliency isomorphism principle:

"The degree of perceptual/ mental saliency of the code tends

to be isomorphic to the degree of communicative saliency of
the message".

The only coding device the antipassive added, above and beyond
what was seen in the passive, is co-lexicalization (36c), whereby an ob¬
ject low in topicality or referentiality may become part of the verbal
word (as prefix or suffix), and thus receive less independent coding. As
examples of English constructions that are functionally equivalent to
the antipassive of ergative languages, consider:

(38) a. Transitive:
He shot the target

b. Antipassive, object incorporation:

He went target-shooting

c. Antipassive, zero object:

He went shooting

3.4.4 7. Stereotypicality, predictability and coding saliency

The very same process of morphological incorporation or zero expres¬

sion ('deletion') governed by principle (37) is seen in the case of in¬
strumentals and manner adverbs . As illustration consider:
118 Mind, Code and Context

(39) Instrumentals:

a. Independent coding:
John fished with a fly-hook

b. Morphological incorporation:
John went /Zy-fishing , .

c. Semantic incorporation (zero coding):

Mary slapped John (implied: 'with her hand')

(40) Manner adverbs:

a. Independent coding:
Mary stitched the pillow cross-wise

b. Morphological incorporation:
Mary cross-stitched the pillow

c. Semantic incorporation (zero coding):

Mary murdered John (implied: 'deliberately')

Here, the element that receives less salient coding is highly predictable
or stereotyped. One may suggest that our code-quantity principle (19) is
involved, by which more predictable information receives less (or less
salient) coding.

3.4.4.8. Linear ordering, proximity and scope

The most common device in grammar for indicating scope relations

and conceptual (or functional) proximity, is via linear ordering and/or
linear proximity. This coding behavior is virtually dictated by the nar¬
row, single-channel nature of the linguistic code. To illustrate this
phenomenon at the syntactic level, consider the rigid ordering of the
pre-verbal auxiliaries in Creole languages.60 The relative ordering of
these auxiliaries tends to be:

(41) (PERFECT) (MODAL) (DURATIVE) VERB

Of the three auxiliaries, 'durative' has the narrowest -- lexical-semantic

-- scope, ranging over the verb itself. 'Modal' ('irrealis') has a wider —

60 See originally in Bickerton (1975,1981), as well as further discussion in Givon (1982b).

 The Linguistic Code: The Iconicity of Grammar 119

propositional -- scope. And 'perfect' ('anterior') has the widest -- dis¬

course-pragmatic scope, ranging over a chunk of discourse larger than
the single proposition. The relative ordering of these pre-verbal aux¬
iliaries in the morpho-syntax of Creoles, thus their relative proximity to
the verb, reflects rather iconically the relative semantic or pragmatic
scope of these operators, and their relative relevance to the verb. This
iconic relation is another manifestation of our proximity principle (27).
Another illustration of principle (27) may be seen in the grammar of
negation in Russian. This example involves the correlation between the
syntactic position of the negative marker, and the contrastive focus as¬
sociated with negation. Thus consider:61

(42) a. Ivan jego ne-ubil

Ivan him NEG-killed
'Ivan didn't kill him' (neutral or V-focus)

b. Ivan ubil ne-jego

Ivan killed NEG-him
'Ivan didn't kill him ' (OBJ-focus)

c. Jego ubil ne-Ivan

him killed NEG-Ivan
'Ivan didn't kill him' (SUBJ-focus)

What one can see in (42) is an interaction between a rather abstract

grammatical norm in Russian, one that requires the placing of a focused
element at the end of the clause, and the cognitively-transparent pro¬
ximity principle (27), through which the negative marker is placed
nearest the element under the scope of contrastive negation. Once again,
the grammatical code integrates norm-dependent and cognitively-
transparent iconicity principles.
To illustrate the contrast between the relative arbitrariness of the
specific word-order norm in Russian negation and the naturalness of
the proximity principle (27), consider the very same process occurring
in a verb-first language, Bikol. In this language, the more abstract gram¬
matical word-order norm places focused elements at the beginning of the
clause. Still, the negative marker in focused negation is placed next to
that fronted element:62

61 After Dreizin (1980).

62 For the Bikol examples and discussion, see Givon (1984a, ch. 9).
120 Mind, Code and Context

(43) a. Affirmative:
nag-gadan 'ang-lalake ning-kanding
AGT-kill SUBJ-man OBJ-goat
'The man killed a goat'

b. Neutral (or V-focus) negation:

da'i nag-gadan 'ang-lalake ning-kanding
NEG AGT-kill SUBJ-man OBJ-goat
'The man didn't kill a goat'

c. SUBJ-focus negation:
da'i 'ang-lalake nag-gadan ning-kanding
NEG SUBJ-man AGT-kill OBJ-goat
'The man didn't kill the goat'

d. OBJ-focus negation:
da'i ning-kanding nag-gadan 'ang-lalake
NEG OBJ-goat AGT-kill SUBJ-man
'The man didn't kill the goat'

Finally, principle (27) is also apparent at the morpho-tactic level,

where it governs the relative order and proximity of grammatical/in¬
flectional and derivational morphemes vis-a-vis lexical stems.63

3.4.5. Some meta-iconicity considerations

Most of the iconicity principles surveyed above, with the exception of

meta-principle (8), are concrete, cognitively transparent rules for assem¬
bling together morpho-syntactic structure. In this section we will deal
briefly with a number of more general meta-iconic considerations.

3.4.5.1. Arbitrary grammatical norms vs. natural iconic principles

As seen in the discussion of negation in Russian and Bikol, above, a

certain relation exists between the abstract, often language-specific
grammatical norms, and the more universal, concrete, cognitively-
transparent iconicity principle. This relation may be expressed as the
following meta-iconic prediction:

63 For many cross-language examples and an extensive discussion, see Bybee (1985).
 The Linguistic Code: The Iconicity of Grammar 121

(44) Meta-iconic principle of universality:

"The more concrete and cognitively transparent a coding

principle is, the more it is likely to be universally manifest in
the grammars of all human languages".

As we shall see further below, cognitively transparent iconicity prin¬

ciples are also more likely to be manifest in pre-human communication.

3.4.5.2. Conflicts between iconicity principles

One of the most concrete, cognitively transparent iconicity principles

involves the isomorphism between the temporal order of experiences
and the temporal order of their coded-expression in communication.64
In this connection, one case where the sentence-internal ordering of con¬
stituents seems to reflect temporal ordering, has been discussed widely
in the psycho-linguistic literature.65 It pertains to the relative ordering of
agent ('cause') and patient ('effect'). As Sridhar (1980) puts it:

(45) The cause-before-effect principle:

"...nominals denoting figure-of-state and agent-of-action

precede those denoting ground and patient..." (1980, p. ix)

Now, of the rigid word-order types found in human language, the most
common ones — in terms of cross-language distribution -- are SVO, VSO
and SOV. In all three, the agent ('subject') indeed precedes the patient
('object'), thus presumably upholding principle (45). However, there
exist two other well-attested word-order types, VOS66 and OVS.67 Of
these two, one (VOS) is just as rigid as can be found.68 The word-order
of VOS and OVS languages, one may now argue, is obviously not
governed by the semantic principle (45). Rather, it is governed by the
pragmatic principle (21/23) above. That principle, recall, orders more
topical ('predictable') information after less-topical/predictable infor¬
mation. Now, the subject is most commonly more topical ('predictable'
and 'important') in discourse than the object - and the agent is assigned
the subject role much more frequently in discourse than the patient.69

64 See extensive discussion in Haiman (1983; to appear), as well as section 3.4.4.1., above.
65 See Osgood (1971) and Sridhar (1980), inter alia.
66 As in Malagasy (Keenan, 1976).
67 As in Carib languages, cf. Derbyshire and Pullum (1981).
68 Word-order rigidity is often a matter of degree. For an extensive discussion, see Givon
(1984a, ch. 6), as well as Chapter 6, below.
69 See discussion in section 3.4.4.3., above, as well as in Givon (ed. 1983a).
122 Mind, Code and Context

Therefore, VOS and OVS languages seem to reflect -- iconically — prin¬

ciple (21 /23). The gist of the argument is this: When two equally natural
coding principles predict conflicting results, the resolution of such con¬
flict in any particular language is not always predictable from general
considerations.

3.5. Some bio-adaptive considerations

3.5.1. The difference of Man70

We have surveyed above clear and ample illustrations of precisely

what Chomsky said is not to be found in human language:

"...the selection of a point along a linguistic dimension determines

and signals a certain point along an associated non-linguistic
dimension..." (1968, pp. 69-70)

We have seen that an extremely small number of cognitively-

transparent principles, pertaining to sequential order, code quan-
tity/saliency, and relative proximity of the code units, underlie the syn¬
tax of natural human languages. We have seen how these principles
govern an array of systematic, isomorphic-iconic relations between the
units of the syntactic code and message properties such as information¬
al predictability, importance, conceptual relevance or precedence.
It is of course true that the grammar of human languages incorporates
many more abstract, arbitrary grammatical norms, and those turn out to
be much more language specific. But such less-iconic norms are well inte¬
grated, in syntax, with the more natural, universal, cognitively-trans-
parent principles. And quite often these seemingly arbitrary norms turn
out to be themselves motivated either cognitively or socio-culturally.71
One may perhaps suggest that the introduction of increasingly non-
iconic ('more arbitrary') codes into communication is the unique evolu¬
tionary hallmark of human language. It can be shown, however, that the
mixing of more cognitively-transparent ('natural') codes with less trans¬
parent, ('arbitrary') codes within the same complex coding scale is a fea¬
ture of primate communication as well. To illustrate this, consider the

70 With indebtedness, for the title, to Mortimer Adler (1967). Like Descartes and Chomsky,
Adler considers human language to be non-contiguous with animal communication, thus
making 'The difference of Man'.
71 See extensive discussion in Slobin (1973), Langacker (1984) or Givon (1984a), inter alia .
 The Linguistic Code: The Iconicity of Grammar 123

scale that codes the dominant-subordinate message in Bonnet Macaques

(after Simonds, 1974):

(46) Dominant-subordinate coding scale in Bonnet Macaques:

6 Attack
5 Lunge
dominant 4 Open-mouth threat
signals 3 Eyelid threat
2 Stare
1 Look

1 Lip-smack slowly
2 Lip-smack rapidly
subordinate 3 Grimace
signals 4 Grimace widely
5 Grimace and present
6 Run off, grimace and tail-whip

On the dominant portion of the scale, an increase in code intensity,

abiding by our importance principle (32), is clearly discernible from
point 1 to 2 to 3, all involving eye-signal intensity. But the transition
from 3 to 4 involves switching to another code element, the open mouth.
And while the escalation is natural and obvious, the integration of eye-
signal and mouth-signal into the same scale involves a higher level of
symbolic — i.e. more arbitrary -- representation.
A similar mix of naturalness and arbitrariness can be seen when one
compares the coding of aggression and submission in dogs and horses.
Dogs raise their head and prop their ears up to display aggression. Hor¬
ses lower their head and flatten their ears back to display aggression.
Both, however, display a similar quantity scale in coding the degree of
aggression.

3.5.2. The biological basis of isomorphic coding

3.5.2.1. Function, teleology and organisms

In sounding out the major themes in the study of organisms, Eckert

and Randall (1978) write in their introduction to animal physiology:

"...The movement of an animal during locomotion depends on the

structure of muscles and skeletal elements (e.g. bones). The movement
produced by a contracting muscle depends on how it is attached to
124 Mind, Code and Context

these elements and how they articulate with each other. In such a rela¬
tively familiar example, the relation between structure and function is ob¬
vious. The dependence of function on structure becomes more subtle,
but no less real, as we direct our attention to the lower levels of or¬
ganization — tissue, cell, organelle, and so on... The principle that struc¬
ture is the basis of function applies to biochemical events as well. The in¬
teraction of an enzyme with its substrates, for example, depends on the
configuration and electron distributions of the interacting molecules.
Changing the shape of an enzyme molecule (i.e. denaturing it) by heat¬
ing it above 40° C is generally sufficient to render it biologically non¬
functional by altering its shape ..." (1978, pp. 2-3; emphases are mine; TG)

The isomorphic coding of functions, at whatever level, is the oldest fact

of biological organization. Even the most primitive organismic fragment
-- the virus -- has at least one coding relation that is fully isomorphic,
matching sequences of nucleotides in its DNA (or RNA for some
viruses) with sequences of amino acid in its protein.72 The critical ele¬
ment that makes something a biological code (i.e. in Peirce's words
"...something by knowing of which we know something more...") is the
presence of some teleology, purpose (or, for a less-conscious organism,
function), i.e. an interpretant. And the notion of function, in turn, is the
sine qua non of the definition of 'biological organism'.73

3.5.2.2. Ontogeny, phylogeny, language change and the

iconicity of language

All biological codes are constructed, and eventually modified and

remodified, by evolutionary processes. And all evolutionary processes
are interactions between an organism and its environment. The context
for evolutionary interaction is always functional-adaptive. The match¬
ing of structure and function in biology is always mediated by evolu¬
tionary change.
The same must hold for the communicative code. Its pervasive icon¬
icity arise -- and is constantly modified -- via three functionally-me¬
diated developmental processes: Diachronic change, language learning
and language evolution. The evolutionary accretion of a communicative
code (the 'S-relation') to a pre-existing perceptual, neural and cognitive
coding of experience (the 'X-relation') is only one episode in a long
biological scenario. Within that scenario, the increased complexity and

72 This seeming single-step coding relation is in fact functionally more complex. In order for
the virus to replicate and perpetuate itself, it must penetrate a host cell, where the virus
DNA code is first interpreted by the host-cell's enzymes, who then translate it into amino
add sequences as well as virus DNA.
73 See further discussion in Chapter 8, below.
 The Linguistic Code: The Iconicity of Grammar 125

abstractness of human communicative functions and their code, is a

natural adaptively extension of pre-existing functions, structures and,
above all, pre-existing coding principles.

3.5.2.3. From icon to symbol

The increasing abstractness of the human linguistic code ~ from

multi-point iconicity to one-point symbolism — has not substantially
transformed its pervasive isomorphism. There is no cogent reason why
this increasing abstraction should not be considered as part and parcel
of function-driven evolutionary change.74 This integration of symbolic
('arbitrary) and iconic ('natural') elements in the human linguistic code
was indeed noted by Peirce:

"...Particularly deserving of notice are icons in which the likeness is

aided by conventional rules..." (1940, p. 105)

While Peirce did not intend his observation developmentally, it is just as

relevant today as when first made. The addition of abstract ('con¬
ventional') elements to our mental and communicative codes does not
alter their inherent iconicity. When confronted with seemingly arbitrary
relations between code and designatum, the human organism — like all
others -- strives to discover, impose or construe a maximal measure of
non-arbitrariness. In following our iconic meta-imperative, we strive to
retain and reinforce the magic that makes it so much easier to store,
retrieve and communicate information within finite time and finite
neural resources -- the magic of an isomorphic-iconic code.

74 In the same vein, the highly iconic second-language Pidgin, and early childhood pragmatic
mode of communication (Givon, 1979, Chapters 5, 7), eventually give rise to the more
abstract - thus more symbolic - syntactic mode. Iconic-pictographic writing systems
eventually evolve into more arbitrary alphabets (see section 3.4.3., above). At the on¬
togenetic, evolutionary or diachronic levels, increased abstraction of the code is indeed a
recurrent theme.
 -<

'

'

.
CHAPTER 4

PROPOSITIONAL MODALITIES:
TRUTH, CERTAINTY, INTENT,
AND INFORMATION

4.1. Propositions, sentences and information

The proposition -- or its equivalent in natural language, the sentences
or clause1 -- has always been considered, whether on intuitive or formal
grounds, the basic information-processing unit in human language.
Words may have meaning, but only propositions carry information.
Even when it appears, prima facia, that a single word bears information,
on further inspection that word turns out to stand in for an entire
proposition. For example, in (1) below the curt one-word response (lb)
stands for the fuller proposition (lc):

(1) a. Who killed the butler?

b. The maid.
c. The maid killed the butler.

In this chapter we will not concern ourselves with the specific infor¬
mation carried by individual propositions. As noted in Chapter 3,
above, propositional information in language is assembled together
from two distinct sources:

(a) The actual lexical words2 used; and

(b) The propositional frame within which the words are as¬
sembled.

Thus, proposition (lc) above combines the lexical words 'maid', 'kill'
and 'butler'. They are assembled into a transitive propositional frame

1 For the purpose of the discussion here I will disregard the difference between 'clause' and
'sentence'.
2 Sentences in natural language contain, in addition to lexical words, also grammatical words
('grammatical morphemes'). Thus, the two definite articles 'the' in (lc) are 'grammatical
words' associated with the discourse-pragmatics of coreference; while the past tense '-ed'
is a grammatical morpheme. For the purpose of this discussion, we need not concern
ourselves with these grammatical operators.
128 Mind, Code and context

that assigns to 'maid' the role of agent of 'kill', and to 'butler' the role of
patient of 'kill'.
We will, from here onward, concern ourselves primarily with certain
meta-operators on propositions, namely the propositional modalities
under whose scope the entire proposition falls. Propositions in human
language, so long as they are embedded within some communicative
context, always fall under some modality (quite often under more than
one). Once communicative context is taken into account, the severe
limits of the traditional logical approach to propositional modalities be¬
come apparent. Thus, in departing from the logical tradition, we will
sketch out a discourse-pragmatic approach to propositional modalities.

4.2. Epistemic modalities: The logico-semantic tradition

In the logical tradition,3 propositions are treated as atomic entities, ex¬
isting apart from speaker, hearer and discourse context. Likewise,
propositional modalities in this tradition pertain ('belong') to the
proposition itself. Further, these modalities are judged to be purely epis¬
temic, i.e. pertaining to the truth of the proposition. This approach can
be traced back at least as far as Aristotle's De Interpretations, where the
following four propositional modalities were recognized:

(2) Epistemic modalities:

(a) Necessary truth: true by definition

(b) Factual truth: true as fact
(c) Possibly truth: true by hypothesis
(d) Non-truth: false

The distinction between (a) and (b) above overlaps to quite an extent
with Kant's distinction between analytic and synthetic truth, respec¬
tively. Roughly speaking,4 a proposition is analytically true if it follows
by definition from the internal properties of the representational system,
such as mathematics, language or a game. Such propositions are, for ex¬
ample:

(3) a. Two plus two equals four

b. A bachelor is an unmarried adult male
c. To win at chess, you must capture the king

3 Cole (edv 1981, Introduction) uses the term semantic for the logical tradition, to contrast
with pragmatic. There is a certain intuitive justification for this, in that semantics in human
language has traditionally dealt with the two least context-dependent realms of the
linguistic message, lexical-semantics and propositional frames (see discussion in Givo'n
1984a, ch. 2).
4 For some detail, see discussion of Kant's epistemology in Chapter 1, above.
 Propositional Modalities 129

In contrast, a proposition is synthetically true if it expresses a non-

analytic fact about the world.
As noted in Chapter 1, above, analytic truth is already a crypto-prag¬
matic modality. This is so because analyticity rests implicitly upon con¬
ventions, which must then be shared ('subscribed to') by the parties to
a contract. Traditional philosophers have masked this inherently social
aspect of language, by arbitrarily restricting the contract's parties to a
single mind. This restriction then made it possible to ascribe all
modalities to the proposition itself.
There is another reason why Kant's epistemic modes should be inter¬
preted as crypto-pragmatic. The analytic and the synthetic are also
modes of knowledge, indicating one's source of information:
Knowledge is obtained analytically through learning the conventions of
language, logic, games. Knowledge is obtained synthetically by dis¬
covering what the world is like. And in natural language, the epistemics
of source shades into the epistemics of evidence, in determining the
communicative mode of propositions.
As noted above, it has been traditionally assumed (as in, for example,
Carnap, 1947) that the four epistemic modalities pertained to the
proposition itself. Even as such, these modes already echo -- faintly, im¬
plicitly -- the four major epistemic modalities of natural language.5 In
human language, however, the four modalities are defined pragmatical¬
ly, i.e. with explicit reference to the speaker, to the hearer, and to their
respective communicative intent.

4.3. Epistemic modalities and the communicative contract

4.3.1. Preamble: Epistemic vs. other modalities

Human communication involves an intricate network of conventions
concerning what speakers and hearers are entitled to expect of each
other when carrying out their respective roles in communication.6
Within this vast network, which is all part and parcel of the pragmatics
of language, only a small portion deals with strictly epistemic matters of

5 There are obviously many more propositional sub-modalities in natural language, some
dearly epistemic. Many of those are explicitly coded in the grammar of particular lan¬
guages. Each of the three 'truth' modes has many sub-modes (or sub-clauses in the
communicative contract).
6 Some of the conventions subsumed under the communicative contract pertain to non-in-
formative speech acts, such as command, request or query (cf. Austin, 1962 or Searle, 1970,
inter alia). Others involve various other aspects of the dedarative speech-act (d. Grice,
1968/1975, Cole and Morgan, eds, 1975 or Gordon and Lakoff, 1971, inter alia). For the
moment, we shall deal primarily with the portion of the communicative contract that
govern the epistemic aspect of the transaction.
130 Mind, Code and context

truth or subjective certainty. Rather, even those modes that are primari¬
ly epistemic tend to shade gradually into other — intentional, social or
manipulative -- modes. This gradual progression may be represented as
follows:

(4) Shading of communicative modalities:

(a) epistemic modes of truth, of probability

(b) psychological modes of subjective certainty
(c) intentional modes of wish, ability or need
(d) social modes of status, authority, power or obligation
(e) action modes of causation or manipulation

The latter modes, (4d) and (4e), shade further into the vast area of non¬
verbal interaction. A clearer cleavage may be perhaps suggested be¬
tween the epistemic-psychological cluster of modes (4a,b,c), the inten¬
tional-social-action cluster (4d,e), and other inter-personal modes; but
such a cleavage is to quite an extent a Platonic mirage. It may merely
highlight, along the graduated modal space, three somewhat idealized
prototype peaks.

4.3.2. The epistemic space and evidentiality

4.3.2.1. The communicative contract

Epistemic modalities, like all other propositional modalities in human
language, are governed by the communicative contract between the
speaker and hearer. At the epistemic extreme of the modal range (see
(4a), above), the contract governs the speaker's responsibilities concern¬
ing the evidentiary justification of propositional information, accord¬
ing to its source and reliability. At the psychological sub-range (4b), the
contract governs the subjective certainty that speakers assign to the
communicated information. At the intentional, inter-personal and social
sub-ranges (4c,d,e), the contract governs the interaction between speak¬
er and hearer concerning the epistemic status of the communicated in¬
formation: The hearer's reaction -- from challenge to tacit assent to affir¬
mation; the speaker's reaction to that reaction; etc. All these facets of
communication are presumably governed by various clauses and sub¬
clauses of the communicative contract. The various clauses and sub¬
clauses are not altogether independent of each other. Often they display
predictable interactions ('redundancies'). Such redundancies make it
impossible to separate sharply epistemic from intentional, intentional
from manipulative, and manipulative from other social, inter-personal
 Propositional Modalities 131

modalities. At their prototypical core, the various modes on the epis-

temic range may display sharp - Platonic ~ categorial peaks. At the
same time, they also display considerable Wittgensteinean gradation
and overlapping similarities.

4.3.2.2. Evidentially and epistemic modes

The study of evidentiality in the grammar of human languages7 allows
us to re-define the four epistemic modalities (2a,b,c,d) in discourse-prag¬
matic terms. Respectively:

(5) Pragmatically re-defined epistemic modalities:

a. Presupposition: Presupposed or backgrounded infor¬

mation
b. Realis-Assertion: Strongly-asserted information
c. Irrealis-Assertion: Weakly asserted information
d. NEG-assertion: Strongly-asserted denial of information

The pragmatics of the NEG-assertion mode (5d), above and beyond its
being a strong assertion, will be dealt with separately below. It is in
some ways a hybrid mode, displaying some of the pragmatic properties
of R-assertion (5b), and Presupposition (5a), as well as some of the
semantic properties of IRR-assertion (5c). We will discuss the other three
modalities below, according to the various clauses of the communica¬
tive contract. First, however, let us consider briefly the way these three
modes are coded — or 'distributed' -- in grammar.

4.3.2.2. The grammatical distribution of the three

epistemic modalities
The distribution of the three modalities - presupposition, R-assertion
and IRR-assertion -- in the grammar of natural languages is quite pre¬
dictable and universal; what follows is a brief summary.8

7 For full discussion and survey of the cross-language evidence, see Chafe and Nichols (eds,
1986) or Givon (1982c).
8 For the original observation, see Givon (1973b); for further detail see Givon (1984a, chs 8,
11).
132 Mind, Code and context

4.3.2.2.1. Presuppositional clauses

The following grammatical9 environments tend to harbor, rather con¬
sistently, presupposed clauses. In some of these environments clauses
are obligatorily presuppositional, while in others they are optionally so.

(6) (a) Restrictive relative clauses:

The man I saw yesterday...

(b) Complements of cognition verbs:

I knew that she was there

(c) Complements of evaluative adjectives:

It's terrible that he drinks so much

(d) Adverbial clauses:

Because she didn't show up...

(e) Participial-adverbial clauses:

Having finished dinner, he got up and left

(f) Nominalized subject clauses:

His killing the chicken surprised me

(g) Nominalized object clause:

I respect his decision to quit

(h) WH-questions:
Who killed John?

(i) Cleft/focus clauses:

It was Mary who killed John

In addition, in many languages there is a high correlation between

presuppositionality of clauses and the use of the perfect ('anterior')
aspect, as in:

(7) a. He knew that she had done it

b. Having finished, she left.
c. His having left surprised no one
d. The man who had done it...

9 While we describe these as 'grammatical' environment, they are equally well also
'functional', given the high degree of systematic pairing, in natural language, of specific
structures with particular functions.
 Propositional Modalities 133

4.3.2.2.2. Irrealis clauses

The following grammatical environments tend to display the scope of

an irrealis modality:

(8) a. The future tense:

She will remember

b. Conditional ('if') clauses:

If he comes back,...

c. Complements of non-implicative modal verbs:

He wanted to leave
She must see him

d. Complements of non-implicative manipulation verbs:

She told him to leave
He asked her to help him

e. Complements of non-factive cognition verbs:

He thought that she didn't love him

f. The scope of epistemic adverbs:

Maybe he did it
Presumably she left

g. Non declarative speech-acts:

Leave immediatelyl
Was she there?

4.3.2.2 3. Realis clauses

The R-assertion mode distributes in grammatical environments con¬
trasting with those given in (6), (7) or (8) above. That is, it typically is to
be found in main, declarative, affirmative clauses with a past/perfec¬
tive tense-aspect, and where neither the irrealis environments in (8), nor
the presuppositional environments in (6) or (7), are involved.

4.3.2.3. Subjective certainty

The evidence from the study of natural language strongly suggests
that our three modes of information are ranked according to the subjec¬
tive certainty of the speaker. The ranking is of course transparent:
134 Mind, Code and context

(9) Epistemic modes and subjective certainty:

presupposition > R-assertion > IRR-assertion

Thus, for example, predicates that code the highest subjective certainty
are also most likely to also be presuppositional, as can be seen from the
following descending scale of subjective certainty in English:

(10) a. I know she was here.

b. I am sure she was here.
c. I think she was here.
d. I believe she was here
e. I see she was here.
f. I hear she was here.
g. I guess she was here.
h. They say she was here.

These verbs, often used with the first person subject, tend to function as
epistemic quantifiers on the subsequent proposition. In many lan¬
guages, these verbs become grammaticalized as evidentiary particles.10
But if any of the verbs on this scale can be used presuppositionally, it is
always the verb 'know', the highest on the scale of subjective certainty.
In some languages, such as KinyaRwanda or Thai, 'know' and 'think'
are the same lexical verb, but a special complementizer particle signals
that the complement clause is presupposed. When that particle is used,
the verb is translated as 'know'. When another particle is used, the verb
is translated as 'think'. To illustrate this, consider the following from
KinyaRwanda11:

(11) a. ya-ri a-zi ko amazi yari mare-mare

he/PAST-be he-believe that water was deep-deep
'He knew that the water was very deep'
(and indeed it was)

b. ya-ri a-zi ngo amazi yari mare-mare

he/PAST-be he-believe that water was deep-deep
'He thought that the water was very deep'
(but it may have not been)

However, at a lower portion of the subjective certainty scale, where no

presuppositionality is involved, the very same contrast between ko and

10 See various descriptions in Chafe and Nichols (eds, 1986).

11 See detail in Givon (1982c) and Givon and Kimenyi (1974).
 Propositional Modalities 135

ngo codes the degree of certainty or doubt the speaker holds vis-a-vis
the quoted information:

(12) a. ya-mu-bgiye ko u-a-kora-ga cyaane

he/PAST-him-tell that you-PAST-work-HAB hard
'He told him that you worked hard'
(and I don't doubt it)

b. ya-mu-bgiye ngo u-a-kora-ga cyaane

he/PAST-him-tell that you-PAST-work-HAB hard
'He told him that you worked hard'
(but I doubt it)

The contrast between ko and ngo complements is thus primarily a matter

of certainty or evidentiality. Only at the very top of the certainty range
do complements coded by ko turn presuppositional.

4.3.2.4. Source of information and defensibility

At the very core of the epistemics of evidentiality in human language

lies the differentiation between various sources of information. In this
section we will survey briefly how this feature applies to the three main
epistemic modalities.

4.3.2 4.1. Presupposition

We will continue to use the traditional name of 'presupposition' to

refer to this epistemic modality, although it is clear that the term unchal¬
lengeable information is more accurate. The communicative contract
treats presupposed information as assumed by the speaker to be either
known to, familiar to or otherwise unlikely to be challenged by the
hearer. The sources of this unchallengeability may be attributed to a
number of sub-clauses, all grouped under the three major sub-divisions
of context ('shared background', see Chapter 3, above).

(13) Sources of presupposed information:

a. Shared situational context:

(i) The information is deictically obvious to the
hearer in the immediate speech environment;
(ii) The speaker was a direct participant — as subject
or object -- in the reported event or state;
136 Mind, Code and context

b. Shared generic context:

(iii) The information is universally shared, by all
members of the speech community, as part of
their knowledge of their world and culture, most
of it (but not all) coded in the shared lexicon;
(iv) The information pertains to agreed-upon con¬
ventions, rules or games (including logic) to
which both speaker,and hearer normally sub¬
scribe;
(iv) The information is given by divine revelation,
with the speaker and hearer subscribing to the
same higher being(s);

c. Shared discourse context:

(v) The information was asserted earlier in the dis¬
course by the speaker, and the hearer did not
then challenge it.

To the contract clauses in (13) above one must add the following escape
clause:

(14) Escape clause for presupposed information:

"The information is presupposed by the speaker — for

whatever reason, to be either:
(i) somehow known to the hearer, by whatever
means;
or
(ii) otherwise acceptable to the hearer without chal¬
lenge, for whatever reason".

This escape clause is necessary to account for intuitive, inspired or

telepathic sources of information about other minds. It is also required
to account for ersatz presupposition, of the type Lewis (1979) describes
as follows:

"...If at time t something is said that requires presupposition P to be

acceptable, and if P is not presupposed just before f, then ~ ceteris
paribus and within certain limits — presupposition P comes into exist¬
ence at f..." (1979, p. 340)

Presumably part of Lewis's 'within certain limits' involved the non¬

challenging behavior of the hearer.
As one can see, the presupposition mode includes the traditional
necessary, analytic and apriori-synthetic truth, as well as presupposition
and more. At the core of this modality is then the speaker's reasonable
 Propositional Modalities 137

belief that the hearer will refrain from challenging the information, and
thus that the information requires no justification, no defense.

4.3.2.4.2. Realis-assertion

Under this mode, information is strongly asserted, yet it remains

open to challenge by the hearer. The speaker must then be prepared to
defend the information, by citing the source of evidence. Most typical¬
ly, the source tends to be one of the following:

(15) Sources of evidence in defense of R-asserted information:

(a) Direct experience, by various sensory modalities;

(b) Hearsay, i.e. the communicated experience of others;
(c) Inference — direct or circuitous -- from direct or com¬
municated experience.

4.3.2.4.3. Irrealis-assertion

Under the scope of the IRR-assertion mode, information is weakly as¬

serted, as hypothesis, possibility, probability, supposition, conjecture,
prediction or guess. The source of the information is thus largely ir¬
relevant, since the speaker does not intend to defend the information
too vigorously against challenge. In fact, often the speaker volunteers
the information under the irrealis mode precisely in order to solicit chal¬
lenge, correction or corroboration.

4.3.2 5. Rules of evidence

The phenomenon of evidentiality in the grammar of natural lan¬

guages most commonly skips over presupposed and IRR-asserted
clauses. It thus pertains primarily to R-asserted clauses.12 One may thus

12 Hargreaves (1983) reports that in Newari the grammar of evidentiality pertains to at least
one irrealis clause, the future tense. This is relatively rare. Further, though the same
grammatical markers may be used, the distinctions become those of degree of subjective
certainty, as in (10) above. The contrast between ko and ngo in KinyaRwanda is another case
in point. A similar case is also found in Sango (H. Pasch, in personal communication), where
the presence of a complementizer particle at the top of the certainty scale codes the
difference between 'know' and 'think'. Toward the bottom of the scale, the same com¬
plementizer codes only degree of subjective certainty.
138 Mind, Code and context

suggest that presupposed clauses are considered -- per the communica¬

tive contract -- self evident.13 IRR-asserted clauses, in turn, are asserted
weakly and without sufficient evidentiary support, and without intent
to strongly defend the verity of the proffered information. Only in the
case of R-asserted clauses is evidence assumed to be both available and
expected.
When a language marks evidentiality in its grammar, most commonly
the following distinction are observed, ranked according to the degree
of evidentiary strength, thus corresponding closely to subjective cer¬
tainty:

(16) Scale of evidentiary strength of source:

a. Direct sensory experience

b. Inference from direct sensory experience
c. Indirect inference
d. Hearsay

In languages which further differentiate among several possible sensory

sources of direct evidence, the grammar of evidentiality tends to rank
the senses according to their reliability as source of evidence:

(17) Scale of reliability of sensory evidence:

a. Visual experience
b. Auditory experience
c. Other sensory experience

In the grammar of evidentiality, one also finds the ranking of the par¬
ticipants in the event according to person:

(18) Scale of participants in event:

a. Speaker
b. Hearer
c. Third party

This ranking exempts from the burden of evidence - and thus from the
grammar of evidentiality ~ reports about events/states in which the
speaker was a participant. By the same token, it is considered odd to in¬
form hearers about events in which they participated — unless the
speech act is not informative, but rather a complaint, accusation or ques¬
tion. Thus, evidentiary particles are most commonly found in clauses

13 See DuBois (1985).

 Propositional Modalities 139

reporting states or events with third-party participants, in particular

third person subjects.
In a number of languages, the grammar of evidentiality also ranks the
spatial proximity of the reported event/state to the speech situation,
again in terms of evidentiary strength:

(19) Scale of spatial proximity:

a. Near the speech situation

b. Away from the speech situation

This scale, in a sense, is not entirely independent of scale (18) above,

since the speaker's proximity to the reported event raises the prob¬
ability of direct sensory access.
Finally, the grammar of evidentiality also tends to rank the temporal
proximity of the reported event to the speech time, in a rather predict¬
able way:

(20) Scale of temporal proximity:

a. Nearer to speech time

b. Farther away from speech time

Thus, of the two major realis tenses, present-progressive and past, the
former is much more often exempt from evidentiality marking, while
the latter is much more likely to require evidentiary marking. Again,
this deictic scale is not altogether independent of scales (18) and (19)
above, since temporal proximity of the event to speech time is often as¬
sociated with spatial proximity to speech place, thus with the speaker's
direct presence at the reported scene.

4.3.2.6. Certainty, challenge and response

As one would expect from the discussion above, the three epistemic
modalities -- presupposition, R-assertion and IRR-assertion -- receive
rather different treatment under the last clause of the communicative
contract, the one dealing with the legitimacy of challenge by the hearer,
and the appropriate reaction by the speaker to such challenge. As noted
above, presupposed information is proffered with the expectation of no
forthcoming challenge -- and thus no need for evidentiary defense. The
reaction to a challenge, if and when one comes, may of course depend
on the contractual sub-clause the speaker assumes has been violated.
When the violated sub-clause involves shared textual information,
about some specific event, the reaction to challenge may range from
140 Mind, Code and context

faint surprise to mild protest to extreme indignation, depending on the

real-world pragmatic consequences of the misunderstanding. To il¬
lustrate this, consider the same propositional information, proffered by
speaker A in (21) as presupposition, and in (22) as R-assertion:

(21) A: Joe is really depressed over Marion leaving him.

B: Hold it -- Marion left Joe? When?
A: You mean you didn't know? I thought you did, it's all
over town; she left him Thursday, they say it was quite a
scene.

(22) A: Marion left Joe.

B: She did? When?
A: Last Thursday, they say it was quite a scene.

A's reaction to B's ignorance will be, typically, much milder in (22) than
in (21).
A misunderstanding arising from the sub-clause pertaining to shared
('deictic') situation could cause quite an irritable reaction. Thus con¬
sider:

(23) A: Pass me that monkey-wrench, will ya.

B: Which one? Where?
A: There's only one monkey-wrench right there next to
you! What's the matter, can't you see?

Considerable irritation may be also provoked by a misunderstanding

arising from the sub-clause that governs generic shared knowledge, as
in:

(24) A: Pass me a medium-size monkey-wrench, will you.

B: Monkey-wrench, monkey-wrench...what's that?
A: A monkey wrench! No, not that, that's a lock-spanner!
Don't you know what a monkey-wrench is?

The examples above are only suggestive, but the challenge and reac¬
tion behavior of speakers and hearers can also be investigated empiri¬
cally. In a recent study, Tsuchihashi (1983) observed speaker's and
hearer's challenge and reaction behavior in transcripts of Japanese con¬
versation. In Japanese grammar, propositions may be marked by either
one of 15 verb-final speech-act particles. These particles rank the
proposition all the way from prototypical strong assertion through
weak assertion, hedged assertion and tentative request for confirmation.
 Propositional Modalities 141

down to a prototypical yes/no question. In her study Tsuchihashi

found the following three correlations:

(a) Propositions on the strong-assertion end had the highest

probability of the speaker being a participant in the re¬
ported event; that probability decreased gradually through
the weak-assertion portion of the scale, and was the lowest
at the question end of the scale.

(b) The probability of assertive intervention from the hearer

was lowest at the strong-assertion end of the scale; it in¬
creased gradually through the weak-assertion portion of the
scale, and was the highest at the question end of the scale.

(c) The probability of the speaker responding in a contrary

manner to the hearer's intervention was highest at the
strong assertion end of the scale decreased through the
weak-assertion portion of the scale, and was lowest at the
question end of the scale.

Tsuchihashi thus confirmed the suggestions made above concerning

the pragmatics of epistemic modalities. In addition, her study also fur¬
nished strong support for another pragmatic challenge to the logical
tradition of propositional modalities: It clearly suggests that these
modalities may be on a graduated scale, rather than discretely cat¬
egorized in a Platonic fashion.14 We will deal with this subject further
below.

4.4. The continuum space of propositional modalities

4.4.1. Preliminaries
So far we have treated the three epistemic propositional modalities -
presupposition, R-assertion, IRR-assertion - as discrete entities, each
with its own behavior vis-a-vis various sub-clauses of the communica¬
tive contract. In this section I will survey evidence suggesting that such
rigid discreteness - while justified in the most gross features - is not al¬
ways compatible with other facts of natural language. We will see how
epistemic modalities shade gradually into one another. We will also see
how epistemic modalities may shade into non-epistemic ones, such as
manipulative speech acts. This gradual shading, so typical of a prag-

14 See also Givon (1982a, 1984e).

142 Mind, Code and context

matically-based system, eventually force us to reassess the nature of the

semantic/pragmatic space underlying all propositional modalities. The
reassessment is, of course, in line with the prototype approach to cat¬
egorization, outlined in Chapter 2, above.
Gradation within as well as between propositional modalities should
be expected, if one accepts the validity of the underlying parameters
suggested in our pragmatic treatment of epistemic modalities, above.
These parameters may be recapitulated, together with a some added
ones, as:

(25) Parameters underlying the mental space of propositional

modalities:

(a) Subjective certainty

(b) Trust in the reliability of evidence
(c) Willingness to entertain challenge
(d) Propensity to respond to challenge
(e) Gradient of authority/power
(f) Strength of intent to act, or to manipulate others to act

Of these six parameters, the first four -- (25a-d) -- involve primarily the
pragmatics of epistemic modalities. The last two -- (25e,f) -- shade into
non-epistemic modes. All six parameters are, at least in principle, non¬
discrete psychological dimensions. As noted in Chapter 2, above,
human categorization, be it in grammar, perception or cognition, is
more often than not a hybrid, a functionally motivated compromise be¬
tween Wittgensteinean gradation and Platonic discreteness. Discovering
a similar balance between discreteness and scalarity in the domain of
epistemic modalities is thus only to be expected.

4.4.2. The non-discreteness of presupposition

4.4.2.1. General considerations

Scalarity within the various types of presupposition may arise from a
large number of sources, depending on the sub-clause in the commu¬
nicative contract through which presupposition arises (see (13) above).
In order:

(a) Deictic obviousness: What is present within the field of

vision may be closer or more remote, more obvious or less
obvious; the hearer may be an acute, a careless, or an im¬
paired observer, raptly attending or momentarily distracted.
Considerable information concerning all this is available to
 Propositional Modalities 143

the speaker, in his/her assignment of probabilities concern¬

ing the hearer's belief.

(b) Speaker participant: The speaker may have been an active,

central, closely-located participant in the reported event, or a
passive, peripheral, remotely-located one. The speaker's
first-hand access to the information may thus vary.

(c) Generic information: Every social group displays some

hierarchic sub-divisions, with sub-groups and sub-sub-
groups. The degree of generality of information within the
entire population thus varies. Speakers are conscious of such
variation, which presumably figures in their assessment of
what the hearer is likely to know on general grounds.

(d) Divine revelation: The Divine is presumably held in ab¬

solute awe, but even there a culture may observe a number
of divinities, ranked as to their strength and omniscience.

(e) Mention in prior discourse: Here there are a number of

reason for scalarity. First, the time-lag within which the
hearer is entitled to legitimately challenge a strong assertion
may be flexibly graded. Second, the time-gap between a
strong assertion and the current time of speech may vary,
giving rise to memory gradation. Lastly, information that is
thematically more central in the discourse is retained longer
than less central information.15 But thematic centrality is
often a matter of degree.

(f) Rules, games, conventions: Rules are typically rigid, but as

Wittgenstein (1953) has pointed out, the scope of their ap¬
plicability may vary, as would the degree of their centrality
to the game. Further, since they are generically-shared con¬
ventions, what was said in (c) above also holds here.

(g) Unpredictable information: The source of specific informa¬

tion the speaker has about what the hearer knows may vary
in reliability. In addition, what applied to prior-text, (e)
above, also applies here.

In sum, then, none of the sources from which presupposed information

arises is immune to scalarity.

15 For an empirical study of this, see Anderson, Garrod and Sanford (1983).
144 Mind, Code and context

4.4.2.2. Truth, Knowledge, belief or familiarity?

The grammar of many languages treats a number of clause-types that
are logically not presuppositional as if they were presupposed. Those
clauses are then contrasted with R-asserted clauses. Most conspicuous
among such clause-types are:

(a) Conditionals
(b) yes-no-questions
(c) NEG-assertions

Concerning the first, Haiman (1978) cites much cross-language evidence

in support of the topical status of hypothetical conditional. Now, since
conditionals are irrealis clauses, they could not be logically presupposed.
Nonetheless, they must be considered background clauses, in the sense
that the information carried in them has been discussed or entertained
as possibility in the preceding discourse. The speaker takes for granted
the accessibility of that information to the hearer; but he does not as¬
sume that the hearer believes in it, nor that the hearer knows they to be
true.16
Concerning the second, Bolinger (1978) has shown that yes/no-ques-
tion, which logically have no truth value (and were listed in (8) above
as a sub-type of irrealis), nonetheless are presuppositional in the follow¬
ing sense: The speaker has a considerable bias toward either the affirm¬
ative or the negative, a bias that is reflected consistently in the gram¬
matical form of the question. Thus, consider:

(26) a. Was Elsa there?

b. Wasn't Elsa there?
c. No, she wasn't there.
d. Yes, she was there.

Bolinger argues, I think convincingly, that in uttering (26a) one is more

prone to expect the negative response (26c); while in uttering (26b) be
one is more prone to expect the positive response (26d).
Concerning the third, it has been shown17 that NEG-assertions, which
have a NEG-truth value, are typically used in discourse contexts where
the corresponding affirmative had either been mentioned, discussed,
entertained, contemplated or raised as a possibility. To illustrate this
briefly, compare (27) and (28) below:

16 Further empirical, text-based support for this may be found in Ramsay (1987), where it was
shown that Haiman's generalization applies to pre-posed ADV-clauses, but not to post-
posed ones. A similar suggestion was made in Givon (1982a).
17 Givon (1979a, ch. 3); see discussion further below.
 Propositional Modalities 145

(27) A: Hey, what's new?

B: Well, let's see -- my boy graduated yesterday.
A: Congratulations!

(28) A: Hey, what's new?

B: Well, let's see -- my boy didn't graduate yesterday.
A: Was he supposed to? I didn't know he was that old.

Affirmative R-assertions, it seems, are expressed on the background of

the hearer's relative ignorance.18 NEG-assertions, on the other hand, are
expressed on the much richer background of expectations, namely that
the corresponding affirmative is at least familiar to the hearer, and more
typically that the hearer is indeed disposed to believe in the corre¬
sponding affirmative. But this is clearly not a presupposition in the logi¬
cal sense (i.e. the speaker 'believing in the truth' of the corresponding af¬
firmative). If that were the case, the speaker would have been caught in
the bind of presupposing the truth of p while asserting its falsity.
The gist of this discussion is as follows: Logicians have traditionally
defined the epistemic mode of presupposition either in terms of the truth
of p,19 or at the very least — if a measure of pragmatics has been allowed
to prevail20 -- in terms of the speaker's belief in the truth of p. When
hard-pressed by the increasingly recalcitrant data pf human language,
logicians embarked upon various rear-guard operations, splitting pre¬
supposition into an increasing number of discrete categories: Logical
presupposition, pragmatic presupposition, conventional implicature,
conversational postulates and what not.21 Natural language, on the
other hand, seems to treat presupposition pragmatically from the very
start, in at least two distinct senses:

(a) Presupposition is a matter of the speaker's belief about the

hearer's state of mind, not about the truth of some proposi¬
tion; and

(b) Presupposition pertains not only to the hearer's strong

beliefs, but also to the hearer's weaker beliefs, assumptions,
predispositions or even vague familiarity with a proposi¬
tion.

18 Information is never transacted on the background of total ignorance, or lack of any

presumptions about the hearer's beliefs. For discussion, see Givon (1984a, ch. 7).
19 As in, for example, Keenan (1969,1971); Horn (1972), Oh and Dinneen (eds, 1979), inter alia.
20 As in, for example, Karttunen (1974).
21 See, Gazdar (1979), inter alia.
146 Mind, Code and context

One could then easily scale various grammatical constructions ac¬

cording to strength of presuppositionality, roughly as follows:

(29) Scale of presuppositional strength:

a. Strongest: Complements of 'regret', 'be happy', 'be

good'
b. Strong: WH-questions, cleft-clauses, relative clauses;
Complements of 'know', 'discover', 'forget'
c. Weaker: Various nominalized clauses;
Realis adverbial clauses
d. Weakest: conditionals, negatives, yes/no-questions

One may similarly scale know, believe in and be familiar with on a sliding
scale. While the top of these hierarchies tends to display the so-called
'logical' presupposition, far below the top some pragmatic presupposi¬
tion may persist.

4.4.2.3. Presupposition vs. backgroundedness

As suggested in (6), above, restrictive relative clauses are, typically,
strongly presuppositional. Consider now the contrast between an RR-
clause modifying a definite head noun, and an RR-clause modifying an
indefinite head noun:

(29) a. Definite head:

The man we hired yesterday came in this morning and
resigned.
b. Indefinite head:
A man we hired yesterday came in this morning and
resigned.

While a sentence such as (29a) is uttered normally under an expectation

that the hearer shares the information that 'we hired that man yesterday',
the same is not true of (29b). This is so because when the speaker uses
an indefinite noun, he does not assume that the hearer is familiar with
that noun.22 Therefore, the hearer could not be familiar with an event --
coded in the RR-clause -- in which that noun was a participant. Still, the
RR-clause in (29b) is not strongly asserted, even though it is clearly new
to the hearer. Rather, the information in the RR-clause is backgrounded.
But why is that information coded by a structure that, in other contexts,
when the head is definite, codes shared background information?

22 See discussion in chapter 6, below.

 Propositional Modalities 147

The common denominator for employing RR-clauses in (29a,b) is this

pragmatic notion of backgroundedness. The more logical sense of presup¬
position arise as a stronger, more restricted special case under some dis¬
course conditions. The pragmatic notion of backgroundedness entails
that the speaker expects that the hearer will not challenge the informa¬
tion.23 In the stronger case of presupposed old information, no challenge
is expected because of presumed shared knowledge. In the weaker case
of backgrounded new information, the speaker invites the hearer not to
challenge the information by coding it as subsidiary, non-central, less
important. The speaker thus gives a cataphoric blank check to the
hearer, promising that the information will indeed not be of great
thematic importance in the subsequent discourse. As we shall see in
Chapter 5, above, marking information in such cataphoric terms is a
major feature of discourse pragmatics.

4.4.3. The non-discreteness of assertions

We have already noted that the grammar of evidentiality scales R-

assertions according to evidentiary strength (and thus also subjective
certainty; see (15) through (20), above). IRR-assertions are likewise
scaled for subjective certainty (see (10), above). In scaling both R-asser-
tions and IRR-assertions in English, a wide variety of grammatical
means may be used:

(30) a. Modal verbs: will, may, must, should, can, might

b. Cognition verbs: believe, assume, presume, guess,
suspect, think, suppose, surmise, doubt, be sure
c. Adjectives: possible, probable, likely, true, false, al¬
leged, presumed,
d. Adverbs: maybe, perhaps, probably, possibly, pre¬
sumably, necessarily, truly, really, most likely, allegedly

Finally, as already noted earlier above, the transition between R-asser-

tion and IRR-assertion is often gradual, and such gradually is ex¬
pressed quite often in the grammar of natural languages. Various tags
may be use to temper the strength of R-assertions. Such tags are often
augmented with intonation, as well as with various irrealis operators.
Thus consider:

23 Lewis (1979) notes the role of this no challenge assumption in the 'creation' of presupposi¬
tion, observing: "...presupposition evolves according to a rule of accommodation specifymg
that any presuppositions that are required by what is said straightaway come into exist¬
ence, provided that nobody objects..." (1979, p. 347)
148 Mind, Code and context

(31) strongest assertion

a. He's home.
b. He's home I think.
c. , He's home, I think...
d. He may be home.
e. He's home, isn't he.
f. He's home, right?
g. He is home isn't he? , .
h. He is home, right?
i. He is home, or is he?
weakest assertion

As we shall see directly below, this gradation from strong to weak

assertion may proceed on toward questions -- i.e. toward non-declara¬
tive speech acts.

4.4.4. Epistemic modalities and manipulative speech-acts

4 4.4.1. Preamble

The traditional discussion of propositional modalities, from Aristotle

down, has taken for granted a clear division of epistemic from non-epis-
temic modes. A cursory look at the distribution of epistemic modalities
in the grammar of human languages, most particularly the distribution
of IRR-assertion (8), reveals that this clean break can only be maintained
by considerable idealization of the facts. As a conspicuous example,
non-declarative speech acts systematically partake in the mode of ir-
realis — even though they are not assertions. In this section we will con¬
sider a range of facts, all pointing to systematic shading from the epis¬
temic through the intentional and toward the manipulative.
In a sense, the basis for such shading has already been laid down in
our discussion of the pragmatics of epistemic modalities, above. No¬
tions such as certainty and probability of challenge already shade into the
province of intent, power and action. This shading is not logically
necessary. Rather, it involves pragmatic inferences about likelihood and
normativity. We will begin with a discussion of the connection between
irrealis and intentionality.

4.4.4.2. From irrealis to intent to power

One of the most typical irrealis contexts is the scope of verbs of intent,
ability and obligation, as in:
 Propositional Modalities 149

(32) a. She wanted to leave.

b. He was able to work.
c. They had to come in.

Take want first. Intent is presumably a purely-intemal motivation. But

intent to change the state of the world from that which exists to that
which does not yet, invites considerations of one's ability, thus one's
power to affect the change.
Next, ability may be either internally engendered, or facilitated by
lack of external restraint. Either way, it implies some -- internal or exter¬
nal -- restraints, as well as the power to overcome them. Further, exter¬
nal restraints -- when consciously acknowledged -- shade into the no¬
tion of obligation.
Finally, obligation may be either internally or externally motivated.
Either way it involves power relations vis-a-vis some other participant.
Within the seemingly most innocuous intentional scopes, then, the no¬
tion of power, thus also of power gradients vis-a-vis others, is already
implicit.
The power gradient vis-a-vis an external agent is even more apparent
in another typical irrealis context, the scope of (non-implicative) manip¬
ulative verbs, as in:

(33) a. She wanted him to leave.

b. He told her to go.
c. She asked him to do it.

Examples (33) all involve -- under the scope of a manipulative verb — an

embedded irrealis clause. And the semantic interpretation of that em¬
bedded clause depends on various socio-pragmatic notions, such as
manipulation, power gradients, obligations, propensity to act etc.

4.4.4 3. Subjunctives: From certainty to manipulation

Consider next the distribution of another wide-spread irrealis clause,

the subjunctive. In Spanish, one finds this clause-type most commonly
in two embedded complement clauses:

(34) a. The subjunctive of uncertainty:

no se si venga
no know-I if come-she-SUB
'I don't know if she's coming'
150 Mind, Code and context

b. The subjunctive of manipulation:

Le dije que viniera

hersaid-I that come-she-PAST/SUB
'I told her that she should, come'

The same grammatical form thus seems to span the space from the
epistemics of uncertainty to attempted manipulation. The very same form
is also used in the polite imperative, a non-declarative speech act:

(35) venga\
come-you/SUB
'Come!'

The same sharing of a subjunctive form can be shown in English, in

the systematic triple-use of 'should' and 'must': First as epistemic-
modal subjunctives; second as manipulative subjunctives; and third in
manipulative speech acts:

(36) (i) Epistemic-modal:

a. He must have done it.
b. She should be here in five minutes

(ii) Obligative-manipulative:
c. She told him that he must leave.
d. He told her she should go.

(iii) Manipulative speech act:

e. You must leave immediately!
f. You shouldn't stand there!

Many other languages show this systematic pattern. In natural lan¬

guage, this systematic sharing of grammatical form is most commonly
the footprint of the historical process of analogic extension. Through such
a process, the functional scope of one structure is extended, to code other
functions that were not previously coded by the same structure. Such
extension almost always involves the judgements of similarity of the two
functions. And when many — typologically and genetically unrelated —
languages exhibit the very same pattern of analogical extension, one is
entitled to conclude that the human mind perceives the two functions
involved as similar or related.24

24 Analogical extension can also proceed by structural similarity. For some discussion of the
historical processes through which systematic coding correlations may be achieved, see
Givon (1984a, Ch. 2).
 Propositional Modalities 151

4.4.5. The speech act continuum

4.4.5.1. Preamble

The speech-act mode of propositions has been a notorious stumbling

block in the logical tradition's attempt to represent human language.
The purposive context cannot easily be represented as a semantic
property of the proposition itself, it fairly reeks of the pragmatics of
speaker, hearer and context. It is thus not surprising to find that pre-
Socratic philosophers, such as the sophists, were both familiar with and
interested in non-declarative speech-acts. In this connection, Haberland
(1985) observes:

"...Protagoras distinguishes four parts of discourse..., namely "wish,

question, answer and command"...Protagoras seems to have been inter¬
ested in speech acts, not sentences in modern parlance. But... it is state¬
ments Plato is interested in..." (1985, p. 381; emphases are mine; TG)

Plato's role in narrowing the focus of philosophical interest to the study

of declarative propositions alone is well documented. Haberland iden¬
tifies the reason for Plato's diminished focus as follows:

"...For Plato...true knowledge — which, as he argues in this connec¬

tion, does not coincide with perception -- cannot aim at context-depend¬
ent truths; the truth of a sentence should not depend on who says it, in
which situation, and to whom, and it should not, more specifically,
depend on what the sentence is an answer to... But this interest of
Plato's in statements...is again only understandable from a series of
premises that are no longer self-evident. The first of these is that truth is
the main concern of the philosopher. The second is that analysis of language
is ancillary to philosophical pursuits. As a corollary from these two
premises, we get that linguistic analysis is mainly concerned with truth as
well. The third premise is that truth is timeless and independent of con¬
text..." (1985, pp. 381-382; emphases are mine; TG).

Eventually, Ordinary Language Philosophers (cf. Austin, 1962),

shifted the focus away from the mere truth of linguistic expressions,
and back to their felicity, or appropriateness in context. Subsequently,
speech-acts have been intensely investigated through 25 years of
modem pragmatics, producing a vast literature.25 In the discussion here,
we will not attempt to recapitulate that literature. Rather, we will con¬
fine ourselves largely to three issues:

25 For more recent discussion, see many of the contributions in Cole and Morgan (eds, 1975)
and Cole (ed., 1981), inter alia.
152 Mind, Code and context

(a) The functional-semantic continuum underlying speech acts

(b) The systematic shading between speech acts
(c) The systematic shading between non-declarative speech acts
and the irrealis epistemic modality.

A growing body of cross-language typological studies indicate that

three or four syntactic constructions consistently code the same three or
four major speech-act prototypes in human languages.26 These pro¬
totypes are:

(37) a. Declarative: Goal = imparting information

b. Imperative: Goal = eliciting action
c. Interrogative: Goal = eliciting information...
(i) WH-question: ...to confirm the identity of an
item
(ii) Y/N-question: ... to confirm the truth of a
proposition

It is hard to find a language in which these four speech-act prototypes

are not coded explicitly, via distinct syntactic constructions. Traditional
speech-act analysis has tended to describe these functions as absolute
and discrete.27 This tradition, of studying the prototype peaks without
paying attention to the graded terrain between them, incurs certain em¬
pirical costs. These costs are especially visible in the analysis of the so-
called indirect speech acts.
Explicitly or implicitly, the literature on indirect speech-acts identifies
those as 'meaning one thing while masquerading as another',28 bizarre
coding infelicities that perform the function of one speech act while as¬
suming the syntactic structure of another. This is in spite of a consider¬
able body of data (see survey in Brown and Levinson, 1978) suggesting
that the so-called 'masquerade' is surprisingly systematic cross-linguis-
tically. That is, that the same range of 'indirect' constructions are used in
unrelated languages to perform the same range of speech-act functions.
Still, the bulk of traditional attempts to deal with this chimera presup¬
poses the discreteness of speech-acts.29
There are strong factual grounds for suspecting that the three-or-four
major syntactically-coded speech acts (37) are just the most common.

26 For many cross-language details, see Sadock and Zwicky (1985). For interrogatives, see
Chisholm (ed., 1984).
27 See for example Austin (1962), Searle (1970), Grice (1968/1975), inter alia.
28 See e.g. Sadock (1970), Green (1970, 1975), Gordon and Lakoff (1971), Searle (1975) or
Davison (1975), inter alia.
29 See discussion in Levinson (1983), pp. 263-278. The lone exception seems to be Lyons (1977,
pp. 753-768), where the shading from epistemic doubt (i.e. irrealis) to interrogative speech-
acts is acknowledged, and where questions are considered 'grammaticalized features of
doubt'.
 Propositional Modalities 153

conventionalized (thus 'grammaticalized') prototype peaks, spanning a

multi-dimensional mental space, one made of several non-discrete socio-
psychological dimensions. This multi-dimensional continuum under¬
lies the entire functional domain of speaker-hearer interaction, i.e. our
communicative contract. Within this continuum, the various speech-act
prototypes shade into each other. In the following sections some of this
shading will be illustrated.

4.4.5.2. From imperative to interrogative

Consider first the continuum in (38) below, between prototypical im¬

perative and interrogative speech acts:

(38) most prototypical imperative

a. Pass the salt.
b. Please pass the salt.
c. Pass the salt, would you please?
d. Would you please pass the salt?
e. Could you please pass the salt?
f. Can you pass the salt?
g. Do you see the salt?
h. Is there any salt around?
i. Was there any salt there?
most prototypical interrogative

The extreme points of scale (38) correspond to the well-coded pro¬

totypes. The mid-points on the scale -- (38c,d,e) -- exhibit intermediate
features both functionally and syntactically. Finally, intermediate (38b)
is more like the imperative prototype (38a), and intermediates (38f,g,h)
are closer to the interrogative prototype (38i).
One may consider the intermediate points in (38) as a continuum of
metaphoric extensions between the prototypes of imperative and inter¬
rogative. Felicity conditions under which such extensions are appro¬
priate have been discussed in the speech-act literature.30 The exact na¬
ture of the socio-psychological dimensions that underlie the continuum
in (38) is yet to be empirically determined, by methods which sooner or
later must transcend the traditional arsenal of linguistics and philos¬
ophy. As a weak hypothesis, the following dimensions may be sug¬
gested:31

30 See e.g. Grice (1968/1975) or Gordon and Lakoff (1971), inter alia.
31 These are similar to various features suggested in, e.g., Grice (1968/1975), Gordon and
Lakoff (1971) or Searle (1975). One should not consider these features to be strictly
linguistic. They are perhaps primarily socio-psychological, governing human interaction
in general. As such, they are of course available to mediate verbal interaction.
154 Mind, Code and context

(39) a. The power (authority) gradient between speaker and

hearer

b. The degree of the speaker's ignorance concerning a

state of affairs about which he wishes to learn

c. The degree of the speaker's sense of urgency or deter¬

mination vis-a-vis the attempted manipulation

All three parameters are scalar, thus representing a multi-dimensional

space. One takes it for granted that not all the logically-possible slots
within that space are actually instantiated in natural languages. The
relatively small sub-set that tends to get coded more-or-less systemati¬
cally are constrained, presumably, in part by cognitive universal, in part
by human-universals real-world adaptive tasks, in part by culture-
specific adaptive considerations.

4.4.5.3. From imperative to declarative

Consider next the scale in (40) below, spanning the continuum be¬
tween the prototypes of imperative and declarative:

(40) most prototypical imperative

a. Wash the dishes.
b. You better wash the dishes.
c. You might as well wash the dishes.
d. I suggest you wash the dishes.
e. It would be nice if you could wash the dishes.
f. It would be nice if someone could wash the dishes.
g. The dishes need to be washed.
h. The dishes are dirty.
i. The dishes were dirty,
most prototypical declarative

Expression (40a) is syntactically a prototype imperative, with the agent

left unexpressed, being predictable from the speech situation. Inter¬
mediates (40b,c,d,e) explicitly code the subject/agent 'you'. The manip¬
ulative force of the speech-act gradually recedes, as the syntactic form is
slowly transformed toward the declarative prototype. In (40f) 'you' is
replaced by 'someone', an impersonalization that further decreases the
manipulative power. The impersonalization is further underscored with
the passive form (40g). Finally, both (40h,i) are, syntactically, fairly
 Propositional Modalities 155

prototypical declarative. And the shift to past tense in (40i) makes a

situation-bound imperative interpretation virtually untenable.32
Of the underlying socio-pragmatic dimensions in (39), the two per¬
taining to imperatives -- (39a,c) -- are presumably also involved in the
continuum in (40). In addition, the following features are likely to also
be involved:

(41) a. The speaker's subjective certainty of the information

b. The speaker's assessment of the hearer's ignorance of
that information
c. The speaker's assessment of the strength of the hearer's
motivation to learn that information

Presumably, degree of similarity in form along the scale (40) repre¬

sents, isomorphically, degree of similarity along the relevant functional
dimensions.

4.4.5.4. From declarative to interrogative

One may as well note that a similar continuum exists between the
prototypes of declarative and interrogative. Consider first the scale in¬
volving Y/N-questions:

(42) most prototypical declarative

a. Joe is at home.
b. Joe is at home, I think.
c. Joe is at home, right?
d. Joe is at home, isn't he?
e. Is Joe at home?
most prototypical interrogative

This continuum involves, among other features, a clear decrease in the

speaker's subjective certainty concerning the information in his/her
possession. It thus involves the irrealis - doubt — modality as an inter¬
mediate point (e.g. (42b) on the scale between R-assertion and Y/N-
question. The number of well-coded morpho-syntactic points on this
scale may be much larger, in English as well as in other languages.33
Finally, consider the continuum between declarative and WH-ques-
tion:

32 This is obviously not the entire story. Other types of thematic and real-world pragmatic
'redundant' information are also involved in determining the likelihood of manipulative
interpretation.
33 See Tsuchihashi's (1983) description of this continuum in Japanese.
156 Mind, Code and context

(43) most prototypical declarative

a. Joe called, and...
b. What's-his-name called, and...
c. Whoever it was that called, tell them...
d. I don't know who called.
e. Who knows who called.
f. Who called?
most prototypical interrogative

A similar gradation is seen here as in (42), from R-assertion ('certainty')

through IRR-assertion ('uncertainty') to question.34
In sum, neither the socio-pragmatic space underlying speech-acts, nor
the syntactic structures that code them, isolate them completely from
other propositional modalities, or from each other. One can indeed dis¬
cern clear prototypes in both the epistemic ('declarative') and manipula¬
tive ('interrogative', 'imperative') regions of our modal space. However,
to account for both the functional organization and syntactic coding of
the space, one must adopt a partially-flexible, partially open-ended
pragmatic approach.

4.5. The pragmatics of NEG-assertion

4.5.1. Preamble

As suggested earlier, NEG-assertion is, in some ways, a hybrid epis¬

temic modality. It shares some features of each of the other three modes.
First, in terms of subjective certainty, thus the pragmatics of challenge
and defense, it resembles R-assertion, being also a strong assertion.
Second, in terms of the referential opacity vis-a-vis nominals under its
scope, it resembles IRR-assertion. Finally, in terms of the pragmatics of
shared background, it resembles the modality of presupposition. We
will begin the discussion here by outlining this last aspect of the prag¬
matics of negation.

4.5.2. Negation as a speech act

The logical tradition has always considered negation to be a purely

semantic operation, one which merely reversed the truth value of a

34 The use of WH forms to code epistemic uncertainty is wide-spread. Thus, for example, in
Ute any WH pronoun can be use in low-certainty declarative sentences, if the speaker is
unsure of either the token-identity of a nominal, or of its type-identity. Such expressions
are usually translated into English as: 'what's-his-name', 'what-cha-ma-call-it', 'who
knows' etc.
 Propositional Modalities 157

proposition. In that way, negation in language resembles the negative

operator in math. As suggested above, however, negation also carries a
distinct pragmatic component. NEG-assertions are in some sense
presup positional, though clearly not logically so. They are used when
the speaker assumes that the hearer either believes in the truth of the
corresponding affirmative proposition; or perhaps entertains the pos¬
sibility of its being true; or that the corresponding affirmative had been
established as background expectation in the discourse. The latter con¬
dition, the weakest of the three, is probably their common denominator.
NEG-assertions are used, most often, as a distinct speech act. While R-
assertions purport to inform, NEG-assertions aim to contradict, correct
or deny. What is denied is the background expectation, the correspond¬
ing affirmative.
Establishing the corresponding affirmative as background for a NEG-
assertion may be accomplished explicitly in the preceding discourse, as in:

(44) Background: Joe told me that he won ten grand in the lot¬
tery...
NEG-assertion: ...tho later I found out he didn't

In (44) the speaker himself/herself sets up the affirmative expectation,

then contradicts it. Background expectations may also be due to the in¬
terlocutor, as in:

(45) Background: A: I understand you're leaving tomorrow.

NEG-assertion: B: No, I'm not. Who told you that?

The speaker may rely, in assuming background expectations, on specific

knowledge about the hearer's state of affairs or state of mind. To il¬
lustrate this, consider the felicity of the three responses to the NEG-
assertion below:

(46) A: So you didn’t leave after all.

B(i): No, it turned out to be unnecessary.
B(ii): Who said I was going to leave?
B(iii): How did you know I was going to?

Response (46Bi) suggests that the hearer accepted the corresponding af¬
firmative as shared information. Response (46Bii) suggests that the
hearer believes the speaker is mislead. In response (46Biii), the hearer
registers surprise at how the information leaked out to the speaker, by
inference thus conceding its truth.
Finally, background expectations to a NEG-assertion can also be trac-
gjj to generically shared information. To illustrate this, consider.
158 Mind, Code and context

(47) a. There was once a man who had no head

b. ?There was once a man who had one head
c. ?There was once a man who didn't look like a frog
d. There was once a man who looked like a frog

The reason why the negative (47a) is pragmatically felicitous is be¬

cause it reports a break from the norm- The reason why (47b) is prag¬
matically odd is because it merely echoes that norm, and is thus tauto¬
logical. Conversely, the negative (47c) paraphrases the norm and is thus
pragmatically odd; while the affirmative (47d) breaks the norm, and is
thus pragmatically felicitous. The background norm thus turns out to be
crucial ingredient for understanding the pragmatic felicity of proposi¬
tions. If we were to live in a universe where men had no heads, or where
they most commonly resembled frogs, both felicity contrasts in (47)
would have been reversed.
But if both the affirmative and negative can be, equally, expected
background, why do we consider NEG-assertions to be somehow spe¬
cial, more presuppositional, marked? In order to answer this, one must
consider the ontological status of non-states and non-events. We will do
that under two separate headings:

(a) Explicit vs. implicit normative expectations

(b) The skewed pragmatics of states and events

4.5.3. Explicit and implicit normative expectations

An affirmative background assumption for (47a), above, is relatively

explicit;35 it is derived from our generic knowledge of 'man' as stored in
our shared lexicon. Namely, All men have one head. The pragmatic felicity
of (47a) is thus achieved on the background of this -- more explicit — as¬
sumption. But there are, in addition, potentially an infinite number of
implicit inferences that one can derive from the prototype 'man'. And
some of those are negative implicit inferences. This explains the prag¬
matic oddity of the NEG-proposition (47c): It paraphrases an implicit
norm. Men do not look like frogs. And it is precisely this implicit norm that
makes the affirmative (47d) pragmatically felicitous.
The derivation of implicit background inferences most likely
proceeds gradually and indirectly. The chain of assumptions in (48)
below may illustrate this likelihood, whereby (48e) is derived by a chain
of inferences:

35 If features of meaning are indeed clustered, prototype-like entities (as suggested in Ch. 2,
above), then the difference between explicit and implicit inferences from 'core' meaning
must be a matter of degree, rather than of kind.
 Propositional Modalities 159

(48) a. Men look, normally, like men.

b. Other species (including frogs) normally look like them¬
selves.
c. Entities are classified -- among other things -- by visual
similarity.
d. Since men are not frogs,
e. therefore men don't look like frogs.

One major difference -- or skewing -- between affirmative and nega¬

tive propositions is that explicit normative information tends to be af¬
firmative. Negative background inferences tend to be more implicit. But
by observing this, we have not answered the question concerning the
skewed status of NEG-asserted information. Rather, we have merely re¬
formulated it.

4.5.4. The ontology of negatives states and events

4.5.4.1. The pragmatic status of negative events

Events are changes in an otherwise inert universe. It is a law of physics

-- inertia -- that motivate this facet of our constructed experience, i.e. the
skewed frequency of happening and not-happening. Events are, prob¬
abilistically, much less expected, much less frequent, than non-events.
Inertia is the background norm, it requires less energy. Breaking the in¬
ertia is the exception, the counter-norm; it requires more energy. It is be¬
cause of this probabilistic skewing that events -- changes -- are more in¬
formative, salient -- foreground -- than non-events. For this very reason,
non-events -- stasis -- tend to be the normative background upon which
events appear salient. Negation is a pun, a reversal, a play upon the
norm. It is used when -- much more rarely in normal communication --
one establishes the event as background expectation. On that back¬
ground, the non-event becomes salient, informative. But this is still the
exception to the norm, a relative rarity in discourse.36
Let us illustrate this with a number of simple examples. Consider first:

(49) a. A man came into my office yesterday and said...

b. *A man didn't come into my office yesterday and said...
c. ?Nobody came into my office yesterday and said...

36 Text counts reveal that in common narrative or conversation, the percent of NEG-assertions
as compared to affirmative assertions seldom exceeds 5 percent (see figures and discussion
in Givon, 1979a, ch. 3).
160 Mind, Code and context

The non-event (49b) is pragmatically -- and indeed grammatically37 --

the oddest. This must be so because if an event did not occur at all, why
should one bother to talk about a referring, specific individual that was
then 'involved' in that non-event?
While more grammatical, (49c) is still pragmatically unlikely. This is
so because the norm of my everyday routine is not that All people visit my
office at all times, but rather that Most people don't ever visit my office. Visits
to my office are thus selective, more rare than non-visits. This is what
makes them events. On the background of such a norm of non-events,
(49a) is indeed pragmatically more felicitous.
Consider next:

(50) a. The man you met yesterday is a crook.

b. ?The man you didn't meet yesterday is a crook.

Normatively, you meet only a few men in a given day. So, to identify a
person by an event — coded in the REL-clause in (50a) — is indeed infor¬
mative, salient, an apt way of distinguishing him from the xillion men
you did not meet that day. Given the norm, (50b) is indeed pragmatical¬
ly odd. Unless the foreground-background relations are reversed, as in,
for example:

(51) You were supposed to meet four men yesterday.

Three showed up, the last one never did.

On the background of (51), the non-event (50b) is indeed salient and

pragmatically felicitous.
Next, consider:

(52) a. Where did you leave the keys?

b. ?Where didn’t you leave the keys?

In general, WH-questions are presuppositional, so that the entire clause,

excepting the WH-element, is taken to be background information. The
affirmative (52a) is pragmatically felicitous because normally there are a
xillion possible places where your keys were not left, but only one place
(at a time) where they were. For that very reason, the negative (52b) is
pragmatically bizarre. Even supposing that the background expecta¬
tions were somehow reversed, say with I didn't leave my keys any place,
(52b) remains odd. This is so because, given that a potentially infinite
number of places would qualify for the correct answer, the purpose of
asking -- to elicit a specific location response -- cannot be fulfilled. In-

37 As noted in Givon, (1979a, ch. 3), referential-indefinite arguments are systematically

excluded from NEG-assertions; see further discussion in section 4.5.5., below.
 Propositional Modalities 161

asking — to elicit a specific location response -- cannot be fulfilled. In¬

deed, (52b) is only pragmatically felicitous as an echo question, where
one heard a NEG-assertion — a denial — but has somehow missed the
location part.

Finally, consider:

(53) a. When John comes. I'll leave.

b. ?When John doesn't come. I'll leave

The affirmative (53a) is felicitous because the time when John comes, on
a particular occasion, can be specified; but the xillion times when John
doesn't come are not exactly denumerable. For that reason, the negative
(53b) is odd -- unless one modifies the background, as in:

(54) I waited and waited there. Finally, when John didn't come, I
left.

What makes (54) felicitous is that it establishes a unique reference point

in time, by which John had not come (in the process masking the mean¬
ing of the plu-perfect with an overt use of the past tense). Once such a
point becomes specifiable, the use of the negative in the time-adverb
clause becomes felicitous.

4.5.4.2. The pragmatics of negative states

It had been observed elsewhere 38 that in paired antonymous adjec¬

tives, most typically of size, extension, elevation, texture, loudness,
brightness, speed, weight etc., the positive member of the pair stands
for both the possession of the property (i.e. the positive extreme) and the
generic designation of the property (i.e. the unmarked). The negative
member, on the other hand, stands only for the negative extreme - of
not possessing the property. This systematic bias is not logically predict¬
able, since in logic both of the following equations are of equal status:

(55) a. not alive = dead

b. not dead = alive

The systematic bias in paired adjectives in human language arises from

an equally systematic skewing of their perceptual saliency: The positive

38 See Bierwisch (1967) and Givon (1970), inter alia; for some psycho-linguistic support, see
H. Clark (1970) and E. Clark (1971).
162 Mind, Code and context

(56) paired adjectives

positive negative perceptual property

big small ease of visual perception
it it ii tt
long short
it it it tt
tall short
it it ti tt
wide narrow
it ti ti tt
thick thin » • *
it n it ti
light dark
high low " " " " above
ground
fast slow ease of visual perception of rate
of change over stationary ground
loud quiet ease of auditory perception
high note low note ease of auditory perception
sharp dull ease of tactile perception
heavy light ease of gravity/tactile perception
rough smooth ease of tactile perception
hard soft ease of tactile perception
spicy bland ease of savory perception
strong weak ease of olfactory perception

Much like non-events, negative qualities tend to form the normative

background — in terms of the frequency of pervasive absence — vis-a-
vis which the more rare presence of qualities is salient.

4.5.5. Negation and irrealis

So far we have surveyed evidence suggesting that the mode of NEG-

assertion shares some of the pragmatic characteristics of presupposi¬
tion, namely some backgroundedness. In this section we will briefly
outline how NEG-assertion also shares some of the semantic properties
of IRR-assertion.
The evidence will be discussed in more detail in Ch. 5, below. It has to
do with referential opacity of propositions, and with the possibility of as¬
signing a non-referential interpretation to nominal arguments within
opaque propositions. Briefly, under the scope of R-assertion and presup¬
position, nominal arguments may only be interpreted referentially. Thus
consider the following examples from English, a language in which in¬
definite NPs may be interpreted either referentially or non-referentially:
 Propositional Modalities 163

(57) a. John saw a movie (R-asserted)

(3 there's a particular movie that John saw)

b. It is good that John saw a movie (presupposed)

(3 there's a particular movie that John saw)

In contrast, under the scope of IRR-assertions one may obtain a non-

referential interpretation of an indefinite NP (as well as, for many IRR-
clause types, also a referential-definite interpretation). Thus consider:

(58) John may go to see a movie tomorrow (IRR-asserted)

(i) There's a particular movie that John plans to see.
(ii) John plans to see some movie, tho neither he nor I have
in mind any particular movie.

Under the scope of negation, a non-referential interpretation of in¬

definites is allowed (as under IRR-assertion), but a REF-indefinite inter¬
pretation is odd. Rather, if an argument under NEG-scope is referential,
it must be definite:

(59) a. John didn't see a movie

(i) there exists no movie such as John saw it
(ii) * 3 there exists a particular movie such as John
didn't see it.

b. John didn't see the movie

(3 there exists a particular movie such as John didn't see

it)

It has been suggested39 that the reason why REF-indefinites are barred
from the scope of negation has to do with the presuppositional status of
negation. Briefly, if a NEG-assertion is uttered in the context of the cor¬
responding affirmative proposition being background knowledge or ex¬
pectation, then a fairly obvious inference must hold:

(60) "If a proposition is familiar to the hearer, the identity of the

referring arguments within that proposition must also be
familiar to the hearer; the argument must thus be definite".

This pragmatically-motivated restriction aside, NEG-assertions still share

with IRR-assertions the semantic property of allowing referential opacity
under their scope.

39 Givon (1979a, ch. 3).

164 Mind, Code and context

4.6. Certainty, power and deference

4.6.1. Preamble
In several sections above we noted that there exist some inferential
connections among various propositional modalities, so that the purely
epistemic modes shade gradually into socio-manipulative modes. These
connections may be summarized as the following one-way inferences:

(61) a. truth z> knowledge

b. knowledge => certainty
c. certainty status
d. status z> power

None of these inferences is logically necessary. Rather, they are prag¬

matic norms associated with the communicative contract. The com¬
municative contract, it seems, is itself embedded within a wider context,
a well-regulated matrix of socio-personal interaction. In this section we
will briefly survey some of the socio-personal dimensions that seem to
be consistently associated with the communicative contract.

4.6.2. Certainty and authority

As suggested above, the communicative contract inexorably ties
together the more epistemic dimensions of knowledge and subjective
certainty, with the more socio-personal dimensions of authority, status
and power. Thus, for example, in the grammar of Japanese (Tsuchihashi,
1983), women are assigned a special verb-marking R-assertion particle,
one that ranks consistently below the prototypical male-used R-asser-
tion particles, in degree of subjective certainty, or determination to beat
down a challenge from the hearer.
More generally, Syder and Pawley (1974) observe that in facing an in¬
terlocutor of higher power ('status', 'authority'), speakers tend to scale
down their expression of certainty, by using hedges that place assertions
in a lower — irrealis — epistemic range. This is not done, necessarily, be¬
cause of perceiving a contrary attitude on the part of the high-status in¬
terlocutor. Rather, the tone-down may be a hedge against the possibility
that the higher authority might hold a contrary belief. Such epistemic
deference to power realities is a pervasive feature of many, perhaps all
cultures.40

40 See Brown and Levinson (1978).

 Propositional Modalities 165

4.6.3. Authority, negation and politeness

As noted above, NEG-assertion is a contrary, denying speech-act. One

would thus expect its use to be extremely sensitive to the perceived so¬
cial position of the interlocutor. This is indeed the case, seemingly
universally, in the use of NEG-assertions in the context when the inter¬
locutor is perceived to be of higher status or power. Under such condi¬
tions, speakers tend to tone down their disagreement, they couch their
contrary opinion in a variety of 'softening' devices. Many of these
devices are sub-varieties of irrealis, as in:

(62) a. Quite, quite.

b. Yes, I see.
c. I see what you mean.
d. I suppose you got a point there.
e. Perhaps not quite so.
f. Perhaps you may wish to consider an alternative.
g. Well, I'm not sure about that, maybe...
h. Now if it were up to me, I would suggest...

In more traditional cultures it is often not easy to find any overtly-

marked negation in speech directed toward perceived superiors or out¬
siders.
Somewhat paradoxically, negation can itself be used as a softening
operator in the face of perceived higher authority. This toning down
function of negation seems to apply to both epistemic and manipulative
modes. Thus consider:

(63) a. Won't you come in please?

(= Do come in)
b. I suppose he isn't done yet.
(= I wonder if he's done)
c. I don't suppose he's done yet?
(= I wonder if he's done)
d. Wouldn't it be better if...
(= It'd be better if...)
e. I suppose you couldn't spare a fiver...
(= I wish you would)

This use of negation -- often in conjunction with irrealis operators, includ¬

ing modals, subjunctives, conditionals and yes/no question markers -- is
wide-spread cross-linguistically, perhaps universal.41 I suspect its polite-

41 See Goody (ed., 1978), especially Brown and Levinson (1978). See also Salisbury (1986).
166 Mind, Code and context

ness, deference value derives from the overlap between negation and ir-
realis, along the psychological dimension of subjective certainty

4.6.4. Certainty, modesty and politeness

Questions of politeness and deference, with their complex culture-

specific -- and presumably some universal -- details, shade rather per¬
sistently into the epistemic range of our modal space. Thus, for ex¬
ample, Syder and Pawley (1974) note that some cultures seem to put a
certain prime on the so-called modesty principle, by which speakers as a
matter of course claim to know less than they do, especially when the
information may reflect favorably on their personal stature. It may well
be that this culture-specific tendency is another reflection of a more
universal principle, already noted above:

(64) Subjective certainty and higher authority:

"In communicating to an interlocutor of higher status, one

downgrades one's own subjective certainty".

Lewis (1979) makes a similar observation couched in terms of a puta¬

tive slave-master relationship.

4.6.5. Knowledge, certainty, responsibility and blame

In many cultures, perhaps in most, claiming direct personal respon¬

sibility for conveyed information may be a serious social error, to be
strictly avoided in any but the most intimate -- thus well protected -- so¬
cial contexts. Strong claims to direct authorship of transmitted informa¬
tion, with the attendant marking of high subjective certainty and strong
evidential support, are all to be avoided. In carrying out communication
under these cultural constraints, a variety of highly conventionalized
strategies are used, including indirection, disclaimer, oblique attribu¬
tion, impersonalization, couching R-assertions as yes/no questions,
negation and irrealis. While the structural devices may be the same as in
some of the cases cited above, the guiding principle is perhaps different.
That principle is roughly this:

(65) The hazardous information principle:

a. Knowledge is power, but power is responsibility.

b. Information may be coveted, it may also be hazardous
and socially destabilizing.
 Propositional Modalities 167

c. Transmitting new information may yield a clear social ad¬

vantage, but it also incurs some risks.
d. Therefore, being identified explicitly as the author of infor¬
mation may be unwise, must be avoided.

The operation of this principle is particularly conspicuous in small,

rural, geographically scattered communities, where residents of isolated
homesteads are adept at cajoling fresh gossip, preferably malicious, out
of the occasional visitor. In spite of their geographic scatter, such com¬
munities are often intimate social units, where one's business is
everybody's business, and where most mundane news disseminates
with lightning speed. The transmission of fresh gossip may indeed be
the real purpose of a visit. Yet the would be transmitter must tread light¬
ly, lest he be soon pointed at — often accusingly and by the very same
host who warmed the information out of him -- as the explicit author.

4.7. The hybrid nature of propositions: Between Tautology

and contradiction

"...The propositions of logic are tautologies.

Therefore the propositions of logic say nothing..."

L. Wittgenstein,
Tractatus Logico Philosophicus (1918, p. 121)

In discussing epistemic modalities earlier above, we proceeded as if

propositions are either asserted (R-asserted, IRR-asserted, NEG-as-
serted) or presupposed. In following this line, we have tacitly shared the
traditional assumptions of logical semantics. But in this respect too, the
logical-semantic tradition grossly misrepresents the facts of use of
natural language. In actual communication, all propositions tend to be
informational hybrids, carrying some presupposed ('old') and some as¬
serted ('new') information. In order to understand why this must be the
case, one must acknowledge two general properties of human dis¬
course:

(a) Every new proposition that is transacted in human discourse

presupposes an enormous amount of shared background in¬
formation, some generic-lexical, some situational-deictic,
some text-specific.
(b) Human discourse is overwhelmingly multi-propositional.
That is, it is seldom the case that the message spans only a
single proposition. Rather, it spans some multi-propositional
168 Mind, Code and context

sequence. In other words, it has multi-propositional

coherence.

Each proposition in discourse must thus be coherent with respect to all

three contexts -- the generic (culturally-shared), the deictic (situationally-
shared), and the textual (discourse-shared). Achieving such coherence
requires that not all the information in the proposition be new. Rather,
some of it must recapitulate -- or allqde.to -- shared background infor¬
mation, in order to cohere with it.
Coherence vis-a-vis a larger system is achieved by partial overlap of
information. The best -- albeit negative -- metaphor for this may be
found in Wittgenstein's (1918) discussion of the two logical extremes of
tautology and contradiction:42

"...A tautology has no truth conditions, since it is unconditionally

true; and a contradiction is true on no condition. Tautologies and con¬
tradictions lack sense..." (1918, p. 69)

A proposition is tautological if it repeats the information already given

in another proposition. A proposition is contradictory if it is logically in¬
compatible with the information given in another proposition. Both ex¬
tremes yield communicative disasters:

(a) Total informational redundancy = tautology = no interest

(b) Total informational incompatibility = contradiction = no
coherence

Human communication is necessarily a hybrid system, a compromise

between the two logical extremes.43
The most common data type supporting this concept of the hybrid
nature of propositions comes from the study of the pragmatic properties
of grammatical subjects. In most languages, the grammatical subject of a
clause must be definite — i.e. presupposed to be known to the hearer.
Even in languages — such as English - that tolerate indefinite subjects,
referential-indefinite subjects are rare in text, seldom above 5%-10%.44
In many languages, direct objects also tend to be definite, 'old'.

42 See further discussion in Chapter 7, section 7.2.2., below.

43 As Wittgenstein (1918) has pointed out, all the propositions of deductive logic can be
reduced to either tautology or contradiction. Deductive logic is thus, in principle, the wrong
means for incrementing new knowledge into a system.
44 See text-counts and discussion in Giv6n (1979a, ch. 2).
 Propositional Modalities 169

'topical' information.45 Text counts in English46 suggest about 50% defi¬

niteness for direct objects.
As we noted earlier above, many grammatical clause-types are largely
presupposed. Such background clauses are not very frequent in text, being
pragmatically marked. But their cumulative frequency in many text-
types, or styles, may easily exceed 50% (as against asserted foregrounded
clauses).
In a recent quantified study,47 the average number of new-informa¬
tion-bearing words (subjects, objects, verbs, adjectives, adverbs) per
clause was calculated for a written English text. The clauses ranged in
length between 3 and 10 lexical words, with the average around 5 words
per clause. The results may be couched in terms of the following prin¬
ciple:

(66) Ratio of incremented information in text:

(a) The one-chunk per clause ratio:

"The average number of new-information-bearing

words per clause is 1.4 words, i.e. approximating one
chunk of new information per proposition".48

(b) The 25% per clause ratio:

"On the average, about 25% of all lexical words in a

clause -- thus in the text at large -- are new information
words; the rest is old information".

These results suggest that human language utilizes a highly incremen¬

tal approach to information processing. The bulk of the information in
the clause ('proposition') is old, background, shared, presupposed. Only a
relatively small portion of the clause is new information.
The functional explanation for such an information processing system
is relatively transparent: The 75% old information packed into most
clauses is used to established the coherence -- and thus relevance -- of
each chunk of new information vis-a-vis the already stored information.
The old information chunk are the addressing mechanism. Such a
processing system is motivated, in turn, by the two main features of
human discourse already outlined above:

45 See Givon (1984d).

46 Givon (1979a, ch. 2).
47 Givon (1984a, ch. 7).
48 This principle was first suggested in Givon (1975c).
170 Mind, Code and context

(a) The multi-propositional nature of human discourse and its

coherence;
(b) The large, complex ever-present context that is required in
order to interpret chunks of new information.

The old information packed into propositions in natural communication

is needed to perform anchoring, coherence, relevance functions vis-a-
vis the multi-propositional prior text, and prior context.

4.8. The negotiation of modality

Even the most pragmatic approach to meaning and communication

often clings to one shibboleth of our venerable logical tradition: Speakers
know what is in their own mind; their lexical meanings are fully
specified; their propositions concerning states and events are explicit;
their modalities are deliberately chosen. We cling to these conventions
as to dear life, often in the face of much prima facie evidence to the con¬
trary. Lewis' (1979) charming anecdotal discussion of 'score-keeping in
a language game' is one account of how one negotiates truth, presup¬
position, definiteness, and incidentally also meaning. Unfortunately, the
implications of Lewis' account for our analytic approach have gone lar¬
gely ignored.
This vast, potentially explosive issue must ultimately be resolved
through the empirical study of actual communicative behavior. At this
point I would like to simply illustrate the problem with two textual ex¬
amples. Both are taken from the same work of fiction. The two par¬
ticipants in the negotiation are Mrs. Phillip J. King and Momma. In the
first passage, the shifting peaks of negotiated meaning are boldfaced;
the manipulated epistemic modality is given in italics:

"...Mrs. Phillip J. King said he had been dashing, but Momma would
not go along with dashing and said to her mind he had been not unat¬
tractive, but Mrs. Phillip J. King couldn't see fit to drop all the way from
dashing to not unattractive, so her and Momma negotiated a descrip¬
tion and arrived at reasonably good looking, which was mutually
agreeable though it seemed for a minute or two that Mrs. Phillip J. King
might hold out to have the reasonably struck from the official version.
But Momma went on to tell her how she thought his nose had a fanciful
bend to it which distracted Mrs. Phillip J. King away from the reasonab¬
ly because, as she told Momma back, she had always thought his nose had
a fanciful bend to it herself. Mrs. Phillip J. King called it a Roman nose
and she said there wasn't anything uppity or snotty about it but it was
purely a sign of nobility. And Momma said he certainly carried himself
like a Roman, which sparked Mrs. Phillip J. King to wonder if maybe he
 Propositional Modalities 171

had n't come from Romans, if maybe that wasn't why he was a
Republican. But Momma said she recalled he was a notable Democrat.
And Mrs. Phillip J. King said, "Maybe he was". And Momma said she
believed so. And Mrs. Phillip J. King said "Maybe he was" again...I was
not present when Mrs. Phillip J. King decided she couldn't let reasonab¬
ly good looking rest peacefully and resurrected the whole business
with the argument that a moustache under that fancifully bent nose
would have most certainly made for dashing. But Momma could not see
clear to allow for a moustache since there had not been one actually;
however, Mrs. Phillip J. King insisted that if Momma could just imagine a
finely manicured and dignified Douglas Fairbanks-style moustache
under that Roman nose then all of the rest of the features would surely
come together and pretty much scream Dashing at her. But even with a
moustache thrown in Momma could not sit still for any degree of dash¬
ing though Mrs. Phillip J. King campaigned rather fiercely for Consider¬
ably Dashing and then Somewhat Dashing and then A Touch Dash¬
ing, so Momma for her part felt obliged to retreat some from reasonably
good looking and her and Mrs. Phillip J. King settled on passably
handsome with Mrs. Phillip J. King supplying the handsome and
Momma of course supplying the passably..."

T.R. Pearson, Short History of a Small Place

(1985, pp. 191-192)

The second passage is noteworthy for the wide range of evidentials,

as well as other epistemic operators, used in the progressive upgrading
of 'truth':

"..."Pepsi Cola" she said. "Yes, I believe it was Pepsi Cola because I'm
near certain it was Mr. Womble who ran the Nehi outfit . And Momma
sat straight up and said, "Helen?"... But Mrs. Phillip J. King just went
straight on and said, "It had to be Pepsi Cola. He owned the bottling
plant you know in Burlington. I mean his daddy, now I don' think he
ever owned it himself, but his daddy did and made a killing putting out
Pepsi Cola until he sold the business and made another killing doing
that. Momma said it was just a ton of money that changed hands. She
was brought up in Burlington you know". "But Helen", said Momma...
"And they tell me his wife was just a gorgeous woman but not from
around here...Momma said he went out and got one all the way from
Delaware or Ohio, she couldn't ever remember exactly which, but I im¬
agine it was Delaware since P.J. tells me...that Delaware is one of your
urban states...and P.J. says there's plenty of money in Delaware mostly
on account of the Duponts, and she might have even been a Dupont
herself, anyway I don't know that she wasn't and she was probably
172 Mind, Code and context

from Delaware I imagine, which is where they all come from..." "Wasn't
it cookies instead of Pepsi-Cola?" Momma wanted to know. "Didn't Mr.
Alton's Daddy make those savannahs with white cream filling and
those little oval shortbread cakes that came in the blue sack?" And Mrs.
Phillip J. King got a little hot on account of the cream-filled savannahs
and the shortbread cakes and she said to Momma, "Now Inez, he might
have dabbled in cookies later but I can tell you for a fact it was Pepsi¬
Cola at the first because Momma said it was Mr. Womble at the Nehi
and Mr. Foster at the Coca-Cola and Mr. Todd W. Smith at the Sundrop
and Mr. Nance at the Pepsi-Cola, and Momma herself told me it was
Pepsi-Cola that made him his money but I don't ever recall a whisper of
cookies passing her lips..."..."

T.R. Pearson, A Short History of a Small Place

(1985, pp. 193-195)
CHAPTER 5

THE PRAGMATICS OF REFERENCE:

EXISTENCE, REFERENTIAL INTENT
AND THEMATIC IMPORT

5.1. Reference and existence*

5.1.1. The Real World vs. the Universe of Discourse

The treatment of reference in linguistics developed historically as the
by-product of a long-established logical tradition. Within that tradition,
one formulation (admittedly now somewhat of a straw-man; but see
Russell, 1905, 1919; Strawson, 1950 or Carnap, 1958) holds that refer¬
ence (or denotation) is a mapping relation between linguistic terms and
entities which exist in the Real World. Therefore, the truth of sentences
containing referring expressions depends, inter alia, on whether the
referring expressions inside them denote or do not denote in the real
world. To illustrate this approach, consider the sentences (1) and (2)
below:

(1) a. The king of France is bald

b. The queen of England is bald
c. I rode a unicorn yesterday
d. I rode a horse yesterday

(2) a. There is a king of France (and only one)

b. There is nothing which is both king of France and bald
c. There is a queen of England (and only one)
d. There is something which is both queen of England and
bald

According to Russell's approach to reference (see Strawson, 1950), in as¬

serting (la) one asserts two contradictory propositions: The false (2a), and

* I am indebted to Martin Tweedale, Frank Lichtenberk, Dwight Bolinger and Paul Otto for
many helpful comments. This chapter also benefited from early presentation at the Geor¬
getown Round Table on Linguistics (Wahington, DC,1984), S.I.L. Grammar Workshop
(Ukarumpa, Papua-New Guinea, 1985), the Linguistics Colloquium at Auckland University
(Auckland, 1986), and the Linguistics Colloquium, Australian National University (Canberra,
1986).
174 The Pragmatics of Reference

the true (2b). Further, the falsity of (2a) is due to failure of denotation. In as¬
serting (lb), on the other hand, no contradiction is present. One asserts
the true proposition (2c) -- true due to proper denotation — and the factually
false (2d). Similarly, (lc) is a false assertion because of failure of denota¬
tion; while (Id) may be true because proper denotation may be obtained.
It is nonetheless remarkable that human languages tend to code the
nominals in (la,b) and (lc,d), respectively, with exactly the same gram¬
matical devices, paying no heed to the fine distinction of real world
denotation and truth. Is the grammar of human language -- and the
mind behind it -- confused, misleading or capricious? Or is it perhaps
marching to the beat of a different drum? Reference in human language
seems to not be a mapping from linguistic terms to individuals existing
in The Real World, but rather a mapping from linguistic terms to in¬
dividuals established verbally -- for whatever purpose — in the
Universe of Discourse.
It is of course true that the Universe of Discourse and The Real World
display a considerable overlap in normal human discourse, which tends
to deal with extant human individuals and their everyday affairs. But
when the two worlds do not overlap, the grammar of reference cheerful¬
ly disregards Real-World denotation, abiding instead by denotation in
the universe of discourse.
The grammar of reference may also disregard denotation altogether,
as can be seen from (3) below:

(3) a. John is looking for a horse; it escaped last Friday

b. John is looking for a horse; it better be white

The horse in (3a) denotes a specific individual entity in the universe of

discourse. The one in (3b) does not. Still, the grammar of English applies
the anaphoric pronoun it to both the real horse in (3a) and the hypoth¬
etical horse in (3b).

5.1.2. Reference vs. referential intent

It is of course possible, for a logic-based description of reference, to
concede the point and make every universe of discourse — in fact, the
multiplicity of possible universes of discourse -- a separate realm of
denotation for linguistic terms. In a very clear sense. Possible Worlds
Semantics1 is a formal attempt to accomplished just that. As we shall
see later on, problems will persist even with this approach.

1 See e.g. Kripke (1963,1972), Cocchiarella (1965), Hinttika (1967), Purtill (1968), Scott (1970),
Montague (1970) or Lewis (1972), inter alia; also discussion in Chapter 1, above. Whether
the 'indexing' that proliferated under Possible Worlds Semantics accomplishes much more
than merely labeling some major areas of pragmatics that are impervious to a truth-condi¬
tional treatment, remains to be seen.
 Mind, Code and context 175

Reference in a Universe of Discourse is already a crypto pragmatic affair.

This is so because every universe of discourse is opened ('established') -
for whatever purpose - by a speaker. And that speaker then intends en¬
tities in that universe of discourse to either refer or not refer. And it seems
that in human language it is that referential intent of the speaker that
controls the grammar of reference. Thus, a horse in (3) above is, by itself, a
referentially opaque term. It could either refer to a horse in the universe
of discourse in (3a), or it could have no such reference (or else refer to the
type horse) as in (3b). And in the absence of The Real World as arbiter of
denotation, the speaker's referential intent seems to be all one may have
to go by.
While crypto-pragmatic in this one sense, a treatment of reference
based primarily on referential intent is still semantic in another sense: It
makes no reference to extra-propositional context, above and beyond
the speaker who utters the proposition. Thus, if the speaker can be
abstracted and thus ignored, as has been done matter of factly in logic
since Plato, reference would remain essentially limited to the scope of
the atomic proposition. In the next section we will discuss reference in
natural language from such a semantic perspective. We will note the
predictable relation between propositional modalities and referentiality
of nominals under their scope. We will also see why, given natural lan¬
guage facts, a purely semantic approach to reference can achieve only
partial characterization of the facts.

5.2. The semantics of indefinite reference2

5.2.1. Referential opacity

The term referential opacity is originally due to Quine (1953), in a con¬
text that actually does not concern us directly here. His original observa¬
tions concerned examples like (4a,b) below3 * *:

(4) a. Managua is
(i) Julia's home town
(ii) the capital of Nicaragua
b. Harry knows that Managua is
(i) Julia's home town
(ii) the capital of Nicaragua

2 The treatment of reference given below was first suggested in Givon (1973b). A somewhat
similar account may be found in Jackendoff (1971).
3 I am indebted to Martin Tweedale (in personal communication) for discussing with me the
history of the treatment of reference in philosophy, as well as the history of the term
'referential opacity7.
176 The Pragmatics of Reference

Assuming that (the speaker knows that) the nominal expressions (i)
and (ii) above refer to the same city (Managua), the substitution of (i)
and (ii) in (4a) does not change the truth value of the proposition. An
environment such as (4a) will thus be called referentially transparent.
In contrast, (4b) is a referentially opaque environment, because the sub¬
stitution of (i) and (ii) there seems to change the truth value of the
proposition. That is, Harry may know those two facts about Managua
without necessarily knowing that they refer to the very same city.
One may argue, something Quine did not choose to pursue, that what
is at the bottom of the referential opacity of (4b) is the interjection of
another mind -- thus another perspective. And only when the
knowledge of that other mind does not match the speaker's knowledge
does one obtain this type of referential opacity.
The term referential opacity was also extended by Quine to the en¬
vironments that concern us more directly here. As Aristotle noted in De
Sophisticis Elenchis (McKeon, ed., 1941), in some linguistic environments
one can obtain an ambiguity between the sensus divisus ('referential')
and sensus compositus ('generic', 'attributive', 'non-referential') of a
nominal expression.4 To illustrate this, consider the following expres¬
sions, the first one by Quine's definition referentially transparent, the
second referentially opaque:

(5) a. John married a rich woman.

b. John wanted to marry a rich woman,
i. ...but she refused him.
ii. ...but he couldn't find any.

The speaker uttering (5a) is committed to the referential existence -- in

the universe of discourse -- of some rich woman that John married. In
other words, the following implication must hold:

(6) If John married one, that particular one must have existed.

Expression (5a) is thus referentially transparent in the sense we will be

using here. Expression (5b), on the other hand, is in this sense referential¬
ly opaque. In uttering it, the speaker may or may not have made a
referential commitment. Two interpretations of a rich woman are possible:

4 Medieval philosophers persisted in using Aristotle's terminology. In the 20th century the
terms de re and de dicto were used to denote 'referential' and 'non-referential, respectively.
Donellan (1966) is responsible for recasting those, at least for some grammatical environ¬
ments, as 'referential' and 'attributive', respectively. Finally, linguists have also used the
terms 'specific' and 'generic', respectively.
 Mind, Code and context 177

(7) a. John — and thus the speaker -- has a particular woman in

mind; he wishes to marry her.
b. John -- and thus the speaker — has no particular woman in
mind; he wishes to marry someone of that type.

Interpretation (7a) is compatible with ending (5bi) above; and Inter¬

pretation (7b) is compatible with ending (5bii).
Since the indefinite noun is opaque in (5b) but transparent in (5a), this
type of referential opacity is not signalled in English by the presence (vs.
absence) of the indefinite article itself. Rather, it is somehow due to the
propositional modality under whose scope the indefinite expression
falls.

5.2.2. Reference and propositional modalities

In this section I will recapitulate the earlier discussion5 concerning
how the second type of referential opacity (i.e. as in (5b) above), more
specifically of indefinite nominals, is systematically predictable from
the propositional modality under whose scope the nominal falls. For
this purpose, the four major propositional modalities discussed in Ch. 4
above, may be grouped into two pairs:

(8) a. FACT: Presupposition

R-assertion
b. NON-FACT: IRR-assertion
NEG-assertion

To the non-fact category one must add two linguistic environments that
were not discussed in ch. 4, above:

(9) a. HABITUAL:
John meets a woman every Tuesday at the pub
b. NOMINAL PREDICATE:
John is a man
John is my friend

The common denominator of irrealis, negation, habitual and nominal

predication is fairly obvious: None of these modes refers to the occurrence
of a particular event at a particular point in time. This is, presumably, the
common feature of non-fact.
The rule for (this type) referential opacity is then given as:

5 This was discussed briefly in Ch. 4, section 4.5.5., above. See also Givon, (1972a, 1973b).
178 The Pragmatics of Reference

(10) a. "Under the scope of FACT modalities, nominals can

only be interpreted as referential".
b. "Under the scope of NON-FACT modalities, nominals
may be interpreted as either referential or non-referen-
tial".

The most common FACT and NON-FACT modalities in human lan¬

guage were listed in Ch. 4, above. Since human discourse, most par¬
ticularly everyday narrative or conversation, seems to employ NON¬
FACT modalities rather sparingly,6 it is possible to render (10) above
more realistically as a markedness expression:

(11) 'Nominals may be interpreted non-referentially only if they

are under a NON-FACT scope. Otherwise they must be in¬
terpreted referentially".

One short note must be added here concerning the markedness --

skewed — relation between the referring and non-referring sense of
nominals. Formal logicians 'handled' the description with two quan¬
tifiers, the existential to mark referring senses, and the universal to
mark non-referring ones. There are two things that are odd in this for¬
malization. First, it considers both the referential and non-referential
equally marked, whereas the distributional facts of natural language
strongly suggest that the non-referential is the marked category. Second,
in using the universal quantifier to mark non-referring nominals,
logicians lump together two distinct senses of 'not referring'. To show
this, consider examples (12a,b,c) below:

(12) a. John thinks a girl in his class is willing to marry him.

b. John thinks some girl in his class will eventually agree to
marry him.
c. John thinks all girls might be willing to marry him.

While 'girls' in neither (12b) nor (12c) refers to a specific individual,

their sense of not referring is rather distinct. In (12b), a possible single in¬
dividual is hypothesized. In (12c) an entire multiple group is referred
to, or perhaps the type. The universal quantifier may perhaps charac¬
terize (12c), but clearly not (12b).

6 We have already discussed in Ch. 4, above, why NEG-assertions are relatively rare in
normal discourse. The irrealis mode is also less common, except in specialized discourse
such as (among others) law codicils, exam questions, or academic discussions.
 Mind, Code and context 179

5.2.3. The grammatical marking of indefinite reference

As we have seen above, English indefinite nouns are potentially am¬
biguous as to whether they refer or do not refer to an entity in the
universe of discourse. The indefinite article simply marks the fact that
the term is being introduced into the discourse for the first time — and
without any speaker's presumption of hearer's familiarity. Such
presumption — or presupposition — is in fact the major requirement for
marking nominals as definite.7 The article system of English marks
rather sharply the contrast between definite and indefinite. In many lan¬
guages, on the other hand, the grammar distinguishes just as sharply
between referential and non-referential indefinites. And in predicting
referential opacity, such languages abide by the very same rules (10,11)
as English. To illustrate this, consider the following examples from
Hawaii Creole, where the numeral one marks referential-indefinite
nouns, while zero marks non-referential nouns:8

(13) R-ASSERTION:

a. i rid wan-buk
he read one-book
'He read a book' (=3 a specific book)

b. *i rid buk
he read book (*3 no specific book)

(14) PRESUPPOSITION:

a. i hapi i rid wan-buk

he happy he read one-book
'He's happy that he read a book' (3 a specific book)

b. *i hapi i rid buk

he happy he read book (*3 no specific book)

7 See extensive discussion in Ch. 6.

8 Following Bickerton (1975). Other languages with a similar contrast marked by one vs. zero
are: All Creoles, Turkish, Sherpa, Nepali, spoken Hebrew, Old English, Old German, Old
Spanish, Old French, Mandarin Chinese, and literally scores of others; see discussion in
Givon, (1978a; 1981b).
180 The Pragma tics of Reference

(15) IRR-ASSERTION:

a. i want rid wan-buk

he want read one-book
'He wanted to read a book' (:d a specific book)

b. i want rid buk

he want read book
'He wanted to read a book' (z> no specific book)

(16) NEG-ASSERTION:

a. i no-rid buk
he no-read book
'He didn't read a/any book' (=> no specific book)

b. i no-rid di buk
he no-read the book
'He didn't read the book' (z> a specific-DEF book)

c. *i no-rid wan-buk
he no-read one-book (*=> a specific -INDEF book)

Some languages mark in their grammar only the contrast of referen¬

tial vs. non-referential, with definiteness left unmarked. Such languages
still abide by rules (10,11) as English or Creoles. For example, in Bemba
(Bantu) the VCV- form of the noun prefix codes referential nouns
regardless of definiteness; while the CV- form codes non-referential
nouns. Thus consider:9

(17) R-ASSERTION:

a. a-a-somene ici-tabo
he-PAST-read REF-book
'He read a/the book' (z> a specific book)

b. ^a-a-somene ci-tabo
he-PAST-read NREF-book (*=> no specific book)

9 For details see Giv6n (1972a, 1973b). Generic subjects in Bemba are considered definite,
and thus marked as referential. This is another argument for the inadequacy of the
traditional lumping together of all 'non-referring' nominals under the universal quantifier.
Bemba, like many other languages, distinguishes clearly between a nominal that has
potential individual reference under the scope of NON-FACT, and a nominal that refers to the
type.
 Mind, Code and context i8l

(18) IRR-ASSERTION:

a. a-a-fwaayile uku-soma id-tabo

he-PAST-want INF-read REF-book
'He wanted to read the/a book' (3 a specific book)

b. a-a-fwaayile uku-soma cz-tabo

he-PAST-want INF-read NREF-book
'He wanted to read a book' (z> no specific book)

(19) NEG-ASSERTION:
a. ta-a-a-somene zcz-tabo
NEG-he-PAST-read REF-book
'He didn't read the book' (z> a specific-DEF one)
(*z> a specific-INDEF one)

b. ta-a-a-somene cz-tabo
NEG-he-PAST-read NREF-book
'He didn't read a/any book' (=> no specific book)

In sum, the correlation between propositional modality and referential

opacity seems -- or at least seemed at the time the original studies were
conducted -- a matter of straight-forward semantic description. Those
studies, however, were conducted on isolated, out-of-context sentences.

5.3. The pragmatics of reference:

Denotation vs.referential import

5.3.1. The numeral ’one’ in Hebrew

Spoken Hebrew is one of the languages to which the description of
Hawaii Creole, above, seemed to apply in all detail. That is, when
Hebrew sentences were studied in isolation, the numeral one seemed to
mark semantically referential indefinites, while zero marked non-
referential indefinites. However, when sentences were placed in dis¬
course context, some semantically-referential nouns were nonetheless
marked by zero. To illustrate this, consider:10

10 Givon (1978a, 1981b).

182 The Pragma tics of Reference

(20)
a.
...az nixnasti le-xanut sfarim ve-kaniti sefer- xad;
then entered-I to-store-of books and-bought-I book- one
'...so I went into a bookstore and bought a book;

ve-ratsti habayta ve-karati oto,

and-ran-I home and-read-I it
and I ran home and read it,

ve-ze beemet haya sefer metsuyan...

and-it truly was-it book excellent...
and it was indeed an excellent book../

b.
...az nixnasti le-xanut sfarim ve-kaniti sefer;
so entered-I to-store-of books and-bought-I book
...so I went into a bookstore and bought a book;

ve-ratsti habayta ve-axalti aruxat erev ve-halaxti li-shon...

and-ran-I home and-ate-I meal-of evening and-went-I to-sleep
and I ran home and ate supper and went to sleep../

In both (20a) and (20b) book is semantically referential. Nonetheless, the

subsequent discourse context makes it more natural to use the REF-in-
definite marker one in (20a), and not to use it in (20b). But why? The
answer is reasonably transparent: In (20a) one runs home and proceeds
to read the book and discuss it. The specific referential identity of the
book matters in the subsequent discourse. In (20b), on the other hand,
one does some book-buying, then goes about one's routine; the book is
never discussed again, its specific ('referential') identity does not matter
in the subsequent discourse.
In discussing (20a,b) above we relied, admittedly, on intuitive judge¬
ment. That intuition was nonetheless suggestive: The grammar of refer¬
ence in language seems all of a sudden to be sensitive not to the
semantics of reference, but rather to its pragmatics. That is, the grammar
marks nominals as 'referential' not merely because the speaker intended
them to exist in the universe of discourse, but rather because their
specific referential identity was important in that universe. In the follow¬
ing sections I will describe briefly three empirical studies that seem to
confirm this early intuition. The first two concern the grammatical con¬
trast between the indefinite one and zero. The third involves the contrast,
in spoken American English, between the indefinite use of this and a.
 Mind, Code and context 183

5.3.2 Methodological preliminaries: The empirical measurement

of the importance of referents in discourse

The importance of a nominal topic in discourse may be assessed em¬

pirically along three distinct lines:

(a) Intuitive judgement: One may solicit the subjective judge¬

ment of a sufficient number of naive judges, asking them to
rank the nominal topics in a discourse according to
centrality or importance. Such informal judgements tend to
be highly uniform.

(b) Text frequency: One may make the reasonable assumption

that in human discourse important participants are men¬
tioned more frequently. The importance of nominal topics
can be thus assessed indirectly, through their frequency in
text.

(c) Psychometric measures: It is possible, finally, to assess the

importance of nominals in discourse by a variety of direct or
less direct experimental psychometric methods. For ex¬
ample, one may measure the amount of attention paid to a
noun as it is processed, or the amount of mental effort ex¬
pended in processing it. One may also measure the recall of
nominal topics, in terms or speed or error, assuming -- with
considerable empirical justification 11 -- that thematically im¬
portant topics are recalled faster and more accurately than
thematically incidental ones.

5.3.3. The numeral ’one’ in Krio

Krio is an English-based Creole language spoken in Sierra Leone.
When studied in isolated sentences, the contrast between the numeral
one and zero in the marking of indefinite nouns in Krio abides, just as in
Hawaii Creole above, by the rules of semantic referentiality (10,11). In a
recent text-based quantified study, the distribution of one-marked and
zero-marked indefinite nominals in four Krio narrative texts was inves¬
tigated. The grammatical sub-categories investigated were:12

11 See Anderson, Garrod and Sanford (1983).

12 From Givon (1985a), where further discussion can be found. The tabulation below is based
on the raw data reported there.
184 The Pragma tics of Reference

(21) a. one-marked REF-indefinite:

...pas nomoh foh wan eykuru dog wey dey-fen yamyam...
pass no-more for one mangy dog that PROG-search food
'...except for a mangy dog that was searching for food...'

b. one-marked NON-REF-indefinite:
...i tan lek wan ol mami dey-redi foh dai...
it stood like one old woman PROG-ready for die
'...it stood like an old woman ready to die...'

c. zero-marked REF-indefinite:
...so di chif put-am na blok...
so the chief put-him in cell
'...so the chief put him in a cell...'

d. zero-marked NON-REF-indefinite:
...di moni wey denh go-gi-am foh grachuiti...
the money that they FUT-give-him for gratuity
'...the money that they were going to give him as
gratuity...'

The following properties were then determined for each one-marked

and zero-marked indefinite noun in the text:

(a) Semantic status:

(i) Referential
(ii) Non-referential
(b) Thematic importance:
(i) Major participant in the story
(ii) Minor participant that:
(A) is related to major one
(B) appears at crucial thematic juncture
(iii) Minor participant
(c) Total occurrence in the text

Thematic importance was judged in this case on purely intuitive

grounds. The results of these measures, in percentages and average
values, are tabulated in (22) below.
 Mind, Code and context 185

(22) Correlation between grammatical marking and semantic

status, thematic importance and text frequency of in¬
definite nouns in written Krio narrative

one-marked zero- marked total

N % N % N %
tokens
in text 20 42% 28 58% 48 100%
Semantic
status:
REF 18 90% 4 14%
N-REF 2 10% 24 86%
TOTAL 20 100% 28 100%
Thematic
status:
MAJOR 16 80% / /
MIN-THEM 4 20% 10 36%
MINOR / / 18 64%
TOTAL 20 100% 28 100%
Average
text
frequency: 15.1 occur. 1.17 occur.

The results can be summarized as follows:

(a) Correlation between grammar and semantic status:

Indefinites marked by one are 90% semantically referential.
Indefinite marked by zero are 86% semantically non-referen-
tial.

(b) Correlation between grammar and thematic importance:

Of the one-marked indefinites, 80% are major participants;
20% are minor participants with some local thematic import;
none are strictly minor. Of the zero-marked indefinites 64%
are minor participants; the remaining 36% are minor par¬
ticipants with some local thematic import; none are major
participants.

(c) Correlation between grammar and text-frequency:

One-marked indefinites appear in the text, on the average,
15.1 times. Zero-marked indefinites appear in the text, on the
average, 1.17 times.
186 The Pragmatics of Reference

(d) The text-frequency of semi-minor participants:

None of the members of this category occurred more than
twice (2) in the text, regardless of their grammatical mark¬
ing. Whatever thematic importance these nouns carry is thus
purely local.

In spite of the considerable overlap -- 80% to 90% -- between semantic

and pragmatic referentiality, the results suggest that the grammar of in¬
definiteness in Krio is sensitive primarily to the pragmatics of reference,
i.e. to referential importance. This is so because of the patterns of ex¬
ceptions:

(a) Semantically referential indefinites that were pragmatically

(thematically) unimportant, were marked by zero — much
like the bulk of the semantically non-referential indefinites.

(b) Semantically non-referential indefinites that were pragmati¬

cally (thematically) important, were marked by one -- much
like the bulk of the semantically referential indefinites.

Thus, in the 10%-20% of cases where the semantics and pragmatics of

reference conflicted, the grammar chose to abide by the pragmatics of
reference, completely disregarding the semantics.

5.3.4. The numeral one in Mandarin Chinese

Mandarin Chinese uses the contrast between one and zero exactly as
spoken Hebrew or Creole, to signal at the semantic level the difference
between referential and non-referential nouns. To illustrate this briefly,
consider the following discourse-less sentences:13

(23) a. R-assertion (FACT):

ta mai-le yz-ben shu (REF-INDEF; one-marked)
he buy-PERF one-CL book
'He bought a book'

b. IRR-assertion (NON-FACT):
(i) ta hui mai yz-ben shu (REF-INDEF; one-marked)
he MOD buy one-CL book
'He may buy a (specific) book"
(ii) ta hui mai shu-de (NON-REF; zero-marked)
he MOD buy book-NOM
'He may buy some book(s)'

13 Huang (1985).
 Mind, Code and context 187

c. NEG-assertion (NON-FACT):
ta mei mai shu (NON-REF; zero-marked)
he NEG buy book
'He didn't buy a/any book'

In such isolated sentences, then, the contrast between one and zero mark-
ing of indefinite nouns seems to be predictable on purely semantic
grounds.
In a recent quantified discourse study,14 one-marked and zero-marked
indefinite nouns in five narrative texts of written Mandarin were com¬
pared as to the following measurable properties:
(a) Semantic referentiality;
(b) Text frequency: Total no. of occurrences in the text;
(c) Persistence: Number of recurrences in the ten clauses di¬
rectly following the first occurrence in the discourse;
(d) Thematic importance: as judged by four native speakers of
Mandarin, provided that the judgement of at least 3 out of
the 4 agreed;
The results are tabulated in (24) below.
(24) Correlation between grammatical marking, semantic ref¬
erentiality, thematic importance and text frequency in
written Mandarin Chinese narrative
one-marked zero-marked
REF-INDEF NON-REF TOTAL
N % N % N % N %

total N in text: 40 100% 20 31% 45 69% 65 100%

semantic status:
ref-indef 39 97.5%
non-ref 1 2.5%
thematic status:
major 19 47.5% / / 2 5% 2 3.5%
minor-thematic 17 42.5% / / 1 2% 1 1.5%
minor 4 10.0% 20 100% 42 93% 62 95.0%
average total
# of occurrences 14.5 2.15 1.75 1.87
in the text:
average persistence 3.10 0.30 0.20 0.23
per 10 clauses after
first occurrence:

14 The results presented below were re-tabulated from the data in Huang (1985).
188 The Pragma tics of Reference

The results tabulated in (24) above may be summarized as follows:

(a) Semantic referentiality:

The overwhelming majority of the one-marked indefinites --
97.5% — are semantically referential. Only a single token was
an exception to this otherwise categorial distribution. In con¬
trast, the zero-marked indefinites contain a significant num¬
ber of referential indefinites -- 31%,

(b) Total occurrences in the text:

On the average, a one-marked indefinite recurred 14.5 times
in the text. In contrast, a zero-marked indefinite recurred on
the average only 2.15 times in the text. And the difference be¬
tween referential and non-referential zero-marked indefinites
is rather negligible (2.15 vs. 1.75, respectively). Grammatical
marking thus correlates well with pragmatic importance,
but not at all well with semantic status.

(c) Persistence in the next 10 clauses after first occurrence:

On the average, a one-marked indefinite appears 3.10 times
in the 10 clauses directly following its first introduction into
the text. In contrast, a zero-marked indefinite appears on the
average only 0.23 times in the 10 clauses directly following
its introduction. Again, the difference between semantically
referential and semantically non-referential zero-marked in¬
definites in negligible (0.3 vs. 0.2, respectively).

(d) Thematic importance:

Of the one-marked indefinites, 47.5% were judged to be major
participants in the narrative. Another 42.5% were judged
minor, but either of thematic importance locally, or related —
as kin, part-of-whole, etc. — to an important participant. Only
10% were judged to be strictly minor. In sharp contrast, of the
zero-marked indefinites, 95% were judged strictly minor. Of
those that were semantically referential, 100% were strictly
minor. And of the semantically non-referential, 93% were
strictly minor. In other words, semantic referentiality does
not at all correlate with zero-marking, nor with thematic im¬
portance. Thematic importance, on the other hand, correlates
rather well -- negatively -- with zero-marking.

It is clear, then, that the grammatical contrast between one and zero in in¬
troducing indefinite nouns for the first time into Mandarin discourse
codes the pragmatics of referential importance, rather than the seman¬
tics of denotation.
 Mind, Code and context 189

Once again, it is not an accident that the overwhelming majority of

one-marked - pragmatically important -- nouns are also semantically
referential. The Mandarin texts studied were folk stories about everyday
life of simple people. But this overlap between the semantics and the
pragmatics of reference should not obscure the fact that when the two
are in conflict, the grammar opts to code the pragmatics.

5.3.5. The unstressed indefinite this in spoken English

The use of the unstressed demonstrative this as an indefinite article in

spoken English has been noted informally for a long time.15 As an in¬
definite article, this contrasts with the unstressed that, which seems to
serve as a definite article. Again intuitively, it seems that the indefinite
this marks primarily semantically referential indefinites. This may be il¬
lustrated informally in:

(25) a. this under the scope of FACT:

I saw this girl yesterday, and...

(the speaker has a particular girl in mind, but does not

assume that the hearer can identify her; i.e. REF-in-
definite)

b. this under the scope of NON-FACT:

I was looking for this girl yesterday, and...

(same inference as in a. above)

c. a under the scope of NON-FACT:

I was looking fo a girl yesterday, and...

( the speaker may either have or not have a particular

girl in mind)

The English indefininite article a(n) (Old English 'one') thus marks
referentially opaque indefinite nominals, regardless of their semantic-

15 Dwight Bolinger (in personal communication) notes that the use of the unstressed this as
an indefinite article did not begin to impinge on his sharp normative ear until the early
1950s. The transcripts of spoken rural Western American English of a 60-year old male
studied in Givon (1983b) show no significant use of the indefinite this. Prince (1981) comes,
on the basis of informal data, to the same conclusion reported here, as does Wald (1983).
190 The Pragma tics of Reference

referential indefinites. It thus seems to penetrate the grammar at the

very same point where one did in Old English, as a marker of referen-
tiality. But is it semantic or pragmatic referentiality?
To illustrate how this indeed marks the pragmatics of reference in col¬
loquial English discourse, consider the following letter that appeared a
few years ago in the Dear Abby column:16

(26) "Dear Abby: There's this guy I've been going with for near
three years. Well, the problem is that he hits me. He started
last year. He has done it only four or five times, but each
time it was worse than before. Every time he hit me it was
because he thought I was flirting (I wasn't).
Last time he accused me of coming on to a friend of his.
First he called me a lot of dirty names, then he punched my
face so bad it left me with a black eye and black-and-blue
bruises over half of my face. It was very noticeable, so I told
my folks that the car I was riding in stopped suddenly and
my face hit the windshield.
Abby, he's 19 and I'm 17, and already I feel like an old mar¬
ried lady who lets her husband push her around. I haven't
spoken to him since this happened. He keeps bugging me to
give him one more chance. I think I've given him enough
chances. Should I keep avoiding him or what?
Black and Blue".

The following features in the use of the unstressed this vs. a(n) in (26) are
striking:

(a) The REF-indefinite participant introduced by this recurs

throughout the text and is obviously the other important
participant (together with the narrator);
(b) The REF-indefinite participant introduced by a(n) never
recurs; his specific identity is obviously incidental to the
story;
(c) The only other indefinite introduced by a(n) is a NON-REF
attributive noun.

In an effort to obtain firmer empirical validation of the anecdotal

evidence, a recent study17 analyzed the recorded speech of six 8-10 year
old native American English speakers, chatting among themselves in-

16 This advice column is one of the few places where one can find this recent innovation of
English grammar in writing, tho only occasionally. Only young people seem to use it.
17 Shroyer (1985). See also Wright and Givon (1987).
 Mind, Code and context 191

formally. Four narratives in all were collected and analyzed, and four
grammatical categories were studied:

(27) a. this-SUBJECT: '...there's these two guys...’

b. f/zzs-NON-SUBJECT: '...he saw this great bear...'
c. «(n)-SUBJECT: '...and there was a fly to third base...'
d. fl(n)-NON-SUBJECT: '...he saw this monkey holding a
candy bar...'

The topic persistence (TP) of referents -- i.e. the number of times a

referent was mentioned in the 10 clauses immediately following its in¬
troduction into the discourse — was measured for the four grammatical
categories (27). The numerical distribution of the four categories in the
combined texts, and their average TP measure are tabulated in (28)
below.

(28) Text distribution of grammatical categories marking in¬

definite nouns, and average Topic Persistence (TP) of
tokens within each category measures

Subject Non-Subject Total

N % TP N % TP N % TP
'this' 28 65% 6.95 15 35% 2.40 43 100% 5.32
'a(n)' 13 12% 1.5 94 88% 0.56 107 00% 0.68

As can be seen, f/zzs-marked indefinites typically:

(a) tend to appear in the subject position;
(b) tend to have a much greater TP value; they are thus,
presumably, more important in the discourse.

In contrast, a(n)~marked indefinites typically:

(a) appear overwhelmingly in the object position;
(b) tend to have a much smaller TP value; they are thus,
presumably, less important in the discourse.

The total text-distribution of high-persistence tokens (TP > 2) and low-per¬

sistence tokens (TP = 0-2) in the four grammatical categories was also
analyzed.18 The results are summarized in (29) on the following page.

18 In plotting the distribution of the various persistence levels for the total population of
indefinite nouns in the text, Shroyer (1985) observed a very sharp break in the distribution
curve, occurring between the TP values of 2 and 3. The cutting point between 'important'
and 'unimportant' indefinites was decided on this heuristic basis.
192 The Pragma tics of Reference

(29) Distribution of the high-persistence (TP > 2) and low-per¬

sistence (TP = 0-2) tokens within the four grammatical
categories

THIS A
Subject Non-Subj Total Subject Non-Subj Total
TP N % N % N % N % N % N %
OT T 14.2 10 66.6 14 32.5 10 76.9 89 94.6 99 92.5
>2 24 85.8 5 33.4 29 67.5 3 23.1 5 5.4 8 7.5

The distribution figures in (29) may be summarized as follows:

(a) f/n‘s-SUBJ-marked indefinites are prototypically -- 85.8% --

high-persistence topics;
(b) a(n)-NON-SUBJ-marked indefinites are prototypically --
94.6% -- low-persistence topics;
(c) this-marked indefinites are in general -- 67.5% — high-per¬
sistence topics;
(d) a(n)-marked indefinites are overwhelmingly -- 92.5% — low-
persistence topics.

The only substantial exception to an otherwise rather striking correla¬

tion between grammatical marking and pragmatic referentiality, are the
32.5% of the this-marked category that exhibit low-persistence. The ob¬
vious, though tentative, explanation is that thematic importance and text
frequency, while correlating well over the long run, never correlate 100%.
Grammatical marking of indefinites by this is ultimately sensitive to the
psychological dimension of importance, not to mere text frequency.
In innovating, fairly recently, the grammatical distinction between
referential and non-referential indefinites, English has chosen — just like
it chose 900 years ago with the numeral one, and much like Mandarin
and Krio have -- to code the pragmatic feature of thematic importance,
rather than the semantic distinction of denotation in the universe of dis¬
course.

5.4. Normative action and referential interpretation:

The context-sensitivity of referential intent

While noting, as in rules (10), (11), that under the scope of non-fact
modalities both referential and non-referential interpretation of in¬
definite nominals (in English) is possible, we haven't quite told the en¬
tire story of the semantics of reference. A distinctly pragmatic flavor is
added even to so-called semantic referential intent when one considers
 Mind, Code and context 193

the question of normativity. Consider first the following contrast be¬

tween two verb-object combinations:

(30) a. He's going to take a woman out to dinner, then...

b. He's going to take a bus back home, then...

The propositional modality - irrealis ~ is the same in (30a) and (30b).

Nevertheless, a woman in (30a) is most likely to be interpreted as referen¬
tial, while a bus in (30b) is most likely to be interpreted as non-referen-
tial. This is so because as a norm a man is more likely to plan an evening
out with a specific woman; while going home we takes whatever bus that
may come along, as long as the destination is right.
Consider next the contrast between two friends of mine, one a learned
philosopher who spends much of his life reading in his chosen subject,
the other a rancher who reads himself to sleep every night with a paper¬
back Western. Suppose each said to me, before turning in for the night:

(31) I'm going to retire now and read a book.

The probability is rather high that I would assign a referential inter¬

pretation to a book in (31) when uttered by the philosopher, and a non-
referential one to a book when uttered by the rancher. I know their per¬
sonal habits, the reason why each reads book, and the role that a book's
specific identity is likely to play in the reading habits of either. I also
know that if I asked my rancher friend what book he read last night,
he'd be hard pressed to remember the title.
In a language like Krio or Bemba (see section 5.2.3., above), the
speaker uttering (30a,b) and (31) above would surely use the ap¬
propriate grammatical marking to cue us into their referential intent.
But even in an under-coded language such as standard English, our
knowledge of normativity guides us in assigning referring on non-refer¬
ring interpretation to nominals in such potentially opaque contexts.

5.5. The non-discreteness of reference and definiteness

5.5.1. Preamble
In the preceding sections we surveyed a range of facts (and argu¬
ments) all pointing out to the conclusion that in human language the
grammatical markers that code referentiality do not abide by the seman¬
tics of denotation, but rather by the pragmatics of referential impor¬
tance. Importance -- or relevance -- is indeed a central notion in prag¬
matics. There is no way for deciding relevance or importance on purely
194 The Pragma tics of Reference

deductive grounds.19 It is simply up to the speaker's context-mediated

judgement.
Another major facet of pragmatics that manifests itself in the gram¬
mar of reference and indefiniteness is non-discreteness. Traditional
deductive logic shudders at the specter of non-discreteness, and for ob¬
vious reasons: Clear-cut, truth-conditional decisions cannot be made
when categories such as denotation or definiteness turn mushy. None¬
theless, the study of human language suggests that such mushiness per¬
sists, at least at the margins of seemingly discrete coding systems.

5.5.2. Non-discreteness of reference: Vague speaker’s intent

Consider the -- presumably referentially opaque — use of indefinite
nominals in the following yes/no questions:20

(32) a. Did you see anything there?

b. Did you see anybody there?
c. Did you see any man there?
d. Did you see some man there?
e. Did you see a man there?
f. Did you see a tall man there?
g. Did you see a man wearing a blue shirt there?
h. Did you see a man there wearing a blue shirt and sitting on
a red barrel and twirling a silver baton in his left hand?

There is a gradation from (32a) through (32h), one that seems to

proceed along a speaker-oriented psychological dimension:

(a) "How strongly the speaker intends to suggest that he/she is

referring to a particular individual"

The gradation may be also described, alternatively, as proceeding along

a probabilistic dimension, concerning what the hearer is entitled to
infer:

(b) "What the probability is that the speaker referred to a specif¬

ic individual"

Be the ultimate dimension as it may, it is clearly non-discrete. Further,

the grammar of English seems to code this progression through the use
of three major devices which, when combined, yield finer gradation:

19 In this sense, both 'importance' and 'relevance' are akin to the pragmatic notion of
similarity, see discussion in Chapters 2 and 3.
20 This gradation was described earlier in Giv6n (1982a).
 Mind, Code and context 195

(33) Coding of the scale of referential strength

(a) The scale of indefinite articles:

any 3 some 3 a
(b) The scale of restrictive modification:
less 3 more
(c) The scale of noun specification:
thing 3 person 3 specific noun

Segment (32a,b,c) illustrates coding device (33c). Segment (32c,d,e) illus¬

trates coding device (33a). While segment (32e,f,g,h) illustrates coding
device (33b). It is clear, further, that devices (33b) and (33c) represent a
single scale of increased semantic specification. And the higher an in¬
definite expression appears on any of the coding scales (33a,b,c), the
more it is likely that the speaker meant it to be non-referential.
Consider next the case of WH-questions in Ute, a Uto-Aztecan lan¬
guage spoken in Utah and Colorado. Most commonly, interrogative
pronouns in Ute distinguish between fype-identity (non-referential) and
token-identity (referential) questions. Thus consider:21

(34) a. Token-identity question:

'a-'ara 'ina ta'waci?
WH-be this man
' Who is this man?'
(= please identify him)

b. Type-identity question:
'ini-'ara 'ina ta'waci?
WH-be this man
' What kind (of a person) is this man?'
(=please describe him)

This neat division is muddied up, however, by a further distinction

within the type-identity question category, that of degree of type de¬
scription. Thus consider:

(35) a. 'ini-'ara 'ina?

WH-be this
'What kind (of a person) is this?'

b. 'ini-kwa 'ara-'ay 'ina?

WH-MOD be-IMM this
'What kind (of a creature) could this possibly be?'

21 For full detail see Giv6n (1980b, ch. 12).

19 6 The Pragma tics of Reference

The difference between (35a) and (35b) has to do with the degree of the
speaker's uncertainty concerning the type-description of the entity
under question. In (35a) the speaker has some hunch but requests fur¬
ther information. In (35b) the speaker is completely baffled.

5.5.3. Gradation of referential intent

* . *

Earlier above we noted that propositional modalities have a sys¬

tematic effect on the referential status of nouns under their scope. Con¬
sider now the effect of various propositional modalities -- all but the last
one irrealis -- on the interpretation of indefinite objects in English:

(36) a. John wants to buy a house, but can't find any

b. John usually buys a house in June; it's cheaper then
c. John usually buys a house in June; it is cheaper then;
d. John usually buys a house in June, then sells it in July.
e. John would buy a house in June, then sell it in July.
f. John bought a house in June, then sold it in July

Semantically, the probability that a house in (36a) refers to a specific

house is minimal, in (36f) maximal. And (36b,c,d,e) seem to represent in¬
termediate degrees of specific reference. But such gradation cannot be
captured by deductive, truth-conditional means.
Reconsider now the contrast between a and some:

(37) a. John wants to buy a house

b. John wants to buy some house
c. John wants to buy some house near his Mom's home

The probability of specific referential intent by the speaker is highest in

(37c), lowest in (37a) and intermediate in (37b).
Consider next the effect of pluralization on the reference of indefinite
objects:

(38) a. John wants to buy a house

b. John wants to buy some houses
c. John wants to buy houses

As noted earlier above, the interpretation of (38a) may be either referen¬

tial or non-referential. But somehow the probability of a referential in¬
terpretation is highest in (38a), lower in (38b) and lowest in (38c).
The effect of pluralization can be further shown in the following con¬
trast:
 Mind, Code and context 197

(39) a. *John buys a house for a living

b. John buys houses for a living

The singular a house is less compatible with an extreme non-referential

interpretation; whereas the use of the plural houses is natural with such
a non-referential interpretation
Consider, finally, the interaction between word-order and pluralization:

(40) a. John buys houses in June; he sells them in July

b. John buys houses in June; in July he sells them.

Somehow, (40a) imparts a stronger probability that the very same houses
are involved, and thus a stronger referential intent by the speaker. On
the other hand, (40b) imparts a weaker probability that the same houses
are involved, thus seemingly a weaker referential intent. Referential in¬
tent, it seems, is not always a matter of either or. Like other psychologi¬
cal dimensions manipulated by the speaker, it is on occasion subject to
finer gradations.

5.5.4. Scales of definiteness: Degree of specification

Somewhat similar scales may be observed in the grammar of definite

description. In Spanish, for example, one finds a gradation in the degree
of genericity of nouns; or, conversely, in the degree of specificity of
definite description. To illustrate this, consider:

(41) a. Maria siempre habla con brujos

'Maria always speaks with sorcerers' (NON-REF)
b. Maria siempre habla con los brujos
(i) 'Maria always speaks with
(the) sorcerers' ('SEMI-DEF')
(ii) 'Maria always speaks with
the sorcerers (DEF)

In (41a) sorcerers is least referential. But (41b) has two distinct interpreta¬
tions: Either fully definite, as in (41bii), where the speaker indeed refers
to specific sorcerers identifiable to the hearer. Or the semi-definite in
(41bi), where the group that may fit the description is smaller than in
(41a) but larger than in (41bii).
A similar scalarity of definite description may be shown in English.
Thus consider:
198 The Pragmatics of Reference

(42) The person who killed Smith is insane

a. I know exactly who killed Smith; that person is insane.
b. Someone killed Smith, I don't know who, all I know is
that the killer is insane

As Donellan (1966) has noted, interpretation (42b) is attributive, thus in

some funny sense less-referential -- in spite of the speaker's obvious
commitment to believe that a specific individual indeed killed Smith.
But interpretation (42b) can be slowly tipped toward definiteness, as in
(42a), by gradually loading more and more descriptive attributes on the
semi-definite noun:

(43) a. The person who killed Smith is insane

b. The woman who killed Smith is insane
c. The young woman who killed Smith is insane
d. The beautiful young woman who killed Smith is insane

With the addition of each further description, the probability of the

speaker intending the definite noun attributively -- rather than referen-
tially -- decreases. The device used to achieve this gradation — further
semantic specification -- is the same seen in (32), (33) above.

5.6. Coreference, topicality and evocation:

The pragmatic grey margins of Russell’s Type Theory

5.6.1. Preamble

In the final section of this chapter we will survey what at first ap¬
peared to be merely one more pragmatic facet of reference. Namely, that
in human discourse the so-called relation of coreference between lin¬
guistic expressions often cannot be defined in strict terms of sharing the
same referent. As it turns out, the pragmatics of coreference in natural
language impels us to re-invoke to the discussion of categorization (see
Chapter 2). And that, in turn, leads in short order to the subject of a
hierarchic theory of types. We will open with a brief exposition of how
pragmatics invades the realm of coreference.

5.6.2. The pragmatics of coreference

We have dealt earlier in this chapter with five parameters of the prag¬
matics of reference. First, we saw how reference was a mapping from
language onto a universe of discourse. Second, we saw how it must in-
 Mind, Code and context 199

volve the speaker's referential intent. Third, we saw how it must in¬
volve thematic importance in the discourse context. Fourth, we saw
how it often involves the pragmatic notion of norm, thus perforce also
its underlying base, frequency. And fifth, we saw how it often involves
gradation. In this section we will deal with another facet of the prag¬
matics of reference, one that manifests itself most succinctly in linguistic
expressions of coreference. A more specific treatment of the pragmatics
of coreference -- particularly anaphoric reference -- can be found in
Chapter 6.
The term coreference is traditionally reserved for instances where the
same entity is referred to, following its first introduction (or subsequent
re-introduction) into the discourse. Coreference in natural language is
thus closely related to anaphora and definiteness. In the clearest — and
presumably most common — cases, a definite description is used after a
prior mention of the entity establishes it as referring in the universe of
discourse. If, on the other hand, the prior mention is non-referential, an
indefinite pronoun may be used, either referentially or attributively.
Thus consider:

(44) John decided to marry a rich girl;

(a) ...but she rejected him. (DEF-anaphoric)
(b) ...and he finally found one. (INDEF-referential)
(c) ...he kept looking for one. (INDEF-attributive)

One might of course insist that of the three examples in (44), only the
anaphoric 'she' in (44a) is strictly preferential. The indefinite 'one' in
(44b) is merely referential. While 'one' in (44c) is non-referential. Logically,
it indeed seems, the pronouns in (44a,b,c) bear different relations to
their antecedent. In (44a) 'she' refers to the same token. In (44b) 'one'
refers to a token of the same type. While in (44c) 'one' refers only to the
same type as the antecedent, without referring to any particular token.
Yet, in some sense, however vague, a rich girl in (44) is the proper an¬
tecedent for all three expressions that follow, it evokes all three equally
well. Yet, can such 'evocation' be specified logically?
One way of handling this type of data while still remaining within the
bounds of some logic, is by means of a Russellian Theory of Types.22
One could, for example, note that the common denominator to all three
types of co-reference in (44) is the relation of type identity. One could
then note the following, eminently sensible, one-way implication:

(45) token identity z> type identity

One could then posit the equally sensible grammatical rule:

22 Whether this Theory of Types is dose to Russell's (1919) remains to be seen.

200 The Pragmatics of Reference

(46) "An anaphoric pronoun will be used only when the antece¬
dent is token-identical. When the antecedent is only type-
identical, an indefinite pronoun will be used".

A cursory look at some more data quickly reveals that the situation is
a bit more complicated: Type-token relations in coreference seem to
work both ways. A non-referential ('type') antecedent may evoke either
type or token coreference; and much the same way, a referential
('token') antecedent may evoke either type or token coreference. Thus
consider:

(47) a. John was looking for a white horse;

b. ...I was looking for one too.
c. ...he didn't like the one he had.

d. John was looking for his white horse;

e. ...I didn't have any horse, so I rode a mule.
f. ...and I was looking for mine.

Rules (45, 46), so it appears, must be modified to allow for fully sym¬
metrical co-reference: From type to token as in (47c); from token to type
as in (47e); from type to type as in (47b); and finally from one token to
another token of the same type, and thus in some sense again from type
to type, as in (47f).
In the same vein, consider (48) below:

(48) John wanted to marry a rich girl,

a. ...but she had to be pretty.
b. ...but the girl he finally married wasn't.

The pronoun she in (48a) is non-referential. And the definite noun in

(48b), while denoting a referential individual, corefers only to the type
of the antecedent girl, especially since the antecedent was itself non-
referential. Still, the non-referring expression a rich girl in (48) is, in some
sense, the proper antecedent to both she in (48a) and the girl in (43b). Our
grammar rule (46) is not much help in accounting for (48a,b).
We might, in a desperate attempt to save the spirit if not the actual let¬
ter of rule (46), construct a more complex rule that will still purport to
abide by some Russellian Theory of Types. But our problems becomes
further compounded when we turn to considers the next set of ex¬
amples:
 Mind, Code and context 201

(49) Mary was looking for a house to rent,

(a) ...but the neighborhood was too run down.
(b) ...but the rents were getting too high.
(c) ...but the landlords were all greedy.
(d) ...but all the yards were too small.
(e) ... her roommate was getting married.
(f) ... her rose garden was overflowing.
(g) ... her boa constrictor was getting too big.

All the anaphoric definite expressions in (49a-g) are referential. They are
all evoked, rather transparently, by the non-referential antecedent a
house. In all cases, one may say that there remains an elliptical thread of
coreference between the antecedent and the subsequent anaphoric ex¬
pression, one that can be reconstructed back to the stereotype --
prototype — notion of rental housing.23 But these threads — or inferences
from the prototype -- are, as Wittgenstein would have observed, on a
descending scale of centrality vis-a-vis the prototype of rental housing,
from (49a) through (49g). Toward the bottom, the inferences seem to
stray rather far from the prototype's core. And what is more, one finds
no logical grounds for deciding where the thread of inferences becomes
remote, feeble, too far fetched. In principle, it seems, any token entity in
the universe, or its type, can be found relevant to — and can thus evoke
-- any other token entity, or its type. The only question is -- how
relevant, how obvious, how prototypical is the evocation to be.
The case is even more vexing with co-reference relations that evoke
the type. Thus, consider examples (50a-j), below:

(50) Mary was looking for a house to rent;

(a) ...but no landlord would accept her references.
(b) ...but she had no money to pay the rent.
(c) ...but there was no bus to the University.
(d) ...she is used to being comfortable.
(e) ...she is a big girl now, so...
(f) ...being a single woman is no picnic, so...
(g) ...she's been on welfare for months, but...
(h) ... cream-cheese has never been her favorite dish.
(i) ... enchiladas rancheras have always been my favorite
dish.
(j) ... enchiladas rancheras have always been Joe's favorite
dish.

23 'Houses are in neighborhoods', 'one pays rent for a rental house'/rental houses have
landlords', 'a house may have a yard', 'one may share a house with a roommate', 'one may
have a rose-garden in the house's yard', 'one may keep a boa constrictor as a house-pet,
respectively.
202 The Pragma tics of Reference

In principle, it seems, any type can evoke any other type, provided
some relevance relations can be established, by some inference, how¬
ever indirect and remote from the prototype. It is all a matter of degree.

5.6.3. Coreference and a non-Platonic theory of types

» • »

The moral of all this for a logical theory of reference is reasonably

clear: The general case — or common denominator -- in all coreference
relations is relevance; coreference is ultimately an open-ended, context-
mediated relation of evocation. A sub-set of evocation relations may
also abide by a stricter condition, that of type coreference with the an¬
tecedent (i.e., either type-token, token-type or type-type co-reference).
Finally, a smaller sub-set of type coreference may also abide by a stricter
condition yet, that of token coreference.
Finally, note that type coreference may be considered a mere sub¬
species of token coreference in a Russellian Theory of Types, with the
types taken to be merely tokens of the higher a higher meta-type.24 This
pattern of concentric inclusion of evocation sub-types can be repre¬
sented as a one-way implicational hierarchy:

(51) a. token coreference 3 type coreference

b. type coreference 3 pragmatic evocation

Pragmatic evocation is thus the general case. Type evocation is a

restricted sub-case, requiring a logical, Russellian Theory of Type. And
token evocation, in the traditional logical sense of 'having the same
denotation', is the most restricted special sub-case of the latter.
The semantic tradition of reference and coreference can handle (51a,)
above. But it seems, somewhat surprisingly, that what keeps the door of
denotational logic forever slightly open to pragmatics, is the fact that
the human mind's concept of type does not necessarily abide by
Russell's (1919) platonic strictures. Rather, the mind seems to insist on
the more flexible -- norm-, stereotype- or prototype-based -- system of
categorization, where membership in a category is mediated through a
cluster of features; where features may be more or less relevant to the
prototype; and where tokens may exhibit graded similarities.

24 See discussion of the arbitrariness of taxonomy in Chapter 1.

 Mind, Code and context 203

In order to account for the evocative -- pragmatic -- reference per¬

vasive in human language, a more flexible, Wittgensteinean Theory of
Types must be adopted. Such a theory would not only recognize central
(highly necessary, mostly sufficient) features that characterize the type in
the main. It must also take account of the more peripheral, inferred, in¬
creasingly context-bound features. Centrality to the definition of a
prototype is a matter of degree, it is also context-mediated. Therefore,
one cannot draw a sharp line to divide type-reference from the prag¬
matics evocation. Within a non-Platonic theory of categories, i.e. of
types, these two are one and the same.
>

...

X
CHAPTER 6

THE PRAGMATICS OF
ANAPHORIC REFERENCE:
DEFINITENESS AND TOPICALITY

6.1. Introduction: Definiteness, presupposition and knowing

other minds *
That definite description is inherently a pragmatic notion, one which
depends on context-derived assumption the speaker makes about what
the hearer is likely to know, was not by any means always taken for
granted in the logical-philosophical tradition. Thus Russell (1905, 1919)
treats definite description as somehow a semantic matter of unique
identification. And it is far from clear how such a notion differs from
mere semantic reference. When other philosophers departed from such
a tradition, as in Strawson (1950),1 they resorted to the notion of presup¬
position, i.e. the celebrated three-valued logic. In such a logic, a
proposition could be asserted as either True or False; or else it can be
Presupposed, meaning that it is taken for granted to be true, as precondi¬
tion for other expressions to be either true or false (see Keenan, 1969,
1971). Within such a framework, the contrast between indefinite and
definite reference is still ascribed to speaker-related features such as
unique reference, more specifically to whether that unique reference is
asserted (for indefinites) or presupposed (for definites). Thus consider:

*1 am indebted to Tim Shopen, Morti Gemsbacher, Frank Lichtenberk, Hartmut Haberland

and Paul Otto for many helpful comments on an earlier version of this chapter.

1 See also discussion in Hertzberger (1971).

206 Mind Code and Context

(1) a. Indefinite reference:

A man died
Asserted:
(i) 'There exists a man'
(ii) 'That man died'
b. Definite reference:
The man died
Presupposed:
(i) 'There exists a man'
Asserted:
(ii) 'That man died'

Treating definite description as a semantic presupposition, as in (lb)

above, represents an obvious advance over the two-value logical treat¬
ment. It is, for one thing, already crypto-pragmatic. Nonetheless, it
remains an attempt to handle the pragmatics of definiteness entirely
within the confines of a speaker oriented logic. As such, this approach
resists the rather obvious pragmatic feature of definiteness (and of
presupposition in general), namely that it involves assumptions the
speaker makes about what the hearer knows, believes in, is familiar
with or can identify. In other words, definite description is inherently
about knowledge by one mind of the knowledge of another mind.
And thus also, as it turns out, about the grounds for one mind making
assumptions about what another mind may know, also about how
securely they may know it. And this feature is anathema to any well-
constrained logic. It is indeed strictly disallowed, at least implicitly, in
Russell's (1919) Theory of Types.

6.2. Sources of definiteness:

Grounds for knowing other minds
Linguists have resigned themselves for a long time to the fact that
definiteness involves the speaker's assumptions about the hearer's
beliefs. One reason -- among others -- why it is exceedingly hard to ac¬
count for such an obvious fact within a logical system is that the
grounds on which the speaker makes assumption about what the hearer
knows, believes, is familiar with, etc. may be rather diverse. These
grounds can be described in terms of sources of shared knowledge in
human discourse; or, alternatively, in terms of the contextual files (as
 The Pragmatics of Amphoric Reference 207

was noted in our earlier discussions of grounds for presupposition).2

These grounds may be divided into three major sources, a division that
follows broadly our earlier discussion of the major categories of con¬
text:3

(1) The Generic File: Knowledge held in common by all mem¬

bers of the language-culture group, including knowledge of
the world, culture, and thus shared lexicon;
(ii) The Deictic File: Knowledge shared by the speaker and
hearer due to being present together at the speech situation;
(iii) The Text File: Knowledge shared by the speaker and hearer
because of what was said earlier in the discourse, or in prior
text

To illustrate these different source-files of definiteness, consider:

(2) Sources of definiteness

example file-source
a. '...I got up this morning and the sun was generic
shining...'
b. '...So I turned the TV on and the President..." generic
c. '...I sat on the chair, but the seat was broken...' generic plus
text
d. '...his father was an honest man...' generic plus
text
e. '...Take this chair and put it there...' deictic
f. '...We went to the bank and the cashier generic plus
cashed my check...' text
g. '...He saw a man and a woman. The woman text
was tall...'

As can be seen from these rather rudimentary examples, some

definiteness arises from the interaction of two source-files. Thus in (2c)
above, the seat is definite because (a) the chair whose part it is has been
mentioned in the previous discourse, and (b) it is general knowledge
that chairs have seats. Similar interactions hold for (2d) and (2f).
The three sources of definiteness illustrated above are only the major
ones, in terms of prevalence in human discourse. But as noted earlier
above,4 the source of speaker's knowledge about the hearer's mind is

2 See Ch. 4, above. For various formulations of 'files' see Chafe (1976,1987), DuBois (1980)
or Givon (1985b). In the context of an experimental study of discourse processing, seeGivon
etal (1985).
3 See Ch. 3, section 3.2.2.; Ch. 4, section 4.3.3.
4 Ch. 3, section 3.2.2.; Ch. 4, section 4.3.3.; Ch. 5, section 5.5.
208 Mind Code and Context

essentially unconstrained. That is, any direct or circuitous inference, in¬

tuition or inspiration, be it deductive, inductive or abductive, may — in
principle — serve as legitimate grounds for knowing other minds. Such
inferences, intuitions or revelations form the context in which speakers
consider themselves entitled to use a definite description, whereby they
assume hearers will go along. In the next section we will discuss a con¬
textual notion closely associated with definiteness, that of topic and/or
theme.

6.3. Definiteness and topicality

6.3.1. Topic and theme

The more contemporary preoccupation with topic dates back to the

Prague School's discussion of theme and rheme, which in turn
presumably harkens back all the way to Aristotle. As noted earlier
above,5 propositions — or sentences -- in discourse tend to contain some
presupposed ('old', 'given') and some asserted ('new') information. The
Prague School6 and related formulations,7 at least under one possible in¬
terpretation, would call the first theme and the second rheme. This in¬
terpretation thus associates the theme/rheme contrast with the presence
vs. absence of an anaphoric ('prior-discourse') context.
Later formulations in the 1970s translated this reading of the contrast
into topic vs. comment, respectively. The Praguean formulation already
allowed for a number of distinct senses of 'theme'. Thus, consider the
following attempts from Firbas (1974):

(3) "...according to one... [of the definitions; TG]... the theme ex¬
presses something that is spoken about; according to the other it
expresses something that is known or at least obvious in the given
situation..." (p. 23)

One could easily show instances in which the two definitions are incom¬
patible. Thus, both the indefinite and definite subjects in (1) above are
equally 'spoken about'.8 But while the definite subject in (lb) is 'known'
or 'obvious in the given situation', the indefinite subject in (la) is not.

5 See Ch. 4, section 4.6.

6 See for example Firbas (1966,1974).
7 As in Bolinger (1952,1954) or Halliday (1967), inter alia.
8 And in very much the same sense as the Aristotelian thema and rhema, which according to
most reasonable interpretations stand for the grammatical notions of subject and predicate,
respectively. I am indebted to Hartmut Haberland (in personal communication) for a
discussion of Aristotle's distinction and its scope.
 The Pragmatics of Anaphoric Reference 209

Other Praguean attempts to define the contrast between theme and

rheme are downright obfuscating. Thus consider (Firbas, 1974):

(4) "...the theme-transition-rheme sequence renders the word-

order non-emotive, unmarked, whereas the rheme-transi¬
tion-theme sequence renders it emotive, marked..." (p. 13)

An attempt to improve the definition by reference to the scalar notion of

communicative dynamism remains equally impenetrable. Thus con¬
sider (Firbas, 1974):

(5) "...By degree of communicative dynamism carried by a linguis¬

tic element, I mean the extent to which the element con¬
tributes toward the development of communication..." (p. 19)

Of the various Praguean senses of 'theme', the two which most com¬
monly carried over to the revived discussion in the mid-1970s were
those of old information and that which is talked about.

6.3.2. In search of the elusive topic

6.3.2.1. Preamble

Beginning with the early 1970s, much of the linguistic work in dis¬
course-pragmatics hinged on the discrete, binary, functional opposition
topic vs. comment, the inheritor of at least one of the Praguean contrast
of theme-rheme. This tradition is still on-going, and has been quite
beneficial in launching linguistics upon a vigorous search for the func¬
tional-pragmatic correlates of grammar. There are, however, a number
of indications that the topic-comment tradition has outgrown its useful¬
ness. In the following sections I will attempt to sketch out why this is so,
and why this important tradition in discourse-pragmatics must be
replaced with a much more complex, cognitively-based notion dis¬
course function. I will open the discussion by citing some of the facts
that our discrete, unitary notion of 'topic' was supposed — but failed --
to account for.

6.3.2 2. Definite, pronoun, agreement and zero-anaphora

The topic literature of the 1970s9 noted that grammatical categories

9 See for example Hawkinson and Hyman (1974) or Givon (1976a), inter alia.
210 Mind Code and Context

such as definiteness and pronouns were somehow associated with

topicality, an association widely expressed through the topic hierar¬
chies:

(6) a. DEFINITE > INDEFINITE

b. PRONOUN > NOUN
c. SUBJECT > OBJECT
It soon became apparent, however, that given the definition of topic as
old ('presupposed', 'given') information, the following grammatical
categories all marked the topic equally well:
(7) a. DEFNOUN: The man came into the house

b. INDEPENDENT PRONOUN: ... (she stayed outside, but)

he came into the house

c. GRAMMATICAL AGREEMENT: ...luego entro en la casa

...then entered-Jze in the house
'...then he entered the house...'
d. ZERO-ANAPHORA: ...(he came,) 0 looked around,...
If all four grammatical categories in (7) equally mark topic, then at the
very least the notion requires further elaboration. And if they differ
from each other in degree of topicality, then what is the substantive value
underlying that graded dimension?

6.3.2.3. Dislocated vs. neutral subject and object topics

Grammatical constructions such as left-dislocation, right-dislocation and

Y-movement have all been called, at one time or another, 'topicalization'.
Constituents subject to these 'movement rules' were considered, in the
topic-comment tradition, to be 'more topical' than their corresponding
neutral-ordered counterparts. They were all said to be old information,
either definite or generic but seldom REF-indefinite. These construc¬
tions are, for example:

(8) a. The man, she saw him (L-Dislocation, OBJ)

b. The woman, she came yesterday (L-Dislocation, SUBJ)
c. She saw him, the man (R-dislocation, OBJ)
d. She came yesterday, the woman (R-dislocation, SUBJ)
c. John she saw right away (Y-movement, OBJ)
d. Mary saw him (Y-movement, SUBJ)

Now, if all these 'moved' subjects or objects are topics, surely further
elaboration must be required in order to distinguish between them. And
 The Pragmatics of Anaphoric Reference 211

again, if they differ in 'degree of topicality' - what is the substantive

dimension involved?

6.3.2.4. Subject and topic

The topic-comment tradition of the 1970's noted that subjects were

more topical — or more likely to be the topic - than objects (see (6c),
above). This was clearly reflected in the much higher text frequency of
definite, pronominal and zero subjects. It is not clear, however, how the
difference between the following three sentences can be accounted for
in terms of "topic":

(9) a. John saw Bill (SUBJ = topic)

b. John, he saw Bill (SUBJ-topic; plus SUBJ = topic?)
c. Bill, John saw him (OBJ-topic; plus SUBJ = topic?)

If in (9a) John, by virtue of being subject, is also the topic, then the same
must be true of John in (9b); but, in addition, is John in (9b) also in addi¬
tion a marked-topic? And if Bill in (9c) is the topic (or marked topic), what
is the status of John there? Grammatically, John is marked the same way
in (9c) as in (9a), so is it still the topic? Or perhaps the unmarked topic?
Again, extra distinctions of either kind or degree are required. And the
suggestion,10 that somehow the subject is a grammaticalized topic, while
diachronically revealing, did not add much to sharpening the functional
definition of either topic or subject.

6.3.2.5. Multi-topic neutral clauses

Consider next the phenomenon of dative-shifting in bi-transitive sen¬

tences, as in:

(10) a. John gave the book to Mary

b. John gave Mary a book
c. John gave her the book
d. "John gave the book her
e. John gave it to Mary
f. "John gave Mary it

In the mid-1970s it was already obvious that there was some 'topicality'
a symmetry between the direct and indirect object; that the DO was
somehow more topical than the indirect object; and that 'promotion to
direct object' involved increased topicality of the erstwhile indirect ob-

10 See Givon (1976a).

212 Mind Code and Context

ject.11 In fact, the following topicality hierarchies were proposed at the

time to express the gradation of both semantic and grammatical case-
roles:12

(11) a. AGENT > DATIVE > PATIENT > OTHERS

b. SUBJECT > DIRECT - OBJECT > INDIRECT - OBJECT

Such hierarchies, if taken literally, suggest that bi-transitive sentences as

in (10) have three topics each; and that those topics are further distin¬
guished by either degree or kind of 'topicality\

6.3.2.6. The scalarity of ’focus’

A similarly vexing problem concerns the relation between the two
atomic notions 'topic' and 'focus', and the seeming scalarity of the latter.
Even a cursory look at grammar data already suggests such scalarity.
Thus consider:

(12) a. It was to John that I gave the book (cleft-focus)

b. To John I gave the book,... (Y-movement)
c. The person I gave the book to was John (pseudo-cleft)
d. I gave the book to John ('neutral')
e. I gave John a book (dative-shifted)

There is a clear gradation in the degree of 'focusness' ('emphasis',

'contrast') vested in John in (12a,b,c,d,e) above, a gradation that was re-
Platonized by in the recent past, by using two or three 'underlying' bi¬
nary features.13 But is it a gradation in the degree of 'focus'? Or, as the
bottom of the scale (12d,e) seem to suggest, a gradation from 'topic' to
'focus'? The two seems to be, somehow, the extreme points on a single
scale. Except that in the case of both cleft-focus (12a) and Y-movement
(12b) there are strong indications that the focused constituent is at the
same time also topical.14 For example, it is odd to start a lecture with a
cleft-focus construction, thought quite natural to start it with a pseudo¬
cleft. Thus compare:

11 See Shir (1979) or Givon (1979a, Ch. 4).

12 See Hawkinson and Hyman (1974) or Givon (1976a).
13 As in, for example. Chafe, 1976; Prince, 1979; Giv6n, 1979a, Ch. 2, inter alia. Such
Platonization of a continuum is reminiscent of the way Generative phonologists used to
handle the continuum of points of articulation. A similar Platonism still haunts - surpris¬
ingly - cognitive psychologists, viz Clark and Malt (1984).
14 See discussion in Givon (1979a, Ch. 2).
 The Pragmatics of Anaphoric Reference 213

(13) a. What I'm going to talk about today is evolution.

b. Tit's evolution that I'm going to talk about today.

The reason for the oddity of (13b) at discourse-initial is presumably due

to the fact that a cleft-focused constituent must have been established as
'topic' in the preceding discourse. A similar observation can be made
about Y-moved constituents, which are both contrastive and 'topical'.

6.3.2.7. Degree of presuppositionality

As we noted in Chapter 4, scalarity also seems to crop up in the de¬

gree of presuppositionality ('backgroundedness') of clauses. Presup¬
positionality of clauses is, presumably, closely related to topicality of
nominal arguments. The following clause-types are clearly ranked with
respect to their degree of presuppositionality:15

(14) a. V-complement: It was bad that she left

b. REL-clause: The woman who left...
c. Negative: She didn't leave
d. Conditional: If she leaves,...
e. 'Neutral': She left

Again, if we are dealing with a scalar dimension, what exactly are its
underlying parameters?

6.3.3. Interim summary: Searching for the cognitive base of topicality

The combined weight of all the facts discussed above must force us to
re-define 'topic' (thus also 'focus') in a way that would satisfy at least
the following requirements:

(a) It must accommodate scalarity;

(b) It must accommodate multi-dimensional space;
(c) It must link that multi-dimensional scalar space to some
plausible cognitive dimensions, ones that are not merely
functional-sounding labels for grammatical categories.
(d) It must allow for an empirical definition — and hopefully
measurement -- of the proposed functional dimensions.

Requirements (c) and (d) above invite further comment concerning

early functionalism in linguistics. Many linguists are in the habit of re¬
christening structural ('grammatical') categories with functionally-

15 See Ch. 4, above, as well as Givon (1979a, Ch.2; 1982a).

214 Mind Code and Context

sounding names (such as 'topic', 'pivot', 'theme', 'focus' etc.). We then

assume, somehow — by a mysterious leap of faith — that our description
has been transformed from mere structuralism to functionalism.16 It is of
course true that the pervasive iconicity of grammars (see Ch. 3,) indeed
entitles one to assume that grammatical categories correspond, at least
in the main, to functional categories. Nonetheless, if one is to
demonstrate a correlation between structure and function, both must be
defined independently. Otherwise 'correlation' is mere tautology.

6.4. The measurement of topicality:

Predictability, importance and attention

6.4.1. Preliminaries: Anaphoric predictability and cataphoric

importance of topics in discourse
Beginning in the late 1970s,17 I have been looking for some scalar
dimension that may correlate with our earlier intuitive notion of
topicality. A dimension -- or dimensions -- that can be measured and
quantified in discourse. In studying the word-order pragmatics of Bibli¬
cal Hebrew,18 the dimension of continuity -- of both the more abstract
'theme' and the more concrete nominal topics ('participants') -- sug¬
gested itself. With regard to the latter, referential continuity was sub¬
sequently equated with accessibility or predictability of reference.19
Finally, a text-based quantified methodology was developed20 to
measure some of the more obvious sources of referential predictability.
Reasoning from general considerations, the accessibility of a referent
to the hearer (or reader) is likely to be affected by the following four fac¬
tors concerning the preceding -- i.e. anaphoric -- discourse context:21

16 Some recent examples may be seen in Dik (1978) or Foley and van Valin (1985), inter alia.
The practice is, however, time-honored, and many of us have indulge in it before learning
to recognize its inherent circularity.
17 Givon (1977,1978a, 1979a).
18 See Giv6n (1977).
19 See Givon (1978a; 1979a, Ch. 2).
20 See Givon (ed„ 1983a).
21 Of the three contextual foci discussed earlier above - generic-cultural, deictic-situational
and discourse-textural - we deal here primarily with the latter. This is not to say that the
first two have no effect on the accessibility of referents in discourse. That effect, however,
is not easy to quantify and measure in text. Most particularly, generic-cultural knowledge
is indispensible for using thematic information to identify referents in discourse. This is
the knowledge of stereotyped 'frames', 'scripts', conventional situations, common
likelihoods etc. (see Schank and Abelson, 1977, inter alia). For some experimental details
see Anderson et al (1983).
 The Pragmatics of Amphoric Reference 215

(15) Factors affecting referential predictability/accessability:

factor cognitive dimension

(a) Referential distance from the memory decay of
previous mention in the discourse referent

(b) Referential complexity of the selective access to

preceding discourse context several referents

(c) 'Redundant' semantic information (same as above)

from inside the clause itself

(d) 'Redundant' thematic information (same as above)

from the preceding discourse

Of the four factors, (15a) and (15b) are easiest to quantify and measure in
discourse. They are the basis of the text-based measures of Referential
Distance (RD) and Potential Interference (PI), respectively. Factors (15c)
and (15d), on the other hand, are much harder to measure in discourse,
although experimental psycho-linguistic work has demonstrated their
clear effect.22
Both the RD and PI measures assess the anaphoric predictability of
referents ('topics') by looking at the preceding discourse context. But, as
noted in Chapter 5, another dimension of topicality is that of referential
importance. And while the importance of a referent is also reflected in
its anaphoric continuity, it is best measured independently in the
cataphoric -- i.e. following -- context. That is, one can presumably assess
a referent's importance in the discourse by measuring how long it per¬
sists once it has been introduced. The text-measure of Topic Persistence
(TP) was devised to do just that.
The most current version of the three text-based measures of
topicality is given in (16) below:

22 For the effect of semantic priming on the recoverability of pronominal co-referents, see
Corbett and Chang (1983). For the effect of thematic priming, see Anderson el al (1983) or
Tomlin (1987).
216 Mind Code and Context

(16) Text-based measures of the topicality:

measure discourse cognitive
(operational definition) dimension dimension

(a) Referential Distance:

"The number of clauses (or predictability memory
elapsed time)23 from in anaphoric decay
the last occurrence in the context
preceding discourse"
(b) Potential Interference:
"The number of semantically predictability competing
compatible referents within in anaphoric referent
the preceding three clauses" context searches
(c) Topic Persistence:
"The number of recurrences importance attention
of the referent in the in cataphoric
subsequent ten clauses" context

One must emphasize that these text-based measures do not assay di¬
rectly the presumed cognitive dimensions suggested in (16) above.[See
FN 23] Rather, they measure some of their more likely correlates in the
text. Using these measures, one can obtain numerical characterizations,
highly replicable both within a language and across languages, of the
topicality of nominal referents marked by particular grammatical
devices. In the following sections we will survey some of the results of
text-based studies employing these measurements.

6.4.2. Predictability and the code-quantity scale

As noted in Chapter 3, there exists a strong inverse correlation between
the degree of predictability of a referent and the phonological size of the
grammatical device used to code it. Thus, for example, the grammatical
devices in (17) below -- all of them anaphoric-definite categories used un¬
der various discourse conditions — scale rather consistently on our
anaphoric predictability measures of referential distance (RD) (16a).

23 Both the mental effort and the amount of attention assodated with referential processing
can be measured experimentally by a variety of direct and less direct means. For some
details see Anderson et al (1983), Dahl and Gundel (1982) or Givon et al (1985), inter alia.
For some more details, see Chapter 7.
 The Pragmatics of Amphoric Reference 217

(17) The code-quantity scale of referential predictability24

most predictable/accessible referent_typical RD value

a. zero anaphora 1 clause
b. unstressed/clitic pronoun/agreement 1-2 clauses
c. stressed/independent pronoun 2-3 clauses
d. full DEF-noun 10 clauses
e. restrictively modified DEF-noun 15 clauses
least predictable/accessible referent

To further illustrate how dependable the RD measure is in scaling gram¬

matical devices, consider the distribution of the RD values for all
pronouns and anaphoric zeros (categories (17a,b,c) above) and all definite
nouns (categories (17d,e) above) in spoken English:

(18) The distribution of referential distance (RD) for pronouns

(including zero anaphores) and definite nouns in spoken
English 25

pronouns DEF-nouns
RD in # of clauses n % n %
1-2 499 86.2 54 26.7
3-8 31 5.7 36 17.8
9-14 8 1.4 17 8.4
15-19 2 0.3 6 2.9
20+ 1 0.1 89 43.9

total: 541 100.0 202 100.0

median RD: 1.0 clauses 12.0 clauses

The principle underlying the correlation between code quantity and in¬
formation predictability, already noted earlier above,26 is reproduced in
(19) below:

(19) The code-quantity principle:

"The less predictable/accessible a referent is the more phono¬

logical material will be used to code it".

24 For the initial study see Givon (ed., 1983a).

25 From Giv6n (1983b).
26 See Ch. 3.
218 Mind Code and Context

Assuming, as one is surely entitled to, that more -- or stronger -- coding

attracts more attention and is lodged more firmly in memory, principle
(19) above may be converted into the general cognitive principle (20):

(20) "Stronger - more salient -- coding will yield a stronger effect

in attending to and memorizing the coded information"

The reason why more predictable information requires less -- or weaker

- coding is transparent: Predictable information is more readily avail¬
able in ('recoverable from') context; or, alternatively, it is more vividly
activated27 in the memory, thus more strongly attended to. It does not -
unlike new unpredictable information -- require strong activation by
massive coding.

6.4.3. Code quantity and referential confusion

The same group of studies28 also noted that the difference in RD

values between unstressed and stressed-independent pronouns — while
reproducible -- was relatively small (1 vs 2-3 clauses, respectively on a
scale from 1 to 20). However, the two types of pronouns are strongly
distinguished by the measure of potential interference (PI), where on a
scale from 1.0 to 2.0 unstressed pronouns score at the very bottom and
stressed pronouns at the very top. In other words, unstressed pronouns
are typically used when no competing referents occur in the immediate
anaphoric environment. While stressed pronouns are typically used
when such competing referents exist. A simple example may illustrate
the use of stressed pronouns in discourse contexts with unpredictability
from a variety of sources:

(21) a. John came home. Later on he left. (unstressed)

b. John talked to Bill. Later on he left. (unstressed)
c. John talked to Bill. Later on HE left. (stressed)
d. John took Mary to see a movie he'd (unstressed)
already seen. SHE hadn't though. (stressed)
e. John and Mary went to see a movie.
HE had already seen it. (stressed)
though SHE hadn't. (stressed)

In (21a) a single referent is introduced and then continued. It is thus to¬

tally predictable, and the pronoun is unstressed. In (21b) two referents
are introduced, but the one coded as subject continues as subject. That is

27 For a compatible notion of activation, see Chafe (1987).

28 Givon (ed., 1983a).
 The Pragmatics of Amphoric Reference 219

still predictable enough, given the strong tendency in connected narra¬

tive to continue the subject (see discussion further below); and so an un¬
stressed pronoun is still used. In (21c), on the other hand, the second
referent -- first introduced as object -- is resumed upon continuation as
subject. Such continuation is less predictable, and a stressed pronoun
must then be used. In (21d) two referents are first introduced, one as
subject and the other as object. The first one is then continued as subject;
that -- as in (21b) -- is still predictable enough, so an unstressed pronoun
is used. Then the second referents takes over as subject. Such a switch is
less predictable for two reasons: First, an erstwhile object is unexpected¬
ly made into subject. Second, a thematic expectation is set up, contrast¬
ing the fact that John had seen the movie with the fact that Mary hadn't
yet seen it. Two expectations of continuity are thus tampered with, and
a stressed pronoun is used. Finally, in (21e) both referents are intro¬
duced as subjects. When John is then continued as subject all by himself;
the expectation of subject continuity is broken, and a stressed pronoun
is used. The code-quantity principle of grammatical coding is thus sen¬
sitive to predictability whatever its source. This is further evident from
surveying the interaction between code quantity and thematic predict¬
ability.

6.4.4. Code quantity and thematic predictability

The sensitivity of the code-quantity scale to differences in thematic

continuity may be further illustrated by contrasting the use of zero
anaphora and unstressed pronouns in English. As was apparent in (17)
above, the referential distance (RD) values for zero-anaphores and un¬
stressed pronouns are virtually the same (1.0 vs. 1.2, respectively, on a
scale from 1 to 20). Similarly, both are equally low on the potential inter¬
ference (PI) scale, scoring at the very bottom of the scale.29 Whatever un¬
predictability controlling the choice between these two grammatical
devices in discourse must therefore arise from some other source. To il¬
lustrate this, consider:

(22) a. He came into the room,

b. 0 looked around,
c. 0 went to the window
d. and 0 looked out.
e. (i) ...He was dog tired.
*(ii) ...0 was dog tired.

The discourse in (22) involves a single participant, and as long as the

29 Givon (ed., 1983a)

220 Mind Code and Context

thematic progression is maximally continuous (thus maximally predict¬

able), zero anaphora is used to code the single referent (22b-d). The use
of the period to end (22d), however, marks a subtle increase in thematic
discontinuity. The (still predictable) referent cannot be coded with zero
anaphora any more, but must be resumed with the phonologically-
larger unstressed pronoun.
A similar phenomenon has been demonstrated experimentally in the
use of pronouns vs. definite nouns. As seen in (17) and (18), above,
pronouns and definite nouns have strikingly different RD values, with
the latter used in discourse contexts of much lower referential continuity.
They also differ, in the same direction, on the scale of potential inter¬
ference (PI).30 However, it has been shown 31 that the use of DEF-nouns
rather than pronouns can be triggered, in environments of absolute
referential predictability, by introducing strong thematic discontinuity.

6.4.5. Subject, object and transitivity: Referential predictability

and referential importance

6.4.5.1. Preamble: Referential importance

The second major cognitive component of topicality is referential im¬

portance. We have already discussed this feature in Chapter 5, in con¬
junction with indefinite reference. This feature also plays a major role in
the grammar of definite reference.

6.4.5.2. Subject and object

The same text-based studies32 that established the code-quantity scale

(17) also revealed a striking difference between subjects and objects of
clauses in terms of their topical predictability. To illustrate this, consider the
topicality properties of agents and patients of semantically-transitive
clauses33 in Chamorro discourse. In this language, the majority of semanti¬
cally transitive clauses in narrative are coded morphologically as ergative.
When the topicality of the agent in however, clauses is radically low, the
clause is coded as passive. When the topicality of the patient is radically
low, the clause is coded as anti-passive.34 Grammatically, passive clauses are
those where the agent has been 'demoted' from the role of subject, and a
non-agent (most commonly the patient) has become the subject of the

30 Givon (ed., 1983a)

31 Tomlin (1987).
32 Givon (ed., 1983a).
33 Clauses are considered semantically-transitive if their verb requires for its normal sense
the participation in the event of both an agent and a patient. For discussion see Givon
(1984a, Chapters 4,5).
34 See discussion in Ch. 3, above; see also Givon (1984a, Ch. 5) and Cooreman et al (1984).
 The Pragma tics of A napho tic Reference 221

clause. Conversely, anti-passive clauses are those where the patient has been
'demoted' from its role as object. The topicality of both arguments was
measured by two of our text-based measures, referential distance (RD) and
topic persistence (TP). A summary of the results is given in (23) below.

(23) Transitivity and the topicality of agent and patient in

semantically-transitive clauses in Chamorro narrative dis¬
course 35

average RD in average TP in
clause type # of clauses # of clauses
agent patient agent patient
ergative 1.49 4.35 2.45 0.81
in-passive 4.06 1.38 1.31 2.00
ma-passive 6.33 3.33 0.44 1.44
anti-passive 1.86 20.00 1.29 0.00

The results reported above demonstrate rather vividly that the subject of
a transitive clause is more topical than the object, both in terms of
anaphoric predictability (RD) and cataphoric importance (TP). When the
agent is 'demoted' from the subject role (passivization), its referential dis¬
tance increases and its persistence decreases. The patient, when
'promoted' to the subject role (passivization), shows decreased referential
distance and increased persistence. Finally, when the patient is 'demoted'
from the object role altogether (anti-passivization), its referential distance
increases maximally and its persistence decreases maximally.

6.4.5.3. Direct and indirect object

We have suggested earlier above that direct objects are 'more topical'
than indirect objects. The striking difference in topicality is measured both
anaphorically, in terms of referential distance, and cataphorically, in terms
of topic persistence. To illustrate this, consider the following results from a
text-based study of the topicality of direct and indirect objects in Nez Perce.

(24) Topicality of direct and indirect objects in Nez Perce 36

average RD in average TP in
# of clauses # of clauses
object type patient oblique_patient oblique

DIRECT OBJECT 1.7 2.2 2.3 1.5

INDIRECT OBJECT 8.3 10.1 0.6 0.4

35 From Cooreman (1982b) and Cooreman et al (1984).

36 From Rude (1985). See also Givon (1984d).
222 Mind Code and Context

The results demonstrate that direct objects are more topical than in¬
direct objects, both anaphorically — by the measure of referential dis¬
tance, and cataphorically -- by the measure of topic persistence. When a
patient is 'demoted' from direct-object, its RD measure increases and its
TP value decreases. Conversely, when an oblique object is 'promoted' to
direct-object, its RD measure decreases and its TP measure increases.
Once again, importance -- as measured indirectly through text persist¬
ence -- proves to be an intimate ingredient of topicality.

6.4.5.4. Grammatical marking of important definites

As noted in Chapter 5, the unstressed demonstrative this in colloquial

English marks thematically important indefinites, contrasting with a
which marked unimportant ones. It may very well be that a similar con¬
trast can be shown between the unstressed that and the article the in col¬
loquial English, with that marking more important definites and the less
important ones.
A reminiscent contrast has been reported (Lichtenberk, 1986) for
To'aba'ita (Oceanic), in the use of two demonstratives in discourse. One
of them (the spatially proximant this) is used to mark thematically im¬
portant definites; the other (the spatially remote that) is used to mark
unimportant ones.
Similarly, in Krio (Givon, 1985a; see also Chapter 5), definite nominals
are usually marked with the proximant demonstrative dis (English
'this'). Definite nouns that are thematically important, in particular
those introduced first into the discourse with the numeral one, are most
likely to be marked, when re-introduced into the narrative after a gap of
absence, by the suffix ya-so (English 'here-so'). Processing instructions
(via grammatical clues) concerning expected cataphoric importance are
thus not given only upon the first introduction of a referent, but also
upon re-introduction following a gap of absence.

6.4.6. Word-order and topicality37

6.4.6.1 The presumed topic-comment ordering principle

Perhaps the area where the Prague-school tradition had registered its
strongest impact on 1970s functionalist work was in the pragmatics of
word-order. Indeed, a received Praguean wisdom prevails to this day,
namely that topics precede comments, theme precedes rheme, old information

37 More detailed discussion and many more cross-language examples may be seen in Givon
(1987b), from which these materials are condensed.
 The Pragma tics of Anaphoric Reference 223

precedes new information and the subject more naturally precedes the verb. A
classical example of this approach at its best may be seen in Bolinger
(1954), where the relative position of the subject in Spanish clauses was
said to abide by this generalization:38

(25) a. Context: Quien canta?

'Who's singing?' (topic = singing)
b. Response: Canta Juan. (VS word-order;
'John is singing' focus = Juan)
c. Context: Que hace Juan
'What is John doing?' (topic = Juan)
d. Response: Juan canta. (SV word-order;
'John is singing' focus = singing)

As recalcitrant as the topic-before-comment generalization seems to

be, detailed -- text-based, quantified — studies show it to be at best
seriously flawed and at worse simply false. In the space below we will
survey some of the evidence showing why this turned out to be the
case. We will see that the topic-before-comment ordering principle holds
true only for the topicality feature of importance. As to the feature of
predictability, the facts suggest exactly the opposite generalization:
Older, more continuous, more predictable information — if at all ex¬
pressed -- consistently follows newer, less continuous, less predictable
information. In other words, the anaphorically less topical information
precedes the more topical.39

6.4.6.2. Word-order flexibility and referential predictability

In (26) below the results of three text-based studies are summarized,

all measuring the referential distance of the subject NP when coded by
various grammatical devices, including when the subject either
precedes or follows the verb (or 'predicate').

38 Bolinger was obviously not alone in this regard. Many of us in the 1970s -and to this day
- have taken this for granted, as in Li and Thompson (1976), or Givon (1976a, 1976b, 1979a),
inter alia.
39 How a false generalization could have been maintained so long and gain such currency is
an object lesson in methodology. Not one of the works purporting to demonstrate this
generalization was based on text-distribution quantified study. All language examples
cited in support of the topic-before-comment principle were either constructed from the
linguist's own intuition, or picked out selectively from texts.
224 Mind Code and Context

(26) Correlation between referential distance (RD) and relative

position of the subject in Written Polish, Biblical Hebrew
and spoken Spanish-English Pidgin40

average RD in # of clauses
topic-coding Spanish- Biblical
device English. Hebrew Polish

zero-subject/ 1.0 1.1 /

verb-agreement

verb-subject 3.4 4.8 3.1

order

subject-verb 13.0 8.4 6.5

order

zero-comment/ 20.0 / /
topic repetition

The data tabulated in (26) suggest that the following pragmatic prin¬
ciple control the relationship between linear order and referential pre¬
dictability:

(27) "A more predictable topic follows the comment; a less-pre¬

dictable topic precedes the comment"

Or, in more general terms of informational predictability:

(28) "Less predictable information is placed earlier in the string"

The results of the Spanish-English Pidgin study, together with data

from two more Pidgins 41 and from spoken English,42 suggest another
possible generalization, one that recognizes the following ranking of
topic-coding devices according to topic predictability:43

40 The Polish data is from Rybarkiewicz (1984). The Biblical Hebrew data is from A. Fox
(1983); the Spanish-English Pidgin data is from Givon (1984b).
41 See Givon (1984b).
42 See Givon (1983b).
43 First formulated in Givon (1983c).
 The Pragmatics of Amphoric Reference 225

(29) Ranking of the most basic topic-coding devices according

to degree of topic predictability

most predictable/accessible/continuous topic

COMMENT (zero topic)

COMMENT-TOPIC
TOPIC-COMMENT
TOPIC (zero comment)

least predictable/accessible/continuous topic

Scale (29) allows one to integrate, tentatively, principle (28) of linear

order and principle (19) of code quantity into a single cognitively-based
principle of task urgency:

(30) The task-urgency principle:

"Attend first to the more urgent task"

Zero topic-coding is used when the topic is most predictable, thus the
task of expressing it is least urgent. Zero-comment (repeated topic) is
used when the topic is maximally unpredictable, thus the task of estab¬
lishing it is maximally urgent. While the two word-order devices -- com¬
ment-topic vs. topic-comment — occupy intermediate points on the
scale, consonant with principle (28).

6.4.6 3. Dislocated order and anaphoric predictability

In a closely related study, the referential distance of subject NPs in

spoken English was measured. Right-dislocated and Left-dislocated
subjects were compared to the neutral SV order, as well as to zero-
anaphora and topic repetition. The results are summarized in (31)
below.

(31) Referential distance (RD) of grammatical devices coding

subject NPs in spoken English44

topic-coding device average RD in # of clauses

zero-anaphora 1.0
R-dislocation 1.0
Neutral SV order 10.1
L-dislocation 15.3
topic repetition 17.4

44 From Givon (1983b).

226 Mind Code and Context

English is a rigid-order language, where SV(O) is the most common,

neutral order. But when dislocated ordering is used, the pre-posed order
(L-dislocation) displays much higher RD values than the post-posed
order (R-dislocation). In other words, pre-posing occurs when referents
are less predictable, post-posing when they are more predictable; and
the neutral order displays intermediate values. Principle (28) is again
reflected in these results. * •

6.4.7. Word order and topic importance

6.4.7.1. Preamble

The fact that grammar is sensitive to topic importance has already

been noted, for indefinite referents, in Chapter 5. Further, in sections
6.3.5.2. and 6.4.5.4, above, it was noted that the use of various grammati¬
cal devices is sensitive to the thematic importance of definite referents. In
this section we will see how word-order, in addition to being sensitive
to referential predictability, is also sensitive to the topic's thematic im¬
portance.
On rather general grounds, one would predict the following correla¬
tion between topic importance and communicative task urgency:

(32) "The more important a topic is, the more urgent is the task of
communicating it"

Given principle (30) above, correlating task urgency with pre-posed

order, above, one would expect more important topics to be pre-posed,
and less important ones to be post-posed. In the space directly below I
will summarize the results of some quantified, text-based studies that
support this prediction.

6.4.7.2. Definiteness and word-order in Papago

Word-order in Papago is completely flexible for both subjects and ob¬

jects. In a recent study, a strong correlation was shown between definite¬
ness and word-order in this language, with the pre-verbal position
favoring indefinite NPs, and the post-verbal position definite NPs. These
results are summarized in (33) below.
 The Pragmatics of Anaphoric Reference 227

(33) Definiteness, contrastiveness and word-order in Papago45

Pre-Verbal Post-Verbal Total

N % N % N %
Indefinite 143 0.89 17 0.11 160 1.00

Definite 7 0.02 277 0.98 284 1.00

"Prag. Marked" 38 0.95 2 0.05 40 1.00

The results indicate that the vast majority of definite -- predictable —

referents are deferred, i.e. they appear post-verbally. The vast majority in¬
definite -- unpredictable — referents are pre-posed. In addition, referents
under contrastive focus -- i.e. those involving either high referential un¬
predictability due to referential clutter or thematic unpredictability46 —
appear primarily pre-verbally. The results so far uphold the correlation
between pre-posed order and unpredictability, i.e. principle (28).
Still, the results are a bit skewed: The post-verbal placement of definite
('given', 'predictable') referents, is almost categorial (96%). The pre-ver¬
bal placement of indefinites ('unpredictable', 'new') referents, on the
other hand, is less categorial (89%, with 11% going post-verbally). When
nouns in this exceptional 11% fraction were compared to those in the
predominant pre-verbal category,47 it was found that the post-posed in¬
definites were all thematically unimportant in the discourse. In other
words, we have here a conflict between the two major cognitive dimen¬
sions underlying topicality — predictability and importance. Indefinite
referents are by definition highly unpredictable, they are thus expected
to be fronted. But the few of them that are of low importance are post-
posed. In other words, topic importance here overrides referential pre¬
dictability in controlling word-order.

6.4.7.3. Word-order and topic persistence in Klamath

Klamath is also a flexible-order language, much like Papago. A recent

study measured the referential distance (RD) and topic persistence (TP)
of pre-verbal and post-verbal subjects and objects in this language. The
results are summarized in (34) below.

45 From Payne (1985).

46 See further discussion below.
47 Payne (1985).
228 Mind Code and Context

(34) Referential distance (RD), topic persistence (TP) and word-

order in Klamath48

RD and TP in # of clauses

Pre-Verbal Post-Verbal
(SV/OV) (VS/VO)
Category RD' TP RD TP
subject-NP (DEF) 7.82 1.98 3.92 1.36

object-NP
DEF. 9.97 1.60 6.04 2.80
INDEF. / 1.22 / 0.27

As one can see, by the measure of referential predictability (RD) pre-ver¬

bal referents are more unpredictable, post-verbal ones more predictable,
both thus abiding by principle (28). In addition -- at least for subjects and
indefinite objects49 -- pre-verbal referents are also more persistent, thus
more important in the discourse. Of more interest is, of course, the behav¬
ior of indefinite objects, where the pre-verbal ones show a much higher
persistence measure than the post-verbal ones. On the basis of referential
predictability alone, one would expect all indefinite object to be pre-ver¬
bal. As in Papago, however, topic importance overrides referential pre¬
dictability.

6.4.7.4. Direct and indirect object

It has been observed that direct object invariably precede indirect ob¬
jects, and this behavior transcends word-order type.50. Now, as shown in
section 6.4.5.3., above, the direct object outranks the indirect object in
both features of topicality, it is both more predictable and more impor¬
tant. But once again we have a conflict here, in the expected correlation
of topicality with word-order: By rule (28), the more predictable direct
object should follow the indirect object. Once again, however, impor¬
tance overrides predictability, and the direct object invariably precedes.

6.4.7.5. Contrastive NPs and pre-posed order

The pre-posing of object NPs to mark contrastiveness or emphasis in

VO languages is one of the most wide-spread word-order devices (viz

48 From Sundberg (1985).

49 The exceptional behavior of definite objects is so far unaccounted for.
50 See Givon (1984d).
 The Pragmatics of Amphoric Reference 229

the Papago distribution in (33), above). As noted earlier, Y-moved NPs

used to be considered topical. And indeed, in terms of their referential
distance (RD) measure, such pre-posed objects turn out to be more pre¬
dictable than post-verbal, neutral-ordered objects. By our rule (28) then,
they should be pos f-posed. However, in terms of their potential inter¬
ference (PI) measure, they seem to be used in environments of high
referential complexity -- i.e. un-predictability.51
As an illustrative example of Y-movement in English, consider:

(35) (a) ...So we wound up owing money to both the German

and the French.
(b) to the German we gave a deferred check,
(b) but the French we paid in cash...

A recent quantified text-based study investigated the discourse distribu¬

tion of pre-verbal objects in two VO languages. Biblical Hebrew and
Mandarin Chinese. The results are summarized in (36) below.

(36) Text frequency, referential distance (RD) and potential in¬

terference (TP) of pre- and post-verbal objects in Mandarin
Chinese and Biblical Hebrew52

Written Spoken Biblical

Mandarin Mandarin Hebrew
Measure VO OV VO OV VO OV
% in text 94 6 92 8 97 3
average RD 10.00 3.00 4.00 2.00 12.10 2.50
average PI 1.09 1.47 1.14 1.80 1.69 2.00

The pre-verbal position of Y-moved objects must be due to either high

referential complexity or thematic contrastiveness, rather than to high
referential distance. Referential predictability is a complex cognitive
dimension, to which a variety of factors can contribute. The grammar of
human language, in this case word-order, is sensitive to the degree of
predictability whatever its source.

51 There are, in addition, some grounds for suspecting that pre-verbal (Y-moved) objects may
code objects that are more important than post-verbal ('neutral-ordered') ones. If that turns
out to be the case, then the pre-verbal position of such object may be ascribed to both low
predictability and high importance.
52 From Sim and Givon (1985).
230 Mind Code and Context

6.4.7.6. Predictability, importance and pre-posed subject

pronouns
The pre-posing of stressed independent subject pronouns in V-first
languages (VSO, VOS) is another example of how contrastiveness, or
referential complexity, overrides referential distance in determining
word-order. Stressed independent pronouns are universally used under
the following discourse conditions:53, .

(a) Fairly low values of referential distance (RD), usually near

the bottom of the scale (2-3 clauses on a scale of l-to-20).
That is, the antecedent has typically been mentioned 2-3
clauses back;
(b) High values of potential interference (PI), usually near the
top of the scale (2 on a scale of 1.0-2.0). That is, another
semantically-compatible (thus often contrasting) referent
had been mentioned in the preceding 2-3 clauses;
(c) High incidence of subject/topic switching from the preced¬
ing clause. That is, the independent pronoun is used to
switch the topic back from the preceding -- now contrasting -
- referent.

As an illustrative example from English, consider:

(37) (a) ...So we wound up owing money to both Tom and Sylvia.
(b) HE agreed to accept a deferred check.
(c) SHE demanded cash...

As noted above. Early Biblical Hebrew (EBH) is a rigid VO language

with flexible subject position. That is, the subject may appear either pre-
or post-verbally. In terms of frequency, post-verbal definite subject nouns
are predominant in the EBH text. On the other hand, independent sub¬
ject pronouns are almost categorially pre-verbal, as seen in the distribu¬
tional figures in (38) below.

(38) Distribution of pre- and post-verbal subjects in Early Bibli¬

cal Hebrew54

pre-verbal post-verbal total

category N % N % N %
Def-Noun 128 32.3 269 67.7 397 100.0
Pronoun 25 92.5 2 7.5 27 100.0
The pre-posing of independent pronouns thus represents another in-

53 See Givon (1983c).

54 From Givon (1977). The text frequency pertains to the book of Genesis.
 The Pragmatics of Anaphoric Reference 231

stance where referential complexity — or contrastiveness — is an impor¬

tant source of unpredictability. In this and many other cases, inde¬
pendent pronouns are also strongly associated with the function of
topic switching.
There is, further, strong evidence suggesting that independent subject
pronoun also rank higher than definite subject nouns in terms of topical
importance. Thus, in another study55 the average topic persistence (TP) of
post-verbal definite subject nouns in Early Biblical Hebrew was found to
be 0.93 clauses. The average TP measure for independent subject
pronouns was 3.26 clauses. The pre-posing of independent pronouns is
thus triggered by both features of topicality — low referential predict¬
ability and high topical importance. A similar distribution has been
reported for spoken Spanish, another VO language with flexible subject
position (Silva-Corvalan 1977).

6.4.7.7. Pre-posed indefinite subjects in verb-first languages:

Where predictability and importance coincide56

Rigid V-first languages are extremely resistant to pre-posing any

nominal arguments, i.e. placing them before the verb. This holds true to
both subjects and objects. There are three pre-posing grammatical
devices that are found even in the most rigid V-first languages: Clefting,
WH-movement, and the pre-posing of referential-indefinite subjects,
otherwise known as the existential-presentative construction.57
Referential-indefinite subjects are by definition maximally unpre¬
dictable, introducing new referents into the discourse. Not only are they
unpredictable, but they are also important in the subsequent discourse.
Otherwise languages tend to introduce unimportant referents in the ob¬
ject position. To illustrate this, consider once again the data from collo¬
quial English (see Chapter 5), concerning the topic persistence (TP) of
indefinite subjects and objects marked by either a or this. Indefinite sub¬
jects marked by either this or a tend to appear in existential-presentative
construction, as in:

(40) a. ...there was this bear who lived in a forest...

b. ...there's a guy I used to know who...

As noted earlier (Ch. 5), this codes more important indefinites than
does a. The higher persistence -- thus importance - of indefinite-existen¬
tial subjects (as against non-subjects) can be seen in (41) below.

55 See A. Fox (1983).

56 For a more expanded discussion, see Givon (1987b).
57 See extensive discussion in Hetzron (1971) and Givon (1976a).
232 Mind Code and Context

persistence (TP > 2) indefinite NPs marked by 'this' and 'a'58

'THIS' 'A'
Non- Non-
Subject Subj Total Subject Subj Total
TP N % N % Tl % N % N % N %

0-2 4 14.2 10 66.6 14 32.5 10 76.9 89 94.6 99 92.5

2 < 24 85.8 5 33.4 29 67.5 3 23.1 5 5.4 8 7.5

TOT 28 100.0 15 100.0 43 100.0 13 100.0 94 100.0 107 100.0

These results recapitulate our earlier discussion concerning the higher

topicality of subjects over objects -- higher in terms of both predict¬
ability and importance. In the case of definite subjects in V-first lan¬
guages, predictability overrides importance, and -- given the rigid
V-first constraint -- the subject follows the verb. However, in the case of
indefinite subjects, both features of topicality work in the same direction,
toward a pre-posed order. The fact that the pre-posing of indefinite sub¬
jects is so pervasive even in the most rigid V-first language, can be thus
ascribed to the conflation of both features of topicality.

6.4.7.8. Cleft-focus constructions: Counter-expectancy

and word-order

Another NP-fronting grammatical device that appears even in the

most rigid V-first language is that of cleft-focus, as in:

(41) a. It's the book I'm looking for (not the cup)...
b. It's a book I'm looking for (not a cup)...

As suggested earlier above, there are grounds for suspecting that the
fronted, cleft-focused referent is also topical in the sense of givenness;
that is, it had been made topic, often quite recently, in the preceding dis¬
course.59
The existence of cleft-focus fronting even in the most rigid V-first lan¬
guages thus represents another case when strong referential and
thematic counter-expectancy overrides 'givenness', and is the apparent
source of unpredictability.

58 From Shroyer (1985); see also Wright and Giv6n (1987).

59 See section 6.3.2.6., above.
 The Pragmatics of Amphoric Reference 233

6.4.7.9. Word order and thematic predictability

In the preceding sections we have seen that word-order in human lan¬

guage is sensitive to predictability arising from diverse -- primarily
referential — sources. In this section we will survey some evidence sug¬
gesting that word-order is also sensitive, in very much the same fashion,
to the level of thematic predictability in the discourse.
As is generally known, the paragraph-initial clause in discourse oc¬
cupies a point of thematic discontinuity, as compared to paragraph
medial or final clauses. In a recent study, the distribution of pre-verbal
vs. post-verbal subjects/agents in Tagalog, a V-first language, was
studied, in paragraph-initial and non-initial clauses. The results, found
to be statistically significant, are summarized in (42) below.

(42) Pre-verbal subjects/agents and paragraph boundaries in

Tagalog60

_Position in the Paragraph_

Initial Non-Initial
(discontinuous) (continuous) Total
_ N % N % N %

AGT-V 61 0.27 165 0.73 226 1.00

V-AGT_50 0.13 338 0.87 388 1.00

Total 111 503 614

As can be seen, paragraph-initial clauses display twice the frequency of

pre-posed subjects/agents as non-initial clauses. In other words, en¬
vironments of high thematic discontinuity ('unpredictability') favor the
fronting of nominal referents.
The same effect was also shown, in the same study, in the distribution
of pre-verbal subjects/agents in clauses preceded by a comma or zero
punctuation, as against clauses preceded by a period. As most literate
people are bound to know,61 commas in written discourse signal higher
thematic continuity ('predictability') than periods. In (43) below the dis¬
tribution of pre-verbal subjects/agents in Tagalog, directly following
either a comma or zero, or following a period, is given.

60 From B. Fox (1985). .

61 Amazingly enough, many linguists, at least in their more guarded professional moments
are blissfully unaware of the rudimentary correlation between thematic continuity and
punctuation.
234 Mind Code and Context

(43) Pre-verbal subjects/agents and punctuation in Tagalog62

_Punctuation Type_
Preceded by Preceded by
Comma/Zero Period
(continuous) (discontinuous) Total
N % N % N %

AGT-V 83 0.37 143 0.63 226 1.00

V-AGT 263 0.68 125 0.32 388 1.00

Total 346 268 614

As the figures suggest, pre-verbal subjects/agents are much more

prevalent in the more discontinuous ('unpredictable') thematic environ¬
ment following a period, than in the more continuous environment fol¬
lowing a comma.

6.5. Discussion

6.5.1. The pragmatic nature of definite reference

In dealing with pragmatics of definiteness and anaphoric reference,

one deals essentially with communicative tasks in discourse. Atomic,
Platonic notions such as 'identifiable', 'known', 'referred to earlier' etc.
go only a short distance toward accounting for the wealth of structural
devices used in the grammar of anaphora and definite reference in
human language, let alone for the use of grammar in natural com¬
munication. A speaker-hearer based functional-pragmatic account is
both more realistic and more far reaching in explaining the facts of
structure and function. The profoundly pragmatic, context-driven na¬
ture of anaphora, definiteness and topicality may be summarized
through its following features:

(a) it involves the pragmatically-based notions of importance or

relevance;
(b) it involves the speakers communicative intent; and
(c) it involves expectations the speaker makes about what
transpires in another mind — the hearer's.

62 From B. Fox (1985).

 The Pragmatics of Amphoric Reference 235

6.5.2. Task urgency, code quantity, linear order and attention

We have noted above that both code quantity and linear order in
human language are sensitive to both major features of topicality -- pre¬
dictability and importance. Both low predictability and high importance
induce an increase in the quantity of the code used to mark a constituent.
Both can induce the pre-posing of a constituent. One may suggest, tenta¬
tively, that both dimensions are merely sub-types of communicative
task urgency. A communicative task in discourse is more urgent if the
speaker deems it to be either less predictable ('less accessible') to the
hearer, or more important. As is well documented in cognitive psychol-
ogy,63 the initial item in a string of information tends to receive more at¬
tention, it is memorized better and retrieved faster, compared to non¬
initial items. The fronted position must therefore be perceptually more
salient. The pre-posing of more urgent information -- either less predict¬
able or more important -- must then be a communicative device
designed to attract more attention, much like assigning a higher code¬
quantity to the information. Our principle (30), above, may be thus
recapitulated as:

(44) "Attend first to -- i.e. code more saliently -- the more urgent
task; that is, the less predictable or more important task".

That (44) is fully consonant with general cognitive and perceptual prin¬
ciples is only to be expected, since grammar is not likely to have
evolved in a complete perceptual and cognitive vacuum.64

63 For studies showing that initial clauses in the paragraph receive more attention, see Cirilo
and Foss (1980), Glazner et al, (1984) and Haberlandt (1980). For studies showing a
slow-down - i.e. more attention expended - in the processing of sentence-initial words,
see Aronson and Ferres (1983), Aronson and Scarborough (1976) and Chang (1980), inter
alia. For studies showing that more mental effort is allocated to sentence-initial words than
to non-initial words, see Cairns and Kamerman (1975), Cutler and Foss (1977) and Marslen-
Wilson et al (1978). The favoring of clause-initial words in comprehension has also been
reported by Gemsbacher (1985). Finally, a more recent study by Gernsbacher and
Hargreaves (1986) shows that the relative early position of a noun within the clause affects
favorably its retrieval, as compared with the second-placed noun, regardless of subject-ob¬
ject role, and regardless of whether the first noun was at a clause-initial position.
64 For reasons that sometime belie one's understanding, some linguists persist in viewing
the organizational principles of language as being so unique and abstract, so as to be
virtually independent of cognitive and perceptual organization. For a fascinating argument
in this vein, see Chomsky's various contributions to Piatrelli-Palmarini (Ed., 1980). In
Chapter 10, below, we will also suggest that such modular separation between grammar
and cognition is incompatible with what we know about evolution.
 .
- , .

'

N,
CHAPTER 7

MODES OF KNOWLEDGE AND

MODES OF PROCESSING:
THE ROUTINIZATION OF
BEHAVIOR AND INFORMATION

7.1. Introduction*

This chapter deals with a complex phenomenon, one that may be ap¬
proached from three distinct entry points. Each entry point corresponds
to the traditions of one of the three cognitive disciplines — philosophy
linguistics, psychology. The latter intersects with a fourth tradition, that
of neurology. In spite of extreme diversity in method and frame of refer¬
ence, these traditions seem to converge at, define -- or at the very least
evoke — a recognizable core phenomenon. Of the three, philosophy's
claim to a share in this convergence is a bit more tentative, perhaps only
suggestive. This is of course no accident, given the traditional remote¬
ness of the mother discipline from empirical concerns.
From the vantage point of philosophy, the core phenomenon may be
given in terms of the three modes of inference -- also modes of
knowledge: Deduction, induction and abduction. From the vantage
point of linguistics, the phenomenon has been described more recently
in terms of the process of grammaticalization, i.e. the rise of grammati¬
cal structure as an instrument of language processing. From the vantage
point of psychology, the phenomenon has been investigated in terms of
the contrast between attended vs. automated processing. This latter ap¬
proach also links the discussion firmly to neurology and motor be¬
havior, perhaps eventually also to artificial intelligence.
The unifying theme here is the routinization -- or automation -- of be¬
havior and information processing as it is relevant not only to higher
cognitive levels, but also to lower sensory-motor functions. Thus, while

*1 am indebted to Morti Gernsbacher, Ray Hyman, Daniel Kahneman, Steve Keele, Mike
Posner, Walter Schneider and Harry Whitaker for helpful comments and suggestions. I have
also benefited from discussing an earlier version of the chapter at the University of Oregon
Symposium on Attended vs. Automated Processing, November, 1986
238 Mind, Code and Context

the discussion at this juncture will focus primarily on the routinization

of information processing at higher levels, one must keep in mind the
enormous ramifications the subject has for behavior and information
processing at all levels of biological organization. We will return to
some of those broader ramifications in Chapter 10.

7.2. The philosophical tradition: Modes of inference

7.2.1. Preamble

The philosophical tradition has confined its discussion largely to two

modes of inference, or sources of knowledge:

(a) Deduction: by which one infers specific instances from

the general rule;
(b) Induction: by which one presumably discovers the
general rule from a representative sample of
specific instances.

In the non-pragmatic reductionist traditions of Western epistemology,

there exists an obvious parallelism between deduction and the
rationalist approach on the one hand, induction and the empiricist ap¬
proach on the other. Still, complete symmetry here turns out to be some¬
what of an illusion. To begin with, the Godfather of Western empiricism,
Aristotle, already defined the third -- pragmatic -- mode of inference:

(c) Abduction: by which one reasons by hypothesis from in¬

stances or general rules to their wider context.

Further, several eminent exponents of inductive reasoning, among them

Peirce and the 'early' Wittgenstein, have noted that induction itself
depends crucially upon an irreducible element of abduction.

7.2.2. The ’early’ Wittgenstein on induction and deduction

7.2.2.1. Preamble

In his Tractatus Logico Philosophicus, Wittgenstein makes a number of

observations concerning the usefulness of the two traditional modes of
inference, deduction and induction. Without explicitly outlining a third
mode, he somehow manages to point out the fatal flaws of both indue-
 Modes of Knowledge and Modes of Processing 239

tion and deduction, in the process laying the groundwork, perhaps

unintentionally, for a pragmatic theory of information.

7.2.2.2. The limits of induction

Wittgenstein treats induction in the Tractatus from a fairly conserva¬

tive deductivist point of view,1 arguing that however many instances
support the same general rule, no deductive certainty ('logical jus¬
tification') can be gained from such empirical support:

"...the procedure of induction consists in accepting as true the simplest

law that can be reconciled with our experience. This procedure, how¬
ever, has no logical justification, but only a psychological one..." (1918,
p.143)

In calling the justification of inductive inference psychological, Wit¬

tgenstein dismisses the possibility of systematic justification of inductive
reasoning. In that he is not alone: In the deductivist's valhalla, whatever
is not 100% certain is not worthy of consideration:

"...It is a hypothesis that the sun will rise tomorrow [based on the in¬
ductive observation of repeated instances of sunrise; TG]: and this
means that we do not know whether it will rise..." (1918, p. 143)

As we shall see in Chapter 8, a systematic justification of inductive

reasoning can indeed be achieved, and it is neither psychological in
Wittgenstein's presumed sense, nor is it logical in the only other sense he
was willing to consider at the time. It is an open question whether
Wittgenstein's psychological was intended to correspond to the silent
pragmatic partner of inductive inference, hypothetical reasoning or ab¬
duction. Certainly, if such correspondence was intended, the term
psychological, in the context of the Tractatus, suggests lowly status, an en¬
tity not worthy of serious philosophical attention.

7.2.2.3. The limits of deduction

Having disposed of induction, Wittgenstein proceeds to expose the

severe limits of deductive logic as a means of transacting new informa¬
tion. The propositions of logic, he observes, yield either tautologies or
contradictions:

1 We return to the same argument in Chapter 8, in the context of Popper's deductivism in

the philosophy of science (Popper, 1959).
240 Mind, Code and Context

"...The propositions of logic are tautologies. Therefore the proposi¬

tions of logic say nothing. (They are the analytic propositions)..." (1918,
p. 121)

"...Tautologies and contradictions show that they say nothing. A

tautology has no truth conditions, since it is unconditionally true; and a
contradiction is true on no condition. Tautologies and contradictions
lack sense..." (1918, p. 69)

Wittgenstein now zeroes in on the use of propositions to convey infor¬

mation. The information-theory context within which he is working is
obviously the Positivist -- Russellian -- one, whereby information is
about objective reality:

"... Tautologies and contradictions are not pictures of reality. They do

not represent any possible situations. For the former admits all possible
situations, and the latter none...” (1918, p. 69)

Wittgenstein next ties up the discussion of tautology and contradiction

to the scale of three epistemic modalities (certain, possible, impossible),
observing that the information-bearing proposition stands at neither ex¬
treme of the scale:

"...a tautology's truth is certain, a proposition's possible, a contra¬

diction's impossible. Certain, possible, impossible: here we have the
first indication of the scale that we need in the theory of probability..."
(1918, pp. 69-71; emphasis is mine; TG)

While Wittgenstein's major preoccupation here is not information

theory but rather probability, the two are intimately connected, as is ap¬
parent from his immediate return to semeiology:2

"...Tautology and contradiction are the limiting cases -- indeed the dis¬
integration — of the combination of signs..." (1918, p. 71)

Wittgenstein never translates his discussion from the logical terminol¬

ogy of tautology vs. contradiction to the information-theoretical context of
totally old vs. totally new information. Still, one may argue that his dis¬
cussion implicitly presages just that:

"...Contradiction is that common factor of propositions which no

proposition has in common with another. Tautology is the common fac-

2 Indeed, Wittgenstein proceeds later on (1918, pp. 71-73) to define the range between
tautology and contradiction in probabilistic terms.
 Modes of Knowledge and Modes of Processing 241

tor of all propositions that have nothing in common with one an¬
other...Contradiction is the outer limit of propositions; tautology is the
unsubstantial point at their center..." (1918, p. 79)

Wittgenstein's discussion may be summarized by setting up a four¬

way parallelism between propositional modalities, modes of inference,
epistemic certainty and modes of information. In doing so, one allows
for the fact that Wittgenstein collapsed together two of the Aristotelian
modes — fact and possible.3

(1) Propositional modalities and modes of inference

propositional mode of epistemic mode of

modality inference certainty information

tautology: deduction certain presupposed

proposition:
fact: induction possible (strong) R-asserted
hypothesis: abduction possible (weak) IRR-asserted
contradiction: deduction impossible NEG-asserted

Wittgenstein's collapsing of the two non-deductive Aristotelian

modalities (fact and hypothesis) is itself an interesting omission. It may
simply be the deductivist's refusal to further discriminate between
modes that attain less than 100% certainty. It certainly reflects
Wittgenstein's lack of familiarity with Peirce's work on abduction. It
may also suggest some intuition on Wittgenstein's part about induction
containing an irreducible core of abductive reasoning. We will return to
this issue further below, as well as in Chapter 8. Be that as it may, modes
of inference can be correlate to informational certainty in a manner that
recalls the discussion in Chapter 4:

(2) Modes of inference and degree of certainty:

"Deductive processing of information is associated with
maximum certainty. Inductive-abductive processing of in¬
formation is associated with lower certainty".

As we shall see further below, this correlation establishes a plausible

connection between modes of inference and modes of information
processing.

3 See Chapter 4.
242 Mind, Code and Context

7.2.3. Peirce and abduction

7.2.3.1. Abduction and hypothesis

The fact that no new knowledge can ever be obtained by deduction

has of course been known before Wittgenstein. Kant's analytic pre¬
sumably includes deduced propositions.4 Peirce begins his discussion
by observing — much like Wittgenstein decades later -- that induction,
taken by itself, does not account for the most crucial step in the process
of gaining knowledge through experience. This is the step of correlating
isolated facts, integrating them into a system of knowledge, a system
within which they find their significance.
Peirce begins his discussion by probing into the early stages of
hypothesis formation:

"...The first starting of a hypothesis and the entertaining of it, whether

as a simple interrogation or with any degree of confidence, is an inferen¬
tial step which I propose to call abduction... This will include a pre¬
ference for any one hypothesis over others which would equally well ex¬
plain the fads, so long as this preference is not based upon any previous
knowledge bearing upon the truth of the hypotheses [hereby ruling out
deduction; TG], nor on any testing of any of the hypotheses, after having
admitted them on probation [hereby ruling out induction; TG]. I call all
such inference by the peculiar name, abduction, because its legitimacy
depends upon altogether different principles from those of other kinds
of inference..." (1940, p. 151; emphases are mine; TG)

Peirce then goes on to observe that the hypothesis referred to here is

not simply a guess that some individual facts ('propositions') are true,
but rather a hypothesized explanation of new facts by some other,
presumably known, fact(s). In other words, abduction is the necessary
inferential reasoning step in explanation:

"...Long before I first classed abduction as an inference it was recog¬

nized by logicians that the operation of adopting an explanatory
hypothesis -- which is just what abduction is -- was subject to certain
conditions. Namely, the hypothesis cannot be admitted, even as a
hypothesis, unless it be supposed that it would account for the facts or
some of them..." (1940, p. 151; emphasis is mine; TG)

4 See Chapters 1,4, above.

 Modes of Knowledge and Modes of Processing 243

Peirce now sets up the abstract general form of abductive inference:

"...The surprising fact C is observed;

But if A were true, C would be a matter of course;
Hence, there is a reason to suspect that A is true..."
(1940, p. 151)

The accretion of knowledge, which necessarily demands abduction, is

thus not a simple addition of isolated facts to an unstructured bin of
other known facts. Rather, it involves correlating and integrating those
facts into a more general description, i.e. in a wider context. The first
pragmatic element of abduction is, therefore, a hypothesis concerning
the existence -- and relevance -- of such wider context, within which
atomic facts may be explained, receive their meaning, make sense.
Peirce next goes on to outline the intimate connection between abduc¬
tion and induction. The connection is seen from three separate perspec¬
tives, which we will discuss in order.

(a) The use of induction in hypothesis testing:

"...The operation of testing a hypothesis by experiment,

which consists in remarking that, if it is true, observations
made under certain conditions ought to have certain results,
and then causing those conditions to be fulfilled, and noting
the results, and, if they are favorable, extending a certain
confidence to the hypothesis, I call induction..." (1940, p. 152)

(b) The residual uncertainty of induction:

"...All induction whatever may be regarded as the inference

that throughout a whole class a ratio will have about the
same value that it has in a random sample of the class,
provided the nature of the ratio for which the sample is to be
examined is specified (or virtually specified) in advance of
the examination..." (1940, p. 152)

In arguing from specific cases (partial samples) to general rule (the en¬
tire population), one can only extend a certain confidence to a hypothesis,
but never deductively prove it. The decision when a certain confidence is
good enough is inherently pragmatic, having to do with how much un¬
certainty one is willing to live with, what the consequences of being
found wrong are, etc. This is the level of confidence dismissed by the
early Wittgenstein as psychological (rather than logical).
244 Mind, Code and Context

(c) The abductive residue of imprecise induction:

Peirce observes that not all inductions are made in a clean ex¬
perimental environment, where units and their ratios can be un¬
ambiguously counted. In everyday life, many inductions involve
estimating the significance of observed characteristics, which is
an abductive judgement concerning the context within which
the observed characteristic is either more or less significant:

"...So long as the class sampled consists of units, and the ratio in
question is a ratio between counts of occurrences, induction is
comparatively a simple affair. But suppose we wish to test the
hypothesis that a man is a Catholic priest, that is, has all the
characters that are common to Catholic priests and peculiar to
them. Now characters are not units, nor do they consist of units,
nor can they be counted, in such a sense that one count is right
and the other wrong. Characters have to be estimated according
to their significance. The consequence is that there will be a cer¬
tain element of guess-work in such an induction; so that I call it
an abductory induction..." (1940, p. 152)

Peirce's argument above is rather straight-forward: Since observed

properties are often non-Platonic -- non-discrete, scalar, involving simi¬
larity judgements -- induction often retains certain abductive ('prag¬
matic') ingredients.

7.2.3.2. Abduction, causation and teleology

The assignment of cause is simply one sub-type of explanation. A causal

relation between facts can never be deduced, nor can it be induced; al¬
though certainly the process of induction -- see Peirce's description, above -
- plays an important role in determining the degree of certainty we attach
to hypotheses about causation. The reason why assigning causation invol¬
ves abductive ('pragmatic') judgement is because, like other species of ex¬
planation, causation is a relationship between some facts and their context.
One sub-species of causality involves the functional context of facts, as¬
sociated with biological and social organisms. We will return to discuss
both causal and functional explanation in Chapter 8.

7.2.3.3. Abduction, similarity and relevance

We noted in chapter 2 that a major principle via which natural --

prototype-like -- categories get organized is that of similarity. We noted,
further, that judgements of similarity -- or analogy -- cannot be made on
either deductive or inductive grounds. This is so because any judgement
 Modes of Knowledge and Modes of Processing 245

of similarity requires a hypothetical leap to determine the appropriate

context within which two entities are said to be similar. And the deter¬
mination of such propriety — or relevance -- is in principle abductive.

7.2.4. The philosophical tradition and information processing:

An interim summary

We are now in the position to summarize the discussion of the three

modes of knowledge ('modes of inference') within the philosophical tradi¬
tion, in a way that would clarify the relevance of this tradition to informa¬
tion processing. This interim summary requires the collapsing together of
two of the traditional modes -- induction and abduction; we then contrast
two extreme poles of information processing: deductive vs. non-deduc-
tive. We have noted ample precedent (viz Peirce and Wittgenstein) for
lumping together induction and abduction. One may still propose that
atomic bits of factual knowledge may be accumulated without recourse to
abduction.5 However, such atomic data cannot be integrated into a general
system of knowledge, cannot display any coherence vis-a-vis a body of
general knowledge, without resorting to abduction. Coherence and in¬
tegration are in principle contextual, pragmatic notions. As noted in Chap¬
ter 4, the very notion of information is context-bound. Information accrues
on a certain background, vis-a-vis which it receives its coherence.
Deductive and pragmatic information processing may be now con¬
trasted as follows:

(3) Deductive vs. pragmatic information processing

feature deductive mode pragmatic mode

categories: rigid, Platonic flexible. Prototype

space: discrete, continuum,

uni-dimensional multi-dimensional

context- context-free context-sensitive

dependence:

informational high low

certainty:

automation: algorithmic non-algorithmic

5 Whether such 'bits' of knowledge are at all meaningful remains in doubt. Their appropriate
filing, in the right contextual slots of the existing knowledge-base, dearly requires abductive
judgement. We will return to discuss this, within the context of philosophy of sdence, in
Chapter 8.
246 Mind, Code and Context

7.3, The linguistic tradition: Grammar

and modes of information processing

7.3.1. Grammatical vs. pre-grammatical speech

Quite independently of the philosophical tradition, the existence of

two extreme modes of organization of natural discourse has been ob¬
served, in the context of studying the role of grammatical clues -- gram¬
matical morphology and syntactic structures — in language processing.
It was first noted that there exists one communicative mode that seems
to dispense altogether with these grammatical devices. That mode can
be found most clearly in the early stages of language acquisition, both
first and second. It had been referred to as the simplified. Pidgin or
pre-grammatical mode. In addition to early stages of language acquisi¬
tion, one may also detect some of the features of this mode in informal,
colloquial, unplanned speech genres. These genres contrast, in this
respect, with more formal, planned, written genres, which tend — on a
scale -- to display more reliance on grammar.6
In order to impart, informally, the flavor of the difference between the
grammatical and pre-grammatical linguistic modes, consider the fol¬
lowing three simulated passages, all telling the same story. The first tells
it in the top-of-the-scale grammatical mode; the second renders it in the
mid-scale, conversational-informal mode; the third imparts it in the
rock-bottom, pre-grammatical Pidgin mode.

(3) Grammatical vs. pre-grammatical language processing modes

a. Fully grammaticalized ('planned') mode:

"...After we parted company I realized that the man you
told me about was the one who stopped me on my way
to work two days before and said that he thought my tie
was on backwards..."

b. Informal-colloquial ('unplanned') mode:

"...Well so we part company, right? Well so right away I
realize, I say to myself, why, that man, I say... well you
told me about him, right? Well, it was him alright, I say,
well he's the one, stopped me on my way to work, it
must've been two days ago, maybe, and he told me, he
said, I think your tie is on backwards..."

6 See initial discussion in Ochs (1979c) and Givon (1979a, ch.5); see also recapitulation in
Givon (1984b).
 Modes of Knowledge and Modes of Processing 247

c. Pre-grammatical ('pidgin') mode:

"...We go, right... time after, time after, you-me go, that
time, right... me see, say me, say, hey, man, that man,
man... you know, man, you tell me man, right... same
man, right, same man... well same man, same, stop, way
go work, man stop, me stop... say, say me, say... maybe
two day, maybe one day more, same man... man stop,
me stop, say, hey, you tie, right there, you tie man, say...
say you tie, me tie, no good on... no good on, crook, tie
crook, me tie, crook..."

As one can see, the pre-grammatical Pidgin mode involves the use of al¬
most only lexical words with no morphology, few pronouns, no com¬
plex syntactic constructions, much repetition and pausing. The message
eventually comes across, halting, fractured, rudimentary. The gradation
implied in having an intermediate step between grammatical and pre-
grammatical modes arises from the fact that grammar has many fea¬
tures of morphology and syntax, and seemingly intermediate speech
modes may use — on a graduated scale -- some of those features but not
others, in many specific combinations or overall densities.

7.3.2. Pre-grammatical rules

It would be somewhat misleading to say that the Pidgin mode follows

no grammatical rules. Rather, the rules it follows are probably the most
basic, universal, ontogenetically and phylogenetically older and, above
all, the most iconic in human grammar.7 For example, in the grammar of
referent tracking,8 the Pidgin mode adheres rather closely to both the
quantity principle and word-order principle correlating with referen¬
tial predictability and thematic importance, discussed earlier:9

(5) Referential predictability and code-quantity

most predictable referent

zero anaphora
pronoun
full noun
repeated full noun_
least predictable referent

7 See Chapter 3.
8 See Chapter 6.
9 See Chapter 6. For early second-language Pidgins see also Givon (1984b).
248 Mind, Code and Context

(6) Referential predictability, importance, and word-order:

"Less predictable or more important referents are pre-posed".

7.3.3. Grammar and automated processing

The pre-grammatical mode of language processing may be contrasted

with the grammatical ('syntactic') mode in a number of formal as well
as substantive properties:

(7) Grammatical vs. pre-grammatical linguistic modes:

grammatical pre-grammatical
properties mode mode

grammatical abundant absent

morphology

syntactic complex/ simple/

constructions embedded conjoined

use of word- grammatical pragmatic

order (subj/obj) (topicality)

processing speed fast slow

processing automated/ analytic/

mode routinized attended

context dependence lower higher

informational high low

certainty

Grammatical morphology (7a) involves, most typically, articles,

gender/classifiers, case markers, pronouns/agreement, speech-act
markers, tense-aspect-modality, complementizers and subordinators.
Embedded syntactic constructions (7b) are most typically relative
clauses, verb and noun complements, subordinate adverbial clauses, or
transitivizing and de-transitivizing constructions. Grammatical word-
order (7c) most typically refers to the relative position of subject, object
and verb. One may refer to all these elements of grammar as automat¬
ization clues (7e) used in information processing via language. Process¬
ing in the absence of such clues (i.e. in the pre-grammatical mode) is
relatively slow (7d). It is more analytic, demanding more attention (7e).
When information processing is closely attended, one presumably
 Modes of Knowledge and Modes of Processing 249

monitors more carefully the fine shades and gradations of the context
(7f). One may thus say that the grammatical mode is relatively more
feedback-free, while the pre-grammatical mode is relatively more feed-
back-dependent (7f). Finally, the more automated, grammatical mode is
used in informational contexts of high certainty, where the applicability
of the linguistic code ('rules') is less ambiguous (7g).

7.3.4. Routinized processing and context

One must note that biological -- and thus linguistic — automated

processing is not entirely context free. Rather, the dependence of
automated processing upon contextual scanning is vastly reduced and
routinized. Rather than scanning the entire contextual range for shades,
degrees and fine detail, the scanner is programmed for only a few, major
processing clues. Such clues are highly distinct, salient, discrete,
categorial. They are located at fixed points of the contextual range. The
scanner is sensitive to those ranges only, disregarding the rest of the
range. It is also particularly sensitive to the coding frequencies
('modalities') of those clues, disregarding all other frequencies.

7.3.5. Interim summary

In spite of the difference in tradition and source, one can note that cer¬
tain broad similarities exist between the philosophical distinction of
deductive vs. pragmatic inference, and the linguistic contrast of gram¬
matical vs. pre-grammatical information processing. These similarities
can be summarized as in (8) below:

(8) Processing modes: Interim summary

deductive/automated pragmatic/analytic
processing_processing

a. fast slow

b. algorithmic/automated non-algorithmic/analytic

c. depends on rigid categories/ dispenses with rigid

structures categories/structures

d. context-free context- dependent

e. high certainty low certainty

 250 Mind, Code and Context

In the following section we will see how a substantially similar distinc¬

tion emerges from the empirical study of information processing in per¬
ception, cognition, motor behavior and neurology.

7.4. The psychological tradition: Attended vs. automated

processing

7.4.1. Preamble

DOONESBURY
By Garry Trudeai.

$// OOOPMOmNO.
/ SIR. HERFS WJ
I USTOF PHOTO
I OPSKRTOUTO
s CW0S6 FROM ,

DOONESBURY © 1987 G. B. Trudeau. Reprinted with permission of Universal Press Syndicate.

All rights reserved.

A general summary of the contrast between attended and automated

processing, a contrast that crops up in widely differing performance
domains of perception, cognition, motor control and neurology, is given
in Schneider (1985):

"...Consider, for example, the changes that occur while learning to

type. At first, effort and attention are devoted to the smaller movement
or minor decision, and performance is slow and error-prone. After ex¬
tensive training, long sequences of movements or cognitive processes
are carried out with little attention... Controlled processing is charac¬
terized as a slow, generally serial, effortful, capacity-limited, subject-
controlled processing mode that must be used to deal with novel or in¬
consistent information... Automatic processing is a fast, parallel, fairly
effortless process that is not limited by short-term memory capacity, is
not under direct subject control, and performs well-developed skilled
behavior..." (pp. 475-476; emphases are mine; TG)
 Modes of Knowledge and Modes of Processing 251

In this section we will briefly survey the literature on this contrast in

cognitive psychology, kinesiology and neurology.

7.4.2. Cognitive psychology: Perception, Memory and attention

7.4.2.1. Consciousness vs. automation

In their review on the role of attention (or 'consciousness') in percep¬

tion, memory and retrieval, Posner and Warren (1972)10 recapitulate the
experimental psychological literature that led to positing the contrast
between automated encoding and conscious processing. They write:

"...How can we study automatic processes as distinguished from

those that involve conscious search? Even the definition of "automatic"
is a difficult matter... Our effort to analyze this problem has involve the
notion of a single limited-capacity central processing system that in¬
tegrates signals from all modalities. When this system is occupied by
any signal, its capacity is reduced for dealing with any other signals or
mental operations that require its use... Many complex mental opera¬
tions that are learned and that require time can be performed outside
this system... For our purpose, the use of this system becomes the
central definition of a "conscious process", and its non-use defines what
is meant by "automatic"..." (1972, p. 34)

The following general features seem to distinguish automated from

conscious/attended processing:

(9) Properties of automated processing:

(a) Informational certainty:

Automate processing is associated with repeated,
rehearsed tasks, or with expert skills acquired through
habitual performance. It thus involves the processing of
highly routinized, conventionalized activities. In terms
of information theory, the automation of processing oc¬
curs when information is more predictable, certain,
consistent.

10 Other studies involved in motivating and elucidating the distinction between

automated/unattended and conscious/attended processing in experimental psychology
include: Attneave (1957), Bartlett (1932), Broadbent (1958), Frost (1971 a, b), Hawkins (1969),
Hintzman (1970), Keele (1969, 1972), Posner (1969), Posner and Boies (1971), Posner and
Keele (1968), Posner and Klein (1971), Posner and Marin (eds, 1985), Posner and Snyder
(1974), Schneider (1985), Schneider and Fisk (1983), Schneider and Shiffrin (1977), inter alia.
252 Mind, Code and Context

(b) Processing speed and error:

Automated processing is more rapid and error free.
Conscious processing is slow and more error-prone.

(c) Memory file:

Automated processing tends to be associated with short
term memory. Conscious processing tends to be as¬
sociated with long term memory.

(d) Categories and structures:

Automated processing relies heavily on the presence of
rigid structures or schemata, thus implicitly on rigid,
discrete categories.

(e) Limiting capacity:

Several automated tasks can be processed in parallel. In
contrast, conscious processing seems to be linear, rely¬
ing on a limiting capacity that occupies the central
processor for a particular slot of time.

Concerning the critical role of certainty or conventionalization (9a),

Posner and Warren (1972) cite studies showing that memory is distorted
systematically toward habituated, conventionalized knowledge:

"...memory is often distorted in the direction of convention... Frost

(1971a) found recognition errors to be in the direction of orientation
most frequent or conventional... (p. 30; emphases are mine; TG)

1 A.2.2. Levels of consciousness and automaticity

The discussion thus far distinguished only between two levels of

processing, attended ('conscious') vs. automated. Several recent studies
suggest that this extreme dichotomy must be revised, perhaps yielding
a scale of degree of conscious processing, thus also it converse, degree
of automaticity. Thus, for example, Posner (1985) cites experimental
evidence that points toward the existence of more than one level of
visual attention. The modality-independent highest ('central') level may
simply occupy the very top of a hierarchic network of progressively
less conscious, less central and more task-specific levels of effortful pro¬
cessing.
In another recent study by Nissen and Bullemer (1986),11 the perfor¬
mance of aphasic and normal subjects in memorizing random vs. struc-

11 See also Cohen and Squire (1980), Squire and Cohen (1984), Mishkin et al (1984), inter alia.
 Modes of Knowledge and Modes of Processing 253

tured digit sequences was compared. The results suggested a similar

hierarchic array of levels of attention. Learning tasks that clearly in¬
volved automation nonetheless required some type of attention; this
was shown by the fact that such learning tasks were interfered with by
attention-demanding secondary probe tasks. Further, the level of atten¬
tion that was involved in these routinized-learning tasks was shown to
be distinct from the 'highest' level of conscious attention. Aphasic sub¬
jects, who could neither remember learning the primary task nor con¬
sciously report its regularities, nevertheless displayed — qualitatively --
the same repetition-induced learning curve as normal subjects. We will
return to this issue further below.

7.4.3. Kinesiology and motor skills

A very similar picture emerges in the study of the acquisition of

motor skills. In reviewing the literature within the context of schema
theory, Shapiro and Schmidt (1980)12 note the role of repetition in
facilitating the acquisition of motor skills. The routinization of motor
skills involves, accordingly, the creation of rigid, complex generalized
schemata that guide performance along habituated pathways. Skilled,
routinized ('schematized') performance is less conscious and non-
analytic; it proceeds faster, and allows parallel processing of other
tasks.
An interesting parallel with the development of language is the exist¬
ence of some critical period in the development of motor schemata:

"...schema development occurs much more easily in children than it

does in adults..." (Shapiro & Schmidt, 1980, p. 17 of ms).

It is certainly true that the acquisition of grammar proceeds faster and

more completely in children than in adults. There is, thus, some critical pe¬
riod phenomenon in the acquisition of routinized language processing.13
Shapiro and Schmidt also note that some components in the acquisi¬
tion of motor skills, specifically the learning of novel tasks, are not
schema-governed. Rather, they require consciousness, analysis and
decision:

12 See there for a general review of the literature; see also discussion in Denier and Thuring
(1965), Grillner (1975), Herman etal (eds, 1976), Schmidt (1975), Shapiro etal (1980), Whiting
et al (1980), inter alia. The literature on the routinization ('automation') of motor behavior
covers a wide range of acquired physical skills, including locomotion, piano playing,
dance, writing and typing, athletics, etc. It also covers a large range of human, mammal,
vertebrate and invertebrate subjects.
13 See for example Ascher and Garcia (1969), Curtiss (1977), Ervin-Tripp (1974), Krashen
. (1972), Krashen et al (1979), inter alia.
254 Mind, Code and Context

"...the learning in the motor-learning experiment can be regarded as

the discovery of the size of the value to be added to the parameters [of
the schema; TG] in order to make the program "work" in this particular
situation. That learning should, however, be situation-specific as opposed
to the schema learning which should be general across all uses of the
generalized motor program..." (1980, p. 45 of ms).

The issue of conscious ('feedback-dependent') vs. programmed

('feedback-free') learning, is also tackled in two studies by Shapiro
(1977, 1978). Schmidt (1980) summarizes those as follows:

"...There is some evidence that if the movement is very long in dura¬

tion (i.e. 1-2 sec), one could control the action with a program set up in
advance, or alternatively by processing the feedback from each sub-action
and making corrective steps..." (p. 125; emphases are mine; TG)

7.4.4. Neurological aspects of automated processing

7.4 4.1. Routinization learning

In a recent review, Schneider (1985) discusses the neurological coding

correlates of the contrast between attended ('controlled') and automated
processing. He first describes the neurology of information transmis¬
sion:

"...Cortical information transmission occurs when a population of

neurons (e.g. a hyper-column) sends a set of firing rates (e.g. a vector of
activation) to another population. This set of the firing rates of the out¬
put neurons can be modulated as a set... (Szentagothai, 1977)... Learning
in the physical system occurs after this set of firing rates comes into a
population and a second burst is output (Levy and Steward, 1983)..." (p.
477; emphases are mine; TG)

Schneider next outlines an in formation-flow model that integrates the

neurological and the performance findings:

"...Communication theory provides optimality considerations regard¬

ing how best to allocate transmission time in a network of vector trans¬
mission units (see van der Meulen, 1977). Communication theory
theorems indicate that if the brain optimally processes information,
there should be two modes of transmission: a serial, time-sharing, con¬
trol-process-type mode, and a parallel, automatic-process-type mode..."
(1985, p. 477; emphases are mine; TG)
 Modes of Knowledge and Modes of Processing 255

A formal model is now described:

"...In the model, controlled processing is conceived of as a limited

central-processing mechanism that gates the transmission of messages be¬
tween units and compares the received messages. The development of
automatic processing is the result of two types of learning. The first, as¬
sociative learning, is the mechanism by which one message is associative-
ly translated to another message. The second type of learning, priority
learning, is the mechanism by which a unit determines how strongly to
transmit a message. The unit-specific message priority determines the
strength of the automatic message transmission. Automatic processing
occurs when priority and associative learning are sufficiently advanced
to allow a sequence of transmissions without any controlled-process
gating the information..." (1985, p. 477)

If one may attempt a paraphrase, the gating function of consciousness

is a higher traffic-control function; it allocates information-flow
priorities at branching nodes where ambiguities — i.e. potential flow-
conflicts — are possible. The associative component of routinization is
presumably the process by which repetition of 'similar' information in a
'similar' context creates an association between the information and its
context (say, in the most minimal term, the preceding or following
chunks of information). Finally, the priority component of routinization
is an automatized replacement of the 'gating7 function of the erstwhile
conscious processing. Once processing is automated, all the competing
branches at a (potentially-ambiguous) node are assigned fixed firing
strengths, in effect thus ranking them in terms of relative priority, per¬
haps also in fixed transmission order.
The learning process that gives rise to routinized processing presumab¬
ly thus involves the creation of firmed-up, relatively rigid neural path¬
ways. They are created through experience during the lifetime of the or¬
ganism, and they may persist for the duration of that lifetime.

7.4.4 2. Genetically determined routinization

There are some indications that some of the routinization that occurs
during the lifetime of the organism is not wholly experience-triggered,
but at least in part wired in, thus presumably triggered automatically
by the process of neurological maturation. Thus, reflexive, uncontrolled,
clearly routinized early ambulation movement of human neonates is
reported14 to first display a four-legged movement pattern, long before
any crawling experience has taken place. Later on, a pattern of bi-pedal
movement is manifest, again prior to any actual walking experience.

14 See Thelen (1984), also Fossberg's (1985) interpretation of Thclen's findings. Similar find¬
ings in other mammals are summarized in Smith (1980).
256 Mind, Code and Context

One must note that information flow in many major lower physiologi¬
cal functions of organisms, including the control of the digestive,
eliminatory, cardio-vascular and pulmonary functions, as well as in
many reflexive sensory and motor functions, is automated from birth, i.e.
genetically pre-programmed.15 A major feature of such reflexive systems
is that the tasks they perform are highly repetitive, predictable; thus,
the degree of informational certainty in these processing systems is very
high. It is reasonable to assume that such automation of processing --
which I propose to call phylogenetic learning — evolved biologically
under the same adaptive constraint that motivate routinization-learning
during the life-time of the organism -- i.e. ontogenetic learning. More
specifically, one may suggest that:16

(a) The general correlation between automation and repetition

('predictability') of information holds for both the onto¬
genetic and phylogenetic routinization-learning; and

(b) There is an obvious -- tough yet to be elucidated -- correla¬

tion between the ontogenetic and phylogenetic processes of
routinization-learning.

We will return to discuss these suggestions in section 7.5.3., as well as in

Chapter 10.

7.5. The rise and function of automated processing

7.5.1. Properties of automated processing: Interim summary

As we have seen in the preceding sections, a number of identifiable

common threads emerge from the comparative study of the automation
of processing in language, cognition, perception, motor behavior and
neurology. One may summarize these common threads as follows:

(9) Properties of automated processing:

(a) Conscious control: Automated processing is less moni¬

tored, thus less controlled, by analytic consciousness.
(b) Channel properties: Automated processing allows non¬
interfering, multi-channel, parallel processing.

15 See more recent summaries in Smith (1980), Grillner (1975), Gurfinkel & Shik (1977), Stein
(1978), as well as older reviews in PaiUard (1960) or Galambov & Morgan (1960), inter alia.
16 Mishkin et al (1984) raise a superficially similar suggestion that turns, on close inspection,
to presuppose a very different evolutionary mechanism. See further discussion in section
7.5.3., below.
 Modes of Knowledge and Modes of Processing 257

(c) Modal specificity: Automated processing tends to in¬

volve more specialized, modality-specific sub-systems.
(d) Processing speed: Automated processing is faster.
(e) Categorial structure: Automated processing subsumes
the development of rigid, discrete categories, as well as
structures (i.e. configurations of categories), which
presumably act as the automated processing clues.
(f) Context dependence: Automated processing is less con¬
text dependent; it is thus, relatively, a feedback-inde¬
pendent, non-monitoring mode of information pro¬
cessing.
(g) Informational predictability: Automated processing
arises under conditions of repetition, habituation, con¬
ventionality, routine; it thus involves higher informa¬
tional predictability.

7.5.2. The functional motivation for automated processing

7.5.2.1. The adaptive tradeoff

The functional context within which automated processing develops,

either ontogenetically or phylogenetically, can be described as an adap¬
tive tradeoff between three of the properties listed in (9) above:
(a) conscious control
(d) processing speed; and
(g) informational predictability
The other dimensions, and their mode of involvement, are to quite an
extent predictable within the context of this tradeoff:

(i) Automated processing:

Repeated, predictable tasks are the only ones that can be

routinized. Routinization is the means for increasing processing
speed. Structurally, routinization is achieved via the buildup of
discrete, categorial structures that act as a network of
automated flow-control cues. Such automation procedure may
be also called an algorithm. (Deductive logic is an extreme case
of the latter). When central, conscious monitoring is not re¬
quired, parallel processing of several automated tasks, each of
them modal-specific, can take place. Parallel processing is
another contributor to increasing processing speed, i.e. an in¬
crease in the economy of the overall processing system.
258 Mind, Code and Context

(ii) Attended processing:

Conversely, attended processing is reserved for tasks that are

novel, less repetitive, less predictable. Such tasks require
analytic monitoring of shades and gradations of the context.
Rigid categorial structure cannot be built up in such a
processing system, since it would sacrifice context-depend¬
ent flexibility. Processing speed cannot be built up either,
presumably because it would create an unacceptable in¬
crease in error in monitoring the context. Detailed, non-dis¬
crete monitoring is necessarily slow.

In sum, the adaptive tradeoff is between conscious, detailed contextual

scan and processing speed. The controlling feature is the degree of pre¬
dictability of the task.

7.5.2.2. Task urgency, predictability and economy

One may describe the pivotal role of predictability in determining

processing tasks that will eventually be automated as a manifestation of
a biological principle of economy. Routinization itself incurs certain
costs, associated with the building of the requisite rigid structures.
These costs are not justified unless the processing system is used rela¬
tively frequently. But under certain adaptive condition this principle of
economy can be, and often is, suspended. Thus, for example, hair-trig¬
ger response to potential predators is presumably an automated, speedy
process, regardless of the level informational certainty. In processing
such urgent survival tasks, the organism presumably biases the
automated cues toward a greater margin of safety, while sacrificing
economy. When survival is less of an issue, normal economy considera¬
tions -- manifest in the routinization of frequent, repetitive, predictable
tasks -- presumably govern more directly the tradeoff between speed
and consciousness.

7.5.2.3. The automaticity continuum

Whenever the processing of some repeated, predictable tasks is

automated, consciousness retains higher governing functions: Overall
flow control, context-dependent monitoring, subtle contingent choices,
goal-governed priority assignment. As suggested earlier, however, the
contrast between conscious and automated processing is not a single
discrete division, but rather a hierarchic, multi-level, scale. The alloca¬
tion of monitoring consciousness works through this hierarchy in a bot¬
tom-up fashion. That is, the highest level of attention is assigned to the
 Modes of Knowledge and Modes of Processing 259

highest node in the hierarchy. Such a node governs tasks that are least
automated, being least predictable. Relatively lower levels of attention
are assigned to more predictable, thus more automated, task-nodes.
Schneider's (1985) gradualistic model of the process of routinization-
learning is, I believe, rather compatible with this hierarchic, bottom-up
model.17

7.5.3. Evolutionary aspects of automated processing

7.5.3.1. Rise of consciousness vs. rise of automation

As suggested earlier above, both the ontogenetic and comparative-

distributional facts of automated vs. attended processes fairly beg for an
integrated behavioral-evolutionary-developmental approach. A stab at
this was made by Mishkin et al (1984), suggesting that automated
processing is phylogenetically prior to consciousness, i.e. that the rise of
conscious, attended processing is a later evolutionary step. There are a
number of reasons why I think this proposal is rather unattractive:

(a) Such an approach implies that information processing in lower

organisms is not subject to the same adaptive constraints that
seem to govern the rise of automated processing elsewhere (i.e.
the adaptive tradeoff bet ween consciousness, predictability and
speed). This would create an evolutionary discontinuity be¬
tween information processing in pre-conscious vs. conscious or¬
ganisms, a rather implausible state of affairs.18

(b) Such an approach would dissociated the ontogenesis of

routinization-leaming (i.e. automation during the lifetime of in¬
dividual organisms) from the evolutionary process. The latter
is described by Mishkin et al (1984) as proceeding in exactly the
opposite direction of the former. Given the high degree of shared
directionality between ontogenesis and phylogenesis in
general, perhaps even downright partial recapitulation,19 such
dissociation is highly unlikely.

17 See in particular Schneider's (1985) discussion of optimization. In its general outline, this
from-the-bottom-up formulation is compatible with the one suggested by Broadbent
(1977), whereby the bottom-up development of automatidty is said to involve the gradual
adjustment of parameters to the values that were first determined by conscious monitoring.
McLeod et al (1985) envision a similar gradual process, which they call encapsulation of
information.
18 See discussion in Chapter 10.
19 See Gould (1977), as well as Chapter 10.
260 Mind, Code and Context

7.5.3.2. Pre-conditions for routinization:

Taxonomy and frequency of experience

An alternative approach to the evolution of automaticity would insist

that the earliest mode of information processing at whatever level is --
and has always been — the attended, conscious mode. This is so because
at the very beginning of the organization — or processing -- of ex¬
perience, all input is by definition novel (with the obvious exception of
genetically-coded automated processes that are themselves the product
of earlier phylogenetic evolution). Repetition of experience in turn in¬
stigates in the organism two separate but related processes, the first an
absolute pre-condition for the second, and both pre-conditions for
routinization:

(a) Taxonomic categorization:

The organism must classify, by whatever means, individual
data ('instances') of experience as to whether they will count
as different tokens of the same type, or tokens of a different
type, of experience. As noted in Chapters 1, 2, such tax¬
onomic decisions are made, in principle, on pragmatic, con¬
text-dependent grounds.

(b) Frequency determination:

Once individual tokens of experience are classified into
types, it is possible to determine which types of experience
are more repetitive, predictable, thus candidates for
routinization, and which ones are novel and unpredictable,
thus to be processed under conscious control.

A taxonomy of experience (a) and a frequency determination applied to

the resulting taxa (b) are two key preconditions for the automation of in¬
formation processing.

7.5.3.3. Bottom-up routinization

As the taxonomic-hierarchic organization of repeated experience pro¬

ceeds gradually from the bottom up, so does routinization. In this
fashion, today's higher node that require conscious monitoring may in
time becomes tomorrow's repeating, predictable, newly-routinized node
- but a node at a higher level, representing more complex experience.
The role of conscious monitoring is, in this fashion, pushed upward
progressively. As this upward move proceeds, the nature of consciousness
— thus of mind — becomes increasingly more complex. Yet all along, con¬
sciousness retains essentially the same functions:
 Modes of Knowledge and Modes of Processing 261

(a) Monitoring experience types that are least repetitive, least

predictable, least categorizable;
(b) monitoring shades and gradations of context;
(c) monitoring the higher controlling nodes ('gates') of paral¬
lel-processed information, whose bulk is handled more rou¬
tinely at lower, task-specific (or 'modality-specific') nodes;
and thus
(d) allocating purpose-mediated priorities among lower pro¬
cessing channels.

Hierarchically-lower levels of attention, however many of those there

might be,20 perform the same monitoring functions downward.

7.5.3.4. The biological unity of automation

The bottom-up developmental model should be equally valid for

describing the rise of automated processing at all three levels of its
development:

(a) Routinized-learning during the life-time of individuals;

(b) The ontogenetic sequence routinization during the em¬
bryo's maturation;
(c) The phylogenetic evolution of routinized processing.

The tantalizing possibility of a causal interaction among these three

levels remained a key issue of evolutionary thinking. First, a possible in¬
teraction between levels (a) and (c) is at the crux of the old Lamarckian
argument. Second, a connection between (b) and (c) is at the crux of
recapitulationism. We will return to both issues in chapter 10, below.

7.5.3.5. Linguistic parallelisms

It is, lastly, of interest to note that the parallelism between the three
biological levels -- individual behavior, ontogenesis and phylogenesis --
is also suggested for the development of grammatical processing in
human language. It has been suggested21 that the process of gram-
maticalization, via which grammatical morphology and syntactic con¬
structions evolve, follows substantially the same route in four domains:
(a) language behavior by individuals
(b) historical change of the communal language
(c) the child's acquisition of grammar
(c) the evolution of grammar in the human species

20 See section 7.4.2.2., above, as well as footnote 17.

21 See Givon (1979a, chs 5,7) as well as Bickerton (forthcoming).
262 Mind, Code and Context

If this parallelism is valid, it may reinforce the view that the develop¬
ment of grammar, as a mode of information processing, is another in¬
stance of the bottom-up shift from attended to automated processing.

7,6. Some experimental evidence on grammar as

an automated language processing mode
% ■ »

We return now to discuss grammar as an automated system of lan¬

guage processing, as outlined in section 7.3., above. The validity of this
view of grammar is supported strongly, if indirectly, by the striking
parallelism between the parameters that govern grammaticalized vs.
pre-grammatical speech, on the one hand, and automated vs. attended
processing, on the other. In this section we will survey the results of a
series of experiments dealing with referential tracking in discourse,
more specifically with the use of anaphoric pronouns vs. definite nouns.
As noted in Chapter 6, anaphoric pronouns (and zero anaphora) are
used in discourse contexts of high predictability. In contrast, definite
nouns are used in discourse context of lower predictability. We may il¬
lustrate this by reproducing here the text-distribution of the referential
distance measure (RD; the gap between the last previous mention of the
antecedent in discourse) for anaphoric pronoun and definite nouns:22

(10) Distribution of referential distance (RD) of anaphoric

pronouns and definite nouns in spoken English narrative
(from Givon, 1983b)

pronouns DEF-nouns
RD in # of
clauses n % n %
1-2 499 86.2 54 26.7
3-8 31 5.7 36 17.8
9-14 8 1.4 17 8.4
15-19 2 0.3 6 2.9
20+ 1 0.1 89 43.9
total: 541 100.0 202 100.0
median RD 1.0 12.0

The median RD value for pronouns is 1.0 clauses to the left. What is
more, this median value is highly stereotypical of the pronoun popula¬
tion: 86% of the population conforms to that stereotype. In contrast, the
median RD value for definite nouns is 12.0 clauses to the left; and the

22 See discussion in Ch. 6.

 Modes of Knowledge and Modes of Processing 263

population is widely scattered over the scale, with 43.9% not even
showing antecedent in the preceding discourse (RD = 20+). This is a
much less stereotyped population, so far as both distance and source of
coreference are concerned.
In a series of experiments, the effect of referential distance on the
amount of attention (or 'mental effort') expended during the processing
of anaphoric pronouns and definite nouns was investigated.23 Written
passages were presented visually on a computer screen, with the sub¬
jects releasing successive words at their own reading pace, by pressing a
lever. A secondary visual-probe task was flashed on the screen, above
the primary text, during the presentation of anaphoric pronouns and
definite nouns. The subjects were asked to react to that secondary probe
by pushing a second key. Both anaphoric expressions occurred at
referential distances of 1-clause and 12-clause from their antecedents in
the text. The probe reaction times (probe RT) and word-release times (key¬
press time) -- both in the presence of the secondary probe — were
measured. Typical text-frames used in this experiment were, for ex¬
ample:

(11) a. 1-clause RD opening:

"THE WOMAN waited patiently near the counter. The
shop was almost empty, nobody seemed to notice her."

b. 12-clause RD filler:
"[(10a) above, plus:]...It was a large store with many
departments spread along the cavernous floor space of
three main levels. The time was early in the afternoon
on a Monday in late spring, and nobody seemed to be
moving around. On weekends the store was always
packed with hoards of bargain hunters milling in the
aisles, holding sale coupons in hand, and craning their
necks in search of the special deal that was somehow al¬
ways two counters away, or so it seemed."

c. Ending:
"...SHE/THE WOMAN took out her compact and
touched up her hair."

From purely general consideration of memory decay, one would ex¬

pect a direct correlation between referential distance and measured
probe RTs. But in fact, the correlation turned out to be direct only for
pronouns, but inverse for definite nouns, as can be seen from the results,
summarized in (12) below:

23 Givon, Kellogg, Posner and Yee (1985).

264 Mind, Code and Context

(11) Probe reaction time as function of referential distance for

anaphoric pronouns and DEF-noun (from Givon, Kellogg,
Posner & Yee, 1985)

The results suggest a grammar-specific effect: Anaphoric pronouns are

processed more efficiently at the minimal referential distance (RD = 1),
the one that corresponds to their median, stereotypical value. They lose
processing efficacy -- i.e. require more attention to process — at the
higher RD. In contrast, definite nouns are processed less efficiently at
the minimal RD value, but gain processing efficacy -- i.e. require less
mental effort — at the higher RD value, the one that corresponds more
closely to their median (though not stereotypical) value (RD = 12). One
may propose that anaphoric pronouns as grammatical cues automate
the speech perceiver's reaction toward expecting a short-distance
memory search for co-reference. Definite nouns, on the other hand,
automate the perceiver's reaction toward a rather different search.
In a related experiment, the word-release ('key-press') times for
pronouns and definite nouns were compared (all in the presence of the
secondary probe), at both 1-clause and 5-clause RD values. The results
are presented below. The key-press values for the probed word
(pronoun or DEF-noun), as well as for the two words preceding and fol¬
lowing it in the text, are given in (12) on the following page.
 Modes of Knowledge and Modes of Processing 265

(12) Mean key-press times for anaphoric pronouns and DEF-

nouns at 1 and 5-clause RD (from Givon, Kellogg, Posner
and Yee, 1985)

The results for pronouns, at both RDs, reveal the presence of a process¬
ing peak ('processing slow-down') on the probed word. In contrast, for
definite nouns that processing peak was delayed until the next word
(n+1).
Since the results can be merely due to the phonological length difference
between nouns and pronouns, another experiment was designed to com¬
pare the key-press times, at 1-clause RD, for indefinite nouns and two
types of DEF-nouns: (a) anaphoric DEF-nouns (as in (11) above; i.e. with
explicit discourse antecedent); and (b) thematic DEF-nouns (i.e. ones
without explicit discourse antecedence). The latter share a feature with
indefinites: They appears in the text for the first time. To illustrate these
categories, consider the following alternative ending to the department-
store frame in (11) above, where (13) below represents the two new op¬
tional endings (in addition to (11c)):
266 Mind, Code and Context

(13) Indefinite and Thematic-definite endings:

"A/THE SALESMAN finally came over and asked her what
she was looking for".

'The Salesman' in (13) is thematically predictable from the conven¬

tionalized department-store frame ('script').24 The results of the last ex¬
periment are given in (14) below.

(14) Mean key-press times for INDEF, anaphoric-DEF and

thematic-DEF nouns at 1-clause RD (from Givon, Kellogg,
Posner and Yee, 1985)

The results may be summed up as follows:

(i) The delayed processing-peak effect observed first in (12) is not a
function of the larger phonological size of nouns, since in¬
definite nouns do not display this effect.
(ii) Definite nouns display the delayed processing-peak effect
regardless of whether their source is anaphoric or thematic.

24 See discussion in Chapter 6, above.

 Modes of Knowledge and Modes of Processing 267

The explanation tentatively suggested in the original study makes use

of the notion of grammar as an automated processing device. It runs as
follows:

(a) Pronouns are an unambiguous processing clue, instructing

the speech perceiver to search for a co-referent within a high¬
ly stereotyped memory range — 1 clause back;
(b) Indefinite nouns are equally unambiguous, instructing the
perceiver to proceed and25
(i) Open a new referent file; and
(ii) Not search memory files for coreference.
(c) In contrast, definite nouns are a highly ambiguous clue as to
the source of shared knowledge. As noted in Chapter 6,
definiteness may arise anaphorically (i.e. the presence of an
antecedent co-referent within the prior text); it may arise
from discourse-thematic knowledge (which usually dis¬
plays some generic elements); or it may arise from generic
shared knowledge. The one-word delay in the processing-
peak observed in our experiments may represent the pro¬
cessing time allotted for a decision about the source ('file') of
the definiteness of the noun. Once that determination is
made, the appropriate memory-file is then searched, result¬
ing in the characteristic slow-down effect of processing.

The experimental study of grammar as an automated language¬

processing device is still in its infancy. The results of experiments such
as the ones cited above are preliminary and tentative. The general
properties of grammatical vs. pre-grammatical speech, and the study of
the contexts under which they occur, strongly suggest that grammar is
indeed a partly automated processing system.26

25 As noted in Chapter 5, above, the grammar of indefinites presents another set of grammati¬
cal dues that automate referential processing. Those clues flag a new referent in the
discourse by a spedal marker when it demands more attention. An experimental study of
the processing of indefinites is now under way.
26 For a similar approach to grammar, with extensive neurological and evolutionary support,
see Lieberman (1984).
 *

■
CHAPTER 8

FACT, LOGIC AND METHOD:

THE PRAGMATICS OF SCIENCE

8.1. Introduction*

8.1.1. Epistemology and organized science

In this chapter we return to consider the role of the three modes of in¬
ference — deduction, induction and abduction -- in the acquisition, in¬
tegration and interpretation of new knowledge. This time, however
what is at issue is not information processing by the biological organism
(i.e. the mind), but rather information processing by the scientist. In the
course of our survey, we will make two -- essentially empirical -- obser¬
vations concerning some striking parallels. The first is epistemological:

(a) The information-processing behavior of organized science

strongly parallels that of the organism ('mind').

The second parallel, itself peculiarly isomorphic to the first, is historical:

(b) The array of traditional philosophical positions in the

philosophy of science essentially recapitulates the array of
traditional positions in epistemology.

In terms of parallel (a), abductive-contextual reasoning is the key to the

scientist's extension of aggregate cultural knowledge, much like it was
shown to be the key to the organism's ability to process new information.
In terms of parallel (b), the discrete cleavage between the two extreme
reductionist schools in epistemology -- rationalism and empiricism -- is
echoed faithfully in the split between rigid deductivism and rigid induc-
tivism, respectively, in the philosophy of science. And, in both epistemol-

*1 am indebted to Henning Andersen, T.K. Bikson, Robert Nola, Paul Otto, Frank Schroeck
and Martin Tweedale for many helpful comments on an early version of this chapter. I have
also benefitted from the opportunity to present earlier versions at the Cognitive Science
Colloquium, University of Oregon (November 1986), the Pragmatics Workshop, Roskilde
University (May 1987), and the Colloquium of the Institute for Theoretical Physics, Univer¬
sity of Koln (June 1987).
270 Mind, Code and Context

ogy and science, it appears, untenable reductionist positions resolve

towards the pragmatic middle ground; that is, through understanding
the role of abductive-contextual reasoning in information processing.

8.1.2. The legacy of reductionism

In his Logic of Scientific Discovery, Karl Popper makes the following
observation concerning the polarization of the philosophy of science, an
observation prompted in large measure by Wittgenstein's Tractatus:

"...The positivist dislikes the idea that there should be meaningful

problems outside the field of 'positive' empirical science — problems to
be dealt with by genuine philosophical theory. He dislikes the idea that
there should be a genuine theory of knowledge, an epistemology or a
methodology*...
"In the two years before the first publication of this book, it was the standing
criticism raised by members of the Vienna Circle against my idea that a theory of
method which was neither an empirical science nor pure logic was impossible: What
was outside of these two fields was sheer nonsense..."
(1959, p. 51; emphases are mine; TG)

Popper's complaint, while directed explicitly at the positivists, can be

read nonetheless as a broader objection to the reductionist view of truth
and knowledge. That view conceded legitimacy to only two of Kant's
modes of truth (analytic and synthetic), arrived at via only two modes
of inference (deduction and induction, respectively).
Reductionism, as attractive as it may seem to the purist, turns out to
be both philosophically bankrupt and methodologically unrealistic. Ex¬
tremists of both philosophical stripes, as Popper demonstrates repeated¬
ly in his own corner of the field, are quite capable of debunking each
others' folly. They do so with insight, eloquence and, above all, with
remarkable gusto. What remains to be done then, once the dust has
cleared, is to constructively identify the many points in the complex
process of empirical science where pragmatic reasoning is relevant, in¬
deed indispensible.

8.1.3. What is an apt paradigm-metaphor for behavioral and

cognitive science?
One of the most persistent legacies of earlier discussions in the
philosophy of science is a strong preference for physics as the paradigm
discipline. Given the relatively clean, formal, law-like nature of
generalizations in physics (as compared to the rather messy study of biol¬
ogy and behavior), such early preference may have been historically jus¬
tified. In large measure, however, it also bears some historical - though
not necessarily logical — responsibility for the persistence of extreme
 Fad, Logic and Method 271

reductionism in the philosophy of science. It is only in a science as

abstract and formalizable as physics that the role of non-deductive, non-
inductive methods could ever be ignored. The methods of the more
'messy' sciences, such as biology, psychology or linguistics, are much har¬
der to reduce into a clean deductive-inductive dichotomy. In such empiri¬
cal disciplines, generalizations are less law-like; they often involve com¬
plex multi-variable interactions; Taws' are almost always mediated by
protracted evolution, ontogenetic development or historical change, all of
which tend to exhibit some off-equilibrium transition phases. During
such phases, the connectivity between individual facts — even when those
facts would ultimately prove to possess strong causal links -- is less than
perfect, often less than obvious. Such evolutionary, developmental or his¬
torical process are, in turn, governed by the functional-adaptive behavior
of an entire population. And such behavior is in turn the product of the
teleology of individual organisms. Thus, while the argument for the in¬
dispensability of pragmatics as method applies to science in general, it ap¬
plies in spades to the bio-behavioral disciplines.

8.2. Reductionism in the philosophy of science

8.2.1. Theory vs. practice in science

"...I think there is a moral to this story, namely, that it is more
important to have beauty in one's equations than to have
them fit experiment..."

P.A.M. Dirac (commenting on the his¬

tory of Schrodinger's wave equations;
cited in Bach, 1965, pp. 113-114)

"...the best and safest method of philosophizing [doing

science]...seems to be, first to inquire diligently into the
properties of things, and of establishing these properties by
experiment, and then to proceed more slowly to hypotheses
[theories] for the explanation of them..."

Isaac Newton (commenting on his own

work; cited in Andrade, 1954, p. 64)

It is much easier to find clear-cut examples of either extreme deduc-

tivism or extreme inductivism in the pronouncements of philosophers
of science, than in the work of practicing scientists. The two quotations
above, from eminent physicists, represent reflective positions rather
than empirical practice. So-called naturalistic accounts of science may
yet succeed in untangling the two. Such accounts seem to emanate from
272 Mind, Code and Context

two distinct sources. First, the positivists suggested a 'naturalistic' ac¬

count, in an attempt to make the study of methodology itself empirical.
Here is how Popper (1959) views their endeavor:

"...only two kinds of statements exist for them: logical tautologies and
empirical statements. If methodology is not logic, then, they will con¬
clude, it must be a branch of some science -- the science, say, of the be¬
havior of scientists at work.." (1959, p. 52; emphases are mine; TG)

Second, Kuhn's (1962) socio-historical account is also 'naturalistic' in

Popper's sense. Popper is considerably more benign in his comments
here, at least implicitly:

"...I am quite ready to admit that there is a need for a purely logical
analysis of theories, for an analysis which takes no account of how they
change and develop. But this kind of analysis does not elucidate those
aspects of empirical science which I, for one, so highly prize..." (1959, p.
50; emphases are mine; TG).

Nola (1986) has recently pointed out that Popper himself has advo¬
cated at various times either an apriori-conventionalist (Nola's 'rule-
conventionalist')1 approach to methodology, or an empirically-based
(Popper's 'naturalistic') approach. The first is of course embedded in
Popper's (1959) central argument: Methodology is a set of conventions

1 Popper also identifies another conventionalist approach, one he ascribes variously to

Poincare and Duhem, H. Dingier, Cornelius, Ajdukiewicz and Carnap. Nola (1986) calls
this 'methodological conventionalism'. Popper opposes it vigorously as an attempt to solve
the problem of induction, an attempt to debunk his falsificationist approach to methodol¬
ogy, as well as a stratagem: "...I regard conventionalism as a system which is self-contained
and defensible. Attempts to detect inconsistencies in it are not likely to succeed. Yet in spite
of all this I find it quite unacceptable. Underlying it is an idea of science, of its aims and
purpose, which are entirely different from mine. I do not demand any final certainty from
science (and consequently do not get it); the conventionalist seeks in science 'a system of
knowledge based on ultimate grounds', to use a phrase of Dingier. This goal is attainable;
for it is possible to interpret any given scientific system as a system of implicit definitions..."
(1959, p. 80). A similar, formal conventionalist argument is attributed by Popper to Carnap,
presumably referring to Carnap's attempt to get away from the indeterminacy inherent in
Popper's falsificationist approach: "...For there is always a possibility of '...attaining, for any
chosen axiomatic system, what is called its "correspondence with reality"'..." (ibid, p. 81; the
quotation is from Carnap (1923)). Popper goes on to suggest that conventionalist
stratagems are particularly useful in times of scientific crisis, in defense of the established
'classic' system, as means of explaining away its inconsistencies: "...Whenever the 'classical'
system of the day is threatened by the results of new experiments which might be
interpreted as falsification according to my point of view, the system will appear unshaken
to the conventionalist. He will explain away the inaccuraccies which may have arisen;
perhaps by blaming our inadequate mastery of the system. Or he will eliminate them by
suggesting ad hoc the adoption of certain auxiliary hypotheses, or perhaps certain correc¬
tions to our measuring instruments..." (1959, p. 80).
 Fact, Logic and Method 273

whose status is presupposed, in advance of undertaking a scientific inves¬

tigations. The second is found in Popper's later appeal (1974) to typical
exemplars of true science', in whose work one may find the workings
of scientific methodology. As Nola (1986) notes, this is an appeal to
some intuitive yet still inherently empirical criteria for deciding what is
'true science'. Such criteria may take into account either the judgement
of history, or the collective opinion of certified scientists.
It is doubtful that either extreme deductivism or extreme inductivism
is ever practiced in a truly empirical science. Even in philosophy they
are on occasion mitigated by a measure of eclecticism.2 Still, it is of some
use to explore, however briefly, the philosophical positions of extreme
inductivists and deductivists.

8.2.2. Inductivism

8.2.2.1. Extreme inductivism in philosophy

8.2.2.1.1. Induction and the discovery of ’laws’

In the intellectual climate of the 1920s and 1930s, when Popper first
elucidated his approach to the method of science, the dominant dogma
in the philosophy of science was an extreme variant of empiricism --
logical positivism. That dogma held that empirical science, and im¬
plicitly also the organism, proceeded to acquire knowledge via the
method of induction. To cite Popper:

"...According to this view, the logic of scientific discovery would be

identical with inductive logic... It is usual to call an inference 'inductive'
if it passes from singular statements (sometime called 'particular
statements'), such as accounts of the results of observations or experi¬
ments, to universal statements, such as hypotheses or theories..." (1959, p.
27)

Popper ascribes these views to Carnap, Russell, Reichenbach and the

'early' Wittgenstein, inter alia, observing that whatever role induction
must play in the acquisition of knowledge, it could not possibly be the
role suggested by the inductivists. Scientific generalizations -- laws,
hypotheses, theories -- cannot be arrived at solely, or even primarily, via
induction. Nor can they be proven by induction.

2 As I suggest further below, there is an important pragmatic element underlying Popper's

conventionalism, one arising from his concepts of goals, aims or purposes.
274 Mind, Code and Context

8.2.2.1.2. Induction and the ’leap of faith’

In arguing against the feasibility of an inductive method in science.

Popper uses a variant of the argument of the incompleteness of facts.
He first points out, following Hume, that the so-called principle of induc¬
tion upon which inductivist methodology is founded has no logical
status whatever, it is merely a leap of faith that asserts (quoting Hume):

"...that those instances...of which we have had no experience [are like¬

ly to] resemble those of which we have had experience..." (1959, appen¬
dix *vii, p. 369)3

And further, again quoting Hume:

"...All probable arguments are built on the supposition that there is

conformity betwixt the future and the past..." (ibid)

As we have seen earlier above,4 the same observation was made by Wit¬
tgenstein in the Tractatus:

"...the procedure of induction consists in accepting as true the simplest

law that can be reconciled with our experience. This procedure, how¬
ever, has no logical justification, but only a psychological one..." (1918, p.
143; emphases are mine; TG)

A similar, even more explicit, observation was made earlier by Peirce,5

who in addition also suggested that induction is a means for raising the
investigator's confidence in a generalization:

"...The operation of testing a hypothesis by experiment, which con¬

sists in remarking that certain conditions ought to have certain results,
and then causing those conditions to be fulfilled, and noting the results,
and, if they are favorable, extending a certain confidence to the hypothesis,
I call induction..." (1940, p. 152; emphasis is mine; TG)6

8.2.2.1.3. Vindicating induction’s ’leap of faith’

There are a number of arguments that can be raised against the

surprising consensus of Wittgenstein and Popper, both of whom dis-

3 Popper is quoting here from Hume's (1739-1740) Treatise of Human Nature.

4 See Chapter 7.
5 See Chapter 7, as well as Peirce (1940, p. 152)
6 Peirce seems to be here at extreme variance from Popper's contention that the mode of
reasoning involved in such experimental procedure is deductive -- by modus tolens - rather
than inductive. As I we see further below, the discrepancy is only apparent. Each refers to
a different phase of testing.
 Fact, Logic and Method 275

miss the leap of faith' associated with induction as a mere psychological

phenomenon.

(a) Epistemology and 'grounds for belief':

It is indeed true, though somewhat trivial, that induction en¬
dows the investigator with a certain degree of confidence,
and thus ultimately involves grounds for belief. But dis¬
missing grounds for belief as 'psychological' — thus outside
the purview of philosophy — may be an overly rigid stric¬
ture. After all, one respectable epistemological interpretation
of the three modes of inference -- deduction, induction and
abduction — is that they specify three different grounds for
belief in the truth or falsity of propositions (see discussion
in Chapters 4 and 7).

(b) Epistemology and abduction:

Popper is indeed right that induction does not serve as the
sole logical tool for arriving at generalization via experience.
Nonetheless, some inductive reasoning is involved, some¬
where, in such a process. However, the crucial 'leap of faith'
associated with induction, via which one proceeds from in¬
dividual cases to general statements, demands a third mode
of inference -- abduction. And since abductive reasoning
falls within the legitimate domain of epistemology, the rejec¬
tion of induction merely because it subsumes abduction
would be surely an unnecessary stricture.

(c) Real-world constraints on generalizing from experience:

More substantively, the expectation that the untested members
of the population approximate the behavior of those members tested
in the limited sample is, on purely logical grounds, indeed a
leap of faith. It is, nonetheless, a very tenable working
hypothesis for biological organisms. Such a hypothesis must
lie at the very heart of the organism's — and the scientist's --
ability to generalize. In generalizing from experience, both
the organism and the scientist must abide by realistic con¬
straints on time, space and means. They hardly have a choice
in this matter.

(d) Induction and natural categories:

The 'leap of faith' that underlies induction rests upon one
fundamental, indeed crucial (though often ignored), tenet of
276 Mind, Code and Context

a pragmatic theory of natural categories.7 One may give this

tenet as (la) below, from which the induCtive-abductive
hypothesis (lb) follows:

(1) Inference from clustering of categorial properties:

(a) "Individual members of a natural category do not share
only a single criterial property. Rather, they most often
share many properties, which are thus the definitional
core of their categorial membership".
(b) "Therefore, if known members of a group exhibit prop¬
erties A, B, C etc., and if a sample sub-group also ex¬
hibits property Z (to a statistically-significant degree),
then it is highly likely that the rest -- untested — members
also exhibit property Z".

(e) Induction and survival imperatives:

The suggestion, by the deductivist Popper and the inductivist
'early' Wittgenstein, that induction was somehow tainted by
its admittedly psychological element, is a bit short-sighted on
other grounds as well. These grounds involve the functional-
adaptive imperatives associated with category-governed be¬
havior in a real-world bio-adaptive environment. Consider,
for example, the categorizing behavior of a herbivorous ani¬
mal engaged in an inductive extension of the range of the
category tiger. Suppose it already knows tigers by a thick
cluster of their prototypical traits, such as shape, color, size,
speed, sound, habitat, nocturnality etc. Suppose now it has
just observed the life-threatening, predatory behavior of one or
two individual tigers. Prior experience that tigers share a great
number of clustered properties (la), is precisely the kind of
solid foundation that allows — indeed impels -- the organism
to take an inductive-abductive 'leap of faith' of the type (lb),
and conclude that other tigers, yet to be encountered in the fu¬
ture, are very likely to match the behavior of the observed
minuscule sample. Such conclusions, however logically falla¬
cious, are essential for the organism's survival.

8.2.2.1.4. The level of generalization in induction

With all that was said above to justify the 'leap of faith' of induction,
one must concede Popper's main anti-inductivist argument: Induction

7 See discussion of prototype categorization in Chapter 2. Within such a framework, the

fewer properties are shared by members, the fewer induction-based prediction can be made
about the behavior of members in various contexts.
 Fact, Logic and Method T71

is indeed not the way one arrives at hypotheses -- neither in science nor
elsewhere. As noted above, even the most innocuous hypothesis of
generalization already depended crucially upon the third mode of in¬
ference, abduction. This is all the more true with more profound ex¬
planatory hypotheses, those that concern relevance, connectivity and
coherence, causality, and function. We will return to discuss those fur¬
ther below.

8.2.2.2. Extreme inductivism in the behavioral sciences

That extreme inductivism did indeed hold sway in early 20th Century
philosophy is amply documented. Equally well documented is the
strong impact logical positivism had upon two nascent behavioral-cog¬
nitive disciplines, psychology and linguistics. In psychology, this im¬
pact was expressed in the long tenure of the American behaviorist
movement, associated with illustrious names such as Watson and Skin¬
ner.8 The latter's stimulus-response paradigm is a straight-forward at¬
tempt to interpret behavior and cognition as cummulation of experi¬
ence-derived inductive generalization. In linguistics, the same approach
was embodied in the equally long tenure of the American Structuralist
(or 'Bloomfieldian') movement.9 The conscious philosophical pronoun¬
cements of Leonard Bloomfield, whose conversion to empiricism was
prompted by exposure to behaviorist psychology, follow a straight and
narrow inductivist line:

"...the only useful generalizations about language are inductive

generalizations..." (1933, p. 20)

The Bloomfieldians ridiculed all attempts to make universal hypotheses

that did not relate directly to observable facts. In fact, they rejected the
making of any general statements beyond the somewhat trivial first-
level generalization. In his review of Sapir's Language, Bloomfield
writes:

"...we must study people's habits of language - the way they talk --
without bothering about mental processes that we may conceive to un¬
derlie or accompany habits. We must dodge this issue by a fundamental
assumption, leaving it to a separate investigation, in which our results
will figure as data along the results of other social sciences..." (1922, p.
142)

8 See Skinner (1938,1957) as well as Chomsky's (1959) review of the latter.

9 See for example Bloomfield (1922,1926,1933), as well as discussion in Bach (1965) or Givon
(1966).
278 Mind, Code and Context

In rejecting the mentalism associated with the 19th Century German

Romanticist Herman Paul, Bloomfield writes:

"...The other great weakness of Paul's 'Principles' is in his insistence

upon "psychological" interpretation... [and on] mental processes which
the speakers are supposed to have undergone... The only evidence for
these mental processes is the linguistic process; they add nothing to the
discussion but only obscure it..." (1933, p. 17)

Bloomfield's extreme inductivism forced him, in turn, to interpret

meaning purely extensionally, much like the positivists. Consequently,
he had no recourse but to exclude the study of meaning from the realm
of linguistics:

"...In order to give a scientifically accurate definition of meaning for

every form of the language, one should have to have a scientifically ac¬
curate knowledge of everything in the speakers' world... In practice, we
define the meaning of a linguistic form, whenever we can, in terms of
some other science..." (1933, pp. 139-140)

Finally, in rejecting Paul's and Sapir's 'mentalistic' approach to mean¬

ing, Bloomfield describes it as an 'untenable delusion':

"...The mentalistic theory... supposes that the variability of human con¬

duct is due to the interference of some non-physical factors, a spirit or
will or mznd...that is present in every human being...[and] is entirely dif¬
ferent from material things and accordingly follows some other kind of
causation or perhaps non at all..." (1933, pp. 32-33)

Fortunately, Bloomfield the descriptive linguist consistently neglected

to follow the dictates of Bloomfield the inductivist philosopher, producing
linguistic descriptions that abounded in higher-level generalization, pos¬
tulated meanings and invisible mental constructs. Historically, Bloom¬
field's extremism was partially prompted by the excesses of a previous
wave of deductivist language philosophy,10 and was in turn the foil for a
subsequent wave of extreme deductivism.

10 Most specifically the German romanticist Herman Paul, whose disciple Bloomfield was
prior to his conversion to behaviorism.
 Fact, Logic and Method 279

8.2.3. Extreme deductivism

8.2.3.1. Popper and the Hypothetico-Deductive (HD) method

"...The game of science is, in principle, without end..."

K. Popper, The Logic of Scientific Dis¬

covery, (1959, p. 53)

8.2.3.1.1. The reduction of science

In his Logic of Scientific Discovery, Karl Popper advances a number of
extreme deductivist positions, later coopted by more rigid exponents.
Still, Popper's position remains somewhat problematic. Given the pre¬
vailing intellectual climate of reductionism. Popper's cogent arguments
against inductivism have been often interpreted as a plea for the op¬
posite extreme. What is more, in making a deliberate choice to focus his
attention on the phase of hypothesis testing. Popper almost invites ex¬
treme interpretations. After all, hypothesis testing is the phase where
the role of deduction is most obvious. Nonetheless, a number of aspects
of Popper's approach to methodology can be interpreted as pragmatic.
Or, put another way, extreme deductivism is not the only alternative to
extreme inductivism.
Popper's deductive-falsificatory approach to the testing of scientific
hypotheses rests on a curious separation between grounds for knowl¬
edge and truth:

"...I propose to look at science in a way which is slightly different

from that favored by the various psychologistic schools: I wish to dis¬
tinguish sharply between objective science, on the one hand, and 'our
knowledge' on the other. I readily admit that only observation can give us
'knowledge concerning the facts', and that we can (as Hahn says)
'become aware of facts only by observation'. But this awareness, this
knowledge of ours, does not justify or establish the truth of any state¬
ment. I do not believe, therefore, that the question which epistemology
must ask is '...on what does our knowledge rest?...' or more exactly, how
can I, having had the experience S, justify my description of it, and
defend it, against doubt... In my view, what epistemology has to ask is,
rather: How do we test scientific statements by their deductive conse¬
quences?..." (1959, pp. 97-98; boldfacing is mine; TG)

Both epistemology and science are thus reduced to one dimension,

hypothesis testing; and this is indeed a rather extreme philosophical
stricture, one that relegates all other phases of science (observation, in¬
duction, abduction) to the realm of disinterest.
280 Mind, Code and Context

Popper does not in fact deny that non-deductive mental processes

play a role in science. He only asserts — by fiat — that those other mental
processes are merely psychologists; they have no logical status, and are
therefore of no interest to a 'true' epistemology. Popper's argument
seems to rest on the following three value-laden assertions:

(i) Induction (and abduction) are unsystematic, unpredictable

psychologically-motiva ted processes;
(ii) Only deductive logic has epistemological status;
(iii) The phases of science (and epistemology) that are of interest to
the philosopher are those that involve deduction, thus truth.

8 2.3.1.2. The falsification method of testing

Having argued that induction is not a useful method for confirming
the truth of scientific statements, and thus that there is no positive proof
in science. Popper proceeds to outline the deductivist alternative — the
method of testing by falsification. This method requires first dividing
the statements of science into two classes -- singular statements of
'individual facts', and general statements of rules, laws, hypotheses,
theories etc. A general statement is tested deductively via modus tolens
in the following way:11

(2) Testing hypotheses through falsifiability:

(a) Take a general statement — hypothesis -- G;
(b) Derive from it deductively a hitherto unobserved sin¬
gular statement P;
(c) Creates an experiment which bring about the condition
under which the facts of P should - so G predicts -- be
observed;
(d) If statement P turns out to be false, then — via modus
tolens -- general statement G must also be false.

8.2 3.1.3. Falsification vs. verification

It is certainly fascinating how two great minds, Peirce and Popper,
could describe the very same process, the experimental testing of a
hypothesis, in such diametrically opposed terms — Peirce as verification
by induction,12 Popper as falsification by deduction. Recall Peirce's
version:

11 Following Popper (1959, p. 76).

12 Henning Andersen (in personal communication) suggests that Peirce was a confirmed
fu Is If leu t io n is t in his philosophy of science. If true, the following passage certainly sounds
like a verificationist's credo. But see also fn. 6, above.
 Fact, Logic and Method 281

"...The operation of testing a hypothesis by experiment, which consists

in remarking that, if it is true, observations made under certain condi¬
tions ought to have certain results, and then causing those conditions to
be fulfilled, and noting the results, and, if they are favorable, extending a
certain confidence to a hypothesis, I call induction...” (1940, p. 152)

In fact. Popper and Peirce are both right. They merely describe two dif¬
ferent phases of the process of testing. Popper describes the deductive
consequences of discovering a counter-example: Once a factual predic¬
tion of the hypothesis is shown by experiment to be false, the entire
hypothesis is rejected. Peirce deals with the cumulative inductive effect
of the repeated failure to falsify a hypothesis. Since, as Popper concedes,
the game of science is endless, recurrent failure to falsify indeed increases
the scientist's confidence in it. And that confidence may indeed be based
upon firm inductive-abductive grounds.
Indeed, Popper concedes Peirce's position implicitly, when observing
that the deductive-falsificatory method may have a severe drawback.
That drawback indeed echoes some of the reasons for rejecting verifica¬
tion as a mode of hypothesis testing: Both processes are potentially end¬
less. In the case of falsification, this is so because if one fails to reject a
prediction via one experiment, the hypothesis from which the predic¬
tion had been derived is not verified. Rather, it is simply not yet proven
false, given one experimental procedure. But:

(a) Other experimental procedures may yet prove it false; and

(b) The number of specific predictions that can be deduced from
a hypothesis, especially from a rich and complex hypothesis,
may be potentially infinite.

It is precisely within the context of such repeated attempts to falsify a

hypothesis that inductive-abductive reasoning comes into play.

8.2.3.2. Pragmatic elements in Popper’s method

In this section we will briefly note a number of areas in Popper's writ¬

ings that may be interpreted as pragmatically inclined.

8.2.3.21. Hypothesis formation

Popper (1959) considers the process of hypothesis formation capri¬
cious, unpredictable, and thus of no epistemological interest. By doing
so, he certainly ignores a major point where abductive inference comes
into play in science. Nevertheless, he does not reject the importance of
the process of discovery per se. He only asserts that it is not a logical
process. To illustrate his position, consider:
282 Mind, Code and Context

"...there is no such thing as a logical method for having new ideas, or

a logical reconstruction of this process. My view may be expressed by
saying that every discovery contains 'an irrational element' or 'a crea¬
tive intuition', in Bergson's sense. In a similar way Einstein speaks of
the 'search for those highly universal laws...from which a picture of the
world can be obtained by pure deduction. There is no logical path', he
says, 'leading to these...laws. They can only be reached by intuition,
based upon something like an intellectual love {'Einfiihlung') of the ob¬
jects of experience..." (1959, p. 32; quoting from Einstein, 1934, p. 125 of
English translation)

8.2.3.2.2. The conventional nature of methodology

A number of steps in Popper's proposed methodology are pragmati¬

cally justified. As Nola (1986) points out. Popper's aprioristic approach
takes the methodology of science to be a presupposed set of conven¬
tions, adopted by agreement, decision or choice, among people sharing
the same goals:

"...My criterion of demarcation will accordingly have to be regarded

as a proposal for an agreement or convention. As to the suitability of such
conventions opinions differ; and a reasonable discussion of these ques¬
tions is only possible between parties having some purpose in com¬
mon. The choice of that purpose must, of course, be ultimately a matter
of decision going beyond rational argument..." (1959, p. 37; boldfacing
is mine; TG

A number of Popper's proposed conventions elaborate the falsifi-

catory method of testing of hypotheses. One meta-convention seems to
possess a more profound status. It involves the 'principle of causation',
which Popper interprets as a claim that 'for every event there is a causal
explanation'.13 As a methodological rule of explanation. Popper renders
it as follows:

"...It is the simple rule that we are not to abandon the search for
universal laws and for a coherent theoretical system, nor ever give up
our attempts to explain causally any kind of event we can describe..."
(1959, p. 61)

Since, as noted earlier above,14 overall coherence and explanation -- be it

causal or otherwise -- are context-dependent notions. Popper here im¬
plicitly concede an important role to pragmatic reasoning.

13 Nola goes on to comment: "...this turns out to be unfalsifiable but a respectable piece of
metaphysics..." (1986, p. 18)
14 See Chapter 7.
 Fact, Logic and Method 283

8.2.3.2.3. Utility and results

Next, Popper contends that his proposed apriori conventions for

scientific methodology can be best judged by their utility in producing
good science:

"...My only reason for proposing my criterion of demarcation is that it

is fruitful: That a great many points can be clarified with its help... It is
only from the consequences of my definition of empirical science, and
from the methodological decisions that depend upon this definition, that
the scientist will be able to see how far it conforms to his intuitive idea of
the goal of his endeavors. The philosopher too will accept my definition
as useful only if he can accept its consequences. We must satisfy him that
these consequences enable us to detect inconsistencies and inadequacies
in older theories of knowledge..." (1959, p. 55; emphases are mine; TG)

Since usefulness of results can only be judged in the context of goals,

this is one more pragmatics element in Popper's approach to method.

8.2.3.2.4. The sociology of science

As pointed out by Nola (1986), Popper's (1974) invocation of ex¬

emplars of great science is a retreat from his earlier 'apriorism' toward a
version of 'naive empirical conventionalism'. Such an approach is in
some respects reminiscent of Kuhn's (1962). To the extent that Popper
regards science as a convention-bound, socially-mediate activity, his
approach is once again pragmatic.

8.2.3.3. Deductivism in a social science15

It is perhaps not an accident that a nascent would-be science, such as
linguistics, become so fatally attracted to recurrent waves of extreme
philosophical reductionism. As elsewhere in the history of ideas, one ex¬
treme tends to beget — as violent reaction — its opposite. And so, the
strict empiricism of Bloomfield was in turn supplanted by the equally
rigid rationalist deductivism of Chomsky. The view of the scientific
method as demanding a choice between only two reductionist poles
may be illustrated by the following passage from Bach (1965), where
methodological inductivism and deductivism are labeled 'Baconian'
and 'Keplerian', respectively:

15 1 have benefitted here from the erudite discussion found in Derwing (1973) particularly
chapter 7.
284 Mind, Code and Context

"...Whereas the Baconian stresses caution and 'sticking to facts' with a

distrust of theory and hypotheses... the Keplerian emphasizes the crea¬
tive nature of scientific discovery, the leap to general hypotheses — often
mathematical in form, whose value is judged in terms of fruitfulness,
simplicity and elegance...The prevailing assumptions of American lin¬
guistics prior to 1957 were essentially Baconian in character... [Chom¬
sky's approach, on the other hand is a] deductively formulated meth¬
od..." (1965, pp. 113-114) * 1

In the same vein, Lees (1957) observes:

"...Once it has developed beyond the prescientific stage of collection

and classification of interesting facts, a scientific discipline is charac¬
terized essentially by the introduction of abstract constructs and theories
and the validation of those theories by testing their predictive power..."
(1957, p. 376; emphases are mine; TG)

Aside from the invocation of the potentially pragmatic 'fruitfulness',16

and apart from the glaring departure from Popper's insistence on fal¬
sification (thus subscribing to the inductivist's confirmatory bias), de-
ductivist linguistics closely recapitulates the salient features of extreme
deductivism in philosophy:

(a) Disinterest in discovery:

Rejecting the naive notion of discovery by induction, deduc-
tivist linguists — in spite of lip service to 'creativity7 — have
tended to consider the process of discovery unconstrained,
unsystematic, not meriting serious attention. The early gen¬
erative attacks on the bogey-man of Discovery Procedure17
were also clear reflection of this.
(b) Formalization:
Since deduction requires an explicit notation, deductivist
methodology is often obsessed with formalism. In linguis¬
tics, this most commonly leads to interpreting language as a
formal rule-system — closed, complete, consistent. The per¬
vasive structuralism that tends to accompany this approach,
and its relative detachment from the study of function and

16 See discussion of the pragmatic elements in Popper's approach, above. 'Fruitfulness' in the
Chomskian tradition turns out to be inseparable from simplicity/economy. This is a
consequence of the notion of 'competence' that pervades the generative approach to data.
See discussion in Derwing (1973, ch. 7,8) or Giv6n (1979a, ch. 1), inter alia.
17 Longacre's innocent field manual. Grammar Discovery Procedures (1964), came in for an
inordinate share of critical abuse, since it dared to suggest that some procedures are more
efficient than others in discovering rules of grammar.
 Fact, Logic and Method 285

functional explanation, are not surprising. Much like ex¬

treme deductivists elsewhere,18 'explanation' within this ap¬
proach has tended to be system-internal, deductive and
circular.

(c) The neglect of primary facts:

In general, deductivist linguistics has tended to downgrade
the value of facts, to suggest that 'facts are cheap'.19
Chomsky's notion of linguistic competence played an impor¬
tant role in circumventing deductive-falsificatory testing.
Observable facts can be dismissed as mere performance. What
matters above all is the elucidation of abstract theoretical
constructs.20

(d) Simplicity, economy, elegance:

In the absence of firm empirical criteria, the role of formal
ones — simplicity, economy, elegance — becomes paramount.
This again constitutes a retreat from Popper's falsificatory
testing method. To illustrate the ascendance of 'simplicity'
over facts in deductivist linguistics, and the subsequent dilu¬
tion of 'empirical validation', consider the following quota¬
tion from some of the more empirical proponents of this
approach:

"...We have argued that there is strong empirical support for the claim
that declaratives have abstract underlying structures and that this sup¬
port derives primarily from considerations of the simplicity and generality
of grammatical rules..." (Bever et al, 1965, p. 289; emphases are mine; TG)

The two historical episodes of extreme reductionism in linguistics,

both equally structuralist, managed between them to present a carica¬
ture of a science: You may either practice dogmatic inductivism and
forego making any but the most shallow generalization; or you may
practice rigid deductivism and give up on the vital connection between
data and theoretical constructs. The middle ground - pragmatic-ana¬
logical reasoning, induction-abduction, or even methodological plu¬
ralism -- was never contemplated.

18 See the discussion of Hem pel and Oppenheim (1948), further below.
19 See discussion in Derwing (1973, pp. 229-230).
20 See discussion in Givon (1979a, Ch. 1). In a recent BBC radio interview, Chomsky observed
(I quote from memory): "...We have enough facts on human language, perhaps too many.
What we really lack are the right kind of abstract constructs with which to build a theory..."
286 Mind, Code and Context

8.3. Pragmatic accounts of the scientific method

8.3.1. Preamble

Until a few decades ago, Peirce's pioneering work on abduction and

abductive-induction stood out as the sole non-reductionist alternative
in the philosophy of science. This picture has been slowly changing.
Perhaps the single most forceful exponent of a pragmatic approach in
the philosophy of science was the late R.N. Hanson. As he noted, deduc¬
tive and inductive methods need not be conceived of as exclusive op¬
posites:

"...the two accounts are not alternatives: They are compatible..." (1958,
p. 70)

In this section we will survey some of the major tenets of a pragmatic

philosophy of science.

8.3.2. The fact-driven nature of hypothesis formation

"...The particular facts are not merely brought together, but there
is a new element added to the combination by the very act of
thought by which they are combined...The pearls are there, but
they will not hang together until someone provides the
string..."

Aristotle, Posterior Analytic

(vol. II, p. 19)

As Hanson (1958,1961) points out, hypothesis formation proceeds by

a systematic type of reasoning, one that is perhaps more elusive than
deduction and induction, yet nonetheless distinct: abductive-analogical
reasoning. Further, this type of reasoning in science proceeds indeed in
the direction suggested by inductivists -- from data to hypothesis:

"...The critical moment comes when the physicist perceives that one
might reason about the data in such and such a way. One might explain this
welter of phenomena P, throw it all into an intelligible pattern, by suppos¬
ing H to obtain. But P controls H, not vice versa. The reasoning is from
data to hypotheses and theories, not the reverse..." (1958, p. 88; emphases
are mine; TG)

In discussing a celebrated example from physics, Hanson observes

 Fad, Logic and Method 287

that it represented (a) reasoning from facts that demand explanation to

an explanatory hypothesis; but (b) neither deduction nor induction:

"...Was Kepler's struggle up from Tycho's data to the proposal of el¬

liptical orbit hypothesis really inferential at all? He wrote De Mortibus
Stelae Martis in order to set out his reasons for suggesting the ellipse.
These were not deductive reasons; he was working from explicanda to
explicans. But neither were they inductive - not, at least, in any form ad¬
vocated by the empiricists, statisticians and probability theorists..."
(1958, p. 85)

As to the deductivist account of hypothesis formation, Hanson obser¬

ves:

"...Disciples of the H-D account often dismiss the dawning of an

hypothesis as being of psychological interest only, or else claim it to be
the province solely of genius and not of logic. They are wrong. If estab¬
lishing an hypothesis through its predictions has a logic, so has the con¬
ceiving of an hypothesis..." (1958, p. 71)

Hanson is equally critical of rigid inductivists:

"...So the inductivist rightly suggests that laws are somehow related to
inference from data. He wrongly suggests that the resultant law is but a
summary of these data...Yet the original suggestion of hypothesis type is
often a reasonable affair. It is not as dependent on intuition, hunches and
other imponderables as historians and philosophers suppose when they
make it the province of genius but not of logic. H-D accounts all agree
that physical laws explain data, but they obscure the initial connection be¬
tween data and laws..." (1958, p. 71; emphases are mine; TG)

And again about extreme deductivists:

"...H-D accounts begin with the hypothesis as given, as cooking

recipes begin with the trout. Recipes, however, sometimes suggest 'first
catch your trout'. The H-D account is a recipe physicists often use after
catching hypotheses. However, the conceptual boldness that marks the
history of physics shows more in the ways in which scientists caught
their hypotheses than in the ways in which they elaborated these once
caught. To study only the verification of hypotheses leaves a vital part
of the story untold..." (1958, pp. 30-31)

And further:

"...Physicists do not start from hypotheses; they start from data. By

the time a law has been fixed into an H-D system, really original physi-
288 Mind, Code and Context

cal thinking is over. The pedestrian process of deducing observation

statements from hypotheses comes only after the physicist sees that the
hypothesis will at least explain the initial data requiring explanation. [The] H-
D account is helpful only when discussing the argument of a finished
research report... the analysis leaves undiscussed the reasoning..." (1958,
p. 71; emphases are mine; TG)

Relying extensively on Peirce's work on abductive inference, Hanson

(1958) quotes Peirce:

"...[induction] sets out with a theory and it measures the degree of con¬
cordance of that theory with fact. It never can originate any idea
whatever. No more can deduction. All ideas of science come to it by way
of abduction. Abduction consists in studying the facts and devising theories
to explain them. Its only justification is that if we are ever to understand
things at all, it must be21 in that way. Abductive and inductive reasoning
are utterly irreducible, either to the other or to deduction, or deduction to
either of them..." (C.S. Peirce, Collected Writings (1934), vol. V, p. 146)

And again quoting from Peirce:

"...Deduction proves that something must be; Induction shows that

something actually is operative; Abduction merely suggests that some¬
thing may be..." (ibid, p. 171)

Following Peirce, Hanson recapitulates the general schema for

hypothesis formation via abduction:

(3) Hanson's account of hypothesis formation:

"... 1. Some surprising phenomenon P is observed.
2. P would be explicable as a matter of course
if H were true.
3. Hence there is reason to think that H is true. H cannot be
retroductively [i.e. abductively; TG] inferred until its con¬
tents is present in 2. Inductive accounts expect H to
emerge from repetition of P. H-D accounts make P emerge
from some unaccounted-for creation of H as a 'high-level
hypothesis'..." (1958, p. 86; emphases are mine; TG)

21 'Must be' here is the mode of hypothesis, not deductive necessity. See directly below.
 Fact, Logic and Method 289

8.3.3. The status of ’observable’ facts

"...Facts are not picturable, observable entities...”

R.N. Hanson, Patterns of Discovery

(1958, p. 31)

As one may recall. Popper -- in describing the process of deductive

falsification — set up a sharp dichotomy between singular statements
('facts'), and universal statements ('generalization'), be those latter ones
rules, laws, hypotheses or theories. Both inductivists and deductivists
have accepted this traditional division, taking for granted the objective,
given, atomic status of 'facts'. The fact vs. generality distinction presup¬
poses, in turn, a distinction in vocabulary, between terms referring to ob¬
servable entities ('non-theoretical terms', see Achinstein, 1965) and
those referring to theoretical constructs. Popper (1959, p. 64) refers to
the latter distinction as 'individual concepts' vs. 'universal concepts'.
Carnap, an inductivist, sets up a somewhat similar distinction between
"...terms designating observable properties and relations..." and "...terms
which may refer to unobservable events, unobservable features of
events...", a distinction that is, up to a point, the same as the one be¬
tween theoretical and non-theoretical terms.22
Within a pragmatic approach, the objective 'givenness' of facts is,
rather, a problematic matter. Feyerabend (1970), for example, casts
doubt on the theory-independent status of clean facts, noting that they
are conceptualized within the context ('framework') of a particular
theory, on whose vocabulary they are at least partially dependent. This
is, partly, a late-Wittgensteinean argument, similar to the one advanced
by Hanson.23 Similarly, Achinstein (1965) points out that between purely
theoretical and purely observable terms, a continuum of intermediates
exists. Generalization is thus taken to be a graded process. To illustrate
this with a simple example, consider the following empirically-sup¬
ported 'fact' of classical astronomy:

22 See discussion in Achinstein (1965, p. 234). In another place, however, Carnap argues that
a distinction between 'individual' and 'universal' terms is context dependent: "...this
distinction is not justified...every concept can be regarded as an individual or universal
concept according to the point of view adopted... ” (1934, p. 313; English Translation supplied
by Popper (1959, p. 67)). Needless to say. Popper disagrees strongly with this pragmatist
lapse of Carnap's, although he finds part of the argument correct as it stands (though
irrelevant). For this reason alone, a full parallel between Popper's division of statements
and Carnap division of terms cannot be drawn, as much as Achinstein's discussion may
suggest otherwise.
23 See Hanson (1958, Ch. 2, pp. 31-34); the following Wittgensteinean quotable is found there:
"...If a distinction cannot be made in language it cannot be made conceptually..." (p. 34).
290 Mind, Code and Context

(4) "The orbit of Mars is elliptical"

Statement (4) is obviously already a generalization, inducted-abducted

from more particular observable facts that conform more closely to
Popper's standards, such as for example:

(5) a. "On a certain date, at a certain hour, minute, second

(etc.), the location of Mars relative to the sun was at cer¬
tain distance and/or at coordinates xyz, etc."
b. "On another date, etc.... the location of Mars was at....
(etc.)"
c. etc.

But are statements (4a,b,c) etc. really about the ultimate observable facts?
How finely must their time-axis be established (day, hour, minute, sec¬
ond?) before such facts are transformed from 'general' to 'particular'?
The answer is, of course, that a sharp boundary cannot be drawn as a
matter of principle; rather, the boundary is pragmatically determined, it
depends on either perceptual calibration, or on the frame/context -- be
that frame perceptual-cognitive (for the organism) or theoretical (for the
scientist) -- within which the observation was made.
The objective status of the 'particular facts' in (5) is further under¬
mined by the total relativity of the spatial coordinates from which the
observations were made (i.e. Earth's position), and via which the posi¬
tion of both planets vis-a-vis the sun was expressed. These 'facts' are ob¬
served — or constructed -- in a frame-dependent fashion, they have no
firm existence outside that frame.
In Chapter 3, above, we noted that in epistemology likewise the status of
'events' is equally frame-dependent, and for similar reason. The argument
there revolved, first, on the spatial framing of events, i.e. what physical
components may be judged as parts of it, which ones should be excluded,
and on what grounds. As we pointed out there, the grounds for inclusion
in the event frame were pragmatic at various levels, having little to do with
either hard fact or deductive logic. The continuum of space, and of the en¬
tities within it, allows no firmer, more principled grounds.
It is easy to show that the time-specificity of so-called observable facts
dissolves in the same way. Consider the following examples, all
presumably statements of particular, facts:

(6) a. Joe saw the mule.

b. Joe kicked the mule.
c. Joe fed the mule.
d. Joe chased the mule.
e. Joe rode the mule.
f. Joe trained the mule.
g. Joe loved the mule.
 Fact, Logic and Method 291

The time-span that is normative or characteristic of the events depicted

in (6a-g) varies gradually. Normatively, seeing is instantaneous. Norma-
tively, loving is protracted. The progression between them is gradual.
Further, (6b) through (6f) are transparently complex, composite activities,
having many sub-parts. And if anything is likely to have been 'observed
fact', it was those sub-parts. And how about the sub-parts of those sub¬
parts? Ad infinitum? A similar argument can be made concerning see (6a),
whose instantaneous status is after all only relative to the perceptual
calibration of our sensory apparatus. Finally, feed, chase, ride, train and
love are frame-dependent, culture-mediated notions, as Wittgenstein
(1953) would have surely pointed out. Being assembled constructs, their
status as observables is only as firm as the cultural conventions -- world¬
views, theories, hypotheses -- within which they are embedded.
Hanson (1958) illustrates the frame-dependence of observed facts
with the following example from Biology:

"...Imagine these two [scientists] observing a Protozoon —Amoeba. One

sees a one-celled animal, the other a non-celled animal. The first sees
Amoeba in all its analogies with different types of single cells: liver cells,
nerve cells, epithelium cells. These have a wall, nucleus, cytoplasm, etc.
Within this class Amoeba is distinguished only by its independence. The
other, however, sees Amoeba's homology not with single cells, but with
whole animals. Like all animals Amoeba ingests its food, digests and as¬
similates it. It excretes, reproduces and is mobile -- more like a complete
animal than an individual tissue cell..." (1958, p. 4)

To sum up, then, facts may be graded on a scale, from those closer to
observation to those most theory-dependent. But the scale itself is con¬
text-dependent, as are the criteria by which facts are placed at particular
points on the scale. In the philosophy of science, such context-depend¬
ence has come to be referred to, in Hanson's term, as the theory laden
status of facts.
One must note, lastly, that in adopting Hanson's notion of theory
laden facts one merely recapitulates the Kantian middle-ground com¬
promise between Platonic and Aristotelian epistemology; whereby one
may read for 'facts' Kant's percepts, and for 'theories' Kant's concepts.

8.3.4. The relevance of data

We have noted above that 'fact' is a loaded, context-determined,

theory-dependent notion. In this section we will briefly discuss another
aspect of the pragmatic nature of facts, one that has an immense practi¬
cal impact on the conduct of empirical investigations. It involves the
choice of facts for which one wants to account via hypotheses. That is:
292 Mind, Code and Context

(7) "Of all the innumerable facts of the world, how does one
know which ones are relevant to:
(a) The general empirical domain?
(b) This particular empirical cycle within the domain?
or
(c) The particular hypothesis?"

The decisions one needs to make in order to answer (7) are all, in prin¬
ciple, pre-empirical and abductive. One makes them without recourse
to either deduction or induction. Innumerable facts about the universe
may be -- ultimately -- related, connected or relevant to the subject mat¬
ter ('domain', 'cycle', 'hypothesis'). Their relevance is a matter of de¬
gree; one uses common sense in ruling out the more remote relations;
one concentrates one's meager resources on those facts that are more
likely to be relevant, whose connection to the domain, cycle or specific
hypothesis is more intimate. For example, if one is puzzled by some
facts in the Swahili verb system (see section 8.3.5., directly below), then
various verbal prefixes or suffixes may be relevant at various phases of
the investigation. But one must rank tense-aspect prefixes as more
relevant than derivational suffixes, or more than subject/object agree¬
ment prefixes; and verbal suffixes would be more relevant than nominal
suffixes. But by what criteria?
Consider next a biologist investigating some puzzling facts about the
human cardiac structure. This biologist would surely judge it more like¬
ly that relevant facts may come from the circulatory, pulmonary or
autonomous-nervous domains, than from, say, the reproductive, skeletal
or olfactory sub-domains. But his decisions in this matter are made prior
to having made an overall theory of the domain; they are prerequisites
for the eventual construction of such a theory. They are abductions about
the likely connectivity within a yet to be elaborated higher context. As
Hanson (1958) concludes regarding the Amoeba example cited earlier:

"...This is not an experimental issue, yet it can affect experiment. What

either man regards as significant questions or relevant data can be
determined by whether he stresses the first or the last term of
'unicellular animal'..." (1958, pp. 4-5).

As Hanson further observes, not only is the relevance of the data an

abductive-pragmatic matter, but also the choice of what would con¬
stitute a significant question. These two issues are intimately con¬
nected, since the choice of particular puzzling facts is inseparable from
the choice of significant questions concerning those puzzling facts. The
puzzlement is indeed, itself, the question. As Bromberger (1966, pp. 79-
81) points out, 'why' questions are only appropriate if the facts ques-
 Fact, Logic and Method 293

tioned are somehow surprising. And one's surprise depends, in turn,

on one's presuppositions about the domain, i.e. on one's prior context.

8.3.5. The impetus for an empirical cycle: Puzzling facts

At the beginning of a typical empirical cycle there stands a puzzle.

The puzzle may be referred to as a distortion, an inconsistency between
some newly observed (or newly noticed) facts and the bulk of current
knowledge. These facts simply are incompatible with current knowl¬
edge, with the existing organized ('theoretical') framework of the
domain. Somehow, these puzzling facts are deemed relevant to the par¬
ticular domain. Still, they don't make sense within the existing
framework, they do not cohere with it.
This distortion, puzzle, incompatibility of facts, constitute the im¬
petus for a typical empirical cycle. They demand explanation, i.e.
demand the construction of a new hypothesis within which they will
cohere. Bromberger (1966) has aptly characterized this impetus as a
'why7 question which demands a 'because' answer. And that 'because'
answer is the emerging hypothesis about how some surprising facts
might cohere.
The puzzle is not inductive, and is only marginally deductive, in the
sense that the incompatibility may -- at least theoretically -- have been
detected through the deductive-falsificatory method. But just as plausi¬
bly the newly acknowledged facts may have fallen out of other phases,
other procedures, other experiments, or other investigations. The puzzle
precipitates the need to construct -- by abduction — a new hypothesis,
explanation, theory. The impetus to formulate a hypothesis that will re¬
impose coherence, however fragile and temporary, is described evoca¬
tively by Popper, as a pre-empirical imperative that is part of the
methodological conventions of science:

"...It is the simple rule that we are not to abandon the search for
universal laws and for a coherent theoretical system, nor ever give up
our attempts to causally explain any kind of event we can describe..."
(1959, p. 61 )24

8.3.6. Intermezzo: A case-study of an empirical cycle

To illustrate how puzzling facts serve as impetus to the abduction of

new explanatory hypotheses, consider the following run-of-the-mill ex¬
ample from historical grammatical reconstruction. In this sub-field of

24 Popper attributes this imperative to H. Gompertz' (1907) Das Problem der Willensfreiheit.
294 Mind, Code and Context

linguistics, the method of internal reconstruction is a standard tool for

hypothesis building. This method relies heavily and conspicuously on
abductive-analogical reasoning.
Internal reconstruction is said to be triggered by an irregularity in the
paradigm, i.e. 'distortion' or 'incompatibility' within the extant body of
knowledge. Consider the following case of the Swahili verbal system.
The bulk of 'regular' verbs in Swahili are bi- or poly-syllabic. In finite
verb paradigms, the verb stem must be preceded by a tense-aspect
marker, which is in turn preceded by a subject-agreement marker:

(8) a. a-li-soma 'He read'

b. a-me-soma 'He has read
c. a-na-soma 'He is reading'
d. a-ta-soma 'He will read'
e. a-0-soma 'He reads'

This system looks rather neat, until the following distortions are
noticed: First, a small group of verbs are monosyllabic, and the
paradigm for these verbs contains an added element -ku-, tucked in be¬
tween the tense-aspect marker and the verb stem:

(9) a. a-li-ku-Za 'He ate'

b. a-me-ku-Zfl 'He has eaten'
c. a-na-ku-Zfl 'He is eating'
d. a-ta-ku-Zfl 'He will eat'
e. a-e-la 'He eats'

Second, an added irregularity involves the fact that -ku- does not ap¬
pear in the habitual (zero) tense (9e).
The first distortion demands explanation of two sub-questions:

(10) a. "What is the reason for having -ku- in monosyllabic

verbs but dispensing with it in bi- or poly-syllabic ones?"
b. "Why -ku- rather than any other phonological se¬
quence?"

The answer to (10a) turns to be relatively transparent, once one recalls

that the Swahili word-stress is overwhelmingly penultimate; that the
lexical stress must fall on the verb stem; and that mono-syllabic verb
stems cannot carry this stress without violating the penultimate stress
rule. One makes the interim, tentative, functional-explanatory hypoth¬
esis (11), through which the hitherto disparate facts of -ku- and the
Swahili stress rules are linked in a single general pattern:
 ,
Fact Logic and Method 295

(11) "The function of -ku- must be to augment the monosyllabic

verb stem so that it may carry the prescribed penultimate
stress"

But why -ku-? In historical linguistics, given what we know about the
process of grammaticalization via which lexical stems give rise to gram¬
matical morphology, this question may be re-cast as:

(12) "What is the source of -ku-?"

One now casts around to see where other instances of a -ku- may be
found, not just mere instances, but ones whose syntactic and functional
distribution may tag them as plausible candidates for being the source of the
-ku- in short-verb paradigms. In short, plausible, relevant candidates. The
first relevant observation is that ku- also marks the infinitive of verbs, as
in:

(13) ku-soma 'to read'

ku-la 'to eat'

The next relevant observation is that -ku-marked infinitival verbs ap¬

pear in complements of modality verbs such as:

(14) a. a-na-taka ku-soma 'He wants to read'

he-PRES-want INF-read

b. a-li-kwisha ku-Za 'He finished eating'

he-PAST-finish INF-eat

At this point an abduction-wise investigator possesses just about

enough facts to attempt an abductive leap to an explanatory hypothesis.
But before considering the hypothesis, let us consider some additional puz¬
zling facts, this time concerning the relative-clause form of Swahili verbs.
In their REL-clause form, verbs in three of the tenses -- excepting the
'perfect7 and 'habitual' ('zero') -- may take an infix relative pronoun be¬
tween the tense marker and the verb stem. For long-stem verbs, the pat¬
tern is as follows:

(15) a. mtoto a-li-ye-sdmfl 'The child who read'

child he-PAST-REL-read

b. mtoto a-na-ye-sdma 'The child who's reading'

child he-PRES-REL-read

c. mtoto a-taka-ye-somfl 'The child who'll read'

child he-FUT-REL-read
296 Mind, Code and Context

In (15c) we now spot one additional puzzle: The future marker -ta- has
somehow acquires an additional syllable in this REL-pattem, appearing
now as -taka-. As we shall see below, this puzzling fact eventually falls
into place, and in fact contributes an important clue for resolving the in¬
itial puzzle. This salutary role of an erstwhile puzzle may be re-cast as a
well-known general principle of empirical work:

(16) "What constitutes an inexplicable puzzle within an existing

context, often turns out to both lead to and be explained by
an expanded, more comprehensive context".

More puzzling yet, the 'perfect' -me- tense, which earlier followed the
pattern of the other three tense-aspects in (15), now joins the 'habitual'
(zero) tense in displaying an idiosyncratic REL-clause form:

(17) a. mtoto amba-ye a-me-soma 'The child who has read'

child say-REL he-PERF-read

b. mtoto a-soma-t/e 'A/the child who reads'

child he-HAB-read-REL

As different as the relativization patterns (17a,b) are, they nonetheless

supply us with an added clue, one that turns out to be crucial for con¬
structing a comprehensive hypothesis: In both, the REL-pronoun is a
verb suffix. In (17a) it is suffixed to a 'relative subordinating verb' -amba
'say'. In (17b) it is attached to the relativized verb itself.
Another piece of evidence is the fact that the four tense-aspects in
(15a,b,c) and (17b) may all follow, optionally, the 'REL-auxiliary' pattern
that is obligatory for the 'perfect' (17a).
At this point the hypothesis fairly leaps out of the page at you,25 espe¬
cially if you know something about the most common diachronic source
of tense-aspect markers: They tend to arise, overwhelmingly, from a
small group of verbs in the complement construction of modality verbs
(as in (12) above), via the process of grammaticalization.26 The verbs
that most commonly contribute to this process are have, be, want, go,
come, and finish. Further, each of these verbs gives rise, most commonly,
to a highly specific tense-aspect:

(18) a. 'be' => 'progressive'

b. 'want', 'go' => 'future'
c. 'have', 'finish', 'come' => 'perfect'

25 This is, roughly, the point where the abductive hypothesis occurred to me, in writing Givon
(1972a, ch. 4).
26 For grammaticalization in general, with many illustrations from the rise of tense-aspect
markers from verbs, see Givon (1973a, 1979a, ch. 5,1982a, 1984a, ch. 6).
 Fact, Logic and Method 297

The assorted facts surveyed thus far now fall into a tentative but un¬
mistakably coherent pattern. This pattern leaves a number of facts still
hanging loose, puzzling, yet to be integrated into the overall coherence.
Nonetheless, the pattern is there, a full-fledged abductive, explanatory
hypothesis:

(19) a. "The four tense-aspects that involve some phonological

material (unlike the 'habitual'/'zero') must have all
arose from erstwhile verbs.

b. Further, their grammaticalization into tense-aspect

markers must have proceeded from the modal-verb con¬
structions as in (14). In such constructions, the current
main verb must have still been the complement of the
modal verb".

Hypothesis (19) immediately explains two major pieces of our puzzle,

first the shape and verb-prefix position of -ku- when it appears in front
of short-stem verbs; and second, the position of the relative pronoun -
following the tense-aspect marker —in the 'infix' constructions (14):

(19) c. "The shape and position of -ku- must be both frozen rel¬
ics, reflecting the syntax at the pre-grammaticalization
phase of the construction, as complementizer".

d. "The 'infix' REL-pronoun must be also a frozen relic re¬

flecting the syntactic stage of the time when the tense-
aspect was the main verb and thus carried the
REL-pronoun as suffix".

Hypothesis (19a-d) must be further augmented, however, with:

(19) e. "The verb-suffix position of the REL-pronoun, as in (17),

must have been the older, original position in Swahili
relativization. It is the process of grammaticalization
from verb to tense-aspect that created the 'infix' REL-
pattern in (15)".

As satisfying as hypothesis (19a-e) is, it still leaves the exceptional be¬

havior of both the 'habitual' and 'perfect' tense outside the new co¬
herence system. The first, it turns out, poses no problem, since it dis¬
plays neither verb-related characteristics (infinitive on the following
verb, infixed REL-pronoun), nor any phonological material. It may be
thus dispatched summarily by the further elaborating our hypothesis:
298 Mind, Code and Context

(19) f. "The 0 tense did not arise from a verb; it must have been
simply a morphologically-unmarked form. The relativi-
zation of verb-stems in the habitual thus follows the
older — suffixal -- pattern (cf. (19e))".

The perfect -me- is a bit more problematic, since it shows one verb-re¬
lated feature (-ku- before short stems) but not the other (the 'infix' REL-
pronoun). As it turned out, however; the answer to the next important
question indirectly also resolves the problem of the 'perfect'.
As hypothesis (19a-f) stands now, it predicts one absolutely vital set
of facts, namely that:

(20) "Each one of the four supposedly verb-derived tense-aspect

markers in Swahili -- li, na, ta, and me — must have had a spe¬
cific verbal origin".

One may consider the testing of this prediction as Popper's deductive-

falsificatory step. That is, if (20) were not the case, hypothesis (19) must
be rejected, by modus tolens.
Of the four marked tenses in Swahili, the first three — with least ir¬
regularities in their pattern -- readily yield verbal etymologies, traceable
to 'be', 'have' and 'want', respectively:

(21) a. a-li mu-ngaanda (from Bemba)

he-be in-house
'He is in the house'

b. a-na chakula
he-have food
'He has food'

c. a-na-taka ku-la
he-PRES-want INF-eat
'He wants to eat'

The perfect -me- turns out to be related to Swahili -mal-zza 'finish'

(Bemba -mal- 'finish'). Its current shape in Swahili is derived historical¬
ly from the so-called modified base form (Bemba -meel-e) by predictable
phonological reduction.
We have then plausible etymological sources; but these sources are not
free of problems:
 Fact, Logic and Method 299

(22) a. Normally 'be' tends to grammaticalized into the 'pro¬

gressive' aspect (as in English); but the Bantu -li- be¬
came 'past' in Swahili;
b. Normally 'have' tends to grammaticalize into the 'per¬
fect' aspect (as in English); but Swahili -na- became the
'present-progressive'.

The answer to these two puzzles, it turns out, has to do with more
idiosyncratic historical developments in Swahili, where -na- indeed be¬
came first the 'perfect', then later changed into the 'present', which was
in turn re-interpreted as 'present-progressive'. And -li- was indeed first
a 'past-progressive' tense-aspect, and only later was re-interpreted as
'simple past'.27
Lastly, we still have the functional-explanatory hypothesis (11) con¬
cerning the appearance of -ku- only before short verb-stems. Given the
added facts we have unearthed since first abducting (11), and the cur¬
rent shape of hypothesis (19a-f) that has emerged gradually as a result,
we must now modify (11) slightly, in order that it fit into the coherence
pattern of (19). That is, it must be now expressed in coherent historical-
developmental terms:

(19) g. "The reason why -ku- was retained before short verb
stems even after grammaticalization of the preceding
verb into tense-aspect, must have been because it carried
the verb's lexical stress and thus behaved like part of
the verbal stem".

Our sub-hypothesis (19g) of course takes advantage of other general

knowledge, concerning phonological reduction ('erosion') that occurs
during grammaticalization, and how it impacts primarily de-stressed
syllables.
We wind up, at the end of a protracted process that involved many
diverse steps, with a general theory of the phenomenon (Swahili tense-
aspect morphology). Within this general theory, the distortion that gave
us impetus for launching the cycle of investigation found its resolution,
coherence, context. The whole process involved the gathering of addi¬
tional relevant facts, the successive fitting of those facts into a unified
overall pattern, the gradual construction and refinement - via successive
abductions ~ of explanatory hypotheses. The process even involved one
crucial falsificatory test. The theory we have constructed explains the
original puzzle, but it also explains other puzzles along the way. It ex¬
plains them all within a single -- albeit complex - hypothesis. And it ex-

27 See details in Wald (1973).

300 Mind, Code and Context

plains them all in conformity with a number of well-known general prin¬

ciples observed independently in human language. Whether our
hypothesis will last further falsificatory testing only time will tell.28

8.4. The pragmatics of explanation

8.4.1. Preamble

"...What is scientific explanation? It is a topically unified com¬

munication, the contents of which imparts understanding of
some scientific phenomenon. And the better it is, the more effi¬
ciently and reliably it does this, i.e., with less redundancy and
higher over-all probability. What is understanding? Under¬
standing is, roughly, organized knowledge, i.e., knowledge of the
relations between various facts and/or laws. These relations are
many kinds -- deductive, inductive, analogical, etc. (Under¬
standing is deeper, more thorough, the greater the span of this
relational knowledge)..."

M. Scriven (1962, p. 102)

The single most crucial notion in science, the one that transforms the
enterprise from a mere gathering of disparate facts to a meaningful ac¬
tivity of expanding our understanding, is the notion of explanation. As
only fitting, it is also the most controversial in the philosophy of science.
The traditional philosophical positions display here a certain asymmetry:
Induction, being a progression from facts to general rule, is in principle
irrelevant to explanation; explanation is not part of induction's domain.
The few avowed inductivists ever to actually practice empirical science
— most conspicuously in psychology and linguistics -- have either dis¬
avowed interest in explanation (viz Bloomfield), or else indulged in pos¬
tulating 'inductively transparent' general principles which somehow

28 Several other 'distortions' in the Swahili verbal paradigms also fit into the general pattern
of hypothesis (19). The presence of -ku- following REL-pronoun infixes of short-stem verbs
augments its plausibility. The appearance of -ku- with the negative of the -me- aspect
(-ja-ku-) turns out to cohere with (19) as well; -ja- is identified as the verb 'come'.
Irregularities in the negative form of the 1st person singular (ni- turning into si-) turn out
to reinforce our reconstruction of -li- as 'be'; the same is true of the form ndi-yo 'yes'
(historically *n-li-yo 'it-be-REL') vs. si-yo 'no'. Historically
it-be-REL
the negative form was *n-li-yo, where the old negative
it-be-NEG-REL
suffix followed the verb "be". (Nasal prefixes regularly zero out in Swahili before voiceless
consonants). The post-verbal negation pattern is still found in some tense-aspects in
Swahili.
 Fad, Logic and Method 301

purported to explain (viz Skinner). Such principles turned out, on closer

examination, to be mere labels for the facts, extending not an inch
beyond those facts.
Extreme deductivists, on the other hand, have laid explicit claim to a
rather formal notion of explanation: deduction from the general rule.
While formally legitimate, this is nonetheless the most impoverished
species of explanation. Still, it does manage to identify, correctly if im¬
plicitly, one important pragmatic aspect of the enterprise:

(23) Explanation as contextualization:

"Explanation, at its most general, is indeed the viewing of
the facts to be explained in a wider context; it fits facts in a
larger pattern".

As noted throughout, pragmatics is the method of contextualizing

hitherto isolated chunks of knowledge, of abducting their likely con¬
nectivity, of discovering how they fit in a wider pattern. Since general
rules, or explanatory hypotheses, are not discovered by deduction, the
most deduction can achieve is the testing of the relationship between a
general rule and the instances it purports to generalize.

8.4.2. Deductivism: Generalization as explanation

As unsatisfactory as it may be as explanation, deduction from a

general rule must be viewed as a weak instance — abstract schema — of
explanation. It is weak, indeed impoverished, since by citing the rule as
part of the explanation one merely acknowledges that the instance
belongs to a class, whose members all abide by the same rule. That
leaves open the more interesting question, namely why is the rule the
way it is?
The most extreme deductivist approach to explanation may be found
in either Popper's (1959, p. 59) or Hempel and Oppenheim's (1948, p.
11) schema of causal explanation, within which a puzzling fact, yet to
be explained, is — causally or otherwise -- linked to another fact (called
antecedent condition), via some general rule. The antecedent condition itself
requires no explanation, since it is already understood (i.e. contextual¬
ized) within the prevailing general pattern. The explanandum is said to
be deduced from the explanans; the latter combines the 'antecedent
condition' plus the general rule:29

29 Simplified from Hempel and Oppenheim (1948, p. 11).

302 Mind, Code and Context

(24) Schema of deductive explanation:

''

antecedent conditions
r-\ > explanans
deduction ^general rule

u puzzling facts explanandum

The potential poverty of deductive explanation of the type (24) be¬

comes obvious when one notes that from the general rule All Greeks are
bald and the antecedent condition Socrates is Greek, one may 'deductively
explain' the newly noticed fact that Socrates is bald. This 'explanation' in¬
deed contextualizes — or connects -- the fact of Socrates' baldness with
the fact of his being Greek. But as to contributing much to our under¬
standing — causal or otherwise -- of Socrates' baldness (or, for that mat¬
ter, of the baldness of all Greeks), this deductive 'explanation' is more
than a trifle disappointing. In a real science, this mode of 'explanation'
is only a preliminary step, indicating a certain -- hopefully temporary —
conceptual poverty, not to mention a measure of downright circularity.30
One may as well note that a species of deductive explanation in fact
crops up occasionally in avowed inductivists' feeble attempts to
'explain'. Thus consider:

(25) individual X is fittest

r > explanans
deduction _ only the fittest survive

^ individual X survived explanandum

The circularity in (25) arises from the fact that the 'cause' and 'effect' are
not independent facts, but merely two names for the same fact.
A somewhat similar conceptual poverty crops up in the behavioral
psychologist's causal explanation of rote-learning:31

30 For discussion of such circularity in deductivist linguistics, see Derwing (1973, Part III), or
Givon (1979a, ch. 1).
31 As noted by Chomsky (1959) in his review of Skinner (1957).
 Fact, Logic and Method 303

(26) X repeatedly performed task Y

r- <
explanans
deduction repetition reinforces learning ,

X learned pattern Y explanandum

In the same vein, the generative grammarian may claim to explain some
facts of language by citing the general formal rule of which they were an
instance, as in:32

(27) structure X is a complex-NP

r— < > explanans
deduction one cannot copy out of complex-NPs

it is impossible to copy out of explanandum

structure X

In summing up the difficulty raised by assuming that deduction from

general rule is much of an explanation, Scriven (1962) says:

"...Hempel and Oppenheim's first mistake, then, lies in the supposi¬

tion that by subsumption under a generalization one has automatically
explained something, and that queries about this "explanation" represent
a request for further and different explanation..." (1962, p. 97)

A more serious criticism of the deductivist's explanation-via-rule is of

course that it mis-represents both the directionality and mode of ex¬
planation in much the same way as it misrepresented the directionality
and mode of scientific discovery: In real science, explanations are con¬
structed abductively, in the absence of a general rule. Indeed, explana¬
tion is the process of hypothesizing the general rule, thus making a
necessary connection (the 'rule') between some puzzling facts (the
'effect') and some old facts (the 'cause'), facts that until then were con¬
sidered unrelated. The deduction schema in (23) is in fact the testing
stage of an explanatory hypothesis, not its discovery.

32 This type of argument remains common in deductivist linguistics. For discussion, see
Givon (1979a, ch. 1) or Derwing (1973, Part III).
304 Mind, Code and Context

8.4.3. Causal explanation: Co-occurrence, dependency

and complex patterns
"...The belief in causality is metaphysical, It is nothing but a typical
metaphysical hypothesization of a well-justified methodological
rule -- the scientist's decision never to abandon his search for
laws..."

K. Popper The Logic of Scientific Dis¬

covery (1959, p. 248)

8.4.3.1. Preamble
Traditional deductive-inductive philosophy of science, with its heavy
reliance on classical physics for examples of scientific practice, naturally
selected simple, chain-like mechanical causation as its paradigm case of
explanation. Here is how Russell (1948) describes it:

"...Inference from experience to the physical world can...be justified

by the assumption that there are causal chains, each member of which is
a complex structure ordered by the spatio-temporal of corn-presence...All
members of such a [causation] chain are similar in structure..." (1948, p.
244; emphasis is mine; TG)

The key term in justifying the inference of a causal relation between

events is Russell's spatial and/or temporal com-presence, which one
might as well re-christen co-occurrence. But as simple as that may seem
at first, causation turns out to be a theory-laden notion, a construct that
is inferred, often in complex and indirect ways, from various co-occur¬
rences and, in particular, from their asymmetries.

8.4.3.2. The context-dependent, theory-laden nature of ’cause’

As Hanson notes, members of a 'simple causation chain' are all,

presumably, "...discrete events bound to neighbouring events very
much like themselves..." (1958, p. 50). In an eloquent dissection of the
ontology of causation, Hanson shows why this simple chain-type model
of causation is a caricature of science, either in physics or in the more
complex biological and behavioral sciences:

"...what we refer to as 'causes' are theory-loaded, from beginning to end.

They are not simple, tangible links in the chain of sense experience, but
rather details in an intricate pattern of concepts..." (1958, p. 54; emphases
are mine; TG)
 Fad, Logic and Method 305

Further:

"...This is the whole story about necessary connection. 'Effect' and

'cause', so far from naming links in a queue of events, gesture towards
webs of criss-crossed theoretical notions, information, and patterns of ex¬
periment...The notions behind 'the cause x' and 'the effect y' are intel¬
ligible only against a pattern of theory, namely one which puts guaran¬
tees on inferences from x to y..." (1958, p. 64; emphases are mine; TG)

8.4.3.3. The multi-variable empirical environment

At the bottom of adopting the simple causal chain as explanation lies

the assumption that events are mono-causal, and that the coherence
structure of complex, multi-factored domains can be reduced to the
simple mechanics of cause-and-effect. But even in the relatively simple
mechanical domain of billiard games, Hanson points out, "...causal-
chain accounts of their performance are oversimplified..." (1958, p. 53).
The idea that a simple causal relation can always be established is an
over-simplification in complex, multi-variable science. In such a science,
often many factors link onto each other in a 'criss-cross pattern'
whereby identifying any one -- or any group -- as the cause of any other
fact or group, simply doesn't reflect the complexity of the pattern. Fur¬
ther, as Hanson (1958) notes, the only reason why a complex science can
sometime give the misleading impression of indulging in simple causal-
chain explanation is because of the practice of controlled experiments.
Controlled experiments are the typical fare of complex, multi-variable
science, where causal connections and other dependencies cannot be
determined without extreme abstraction from real-world conditions. In
a controlled experiment, all factors except two are held constant, effec¬
tively removed from consideration. One of the remaining two factors is
then manipulated (i.e. its value varied), and the effect of the manipula¬
tion on the value of other variable is recorded. If the results are asym¬
metrical in a way that manipulating x registers an effect on y, but not
vice versa, one is justified in calling x the independent variable
('cause') and y the dependent variable ('effect').
In the next experiment, the procedure is repeated, but at least one of
the variables is replaced by a hitherto untested one. Hanson further
notes:

"...To characterize such an enterprise as 'this happens, then that, then

those things take place, which results in...', is a bad caricature..." (1958,
p. 67)

Consider, lastly, the complex problem of the Swahili verb paradigms

306 Mind, Code and Context

(section 8.3.6., above). In what sense is any one of the facts unearthed in
the course of trying to explain the initial 'puzzling fact', the 'cause' of
that initial fact? The answer is that the dependency relations among all
those facts are much too rich and complex to be expressed as a simple
causal chain. This is not to say that we did not make a number of causal
hypotheses within the general pattern. But the general pattern did not
consist of a single chain of causal relations.
\ . »

8.4.3.4. The bio-ontology of causal reasoning

As noted above, there appears a curious agreement across the

philosophical spectrum concerning the supposed metaphysical status of
the notion 'cause'. All philosophical schools seem to agree that 'cause' is
neither inductively nor deductively supported. In this section I would
like to outline briefly why the belief in causation is neither irrational, nor
merely a convenient 'psychological' metaphor. Rather, it is a complex
bio-adaptive abduction that organisms must make, in their attempt to
predict the behavior of animate and inanimate entities in the environ¬
ment, entities whose behavior bears crucially on the organism's survival.
Much like the scientist's search for cause, as given in Popper's meth¬
odological imperative, the organism's causal reasoning is a survival im¬
perative, a necessary component in an empirical interpretation of
information. Much like the scientist, the organism's causal reasoning dis¬
plays an eclectic mix of inductive, deductive and abductive ingredients.
The irreducible inductive core of causal reasoning is the observation
concerning a recurring pattern of temporal-sequential order of types of
events (or entities stereotypically associated with them). This is indeed
Russell's (1948) com-presence, except that is rather pre-presence.

(28) Temporal precedence:

"In order for event-type A to be the 'cause' of event-type B,
tokens of event-type A always precede those of event-type B".

Sequentiality is, however, a necessary but not sufficient ingredient.

The next necessary ingredient in causal reasoning is the observation of a
certain logical dependency relation between the occurrences of tokens
of A and B. This involves the asymmetrical conditional distribution in
the temporal relationship between the tokens of event-types A and B:

(29) Skewed conditional distribution:

" In order for event-type A to be the 'cause' of event-type B, it
must be the case that whenever a token of B occurs, a token
of A has always preceded it; but sometimes a token of A oc¬
curs and no token of B follows".
 Fad, Logic and Method 307

The two event-types now display the deductive-logical relationship of

dependency, whereby B is dependent on A, but A is independent of B.
But causal reasoning carries added, complex, abductive-inductive
analytic ingredients, whose survival-based motivation is readily ap¬
parent. Most central is the division of entities in the environment into
animate-agents and inanimates. The latter are bound by the (essentially
Newtonian) laws of inertia. The former are not. The next ingredient is
an abduction concerning internal vs. external motivation for events (i.e.
observed changes in the environment) that involve both types of en¬
tities:

(30) Agency and motion under own power:

(a) "All entities that move -- i.e. are involved in salient

events/changes -- without any external 'cause' to
motivate their motion, must be agents, capable of ac¬
ting/ moving under their own power".

(b) "Otherwise entities are inanimate, i.e. incapable of in¬

itiating their own motion/action. When such entities
are involved in events/changes, an external cause must
be postulated".

Von Wright (1971) notes that our notion of the 'cause' of behavior
rests upon introspection. This is, strictly speaking, a more appropriate
observation about our notion 'agent': It is unlikely that 'intent', the
central ingredient of 'agent', could ever be induced from external obser¬
vations. But the interplay between this introspection and an inductive-
abductive Newtonian notion of inertia is a crucial analytic step:

(31) Inertia vs. motion: Norm and counter norm:

(a) "The majority of entities are inanimate and obey the

law of inertia. They are the normative ground. When
they move, an external cause must be sought to explain
their motion".

(b) "Only a relative minority are agents, capable of counter¬

inertia behavior. They are the counter-normative
figures. When they move, no external cause need be
sought".

A philosopher may of course wish to argue that there is no logical

motivation for such observations. For the bio-organism, however, the
adaptive, survival value of this progressive abductive-inductive anal¬
ysis is rather obvious:
308 Mind, Code and Context

(32) Agency, counter-inert motion and survival:

(a) "Animates-agents are either potential predators or prey.

One must be on one's guard whenever agents are
around, since their behavior ('motion', 'action') cannot
be predicted from the observable environment".

(b) "Inanimates are safer, they tend to stay put, or else their
limited motion is predictable through normative inertia
considerations, and from more visible environmental
clues".

The introspective judgement concerning intended acts, the division of

events into externally motivated and intended events, and the division
of entities into agents and non-agents, are all motivated by survival im¬
peratives of the biological organism. The scientist's information-process¬
ing behavior merely recapitulates analytic patterns grounded in the very
dawn of biological evolution.

8.5. Functional explanation

"...if a piece of wood is to be split with an axe, the axe must of

necessity be hard; and, if hard, must of necessity be made of
bronze or iron. Now exactly in the same way the body, which like
the axe is an instrument - for both the body as a whole and its
several parts individually have definite operations for which they
are made; just in the same way, I say, the body if it is to do its work,
must of necessity be of such and such character..."

Aristotle, De Partibus Animalium (in Mc-

Keon, ed., 1941; emphases are mine;TG)

8.5.1. Preamble

The process of constructing an explanation -- even when partially

masked by the deductivist's strictures — is a profoundly pragmatic
enterprise. It depends crucially on abducting the proper context within
which facts fit, fall into pattern, make sense. Seeking the proper context
for hitherto isolated facts is, in principle, an enterprise that cannot
depend on either deductive or inductive logic. It demands abducting
connections outside the extant organized domain, across uncharted
regions, or into other sub-systems whose relevance to the domain is yet
to be shown. The contextual-pragmatic nature of explanation becomes
 Fact, Logic and Method 309

even more obvious when one transcends the pristine realms of physics
and chemistry, with their relatively simple causality, and moves on into
the messier study of bio-behavior.
The special challenge of functional explanation in biology and behav¬
ior has not gone unnoticed by deductivists. In non-biological science, the
deductivists take it for granted, why X? questions could only mean what
was the cause of x? In the biological domain, however, explanations that
cannot be so easily reduced to simple causation have been considered
meaningful, as a matter of course, for two millennia. Here is how Hem-
pel (1959) describes, as neutrally as possible, the view advocating the
uniqueness of functional explanation in bio-behavioral science:

"...In the exact physical sciences, according to this view, all explana¬
tion is achieved ultimately by reference to causal or correlational antece¬
dents; whereas in psychology and the social and historical disciplines --
and, according to some, even in biology — the establishment of causal or
correlational connections, while desirable and important, is not suffi¬
cient. Proper understanding of the phenomena studied in these fields is
held to require other types of explanation. Perhaps the most important
of the alternative methods that have been developed for this purpose is
the method of functional analysis, which has found extensive use in biol¬
ogy, psychology, sociology and anthropology..." (1959, p. 121; emphasis
is mine; TG)

The description of the 'method' of functional explanation offered by

Hempel is loaded in one crucial respect: It implies that somehow func¬
tional explanation is itself a different species of explanation from 'the es¬
tablishment of causal or correlational connections'. A biologist or be¬
havioral scientist may as well counter as follows:

(33) Species of explanations:

(a) Causal connection are just one species of connection ob¬

taining between phenomena and their wider context;

(b) Correlation is merely the heuristic from which the scien¬

tist proceeds to abduct an explanatory connection; it is
another name for Russell's observed spatio-temporal corn-
presence;

(c) Functional explanations are different from causal ones

merely in that the context to which a phenomenon is
linked via such explanations is some biological or be¬
havioral function.
310 Mind, Code and Context

8.5.2. Structure and function in biology

Since, as Aristotle observed in rejecting structuralist interpretation of

biology, the notions function and instrumentality are peculiar to the bio-
behavioral domain, a few words of introduction to that domain are in
order. In physics or chemistry, the notion of a complex system, made up
hierarchically out of smaller sub-parts is of course familiar. One strives
to elaborate the connectivity within such a system, quite often discover¬
ing causal connections among its sub-parts.
Causal explanation of this simple type is also well known in biology,
and is practiced extensively at the more mechanical, physico-chemical,
lower levels of biological organization. In both physics and biology, thus,
one may legitimately ask why X? with the sense of what caused X?. In ad¬
dition, however, in the biology one may ask why questions in yet
another sense, one that is without precedence in physics or chemistry.
This sense may be given in the least loaded terms as follows:

(34) "Why is organ — or organism -- X structured the way it is?

That is:
(a) What functional requirements motivate — i.e. 'correlate
to' -- that structure?
(b) How does in perform its function?
(c) How does its function constrain its structure?"

These are the most common questions, standard fare, in the ex¬
planatory study of biological design.33 They go to the heart of the com¬
plex relationship between environmental adaptation, individual be¬
havior, population dynamics, and the evolutionary process through
which bio-organisms - and socio-cultures - are shaped and re-shaped.
The hidden reference buried inside function-oriented questions such
as (34), as Hempel and Oppenheim (1948) correctly note, is of course to
a certain implicit teleology. While inorganic physical facts are simply
there, or are there for causes, bio-behavioral facts are there for reasons.
And those reasons may not be dismissed as either 'metaphysical' or
'anthropomorphic imputations', existing only in the interpreter's mind
(see section 8.5.4., below). Rather, they are the theoretical construct
peculiar and appropriate to the phenomenological domain of bio-be¬
havior, explanatory hypotheses about the adaptive behavior of individual
organisms or species.

33 And if one substitutes socio-cultural pattern7 (’culture7) or ’grammatical structure7

('grammar') for 'organ7 ('organism'), these become immediately the kind of explanatory
questions relevant to anthropology or linguistics.
 Fact, Logic and Method 311

8.5.3. Some features of bio-behavior

8.5.3.1. Adaptive tasks and survival imperatives

The adaptive behavior of organisms must be viewed ultimately

within the context of one governing 'meta-goal': The survival impera¬
tive. Organisms, species and socio-cultures seem to behave as if they
pursue this ultimate meta-goal. In pursuing their meta-goal, organisms
(and socio-cultures) proceed to create a complex system. That is, they
break the meta-goal down into a hierarchy of sub-tasks. Higher sub¬
tasks are, for example:

(35) Higher biological sub-tasks:

(a) the survival of the individual (metabolism, defence);
and
(b) the survival of the species (procreation).

These higher sub-tasks in turn are further broken down into their sub¬
component such as, for example:

(36) Lower sub-tasks ('functions'):

(a) ambulation (defensive, foraging, social)
(b) feeding (foraging, ingestion, digestion)
(c) synthesis (energy, growth, maintenance)
(d) waste elimination (solid, liquid)
(e) transmission (breathing, circulation)
(f) information processing (perception, cognition)
(g) reproduction (sexual, vegetative)
(h) defense (protective, offensive)
etc.

Roughly at this level in the hierarchy of functional organization, specific

anatomical and histological structures — organs -- begin to display their
systematic correlation -- sometimes even isomorphism -- to particular
functions ('sub-tasks').

8.5.3.2. The law-governed behavior of bio-organisms:

Uniformity and adaptive choice
A population of inorganic physical entities of the same kind - say
molecules of the same chemical compound or balls of the same material,
size, weight and shape -- obeys the laws that govern their behavior in a
relatively uniform, seemingly exceptionless fashion. Every member of
such a population, when put under identical conditions, is expected to
312 Mind, Code and Context

'behave' the same way. This exceptionless obedience to Taws' also char¬
acterizes the workings of the lower levels of complex bio-organization,
where similar causes uniformly produce similar effects.
There is a radical gap between this type of mechanical causation and
the way organisms -- or their populations -- abide by functional-adap¬
tive Taws'. They behave as i/they make choices. Sometimes they choose
to 'disobey7 a functional-adaptive Taw', and consequently 'take the
penalties' (become mal-adaptive or extinct). Often, however, there are
more than one structure-type that can perform the same function; or
more than one behavior-pattern capable of yielding the same adaptive
ends. Both structurally and behaviorally, bio-organisms seem to have a
certain range of adaptive options. Let us illustrate this with a few
simple examples.
The sub-task of reproduction may be achieved, structurally and be¬
haviorally, either vegetatively ('cloning') or sexually. And within the
two major options a great number of minor variations are possible and
indeed attested. Some lower-level organisms alternate between the two
modes, depending on environmental -- contextual — considerations.
Such choice may be labeled as purely deterministic. But other or¬
ganisms have 'opted for' either one mode or the other. And while the
history of evolution may yet reveal the reasons that affected such a
'choice', a strategic choice among available options was nonetheless
made, based upon an evaluation of the options within some environ¬
mental context.
The function of ambulation outside water may be achieved by either
walking, crawling or flying, and each one of those in turn allows a great
number of structural-behavioral variants. Did birds 'have to' take to the
air? Were snakes 'impelled' to atrophy their limbs and slither? Simple,
deterministic 'causal' interpretation cannot quite explain why close rela¬
tives have opted for other alternatives.
The function of self-defence may be performed in a myriad of struc¬
tural-behavioral ways, some offensive, other defensive, some active
others passive, some strategic others tactical. Closely-related species
often chose widely divergent structural or behavioral ways of perform¬
ing this function.

8.5.3.3. Bio-organisms as interactive systems:

Mutual dependencies and cross-constraints

When an organism 'opts for' a certain structural-behavioral solution

for performing one sub-task, other sub-tasks within the organism, even
if not directly associated with the performance of the same task, seldom
remain wholly insulated from that choice. Rather, they tend to interact
with it, or constrain it, in ways that are peculiar to bio-systems. Unlike
 Fact, Logic and Method 313

their non-biological counterparts, be those complex and hierarchic, bio¬

systems can seldom be described as predictable sums of their parts.
Their sub-parts are most often functionally interactive, a mode of inter¬
action that has no exact analog in non-biological systems, however com¬
plex, hierarchic or causally-connected they may be. Further, the func¬
tional interaction between sub-parts of a bio-system is a matter of
degree, depending on the relevance relations between the various sub¬
tasks.
Consider, for example, aviary flight-motion. It is usually correlated
with a certain streamlined body surface, aero-dynamic shape, adapted
limb design, weight distribution, bone structure, feeding habits, meta¬
bolic rate, breathing rate and heart-beat rate, defensive and reproduc¬
tive behavior. Some of these features are more directly relevant to flight;
others are less relevant to flight but more relevant to other sub-tasks.
They all interact and constrain each other. And, above all, they are rigid¬
ly constrained by their general context -- vertebrate design.
In contrast, other sub-tasks -- say, sugar metabolism or the liver's syn¬
thetic function — are much less relevant to flight, thus much less likely
to interact with or constrain it. And when flight-motion is embedded in
another bio-design context -- say insects -- different systemic constraints
exert their pressure. The insect flight-system and its mutually-constrain¬
ing contributory parts present at best a weak analog to aviary flight, a
radically different 'solution' — functionally as well as structurally -- to
the same general task.

8.5.3.4. Higher-level interaction governing bio-behavior

What makes bio-behavioral systems even more complex, as a domain

for empirical investigation, is the involvement of several pair-wise
meta-interactions. Among those are:

(37) Meta-interactions in bio-behavior:

(a) Organism vs. the environment
(b) Organism vs. the social group
(c) Organism (or social group) vs. the species
(d) The metabolic body (phenome) vs. its genetic code
(genome)
(e) Lifetime learning vs. evolutionary change
(f) Evolutionary change vs. ontogenetic development
(f) Physical structure vs. user's program
(g) Individual behavior vs. the group's culture

All these pair-wise meta-interactions, and the complex multiple inter¬

actions among them, set the bio-behavioral empirical domain even fur-
314 Mind, Code and Context

ther apart from the non-biological sciences. They make it impossible to

interpret bio-behavior as a variant — or analog -- of the non-biological
physical domain and its simple mechanical causation. S.J. Gould com¬
ments on this point in his critique of those who consider adaptive be¬
havior as a matter of mere genes:

"...the fascination generated by Dawkins'34 theory arises from some

bad habits of Western scientific thought -- from attitudes (pardon the
jargon) that we call atomism, reductionism, and determinism. The idea that
wholes should be understood by decomposing into "basic" units; that
properties of microscopic units can generate and explain the behavior of
macroscopic results; that all events and objects have definite, predict¬
able, determined causes. These ideas have been successful in our study
of simple objects, made of few components, and uninfluenced by prior
history.... But organisms are much more than amalgamations of genes.
They have a history that matters; their parts interact in complex ways.
Organisms are built by genes acting in concert, influenced by environ¬
ments, translated into parts that selection sees and parts invisible to
selection. Molecules that determine the properties of water are poor
analogues to genes and bodies. I may not be a master of my fate, but my
intuition of wholeness probably reflects biological truth..." (Gould, 1980,
pp. 77-78)

In sum, biological adaptation to survival imperatives is a complex,

hierarchic, interactive, multi-factored process; the way bio-organisms
are said to 'obey' functional 'laws' is much more complex, much less
mechanistic, and not as strictly exceptionless, as the way inorganic ob¬
jects or molecules obey physical or chemical laws. But this difference,
and the enormous complexity involved, should not obscure the fact that
adaptive motivation and functional explanation are just as legitimate
hypotheses in the study of bio-behavior as the laws of physics or
chemistry are in their respective domains.

8.5.3.5. Some examples of law-governed behavior in bio-culture

To illustrate how function-mediated bio-behavior may abide by

general Taws' without obeying them in the strictest sense, consider the
following brief examples from language.

8.5.3.5.1. Local vs. global adaptive behavior

In general, grammatical morphology within the word is overwhelm¬

ingly linear at its historical point of origin. Morphemes precede or fol-

34 Gould was criticizing R. Dawkins' (1976) The Selfish Gene.

 Fact, Logic and Method 315

low each other in order, each signalling a single meaning ('function').

This reflects two general functional-adaptive 'laws' of natural com¬
munication:

(38) a. Code transparency:

"One form, one meaning"

b. Linear processing channel:

"Process one meaning-bearing element at a time"

Both Taws' are motivated by some basic constraints on the various com¬
ponent of the communication system — neurological, perceptual, cogni¬
tive, linguistic. They are 'obeyed' in the main, with numerous exceptions.
Some languages, such as Turkish or Swahili, tend to obey these laws
rather strictly, as evident in the highly regular Swahili verb paradigm:

(39) a-li-on-a
he-PAST-see-INDIC
'He saw'

Others, such as Hebrew, collapse pronoun, tense and verb stem into a
single portmanteau morpheme:

(40) ra'a
he/see/past
'He saw'

And when one begins to sift through exceptions to Taws' (38a,b), one
recognizes a certain pattern: The 'violations' of such laws tend to arise
over time, as historical developments that are motivated by other
Taws'. Those other Taws' are in turn just as functionally motivated, but
by other communicative functions. When a communicating organism
strives to fulfill those other functions, the end product is often in con¬
flicts with Taws' (38a,b) above.
Put another way, the organism orients its behavior locally when solv¬
ing particular functional problems, hinged in particular sub-domains of
the complex system. But the results eventually reverberate globally,
where they may also display counter-adaptive effects in other parts of
the system. At the time of making an adaptive choice in one functional
sub-domain, the organism is unaware of potential counter-adaptive
ramifications in other sub-domains. But should this make general func¬
tional Taws' such as (38a,b) less valid or less general? Are such Taws'
mere inductive generalization with no theoretical, explanatory status?
Hardly. Rather, the complex, multi-variant nature of the communicative
316 M ind, Code and Context

system, much like elsewhere in the bio-organism, conspires to make

generalizations appear less than law-like, as compared to physics.35

8.5.3.5.2. Conflicting motivations

Consider next the status of typological generalizations in language.

Recall the functionally-motivated generalization discussed in Chapter 6:

(41) "More important information precedes less important infor¬

mation".

As noted there, this 'law' is motivated by a cognitive principle that

governs the allocation of attention to string-initial element, a principle
backed by a wealth of experimental evidence. 'Law' (41) seems to apply
most generally to languages with flexible word-order. But one may
show that it is also reflected in rigid-order languages, such as SVO
(English), VSO (Mayan) or SOV (Japanese). In all these language types,
the subject (S) precedes the object (O); and the subject indeed tends to
code the more important ('topical') referent in the clause. But how about
the -- admittedly few -- languages that exhibit the rigid orders VOS
(Malagasy) or OVS (Hixkaryana) — where the subject follows the object?
They surely 'violate' our functionally-motivated Taw' (41). However,
these language-types abide by another, equally well motivated, order¬
ing principle, one also discussed in Chapter 6:

(42) "Less predictable information precedes more predictable in¬

formation"

Now, since the subject is both thematically more important and anaphori¬
cally more predictable, its relative ordering seems to engender a conflict
between Taws' (41) and (42). SOV, SVO and VSO languages solve this
conflict by opting to go with the importance principle (41). While OVS
and VOS languages solve it by opting to go with the predictability prin¬
ciple (42). And the general validity of the two laws need not be dimin¬
ished as a result.

8.5.3.5.3. Historical change in law-abiding behavior

Consider next a related case, that of Biblical Hebrew. It had flexible

ordering of the subject, abiding primarily by principle (42). Thus, for ex¬
ample, definite subjects in early Biblical Hebrew appeared overwhelm-

35 For many more examples of this phenomenon in grammar, see Giv6n (1979a, ch. 6). For its
close equivalents in phonology, see Hyman (1973).
 Fact, Logic and Method 317

ingly after the verb, while indefinite subject appeared predominantly

before it. As a corollary, Biblical Hebrew made use of neither the numeral
'one' nor the existential-presentative construction to mark indefinite
subjects. It simply relied on principle (42). In other words, of two struc¬
tural alternative available to code indefinite subjects -- word-order and
morphology -- Biblical Hebrew 'chose' the one over the other.
In the course of history. Modern Hebrew wound up with the rigid
word-order SVO -- which now does not allow the coding of indefinite¬
ness by word-order. Instead, the language developed the use of the
numeral 'one' — and an existential-presentative construction -- for this
purpose, following the pattern shown for Krio and Mandarin (see
Chapter 5). Is principle (42) now less 'law', when Hebrew abides by it
no more? Was it more 'law' when Hebrew abided by it earlier? And
since, as an SVO language, Hebrew now abides more by principle (41),
is that principle now more 'law', and was it less 'law' when Biblical
Hebrew did not abide by it?

8.5.4. The reductionist attack on functional explanation

8.5.4.1. Motive as ’cause’ in socio-behavior

The notion of 'function' without the teleology of a willful agent has

been — and indeed should be, if one agrees with Gould — an anathema
to reductionist philosophers. In attempting to exorcise the teleological
ghost out of functional explanation, Hempel and Oppenheim (1948)
chose to split the domain into two parts. The first encompasses the
socio-behavioral sciences, and is said to deal with intentional behavior
by identifiable agents, to which conscious purpose, motive or intent
can be attributed. The agent's motives, it is contended, are simply a sub¬
species of the cause component of the explanans in the deductive causal-
explanation schema (see (24) above):

"...motivational explanation, if adequately formulated, conforms to

the conditions for causal explanation, so that the term "teleological" is a
misnomer if it is meant to imply either non-causal character of the ex¬
planation or a peculiar determination of the present by the future. If this
is borne in mind, however, the term "teleological" may be viewed, in
this context, as referring to causal explanations in which some of the an¬
tecedent conditions are motives of the agents whose actions are to be ex¬
plained..." (1948, p. 16)

The problem with this reduction of function to motive is of course that it

excludes from legitimate functional explanation a wide range of cul-
318 Mind, Code and Context

tural, social, and cognitive behavior that is altogether independent of

conscious motivation by the behaving agent. Such behavior is clearly
functional-adaptive in the most simple biological sense, and one cannot
explain it in any but a functional context.
In the vast socio-cultural domain, for example, individuals and cul¬
tural groups behave very much like biological species — as if they have
purpose. As a scientist, one abducts the functional motivation that must
underlie observed social structures and cultural behavior patterns. Such
abducted functions are unimpeachable scientific hypotheses, dealing
with invisible entities that explain visible patterns. Yet there is no
evidence to suggest that individuals or socio-cultures are aware of func¬
tional considerations that motivate much of their behavior. The patterns
evolve historically, as a conflation of myriad small, step-wise, local and
often sub-conscious decisions. Individual local steps must be explained
at a variety of levels within the complex process. The global pattern of
such explanations reveals the mechanisms via which complex structure
and complex function become matched through protracted evolution
('history7). But complex global functions, however legitimately ab¬
ducted as explanation, cannot be reduced to conscious motives of the
'agents'.
Further, as noted in Chapter 7, many physiological, sensory-motor
and cognitive processes in the behavior of higher organisms -- including
humans -- become automated, either during phylogenetic evolution, or
during the individual's lifetime learning. Such functions are clearly
'performed', by or within the organism, without conscious intent. Two
examples from anthropology and linguistics will illustrate this.
Consider first one function that may be performed by at least some
kinship systems,36 the provision for exogamy (avoidance of in-breeding).
It is indeed a theoretical construct, much like its close analogs in biol¬
ogy. But unlike biology, where purely physical structures are often in¬
volved in blocking in-breeding,37 in human socio-culture the task is
achieved via 'symbolic' categorization of kins, a categorization that
guides the mating behavior of individuals. One thus can - and indeed
should — abduce this functional explanation without necessarily assum¬
ing that memebers of the practicing culture are aware of this -- and
other -- functions of the kinship system.
There exists an obvious analogy between the abduced function of ex¬
ogamy of human kinship system, and in particular the way it correlates
with social structure, and the match-up of function with structure in

36 See discussion of the function of kinship systems and related controversies in chapter 9. It
is of course naive to assume that direct biological motivation, such as avoidance of
in-breeding, can furnish the entire interpretation of kinship systems.
37 Some of the blocking devices in plants are already 'behavioral' rather than purely
'structural'. Thus, for example, avocado species vary the time of opening their male and
female flowers 'in order to' insure cross-pollination.
 Fad, Logic and Method 319

pre-human biology. Many plant and animal species display remarkably

similar correlations between the putative exogamy function, and an
array of structural and behavioral 'means' 'designed' to 'achieve' it.
Much like members of human cultures, these organisms are blissfully
ignorant of their 'goal'.
Consider next a simple example from language, where one may
legitimately abduce that the function of having a regular consonant-
vowel alternation in speech, or more generally, the function of dis-
similatory processes in phonology, is to facilitate both speech percep¬
tion and articulation. One could indeed demonstrate experimentally the
counter-adaptive effect of vowel-free or consonant-free phonology on
speech perception, articulation, and even on lexical memory. One can --
and indeed must (if one is to adhere to Popper's meta-imperative of
science) indulge in such explanatory abductions; one can construct ex¬
perimental procedures to test them. One can and indeed must do all
that in the relatively secure knowledge that:

(a) Such functional explanations of dissimilatory processes are

legitimate explanatory abductions in a biological-behavioral
science; and
(b) The speakers involved in the daily performance -- or histori¬
cal development -- of such dissimilatory processes are bliss¬
fully unaware of their 'motives'.38

8.5.4.2. The denial of bio-teleology

In turning to deal with functional explanation in biology, where a
conscious willful agent cannot be invoked, Hempel and Oppenheim
(1948) dismiss the enterprise as an attempt to ascribe anthropomorphic
behavior to non-human systems. To this practice they concede, at best, a
salutary heuristic value:

"...One of the reasons for the perseverance of teleological considera¬

tions in biology probably lies in the fruitfulness of the teleological ap¬
proach as a heuristic device: Biological research which was psychologically
motivated by a teleological orientation, by an interest in purposes in na¬
ture, has frequently led to important results which can be stated in non-
teleological terminology and which increase our scientific knowledge of
causal connections between biological phenomena...Another aspect that
lends appeal to teleological considerations is their anthropomorphic char¬
acter. A teleological explanation tends to make us feel that we really "un-

38 As noted in Chapter 7, grammar is also an automated instrument of information process¬

ing, an instrument whose users are unaware of specific, 'local communicative sub-goals.
320 Mind, Code and Context

derstand" the phenomenon in question, because it is accounted for in

terms of purpose, with which we are familiar from our own experience
of purposive behavior. But it is important to distinguish between under¬
standing in the psychological sense of a feeling of empathic familiarity
from understanding in the theoretical, or cognitive, sense of exhibiting
the phenomenon to be explained as a special case of some general
regularity..." (1948, p. 17 emphases are mine; TG)

This condescending rendition of the motivation behind functional ex¬

planation in bio-behavior may be summarized as follows:

(a) The hidden motives of the scientist indulging in such ex¬

planations are psychological rather than scientific;
(b) Functional hypotheses somehow cannot aspire to the status
of 'general regularity' that more mechanical rules or causal
explanations have attained;
(c) The only worthy type of explanation -- i.e. of contextualiza-
tion — of facts in science is the deductive variety, i.e. the
viewing of the explanandum 'as a special case of some general
regularity'.

What Hempel and Oppenheim seem to be doing is vent their method¬

ological prejudice. They offer no coherent guideline for constructing an
empirical methodology for bio-behavioral science.

8.6. Closure: Explanatory domains, ranges of facts

and the legacy of structuralism
In choosing — via abduction' hypothesis and experience-derived com¬
mon sense — the likely explanatory domains that will serve as context
within which one purports to interpret puzzling facts, one automat¬
ically also chooses the relevant data ranges. This is, one suspects, one
more facet of Hanson's (1958) theory-laden facts. Suppose, for example,
that one aims to explain why the human body is structured — in its
anatomy and histology - the way it is. One then seeks first the most ob¬
vious domains within which those structures interact, or are embedded:

(42) Explanatory domains in biology:

(a) The external environment
(b) Survival imperatives
(c) The genome
(d) Functions performed by the structures
(e) Phylogenetic evolution
(f) Ontogenetic growth
(g) The social group
 Fact, Logic and Method 321

Of these, domains (a) and (b) turn out not independent of each other,
but rather make up a mutually-dependent complex. Domain (c), in biol-
ogy as in linguistics, tends out to be somewhat of a red herring. It mere¬
ly bounces the why question to its rightful next level - domains (e) and
(f). Domain (d) is the bread-and-butter of understanding biological
structures — within the teleological constraints of (a) and (b). Finally, in
many species functional-adaptive physical structures and behavioral
patterns (d) — most conspicuously sexual reproduction, feeding and
defense — have evolved within some social-interactive context (g), and
are constrained by it.
Given such decisions on relevant explanatory domains, one now
proceeds to seek facts from the very same domains: Ecology (a), global
and local function (b, d), genetics (c), evolution (e), embryology (f), and
group behavior (g). In this fashion, pre-empirical, common-sensical, the¬
ory-bound decisions about the relevant explanatory domains automat¬
ically define the range of relevant facts, to be gathered, sifted, correlated
and eventually used for constructing explanation for the initial puzzle.
But one could also choose, if one is so inclined, an alternative method.
One could begin by severely restricting the range of explanatory
domains vis-a-vis which the facts are to be analyzed. In so doing, one
automatically also restricts the range of facts deemed relevant, (a) to the
entire domain, (b) to the abduction of specific hypotheses, and (c) to
deductive-falsificatory testing. In the physical and biological sciences,
the decision to restrict the range of facts is a methodological con¬
venience, used in preliminary, in vitro, controlled experiments. In the so¬
cial sciences, this pre-empirical decision has on occasion been invested
with a theoretical aura, and extended illicitly to the explanatory domains.
Psychology, linguistics and anthropology have all, at one time or
another, fallen victim to recurrent waves of structuralism. In rejecting
functional explanation, structuralists in the social sciences have, at one
time or another, adopted the philosophical stance of either reductionist
school. Inductivist such as Bloomfield or Skinner, in ignoring Popper's
meta-imperative, saw nothing to explain. Deductivists, like Lounsbury
or Chomsky, saw structure as its own explanation. Since, to recast
Kant's dialectic, data not defined by theory is empty, and theory not driven by
data is blind, structuralism tends to whittle empirical science from both
ends.
 '

» .

: -
CHAPTER 9

LANGUAGE, CULTURE AND

TRANSLATION

9.1. Introduction*

"...to imagine a language means to imagine a form of life..."

L. Wittgenstein, Philosophical Inves¬

tigations (1953, p. 8)

Throughout the discussion in various chapters above, we saw how

the interpretation ('meaning' or 'function') of linguistic expressions al¬
ways depends, at least to some extent, on the context within which
those expressions are used. This is true even of the most semantic fea¬
tures of grammar. As Ron Langacker has observed, semantics is conven¬
tionalized pragmatics. The degree to which meaning is conventionalized
may vary, and this is true for both lexicon and grammar. But in human
language no meaning can be one hundred percent conventionalized, for
reasons that have to do with the unpredictability and open endedness of
context. Thus, a residue of context dependence in interpreting linguistic
expressions always remains, regardless of how conventionalized they
may be.
As suggested at various points throughout (but see in particular
chapters 3 and 6), the contexts within which human language is inter¬
preted fall roughly into three major divisions:

(a) The shared deictic context:

(i) Speaker-hearer relations: goals, speech-act value, social
status/power relations, mutual knowledge and recog¬
nized mutual obligations;

*1 am indebted to Dwight Bo linger, Pete Becker, George Grace, Kenneth Pike, Michael Shapiro,
Anna Wierzbicka and - last but not least - my colleague Phil Young for many helpful
comments and suggestions. I have also benefited from the opportunity to present some of the
ideas discussed here to critical audiences at SIL Papua-New Guinea (Ukarumpa, 1985),
Auckland University (Auckland, 1986), Australian National University (Canberra, 1986), the
University of Texas at Arlington (Dallas, 1987) and Universitat zu Koln (Koln, 1987).
324 Mind, Code and Context

(ii) Deixis of the speech situation: you & I, this & that, here &
there, now & then.

(b) The shared discourse ('text') context:

The knowledge shared by speaker and hearer due to prior
transacted information -- the memory-stored discourse.

(c) The shared ('generic') world-vi^w context:

Shared knowledge of the physical and cultural universe, as
coded in:
(i) The (encyclopedic) lexicon;
(ii) Shared conventions of behavior & communication.

In this chapter we will concern ourselves with the third division of con¬
text, the shared generic knowledge, implicit in the fact that the speaker
and hearer are members of the same speech community -- or culture.
The issue of translation -- i.e. the transfer of knowledge across major
linguistic-cultural boundaries -- invades the discussion rather naturally.
Translation -- or paraphrase -- is indeed one of the best test cases we have
for understanding linguistic expressions. The argument for an intimate
connection between translation, understanding and culture runs rough¬
ly as follows:

(1) Translation and cultural world-view:

If understanding a language entails understanding the cul¬

ture's world-view, then cross-language translation is possible
only to the extent that cross-cultural translation is possible.

The inquiry into the possibility of cross-language — thus cross-culture

-- translation opens up an even more vital question, having to do with
the universality of language and thought:

(2) Translation and universality:

If understanding a language requires understanding the cul¬

tural world-view, then cross-linguistic translation is only
possible if cross-cultural differences are relatively shallow,
that is, if the world-view held by different human cultures is
largely universal.

It is thus not an accident that the discussion of translatability coincides,

to quite extent, with the discussion of linguistic and cognitive universal.
In an intellectual climate that fosters extreme reductionism, the ques¬
tion of translation and universality has been approached in stark
 Language, Culture and Translation 325

categorical terms of: Human thought is either universal or not, cultures

either overlap in all but their decorative trivia, or diverge unpredictably.
Translation is either a totally viable proposition, or an altogether futile
mirage. As elsewhere, the pragmatist's position in considerably less
than enviable, demanding the recognition of the murky, impure middle
ground. We will begin our discussion with a survey of reductionist ap¬
proaches in philosophy, anthropology and linguistics.

9.2. Modern extreme reductionists

9.2.1. Preamble
A certain measure of isomorphism again obtains between the tradi¬
tional debate of rationalist vs. empiricist in epistemology, and the debate
on translatability and universals. But the isomorphism is far from com¬
plete, and the relationships between the protagonists are on occasion not
strictly predictable from their proclaimed philosophical roots.
As one may recall, both Plato and Aristotle held unabashed univer-
salist positions concerning the human mind, albeit for radically dif¬
ferent reasons. Plato's universalism sprang from his belief in the
primacy of a universal set of innate mental categories. Aristotle's
universalism was founded upon the postulated primacy of a fixed set of
underlying forms of the external world. Since the human perceptual fil¬
ter was, for Aristotle, a faithful mirror, universality of world gave rise to
universality of mind.
Both Plato and Aristotle, in the best pre-empirical Greek tradition,
discoursed upon universals of language (and mind) on the basis of a
sample consisting of a single language and culture. Neither was overly
concerned with the implications of linguistic or cultural diversity. When
such diversity was eventually acknowledged, a considerable reassess¬
ment of the notion of universals took place among Plato and Aristotle's
spiritual descendants.

9.2.2. The anti-universalism of modern empiricists

"...The world is all that is the case. The worlds is the totality of
facts, not things...The limits of my language mean the limits of my
world..."

L. Wittgenstein, Tractatus Logico

Philosophicus (1918, pp. 7,115.)

Somewhat characteristically, the 'early' Wittgenstein stakes out an

enigmatic high grounds, one that remains open to considerable claim-
326 Mind, Code and Context

jumping by latter-day interpreters. The identification of language with

world may be interpreted as conservative Aristotelian empiricism, akin
to Russell's or Carnap's referential approach to meaning. Given such
reading, the diversity of the physical habitat would presumably deter¬
mine -- in a Darwinian fashion -- a diversity of human languages. One
could certainly interpret the traditional Sapir-Whorf discussion of the
Eskimo's snow vocabulary and the Bedouin's camel terminology as
reflecting such an empiricist approach-
One could, however, read Wittgenstein's 'limits' as the flip side of the
Sapir-Whorf coin: Language -- or cultural world-view — imposes the
limits on the way one could perceive (or 'coherently interpret') the
world. This second interpretation is plausible because of Wittgenstein's
insistence that the world is not the totality of things but rather of facts:1
It is a bit easier to demonstrate the relativity -- i.e. context dependence --
of 'fact' than of 'entity'.2
The linguistic anthropologists of the early 20th century in fact
diverged as to which possible reading of Wittgenstein they were com¬
patible with. In the manner of late 19th Century post-Darwinian social
science, they were rather impressed with the seeming correlation be¬
tween environmental and racial diversity. Further, they were deeply in¬
volved in the study of an immense, baffling diversity of human
languages and cultures. Both the empiricist among them, such as Bloom¬
field, and the rationalists, such as Sapir, of necessity diverged from Aris¬
totle and Plato, respectively, on the question of universality.
Bloomfield himself did not couch his comments in terms of culture or
mind, but rather in terms of linguistic structure, as is apparent in his in-
ductivist credo:

"...The only useful generalizations about language are inductive

generalizations..." [1933, p. 20]

In criticizing the mentalistic universalism of H. Paul and his associates,

Bloomfield writes:

"...even the fundamental features of Indo-European grammar, such as,

especially, the parts of speech system, are by no means universal in the
human speech. Believing these features to be universal, they [H. Paul et
al; TG] resorted, whenever they dealt with fundamentals, to philosophi¬
cal and psychological pseudo-explanations..." [1933, p. 17]

In the classical anti-universalist argument from diversity, Bloomfield

then observes:

1 Popper (1959, p. 35, fn. *1) commended Wittgenstein on this 'slip' of his inductivist habit.
2 See discussion of the 'objectivity' of 'fact' and 'event' in Chapters 3 and 8.
 Language, Culture and Translation 327

"...North of Mexico alone there are dozens of totally unrelated groups

of languages, presenting the most varied types of structure. In the stress
of recording utterly strange forms of speech, one soon learns that
philosophical presuppositions were only a hinderance..." [1933, p. 19]

9.2.3. The anti-universalism of modern mentalists

Unlike Bloomfield, Benjamin Lee Whorf was an unabashed, dyed-in-

the-wool mentalist, and thus, one would presume, a philosophical
Rationalist in the Herman Paul (i.e. German Romantic) tradition. None¬
theless, the so-called Sapir-Whorf Hypothesis on language and culture3
boils down to an extreme relativist ('anti-universalist') position. One
version of the hypothesis, further, is inherently an empiricist position:

(3) Sapir-Whorf I:

(a) Diversity:
(i) Language reflects culture;
(ii) Culture reflects environment;
(iii) Environmental diversity accounts for cultural,
thus linguistic, diversity.

(b) Translation:
(i) Cross-linguistic translation requires cross-cultural
translation;
(ii) but cultures -- thus languages -- diverge immense¬
ly and unpredictably;
(iii) Therefore, neither universality nor translatability
is viable.

One may, however, read another -- early Wittgensteinean — inter¬

pretation into the Sapir-Whorf Hypothesis, one that is indeed more
compatible with philosophical Rationalism:

(4) Sapir-Whorf II:

(a) Diversity:
(i) Cultural world-views, thus also the languages that
codes them, diverge immensely;
(ii) Linguistic categories pre-determine the way we
perceive/construe the world (a Rationalist or Kan¬
tian hedge);

3 See Whorf (1950,1956); Sapir (1949).

328 Mind, Code and Context

(b) Translation:
Given the manifest cultural-linguistic diversity, 'real'
translation is not possible.

Perhaps the boldest expression of cultural-cum-linguistic relativism

can be found in Whorf's discussion of the Hopi Indians' sense of time
and space:
t • *

"...Just as it is possible to have any number of geometries other than

Euclidean which give an equally perfect account of space configura¬
tions, so it is possible to have descriptions of the universe, all equally
valid, that do not contain our familiar contrasts of time and space..."
(1950/1956, p. 58)

The description is startling because it touches upon probably the most

universal, evolutionarily oldest, biologically most in-wired features of
organismic -- and human -- cognition. These are the very features that
Kant recognized as our in-built mental trap, our species-determined
point of view.4 Whorf goes on to explain what prompted him to chal¬
lenge the universality of our time-space schema:

"...[the Hopi] has no general notion or intuition of time as a smooth

flowing continuum in which everything in the universe proceeds at an
equal rate, out of a future, through a present, into a past; or, in which, to
reverse the picture, the observer is being carried in the stream of dura¬
tion continuously away from the past and into the future.
After long and careful study and analysis, the Hopi language is seen
to contain no words, grammatical forms, constructions or expressions
that refer directly to what we call "time", or to past, present, or future, or
to enduring or lasting, or to motion as kinematic rather than dynamic
(i.e. as continuous translation in space and time rather than an exhibi¬
tion of dynamic effort in a certain process), or that even refer to space in
such a way as to exclude that element of extension or existence that we
call "time"...Hence the Hopi language contains no reference to "time",
either explicit or implicit..." (1950/1956, pp. 57-58)

We know now that the analysis of Hopi grammar and lexicon upon
which this radical assertion is based was woefully superficial and mis¬
leading. It depended, in fact, on a rather narrow culture-specific criteria
of what could constitute explicit or implicit reference to "time". None¬
theless, mentalistic cultural relativism of this kind has been extremely
influential in both linguistics and anthropology.

4 See discussion of Kant's Transcendental Schema in Chapter 1.

 Language, Culture and Translation 329

9.2.4. The universalism of modern Rationalists

The rationalist line of transmission -- from Plato via Descartes to
Chomsky and Katz — is more predictable on the issue of translatability.
Modern Rationalists have preserved the Greek-Latin-French tradition of
expounding upon linguistic universals on the basis of a woefully
restricted sample of the world's linguistic diversity. As far as modem
Platonists are concerned, understanding linguistic expressions requires
no reference to context. And, in defiance of one hundred years of cross-
cultural linguistic anthropology, the cultural context -- thus world view
-- is taken as a matter of course to be uniform and unproblematic.
As George Grace (1986) has noted, using terminology adapted from
Austin, some modern rationalists adopt a strict locutionary approach to
translatability, disregarding even structural cross-language diversity. An
example of this is Katz's effability principle:

"...Every proposition is the sense of some sentence in each natural lan¬

guage..." (1976, p. 37)

In an earlier rendition, Katz identifies most succinctly the particular

brand of universalist motivation that prompts his interest in translation:

"...Effability makes a very strong claim about natural languages. What

is so important about this claim that one should take on its defense?
First, there is the fact that without it there is no basis for an interlinguistic
notion of proposition on which to base hypotheses about the relation of
logic and language..." (1972, p. 22; see also Katz, 1978, pp. 219-220; em¬
phases are mine; TG)

As Grace (1986) points out, universalists of Katz's stripe assume that

grammar is a relatively trivial matter, highly interchangeable across lan¬
guages, thus posing only a weak barrier to translation. For every
'sentence' in one language one can find some meaning-equivalent
'sentence' in any other language. As we shall see below, it is relatively
easy to show that this extreme position is untenable.
Grace (1986) goes on to discuss another universalist approach, that of
perlocutionary translation. This approach concedes the existence of
non-trivial structural differences between languages. It then allows for a
sentence of one language to be translated into any 'expression' of
another language, be that expression a single sentence of the same struc¬
ture, a single sentence of a different structure, or a number of sentences
assembled together in whatever configuration. Such a position is char¬
acterized well in the following quotation from Osgood and Scbeok (eds,
1954):
330 Mind, Code and Context

"...This comes out clearly in statements such as "anything can be ex¬

pressed in any language, but the structure of a given language will
favor certain statements and hinder others..:" (1954, p. 193)

Edward Sapir's position on the issue, while not crystal clear, imparts
a surprisingly universalist flavor, given what one would have inferred
from Whorf's extreme relativism: , .

"...It would be absurd to say that Kant's 'Critique of Pure Reason'

could be rendered forthwith into the unfamiliar accents of Eskimo or
Hottentot, and yet it would be absurd in but a secondary degree. What
is really meant is that the culture of these primitive folks has not yet ad¬
vanced to the point where it is of interest to them to form abstract concep¬
tions of a philosophical order. But it is not absurd to say that there is
nothing in the formal peculiarities of Hottentot or Eskimo which would
obscure the clarity or hide the depth of Kant...It is not absurd to say that
both Hottentot and Eskimo possess all the formal apparatus that is re¬
quired to serve as a matrix for the expression of Kant's thought. If their
languages have not the requisite Kantian vocabulary, it is not the lan¬
guages that are to blame but the Eskimo and Hottentot themselves..."
(1949, p. 154; emphases are mine; TG)

First, Sapir seems to view cross-language differences in grammar --

'formal peculiarities' — as relatively trivial, in so far as translatability is
concerned. One would presume a perlocutionary approach to trans¬
latability is what he may have had in mind. Second, Sapir seems to con¬
sider the main obstacle to translation, the lexicon-culture complex, as a
relatively superficial matter of technical vocabulary ('abstract con¬
ceptions'), whose paucity can be easily remedied. Finally, Sapir seems to
ascribe the cultural barriers to translation to primarily the natives' mo¬
tivation, something that is again presumably easy to remedy, given suf¬
ficient incentive.
In the next section I propose to summarily dispense with the stronger
— Grace's locutionary - approach to translation, by showing that cross¬
language grammatical diversity renders this type of translation unten¬
able.

9.3. Grammar and translation

In this section I will attempt to show that if cross-language translation

is possible, it is only possible in Grace's more lax perlocutionary sense.
That is, it is not always the case that a single sentence in one language
can be translated into a single sentence in another language. And in par-
 Language, Culture and Translation 331

ticular, the Western logician's concept of 'proposition' does not always

match the structural facts of sentencehood.5 In attempting to demon¬
strate this, I will survey, in order, a number of test cases.

9.3.1. Serial verb constructions

The phenomenon of serial-verb constructions involves a certain well-

documented diversity in the way languages code unitary 'events'. A
'simple event' that English tends to code with a single verb — thus
presumably single sentences/propositions whose core is that verb -- is
coded in other languages by a 'serial construction' of two or more verbs.
To illustrate this, consider the following contrast between English and
some West African languages:

(5) Serial verb constructions6

a. iywi axva utsi iku (Yatye)

boy took door shut
'The boy shut the door'

b. mo fi ade ge naka (Yoruba)

I took machete cut wood
'I cut (the) wood with the machete'

c. nam utom eemi ni mi (Efik)

do work thisgme me
'Do this work for me'

d. 9 gbara gaa ahya (Igbo)

he ran go market
'He went to the market'

As far as translatability -- and the universality of the notion 'prop¬

osition' - is concerned, serial-verb constructions may be viewed from
two perspectives:

(a) One could consider verb serialization as a relatively trivial

matter of grammatical structure and lexicalization, as sug¬
gested in Givon (1975a). The multi-verb serial constructions in
(5) will be thus viewed as single sentences, coding single men-

5 This is not to suggest that other problems for translation do not arise from the cross-Ian
guage variability of grammar. Indeed, they often do. For some discussion, see Keenan
(1978) or Givon (1978b).
6 From Givon (1975a).
332 Mind, Code and Context

tal propositions; their divergence from their English counter¬

parts is a relatively superficial fact of structure/grammar.
Given this view, Katz-type ('locutionary') translatability is un¬
problematic.

(b) Alternatively, one may view verb serialization as a profound

cross-cultural cognitive difference in event perception, as
suggested in Pawley (1976, 1987), whereby cultures that
practice this type of event coding indeed perceive our uni¬
tary event as a chain of multiple events. Given this view, 'loc-
utionary' as well as 'perlocutionary' translation are prob¬
lematic.

Even when one finds general empirical grounds, independent of

grammar, for rejecting interpretation (b) of verb serialization,7 there re¬
mains the fact that serial-verb constructions sometime display the gram¬
matical-structural trimmings of multi-sentence ('multi-propositional'?)
constructions. Thus, consider the following example from Kalam, a
clause-chaining Papuan language, with a high incidence of verb
serialization.8 While most serial verbs in this language are stripped bare
of any grammatical morphology, some require typical medial verb mor¬
phology on both verbs in the construction, thus technically signalling
two consecutive clauses within the thematic clause-chain:

(6) ...mon dangiy-ek yin-ek sagdi-sap...

fire light-DS burn-DS leave-PRES
'...She lights the fire and leaves...'
(lit.: '...she lights the fire (till) it bums (and) she leaves...'

The subject of the first verb in (6), chain-medial 'light', is 'she', and the
verb is marked by DS ('different-subject', cataphoric switch-reference)
morphology. The subject of the second verb, chain-medial 'burn', is
'fire', and the verb is again marked by DS morphology. The subject of
the third verb, chain-final 'leave', is once again 'she', and this verb is
marked by finite morphology. The grammar treats the first two clauses
as independent. But in text they and their equivalents appeared
predominantly as fixed expressions, semantically co-lexicalized, much
like the Mandarin resultative verb compounds that display no verb
morphology (Thompson, 1973).

7 See Givon (1987).

8 See Pawley (1966) for a general description of Kalam. The data is from my own field notes,
with generous help from Lyle Scholz and Andy Pawley.
 Language, Culture and Translation 333

9.3.2. Pre-posed adverbial clauses in strict verb-final language

As noted recently by Thompson (1985) and Ramsay (1987), the dis¬

course function of pre-posed adverbial clauses in English is very dif¬
ferent from that of post-posed adverbial clauses. Typical examples of
the two ordering types are:

(7) (a) Pre-posed:

...To do that, you pull the lever and release the bolt...

(b) Post-posed:
...you pull the lever to release the next bolt, then...

The functional differences between constructions such as (7a) and (7b)

above may be interpreted, among other things,9 as reflecting the degree
of integration ('coherence') of the purpose clause vis-a-vis its main
clause. What is important for the present discussion is that pre-posed
adverbial clauses seem to have stronger referential and thematic con¬
nections to the preceding discourse, and a much looser referen¬
tial/thematic connection to their main clause (in spite of their obvious
semantic relation to the latter). Pre-posed adverbial clauses thus display
what Haiman (1978) has termed -- for 'if' clauses -- topic characteristics.
Further, structurally, pre-posed ADV-clauses tend to be separated from
their following main clause by a comma (i.e. pause).
Post-posed adverbial clauses, on the other hand, are much more in¬
tegrated — thematically, referentially and semantically — with their
preceding main clause. They are also more integrated with it structural¬
ly, and usually follow it without a pause. One could thus argue that
both thematically and structurally, pre-posed adverbial clauses are
'more independent', while post-posed ones are more 'part of the main
clause'.
The cross-linguistic distribution of the ordering of adverbial clauses is
skewed in the following way:

(a) All language seem to have pre-posed ADV-clauses;

(b) Many strict verb-final languages10 allow no post-posed ADV-
clauses.

Thus, the meaning (or 'function') of the more-integrated post-posed

9 See for example Ramsay (1987) as well as discussion in Givon (1986a).

10 Such as, for example, the clause-chaining Tibeto-Burman languages of Nepal (Newari,
Sherpa), or New Guinea Highlands languages such as Hua (Haiman, 1979) or Chuave
(Thurman, 1978), inter alia.
334 Mind. Code and Context

ADV-clauses, as in (7b) above, cannot be conveyed in languages of type

(b). Their grammar simply makes no provision for it.

9.3.3. Verb complement structure

In English, there is a systematic meaning difference between the com¬
plement-taking constructions of 'want' and 'wish', as in:11

(8) a. She wanted him to leave

b. She wished that he would leave

The syntactically more integrated single-clause-like expression in (8a)

usually implies some direct verbal contact between the subject of 'want'
and the subject of the complement clause, thus an attempted verbal
manipulation. On the other hand, the more loosely-bound two-clause
structure in (8b) is more likely to characterize a private wish, with no
direct contact and no attempted manipulation. In some languages
'want' is a single lexical item that can take only the looser construction
type (8b), never the tightly-bound (8a). Spanish and Israeli Hebrew are
such languages:

(9) a. querfa que se-fuera (Spanish)

wanted-3s that REFL-go-SUBJUNCT-3s

'She wished that he would leave'

b. *le querfa (a) ir

him wanted-3s (to) go/INF

c. hi ratsta she-hu yelex (Hebrew)

she wanted-3sf that-he 3sm/FUT/leave
'She wished that he would leave'

d. *hi ratsta oto/lo la-lexet

she wanted him/DAT-him to-go/INF

Now, even if this restriction in Spanish and Hebrew is only a superficial

fact of grammar, and if (9a) and (9c) above are in fact ambiguous and
code the meaning of both English (8a) and (8b),12 still we face a clear-cut
situation whereby the single-clause English construction (8a) finds its

11 See discussion in Chapter 3, above, as well as Givon (1980a).

12 There are indeed grounds for believing that Spanish and Hebrew simply cannot express
the shade of meaning of the English (8a) with the verb 'want', as distinct from 'She told him
to leave'.
 Language, Culture and Translation 335

Spanish and Hebrew translation in much looser two-clause construc¬

tions. In other words, locutionary translation is not possible.

9.3.4. Co-lexicalized complement-taking verbs

In some languages, modality and manipulative verbs do not appear

as separate words. Rather, they are co-lexicalized together with the com¬
plement verb, appearing together as a single word. This represents one
further step of structural condensation beyond the English example
(8a), above. As illustration, consider the following from Yaqui (Jelinek,
1986):

(12) a. inepo siim-pea

I lea v e-wish
T wish to leave'

b. aapo Mary-ta siim-'z'a

he Mary-OBJ leave-want
'He wants Mary to leave'

c. inepo Mary-ta siim-sae-n

I Mary-OBJ leave-feZZ-PAST
'I told Mary to leave'

d. apo'ik-ne siim-tua-k
him-I leave-ma/cc-PERF
'I made him leave'

e. apo'ik-ne siim-tua-tevo-k
him-I leave-ma/ce-flsfc-PERF
'I had him leave'

f. enci-ne yi'i-mahta-k
you-I dance-teac/z-PERF
'I taught you to dance'

g. inepo 'a-yi'i-inc fl-k

I him-dance-see-PERF
'I saw him dance'

Clearly then, clause-per-clause (Tocutionary') translation from either

English, Spanish or Hebrew into Yaqui is not possible.
336 Mind, Code and Context

9.3.5. Non-embedded background clauses

In languages such as English, restrictive relative clauses are em¬

bedded within the noun phrase whose head noun they modify. Thus,
they form a single, complex sentence together with their main clause. In
many other languages, restrictive relative clauses cannot be embedded
at all. To illustrate this, consider the following examples from Chuave, a
New Guinea Highlands language (Thurman, 1978). In this language,
restrictive REL-clauses, unreduced verb complements and ADV-clauses
are coded structurally the same way, as backgrounded topic clauses
that must precede the main clause:

(13) a. REL-clause:
gan moi-n-g-u-a,
child be-he-TOP-him-NONSIM
'The child (who) is here,

Gomia tei awi d-i

Gomia there send leave-IMPER
send (him) to Gomia!'

or: 'Send the child who's here to Gomia!'

b. V-complement:
kasu di-in-g-a,
lie say-they-TOP-NONSIM
'That they told a lie,

fai-ke-0-m-a
right-NEG-PAST-it-EMPH
(it's) not right'.

or: 'It's not right that they told a lie'.

c. ADV-clause:
ne iki-num moi-n-g-i,
you house-your be-you-TOP-SIM
'While you are in your house,

tei u-na-y-e
there come-FUT-I-DECLAR
I will come there'.

or: 'I'll come over when you're at home'.

 Language, Culture and Translation 337

d. Topic clause:
koma du-pun-g-a-rai,
before say-we-TOP-NONSIMULT-that
'Concerning that talk we had before.

niki do-0-m-e
bad be-PAST-it-DECLAR
it was no good'.

Clearly, a single though complex sentence of English in (13a,b,c) must

be rendered into Chuave as two separate sentences.

9.3.6. The grammatical coding of evidentiality

In a recent paper, Thompson (1987) shows that the marking of eviden¬

tiality (see discussion in Chapter 4) in English is often achieved via main
clauses containing cognition verbs, as in expressions such as:

(14) a. I think she did it

b. She was there, I hear
c. They say it was a disaster
d. It was a disaster, I understand

But in many languages the very same distinction are coded by gram¬
matical particles, most commonly attached to the verb (Nichols and
Chafe, eds, 1986). Thus, consider the following example from Turkish
(Slobin and Aksu, 1982):

(15) a. Kemal gel-di

Kemal come-PAST/direct experience
'Kemal came' (I have direct evidence)

b. Kemal gel-mi§
Kemal come-PAST/indirect experience
'Kemal came' (I heard, or I infer)

The choice of evidential status is obligatory in the use of the past tense
in Turkish. Locutionary -- clause-per-clause - translation between
English and Turkish is not possible here. If the evidential force of
Turkish propositions is to be rendered into English, it must be coded
there by a separate verbal clause.
338 Mind, Code and Context

9.3.7. Manipulative speech acts in Korean

In English, weak manipulative speech acts involving obligation or

permission are rendered as a single clause with a modal verb, as in:

(16) a. You must do this.

b. You may leave.
c. You should listen more carefully,
etc.

In Korean, on the other hand, such speech acts may only be rendered by
a concatenation of two clauses, the first of which is a pre-posed condi¬
tional clause, followed by a main clause that expresses an evaluative
judgement. Thus consider the following examples (Kim, 1986):

(17)
a. i ch'aek-un an ilk-o-myon, an twe-n-ta
this book-TOP NEG read-MOD-if NEG become- PRES-PRT
'You must read this book'
(Lit.: 'If you don't read this book, it won't be OK')

b. i ch'aek-un an ilk-o-to, twe-n-ta

this book-TOP NEG read-MOD-even/if become- PRES-PRT
'You may not-read this book'
'You don't have to read this book'
(Lit.: 'Even if you don't read this book, it'll be OK')

c. i chaek-un an ilk-o-ya, twe-n-ta

this book-TOP NEG read-MOD-only/if become- PRES-PRT
'You must/should/may not read this book'
(Lit.: 'If you only not read this book, it'll be OK')

d. i chaek-un ilk-o-myon, an twe-n-ta

this book-TOP read-MOD-if NEG become- PRES-PRT
'You must not read this book'
(Lit.: 'If you read this book, it won't be OK')

Once again, single-clause expressions in one language can only be

rendered as two-clause expressions in another language.13

13 There are grounds for suspecting that the difference between Korean and English here is
not merely in grammar, but also in culture. As Brown and Levinson (1978) have observed,
more polite, deferent manipulative speech acts tend to be coded by longer, more circuitous,
multi-clausal linguistic expressions. The fact that Korean cannot code these manipulative
expressions as single-clause expressions may be one indication of the level of deference
that must involve their use.
 Language, Culture and Translation 339

9.4. Meaning, culture and translation

9.4.1. Preamble: Meaning, understanding and context

"...Understanding is, roughly, organized knowledge, i.e.

knowledge of the relations between various facts..."

M. Scriven, Explanations, predic¬

tions and laws (1962, p. 102)

Until now we dealt with the easier problem of translation, the one
that centered on the cross-language differences in grammatical struc¬
ture. As we have seen, such differences render the more extreme
doctrine of clause-for-clause locutionary translation, untenable. In the
course of the discussion we touched relatively little upon meaning or
culture. This is not because grammatical organization never reflects cul¬
tural-semantic perspective, but rather because the bulk of cross-cultural
differences are reflected in the organization of lexical meaning.
As one may recall, in the lax sense of translation, Grace's perlocutionary,
identity of form may be dispensed with but identity of meaning is
preserved. In this section I will attempt to show why even this weaker op¬
tion is not viable, leastwise not in a form that would preserve Katz's ef-
fability hypothesis - and especially its theoretical presuppositions - intact.
We will consider two alternative versions of a universalist trans-
latability hypothesis, subject both to the empirical weight of specific
case studies, and show how they crumble under that weight. In the
process, we will note than attempts to develop missing vocabulary for a
target language, with which to translate concepts from a source language,
may yield three possible results:

(18) Options of cross-cultural translation:

(a) Failure of cross-cultural translation;

(b) Profound readjustment in the target populations' cul¬
tural world-view; or
(c) The maintenance, in the mind of the target population,
of two -- well segregated but conflicting -- cultural
world views.

Paradoxically -- but in keeping with the pragmatic middle ground - we

will then suggest that these results do not necessarily militate for ex¬
treme cultural relativism. Rather, the balance between cultural univer¬
sality and cultural specificity - the possible overlap of world-views ~ is
a matter of degree. It may vary from one source:target dyad to the next;
and within each dyad, from one cultural sub-domain to the next. And,
340 Mind, Code and Context

like similar issues in the typological comparison of language structure,

the matter must be resolved empirically.
It is, I believe, relatively safe to concede that the major impediment to
cross-language — thus cross-culture -- translation is meaning, as coded
primarily in the lexicon. This is so because 'missing vocabulary' does
not entail trivial, accidental gaps in a lexicon construed, as deductivists
have construed it all along, as a mere list of atomic concepts. Rather, the
vocabulary of a language — that massive repository of the cultural
world view of a speech community -- is a highly structured network.
Within such a network, individual words ('concepts') are not defined
purely componentially, i.e. by the mere inclusion of some universal,
Platonic 'semantic features'.14 Rather, they are defined, in large part,
relationally, relative to other concepts partaking in the network. Mean¬
ing, as noted in Chapter 2, and thus understanding, is at least in part con¬
textual and relational. (As Scriven notes, following Wittgenstein, un¬
derstanding must pertain to relations rather than merely to items and
their sub-components). And it is this fundamental feature of meaning --
thus of understanding — that conspires to make cross-cultural transla¬
tion such a thorny issue. And this is true even for those of us who find it
meaningful talk about universals of language, mind and culture.

9.4.2. Translating across lexical gaps: Case studies

In this section we will consider a number of test cases in the transla¬

tion of vocabulary. In each case, vocabulary items will be revealed to
project deeply into the cultural-semantic network of world-view. In
each case, either cross-cultural translatability is impossible, or some de¬
gree of non-trivial change of the target culture must first occur, before
meaningful translation can take place.

9.4.2.1. The red herring of technological vocabulary

Consider the following passage from Carroll (1953), where the prob¬
lem of cultural differences is (a) identified as the main barrier to transla¬
tion, (b) correctly lumped together with the problem of technology, but
(c) at the same breath trivialized:

"...In general (that is, aside from differences arising from culture and tech¬
nology) contrary to the popularly held misconception, anything that can
be said in one language can be said in any other language..." (1953, p. 47;
emphases are mine; TG)

14 As in Katz and Fodor (1964).

 Language, Culture and Translation 341

One seems to discern a similar trivialization in Sapir's discussion of

how one may translate Kant into Eskimo or Hottentot:

"...it is not absurd to say that both Hottentot and Eskimo possess all the
formal apparatus that is required to serve as a matrix for the expression
of Kant's thought. If their languages have not the requisite Kantian
vocabulary, it is not the languages themselves that are to be blamed but
the Eskimo and Hottentot themselves..." (1949, p. 154; emphasis is mine;
TG)

Sapir seems to consider vocabulary gaps as accidents that can be easily

remedied by the speakers, given sufficient incentive. The target lan¬
guage — the system — is not to blame, its formal structure is no impedi¬
ment to translation, it is inherently just as capable as the source
language of producing fresh technical vocabulary — new atomic slots of
meaning -- upon demand.
There are, of course, perfectly good cases where new technology --
thus new technical vocabulary -- may be introduced into the culture
without creating profound conceptual dislocation. But when this oc¬
curs, one usually finds that the functions performed by the new technol¬
ogy are already extant in the culture; they have been recognized within
the cultural word-view. What takes place is simply a substitution of one
manner of performing those functions by another. To illustrate this, con¬
sider the acquisition of firearms -- technology and vocabulary -- by the
Ute indians of Colorado. A missive-shooting weapon, the bow-and-
arrow, had already been in existence. The new weapon, initially the flint
gun, was named tupuy, literally 'rock', 'flint'. Gunpowder was then
named tupuy-kuca-ti-pu, literally 'that which makes the gun start' (i.e.
'gun-starter'). The verb 'shoot' is currently kukwi-, most likely related to
kukim'-vi 'charcoal'; this suggests a likely connection with black powder.
Finally, the newly introduced ammunition, 'bullet', was referred to as
either 'uu- 'arrow' or its derivative 'uu-aga-tu, lit. 'that which has arrow'.
Up to this point, the new technology fitted into pre-existing slots,
roughly as follows:15

(19) Borrowing into an existing technological structure:

slot pre-existing new adaptation

instrument bow 'flint' (=gun)
missile arrow 'arrow' (=bullct)
propellant string 'gun-starter' (=powder)
act pull/release 'charcoal' (=shooting)

15 For the Ute vocabulary see Givon (1979b).

342 M ind. Code and Context

Beyond a certain point, however, there are clear limits to this seemingly
successful technological borrowing, and its concomitant lexical adapta¬
tion: The conceptual apparatus -- or context ~ needed to elucidate what
precisely causes the missive to propel is simply not there. This limit is ap¬
parent from naming powder gun-starter, then deriving the act of shoot¬
ing from charcoal. Both derivation are accurate observation on the causal
chain involved -- accurate as far as they go. From that point on, they
also reveal the limits of the borrowing, namely the failure to borrow the
conceptual apparatus which may explain the actual chemical process in¬
volved in the new technology. This crucial correlate of the new technol¬
ogy was not borrowed into Ute culture of that time. And rightly so,
since that is precisely the part that would have required a considerable,
far-from-trivial restructuring of the cultural world view. And Ute cul¬
ture at the time of borrowing firearms had no pressing need to engage
in such conceptual restructuring. And when the chemical concepts un¬
derlying gunpowder are finally borrowed into a bow-and-arrow civi¬
lization, their incorporation into the culture's world view demands
massive re-structuring and non-trivial cultural change.16

9.4.2.2. Cosmology and the physical universe

Consider a pre-Copernican speech community X, such as many, in

which the concept 'sun' entails, among other things, a physical phe¬
nomenon of a certain size, luminescence, warmth, east-to-west trajec¬
tory, daily and yearly cycle, etc. In addition, it is also the major deity,
male in gender (as contrasted with the female moon), the creator, to be
thanked for the gift of life, to be propitiated against all manner of mis¬
fortune and natural calamity. Culture X has a distinct word for the sun
(also meaning 'day'. Tight' and 'God'). That word carries with it an in¬
tense affective and spiritual load, even when 'merely referring' to the
physical phenomenon. The question now is -- in what sense is the word
'sun' in culture X translatable into present-day English, and vice versa?

16 While the concepts of modern chemistry did not find their way into Ute culture with the
borrowing of fire-arms, the advent of warfare with an invading culture did not leave the
Ute world-view unscathed. Traditional Ute warfare was seldom aimed at total extermina¬
tion of the enemy -- either warrior or home population. Ute warfare — much like in
pre-Napoleonic time (see Rapoport, 1968) — was aimed at demonstrating one's superiority,
then extracting from the enemy maximal benefits. Within this framework, war often took
a somewhat symbolic guise, with the warrior's power ('Medicine') being overwhelmingly
spiritual. The advent of firearms, their exterminatory potential, and the martial practices
of their white bearers, exerted a profoundly depressing effect on the Utes' concept of
'Medicine Power' (puwdv). Tried and true methods of making oneself powerful ceased to
work, rendered inexplicably useless in the face of thick volleys of cavalry bullets. The
cultural-spiritual ramifications for this -- to the Utes inexplicable — failure were enormous,
resulting in withdrawal, despair and recourse to a rapid succession of non-native religious
modes (see Jorgensen, 1972).
 Language, Culture and Translation 343

In attempting to answer this question, we will set up three alternative

hypotheses, beginning with:

(20) TRANSLATABILITYI: The referential hypothesis

"(As a strict logical positivist may argue:) Whatever extra

semantic-cultural baggage the two cultures may load upon
'sun', the word still refers, in both cultures, to the same
physical object, thus presumably to the same experience.
The rest is frills, obfuscation or prejudice".

Hypothesis (20) is admittedly somewhat of a straw-man. Meaning is in¬

deed a matter of experience,17 but experience is in principle not a mere
matter of 'objective' reference. Consider, for example, the following
three references to 'the sun', uttered at the winter solstice:

(21) a. The sun is not very hot today.

b. The sun is angry today.
c. The sun's rays are least dense today.

Of these three references to 'sun', only (21a) is translatable both ways.

(21b) is comprehensible in X but senseless in English. And (21c) is com¬
prehensible in a sub-culture of today's English but senseless in X.18
One may of course proceed to learn about culture X, then understand
(21b) to mean:

(22) a. The speaker believes the sun is a powerful deity,

b. and asserts it is angry today.

Has successful translation from X to English then taken place? Perhaps,

but a number of provisos must be considered.
First, the meanings — in terms of truth conditions -- of (21b) and (22)
are not the same. The speaker of X, felicitously uttering (21b), believes
that 'The sun is a deity7. But the speaker of English felicitously uttering
(22) does not; he/she only believes (22a) - i.e. that 'The speaker of X
believes that the sun is a deity'. Further, the speaker of X asserts (21b) -
'The sun is angry'. But the speaker of English asserts (22b) -- i.e. 'The
speaker of X asserts that the sun is angry7.

17 As one of Anthony Powell's characters observes in Books do Furnish a Room: "...It is not what
happens to people that is significant, it is what they think happens to them..." (1971, p. 145)
18 Phil Young (in personal communication) notes that an adjustment in an individual s
world-view does not necessarily change the culture. Still, all cultural change must begin
with the world-view of some individual member(s).
344 Mind, Code and Context

Second, the cultural context -- belief, world-view. -- was supplied as a

necessary ingredient in order for even the less-than-adequate (22) to serve
as translation of (21b) into English. Without such context no translation
was possible at all. Understanding someone else's cultural world-view is
akin to having one's own enlarged. This is so because the exclusive world
views of either X or English, before they came to grapple with how to
translate (21b,c), precluded the existence of any other world view. Admit¬
ting that another world view exists, even without granting it legitimacy,
already carries with it the germ of profound cultural change.19
Third, does the admission of existence of another world view, i.e.
granting (22a), really guarantee an English speaker an understanding of
(21b)? Consider: Anger has consequences in terms of other beliefs held
by the speakers of X, of certain expectations, fears and concomitant be¬
havior. These other beliefs, expectations, fears and normative behavior
are part of the meaning of what it means for a speaker of X to under¬
stand (21b). What is at issue is a whole range of cognitive and affective
habitual associations of 'sun'. And these associations are not accessible
to speakers of English, unless they themselves come to regard the sun as
a deity. Sharing 'the same meaning' thus lasts only as long as the state of
'sharing the same beliefs' can be maintained. In other words, meaning¬
ful cross-culture translation may require bi-culturalism. And bi-cul-
turalism, in turn, entails holding simultaneously two sets of incom¬
patible beliefs. And, switching back and forth between two sets of
incompatible beliefs may be a prerequisite for understanding meanings
in another culture.
Logicians and deductivist linguists tend to take for granted that peo¬
ple do not hold contradictory beliefs. This presupposes that the entire
body of knowledge stored in the human mind is always accessible for
weeding out contradictions. But as noted in Chapter 7, massive
evidence from cognitive psychology suggest that only a tiny proportion
of a person's knowledge-base is accessible — or placed in the focus of at¬
tention - at any given time. Holding two cultural world-views that can
be accessed selectively, in the appropriate context, is probably the norm
for (at least) complex, fractionated cultures such as ours. Equally mas¬
sive experimental evidence also suggests that the semantic organization
- i.e. knowledge structure or world view - of technical experts is
qualitatively different, usually much more elaborate and detailed with a
different pattern of inter-connections, than the world-view of naive
members of the culture.20

19 It is by no means dear that the majority of speakers of American English embrace a model
of the universe that even remotely approximates the sdentists'. See further discussion of
this in section 9.4.2.4., below.
20 See extensive review in Ericsson (1985), as well as the earlier work by Chase and Simon
(1973) and Chase and Ericsson (1981, 1982), inter alia.
 Language, Culture and Translation 345

Let us consider now an alternative hypothesis, a slight modification

of (20) above:

(24) TRANSLATABILITYII: The 'universal core' hypothesis

"(As a more reasonable universalist may wish to argue:) The

concept 'sun' in the two cultures -- X and English — overlaps
partially. That overlap may be termed the core ('universal')
element of 'sun'. That element, at the very least, is trans¬
latable from X to English and vice-versa”.

Hypothesis (24) concedes the arguments against hypothesis (20), con¬

tending itself with only partial translation whenever there is sufficient
overlap. This certainly sounds reasonable, and presumably the degree
and actual identity of the overlapping 'universal' elements can be deter¬
mined empirically. Still, a number of arguments against hypothesis (24)
remain.
First, perhaps somewhat trivially, the world-view overlap between X
and English may not be the same as between X and other languages, or
English and other languages. The question of universality may turn out
to involve a considerable muddle.
Second, for hypothesis (24) to be a viable doctrine of universality and
translation, a strictly Platonic view of meaning must be entertained, one
in which atomic units of meaning may be found in different languages
and have the same semantic value in all languages, regardless of the diver¬
gent networks/contexts within which they are embedded. As we have ar¬
gued throughout, the plausibility of such a state of affairs is indeed low.

9.4.2.3. Spatial orientation in Guugu Yimidhirr

Miller and Johnson-Laird (1976) have claimed that some features of

spatial organization -- or space perception — such as front-back and left-
right distinctions, are cognitive human-universals. And that in contrast,
cardinal compass points (North, South, East, West) orientation is a rela¬
tively late derivative of industrial-scientific cultures. In the cultures/lan¬
guages of Aborigine Australia, the opposite seems to be true. For
example, Haviland (1986) observes that in one such culture, Guugu
Yimidhirr, no right-left or front-back designations are ever used to indi¬
cate the position of objects in geographical space, relative to the speaker,
hearer, or any other reference point. Rather, the cardinal compass points
are invariably used whenever objects are identified in space. Near ob¬
ligatory reference to those absolute compass points is made whenever
locative post-positions are used. Further, when entities are referred to as
subjects or objects of clauses designating states or events, a frequent use
346 Mind, Code and Context

of the relatively few deictic markers of the language can be seen, and
these markers are nearly always further marked by the compass-points
markers. Even the gestures accompanying the descriptions of physical
entities within geographic space are oriented toward the absolute com¬
pass points. To illustrate the grammar of these distinctions briefly, con¬
sider the following text-derived examples from Haviland (1986):

(25)
a. ...yarra dyiba-rra waguur ganbarr-in...
yonder south/to-FOC outside jump-PAST
'...then (they) went inland yonder to the South../

b. ...coconut nhyayun dyiba-almun galmba dagaarrgarr-in...

coconut that/from south-FOC also grow/REDUP-PAST
'...coconuts used to grow there on the South side
(coming from an unspecified distance to the South)...'

c. ...nyundu nhila nhaamaalman Bala yarrba gunggaalu...

you now see/REDUP/ASP Bala this/way north/from/FOC
'...if you look nowadays this way North from Bala...'

d. ...guwa-alu mission-bi gaday...

wesf-FOC mission-to come/PAST
...they came to the mission (station) to the West...'

e. ...ngayu gurray nyundu yarrba ganbarra ngaal-nguurr...

I say/PAST you this/way jump/IMPER east-FOC
'...I said: "You jump in there on the East side...'

It can be shown, further, that Guugu Yimidhirr speakers who learn

English have no trouble manipulating correctly left-right and front-back
spatial references. But would the introduction of such distinctions into
the Aborigine language be a purely trivial matter? Haviland (1986) and
others who have studied Aborigine cultures think not.21 The ever-pre¬
sent coding of physical space in terms of compass points has profound

21 Ken Liberman (in personal communication).

 Language, Culture and Translation 347

ramifications in Aborigine culture. A competent member of the culture

is compelled to be forever oriented vis-a-vis those points in navigation,
movement and the conceptual organization of the habitat. The survival
value of the system is rather transparent. It also has considerable bear¬
ing upon the people's emotional and spiritual life, their connected, con¬
textualized self-image.22 Right-left and front-back distinction would ac¬
complish nothing in conveying these multiple network connections.
Conversely, the mere translation of the Aborigine compass-point ex¬
pressions into English would convey little if any of the immense cul¬
tural significance they carry for the Aborigine. The references would be
dead, comprehensible only in the most limited sense. This is because in
our present culture the cardinal compass points are not embedded with¬
in the same context — network — as they are for the Aborigine.
Haviland (1986) notes another case of translation difficulty. A three-
way grammar-coded distinction is made in the use of the compass-point
makers in expressions depicting directionality or motion, one that has to
do with the deictic focus of the event:23

aboriginal feature English rendition

a. Focus on the departure point ("near")

b. Focus on the goal point ("remote")
c. Lack of clear point focus
(or focus on the ongoing process) ("intermediate")

Such an array is quite familiar in the analysis of the deixis of verbs, in

English and elsewhere. Thus, for example, the verbs 'go', 'come' and
'be' in their spatial use tend to pattern the same lines as (26a,b,c) above.

22 John Haviland (in private communication) has noted the extreme anxiety and confusion
produced in traditional Aborigines when they are brought into an environment where their
compass-point orientation is confused or contradicted. Other land-based cultures have
been known to display similar geographical-spiritual attitudes. Thus, in his evocative
account of the Kaluli gisaro ceremony, Edward Schieffelin (1976) points out that the subject
matter of the performance by the visitors' troupe in the ceremony, a performance that
moves the Kaluli audience to tears - indeed to cathartic frenzy - often consists of 'mere'
detailed description of the well-known topography of the host audience's country.
23 The distinction is superficially similar to Langacker's "profile" (see Langacker, 1987).
However, one could separate it from "profile" in view of the following facts. Bemba (Givon,
1972a) has two verbs "come" and two verbs "go", thus creating both "profile" and "deictic
focus" contrasts:
verb ’■profile' 'deictic foi
-isha 'come to' to toward
-fuma 'come from' from toward
-ya 'go to' to away
-shya 'go from' from away
348 Mind, Code and Context

respectively.24 Further, when these three verbs are re-interpreted in

terms of temporal-aspectual space, the spatial-to-temporal reanalysis
retains the deictic pattern:25

(27) a. 'go' => FUTURE = clear initiation point (but hazy ter¬
mination point)
b. 'come' => PERFECT(IVE) = clear termination point (but
‘hazy initiation point)
c. 'be' => PROGRESSIVE = clear ongoing process (but
hazy initiation & termination
points)

In spite of this highly universal potential, when English-speaking

Aborigines try to explain the distinctions in (26), they invariably convert
them into a three-way distance gradation: near, intermediate, remote.
The reason why they choose this particular pattern of substitution is it¬
self transparent: In the more normative case, when the speaker is at the
reference point (Haviland's 'anchor' of the compass points), (26a) in¬
deed tends to also involve proximity, (26b) remoteness and (25c) inter¬
mediate distance. But the English-speaking Aborigines insist on their
'mis-translation' even when clear inconsistencies are pointed out to
them. This is, presumably, because for them the deictic focus distinc¬
tions are contextualized within the network, in relation to the compass-
point orientation system. In English the distinction is available either in
the tense-aspect system ('gonna' vs. 'have' vs. 'be') or in the verb lexicon
('go' vs. 'come' vs. 'be'). Still, this seemingly universal distinction does
not seem to transfer well into a target language, when the new context is
sufficiently dissimilar from that of the source language.
The subtlety of both the compass-point and the focus grammar is im¬
mense, given that the point of reference ('anchor') can be here-and-now,
or moved to other reference points in space/time. Again, conceptual
space in one language is not made of atomic slots that can be filled in
during translation. English speaking Aborigine cannot reproduce these
'slots' readily in a new cultural context. Neither our Hypothesis I (20) nor
Hypothesis II (24) can handle this. Let us entertain one more alternative:

(28) TRANSLATABILITY III: The split-view hypothesis:

(a) Translating from one language to another can only be

done by a bilingual individual;
(b) Therefore, translation from one culture to another can
only be done by a bi-cultural individual;

24 See DeLancey (1983).

25 See Givon (1973a).
 Language, Culture and Translation 349

(c) Being bi-cultural means holding two incompatible

world-views;
(d) In translating words from one world view into another,
one changes their meaning by embedding them in a dif¬
ferent context.

Guugu Yimidhirr speakers may be fully bilingual/bicultural; but their

two world-views are kept apart, they remain segregated, considerable
portions of the two do not overlap.26

9.4.2 4. Science and popular culture: Righteous food

One need not go to Australia to find examples of coexisting multiple

world-views. Our own culture, in its split between the scientific and the
traditional perspective, affords us many examples of this phenomenon.
As one instance, consider people's food beliefs.
There is a surprisingly large segment of the American population cur¬
rently disposed to believe -- and not necessarily on religious grounds —
that, in Victor Lindlahr's words,27 you are what you eat. This belief flies in
the face of all we know, through science, about the biochemistry of
digestion and metabolism, and the synthesis of bodily substances.
Nonetheless the belief persists, indeed in a variety of versions. For ex¬
ample, the Zen Macrobiotic diet fad of the 1960s strove to eliminate
mucous substances from the body by restricting their ingestion. Dieters
who feed on an exclusive protein diet actually expect their body to
produce no fat. And people who eat unrefined sugar somehow believe
in thus escaping the harmful effects of carbohydrates. In a recent study,
Sadalla and Burroughs (1981) show that Americans' food preferences
are intimately associated with social and personal value projections.
These value projections have relatively little to do with what educated
Americans know in their capacity as members of a science-oriented cul¬
ture. It is indeed true that the major split in popular food values closely
parallels the educated vs. non-educated socio-economic dichotomy. But
groups on the opposite sides of this major cleavage display, equally, ir¬
rational food-value projections that have little to do with the bio¬
chemistry of nutrition and metabolism. Now, in holding, in addition to
the scientific world-view, another system of food-beliefs that is incom-

26 The idea that humans - even mono-cultural ones - are routinely capable of holding several
incompatible world-views is rather respectable in the psychological study of self-decep¬
tion. For several discussions, see J. Elster (ed., 1985) The Multiple Self, whose title is of course
somewhat of a giveaway.
27 Lindlahr's (1940) summary of his food beliefs somehow manages to transform the logical
relation between 'food' and 'self' from causal connection to identity: "...As far as nutrition
is concerned, you are the result of your deeds. In other words, "you are what you eat!''...'' (p.
74).
350 Mind, Code and Context

patible with science, educated secular Americans closely resemble edu¬

cated devout Catholics who partake in the communion sacrament. In
both cases, two incompatible belief systems dwell in a person's mind
without necessarily causing any destructive consequences. The reason
this is possible -- and in fact characteristic -- of the human mind, is be¬
cause the two system are compartmentalized, i.e. contextually seg¬
regated; they are never deployed -- accessed — in the same context. Still
-- is translation from the one context into the other possible? This ques¬
tion is better re-phrased as:

"Are the contexts interchangeable?"

9.4.2.5. Sin, taboo and Christianity

Smalley (1959) discusses the pitfalls facing Christian missionaries in

preaching the Gospel to the Senoufo people in West Africa. W'hat he per¬
ceives as particularly problematic is the fact that in English one can dis¬
tinguish between 'taboo' (or 'forbidden') and 'sin'. In the Bambara-
based contact language used in the region, one can apparently make the
same distinction. But in Senoufo itself there is only one word that
covers, rather inadequately, part of the range of the two infractions --
kapini 'taboo'. As Smalley observes:

"...Taboos among the Senoufos are innumerable. It is taboo for a man

to see his wife sewing, or to hear her sing the "marriage song". It is taboo
for a man to whistle in a field except when he is resting, or for a woman
to whistle except to make a little wind when trying to blow the chaff off
the grain on a still day. It is taboo for women to see a certain fetish, or
for anyone to watch the old women when they go out to perform a cer¬
tain ceremony..." (1959, p. 62)

While the emotional attitude toward breaking taboos roughly evokes

the Christian attitude toward serious sin, the actual behaviors covered
under kapini are rather different from those covered under 'sin'. Smalley
continues:

"...When we inquired of the Senoufos about things commonly con¬

sidered sinful by us, we found that before they ever saw the white man
those who habitually practiced adultery, lying and stealing were called
silegebafeebi, or 'without-shame-people'..." (1959, p. 62)

But using silegebafeebi for 'sin' does not quite carry the requisite emo¬
tional weight. After weighing a number of other translations of 'sin', in¬
cluding kapini itself, Smalley winds up rejecting all solutions that
 Language, Culture and Translation 351

involve identifying 'appropriate vocabulary'. What is left at the end is

the following admission:

"...The solution to intercultural communication is not so much to

translate as to re-express in another language and cultural system..."
(1959, p. 64)

The strong inference is that only by imposing cultural change on the

Senoufo can the Christian concept of 'sin' be taught:

"...What do we want to say to the Senoufo? Do we want to tell them

that they should not steal? Then let's say it..." (1959, p. 64)

In an equally frank treatment, Reyburn (1958) shows how among the

Kaka people in the Camerouns, some forms of adultery are built into the
kinship system, while others are strongly embedded into the social
fabric, together with pre-marital sex and polygamy:

"...This is in essence how a Kaka philosopher might state the case.

Consequently, pre-marital relations, extra-marital relations, extreme
sexual indulgence, plurality of wives and many children are all aspects
of the good life..." (1958, p. 96)

The inescapable conclusion is, again, that there is no translation possible

without profound, intrusive cultural change.

9.4.2.6. Ute leadership

The concept of 'leader' or 'chief' is a notorious problem for cross-cul¬
tural translation, depending upon the elucidation of an entire system of
social organization. The fortunes of many pre-Westem cultures have
foundered upon this harmless theoretical predicament. The Utes of
Colorado and Utah are a case in point. Hunters-gatherers till herded
into reservations in the late 1800s, the Utes made do traditionally with a
rather loose social organization, flexible bands whose leadership struc¬
ture and decision-making tradition were radically different from our
own. Two types of leadership positions were recognized and lexically
coded:

(29) a. ta'wd-vi 'chief' (from ta'wa-ci 'man')

b. puwa-gatu 'medicine-man' (lit.: 'he who has Power').

Neither designation was hereditary, nor were they self-selected, nor

elective, nor temporally defined. Rather, both designations were contin-
352 Mind, Code and Context

gent upon circumstances, and upon recognition by the community. Au¬

thority — and decision -- was never binding. In particular, the more
'secular'28 authority of the ta'wa-vi, was purely consensual. Decision
making was collective, ponderous, and above all spiritual rather than
programmatic. Deliberation may last for days, speeches might drag on
interminably, competition for the floor was minimal. The orator's goal
was never to 'convince' others of the 'rightness' of his argument, but
rather to establish his spiritual bona fide. The latter amounted to a
demonstration, by circuitous rather than direct verbal means, that the
speaker was of the same spirit with the group. Collective action was only
taken under such conditions of spiritual consensus. To this day Ute
deliberations, carried out in strict adherence to Roberts' Rules of Order,
reflect, rather transparently, this deliberative style.
With all the above in mind, how is one to translate into Ute:

(30) a. 'Take me to your leader!'; or

b. 'Who is your chief?'

In its rightful context, the question was far from academic. In the late
1800s, U.S. authorities sought to negotiate territorial concessions with
the Utes. Repeated efforts to identify the proper leadership were in¬
variably frustrated, and eventually 'Chief' Ouray was appointed. Ou¬
ray, half Apache himself, proceeded to negotiate and sign treaties on
behalf of the entire Ute nation (all seven bands), eventually ceding the
bulk of the traditional Ute habitat, roughly half of Colorado and half of
Utah. Were those treaties binding? According to the U.S. government
(who soon proceeded to break them), they were indeed. Ute opinion on
the subject was not solicited. It was taken for granted that a 'chief' may
speak for his people, that 'binding' instruments -- 'treaties' — may be
'signed', and then be somehow invested with 'legal' authority, instru¬
ments whose consequences extended onto 'perpetuity', and among the
consequences of which land was 'ceded'. The Ute nation came to sub¬
stantial grief on the simple matter of the cross-cultural intranslatability
of 'leader', 'chief' and the attendant conceptual baggage.

9.4.2.7. You ’owe’, I ’pay’: Obligation and exchange in Melanesia

Most languages have lexically-coded notions such as 'give', 'receive',

'trade' or 'exchange'. In addition, some such as English also have lexi¬
cally-coded notions for 'loan'. Tend', 'borrow', 'owe' or 'debt'. In our
culture, 'owing' is intricately associated with the semantic field of

28 The separation of 'secular' from 'spiritual' authority is another cross-cultural translation

problem, since a Ute leader s authority was always spiritual, even when the domains were
he was called upon to perform would have seemed to us 'secular'.
 Language, Culture and Translation 353

'promise' and 'obligation'. A promise may be sealed as a legal 'contract',

it may be a verbal 'assurance' or 'word of honor'. In addition, owing
may be due to other culturally-recognized sources of obligation, ones
that do not require any specific act of exchange. Thus, for example, what
middle-class American parents feel they owe their children quite often
involves little if any expectation of reciprocity.
In the Trobriand culture described by Malinowski (1935), the semantic
field of 'owing7 is organized rather differently. An institutionalized in¬
strument, the urigubu, binds members of the society in intricate ways.
The directionality of giving and receiving urigubu changes during the
lifetime of a male member of Trobriand society. As a child, he benefits --
as resident of his father's hamlet -- from gifts paid by his mother's
brothers. As an adult, he gives urigubu — as a resident of his maternal
uncles' hamlet — to his own father, as well as to the husbands of his
sisters. Leach (1958) summarizes the institution of urigubu as follows:

"...All the affines [i.e. relations by marriage; T.G.]...are potentially hostile

'aliens' whose relationship is modified into a kind of treaty of
friendship by the fact of marriage and urigubu gift giving..." [p. 136]

Translating expressions such 'owe' or 'pay' between English and

Trobriand would require the elucidation of an entire social structure
which supports -- and is supported by -- such customs. The mere obser¬
vation that 'owing' is a matter of 'obligation' (leaving the social sources
of 'obligation' unspecified) is clearly not much of a cross-cultural trans¬
lation. The social configurations and activities that produce an acknow¬
ledged obligation are radically different in the two cultures. The
institution of urigubu in the Trobriands is the main venue of obligations
that result in exchange of material gifts. Its connections are highly
specific to Trobriand culture,29 it evokes highly specific nodes within the
cultural-semantic network. And those nodes in turn evoke other nodes.

Perhaps the best testimony to the profound social specificity of cul¬

tures such as the Trobriand may be seen when such cultures come into
contact with -- or, most commonly, are massively impacted by -- our own
urban, nuclear-family oriented social system. Extremely destructive con¬
flicts of values arise under such conditions, most conspicuously in the
areas where the two cultures have radically different interpretations of -
i.e. semantic field or cultural contexts for -- 'owing', 'obligation' and
'exchange'.30

29 The institution is broadly similar to the exchange system in the Melanesian basin. For a
more typical non-Austronesian Papua-New Guinea variant, see Schieffelin s (1976)
description of exchange relations among the Kaluli.
30 Such conflicts are currently raging in the contact interface in Papua-New Guinea, due to
partial education and urbanization.
354 Mind, Code and Context

9.4.2.8. Culture and grammar: Deference

One may as well note, albeit briefly, that lexicon is not the only place
where culture-specific notions are coded in language. Some culture-de¬
pendent domains are occasionally also coded in grammar. A prime ex¬
ample of this may be found in the grammar of speech acts, inter-per¬
sonal relations and deference.31 As a quick illustrative example we may
re-examine the case of the Korean manipulative speech acts cited in sec¬
tion 9.3.7., above. We said that Korean' codes such expressions with two
clauses, while English codes them with one, as in:

(31) 'You must read this book'

While we earlier considered this difference between English and Korean

to be trivially structural, there is good evidence that it goes beyond struc¬
ture, and is culturally motivated. Briefly, Korean culture is much more in¬
clined to code grammatically social functions such as deference.
The systematic lengthening of manipulative speech acts in a well-
known device for 'softening', 'toning down', or deference. In addition to
such indirection, Korean has an extensive use of grammatical markers
that code deference to both the hearer and the referent. As an example
consider (from Kim, 1987):

(32) Yi-kyosu-nim-i si-rul

Yi-professor-HON/R-SUBJ poetry-OBJ

karuchi-si-oss-o-yo
teach-HON/R-PAST-ASSERT-HON/H

'Professor Yi teaches poetry'

The referent honorific (HON/R) is coded both on the subject and on the
verb. The addressee honorific (HON/H) is coded on the verb. But, like
many European languages, it may also be coded on the second person
pronoun, as in:

(33) Yi-kyosu-nim, kkeso-n-un si-rul

Yi-Pressor-HON/R you/HON/H-TOP-SUBJ poetry-OBJ

karuchi-si-oss-mni-kka?
teach-HON/R-PAST-HON/H-Q

'Prof. Yi, did you teach poetry?"

31 See some discussion in Chapter 4, as well as extensive discussion in Brown and Levinson
(1978).
 Language, Culture and Translation 355

One might of course argue that languages simply have the universal
category 'deference', and the cultural world-view then dictates the con¬
text where it is applicable. However, in languages such as Korean the
choice of various honorific markers varies with the specific status rela¬
tions between the speaker and hearer, and between the speaker and
referent. In order to appropriately control the choice of the honorific
grammatical particles, speakers must be conscious at all times of an in¬
tricate, hierarchic, often subtle network of social relations. Thus, transla¬
tion between English and Korean requires not only finding appropriate
'equivalent' deference expressions, but also understanding the different
networks of social organization.

9.5. The kinship controversy: An old battlefield revisited

9.5.1. Preamble
The debate between extreme - 'cognitive' or 'biological' - universal-
ists and extreme -- 'cultural' -- relativists has been waged most bitterly
and most revealingly, at least in Anthropology, upon the old battlefield
of kinship terms. In the course of the various skirmishes dating back al¬
most to the turn of the century, two other issues have been commonly
invoked. Both concern the interpretation of human cultures, and the
basis for cultural universals, if such universals are to be found:

(a) The contrast between functionalism and structuralism

(b) The role of biology in determining cultural universals

In the following sections we will attempt to probe into the controversy's

core.

9.5.2. Kroeber’s incipient structuralism

In a pioneering work that laid down the foundations for a compara¬

tive study of kinship systems, Kroeber (1909) outlined the universal
parameters that underlie kinship terminology. Those were, in an im¬
plicit order of generality:
356 Mind, Code and Context

(34) a. generational stratification

b. the blood vs. marriage distinction
c. the lineal vs. collateral distinction
d. sex:
(i) of the kin person
(ii) of the mediating kin
(iii) of the speaker
e. age stratification (within a generation)
f. the life condition of a mediating link

In a manner remarkably prescient of the typological-universalist ap¬

proach in current linguistics, Kroeber pointed out how cross-cultural dif¬
ferences in kinship systems may be seen as differences in the degree of
completeness, elaboration and consistency of applying the universal
parameters in (34) to the various kin categories of the language. Illustrat¬
ing his point with Amerindian data, he shows a descending implication-
al hierarchy from (34a) down, one that is again reminiscent of typological
universals of language. The point where Kroeber's realistic -- rather than
extremist -- universalist approach inexplicably crumbles, is where he en¬
deavors to explain the observed balance between universality and
specificity. At that point, Kroeber rejects out of hand any appeal to socio¬
functional explanation:

"...The causes which determine the formation, choice and similarities

of terms of relationship are primarily linguistic. Whenever it is desired to
regard terms of relationship as due to sociological causes as indicative of
social conditions, the burden of proof must be entirely with the
propounder of such views..." (1909, p. 26; emphases are mine; TG)

The term 'linguistic' apparently meant for Kroeber a structural-cognitive

parameter wholely independent of socio-cultural function. He proceeds to
argue, somewhat circularly, that the perceived 'similarity' among kins
grouped together via a common kin term is merely a function sharing a
larger number of underlying features from the list in (34):

"...A woman and her sister are more alike than a woman and her
brother, but the difference is conceptual, in other words linguistic, as
well as sociological. It is true that a woman's sister can take her place in
innumerable functions and relations in which a brother cannot; and yet
a woman and her sister, being of the same sex, agree in one more category
of relationship than the same woman and her brother, and are therefore
more similar in relationship and more naturally denoted by the same
term..." (1909, p. 26; emphases are mine; TG)
 Language, Culture and Translation 357

If such a cognitive-universalist approach is to be adopted, then one

wonders how cultures that group 'mother' with 'mother's brother'
together under the same kinship term can be interpreted: As
demonstrating another linguistic-psychological organization? But why?

The equation of 'linguistic' with 'psychological' remains problematic

when it merely serves to close the search for further explanation. In the
absence of socio-cultural -- i.e. functional — explanation, one must con¬
clude that the 'similarities' invoked by Keroeber are vacuously struc¬
tural, perhaps accidents of neural organization, perhaps bio-perceptual
in motivation. Kroeber goes on to reiterate his anti-functional univer-
salism as follows:

"...It has been an unfortunate characteristic of the anthropology of

recent years to seek in a great measure specific causes for specific
events, connections between which can be established only through
evidence that is subjectively selected... Terms of relationship reflect
psychology, not sociology. They are determined primarily by language and
can be utilized for sociological inference only with extreme caution..."
(1909, p. 27; emphases are mine; TG)

One senses here an underlying current of empiricist, inductivist,

positivist dogma, reminiscent of Bloomfield's methodological strictures
of 15 years later.

9.5.3. Malinowski and Leach: Naive (’biological’)

vs. complex (’socio-cultural’) functionalism
The position of B. Malinowski in the kinship controversy is somewhat
paradoxical. On the one hand, he was, together with Radcliffe-Brown,
an early exponent of functionalism in Anthropology. This is evident in:

"...To explain any item of culture, material or moral, means to indicate

its functional place within an institution..." (1926, p. 139)

On the other hand, in searching for functional explanations, Malinowski

often invoked naive bio-survival explanations. This is apparent when
he ascribes 'an immense biological value' to magic rituals, which enable
a man:

"...to maintain his poise and his mental integrity in fits of anger, in the
throes of hate, of unrequited love, of despair and anxiety. The function
of magic is to ritualize man's optimism, to enhance his faith in the vic¬
tory of hope over fear..." (1954, p. 90)
358 Mind, Code and Context

Malinowski's propensity for seeking transparent biologically-based

explanations for cultural phenomena is nowhere more evident than in
his treatment of the Trobriand kinship system. As Leach (1958) points
out in his celebrated re-analysis, Malinowski (1932, 1935) chose to ig¬
nore his own data concerning the social structure, residency distribu¬
tion, and the directionality of the urigubu exchange system of the
Trobriands, electing instead to interpret the kinship system as one based
essentially on biological descent and mating restrictions. The use of
the term tama for both the biological 'father' and 'other males of father's
clan' is ascribed by Malinowski to linguistic extension by similarity.
The Trobriand social structure, residency pattern and the urigubu ex¬
change system are thus largely incidental to Malinowski's analysis.
Leach demonstrated, I think rather brilliantly, how the kin term tama is
motivated by the interaction of all these cultural factors, with tama
grouping together 'males domiciled at ego's father's hamlet'. This
analysis also accommodates the relational changes ensuing from the
male child's maturation, change of residency, and marriage.
The crux of Leach's re-analysis of Trobriand kinship involved
Malinowski's inexplicable category tabu, a conjunctive grab-bag at the
'core' of which Malinowski detects 'lawful (marriageable) woman'.
Malinowski was forced to invoke either homonymy, or polysemy and
metaphoric extension in order to explain away facts that did not fit.
Leach's re-analysis is much more systemic and general. Perhaps most
attractively, it offers an integrated socio-cultural motivation. In seeking
to understand why Malinowski adhered to his analysis in the face of
numerous discrepancies, Leach writes:

"...Why should Malinowski have been so keen to insist that various

meanings of the word tabu are wholely unrelated? Why, when he him¬
self laid such stress on the taboo between a man and his sister, should he
repudiate the logic by which a boy regards his father's sister as tabu?
The answer seems to be that it was because he took over uncritically
from his predecessors the bland assumption that the key to the under¬
standing of any system of kinship terminology must always be sought
in rules of preferred marriage..." (1958, p. 144; emphasis is mine; TG).

In recapitulating the social dynamics of tabu, Leach concludes:

"...How could it possibly be that the term which thus describes

'lawful woman' should also mean 'forbidden', 'dangerous,' 'sacred'?
The only possible explanation for Malinowski was that we are here
dealing with two or more entirely different words [i.e. homonyms; TG].
What he failed to notice was that when a man does marry a tabu relative
either close or remote, she and her immediate kinsmen forthwith cease
 Language, Culture and Translation 359

to be tabu, and come into the much more closely bonded categories of
lubou and yawa. In other words, marriage is a device whereby the
dangers of tabu are for the time being exorcised..." (1958, pp. 144-145).

A closely related discussion may be found in Hocart's (1937) com¬

parison of genealogical and classificatory kinship systems. Hocart begins
by challenging the significance of the distinction. Biology-bent function¬
al universalists, after Malinowski, had been seeking to interpret the so-
called 'classificatory' kinship systems of Austronesia, Amerindia, Africa
as an extension via similarity of the biologically-based, supposedly
human-universal 'genealogical' system. Hocart turns the argument up¬
side down, suggesting that non-genealogical systems may in fact be
older, and that specific changes in social organization have motivated
later evolution of seemingly biologically-based kin systems.

9.5.4. Structural universalism: The kinship semantics of Scheffler

and Lounsbury

It did not take long for deductive universalists in anthropology to go

on the counter attack, seeking Leach's intellectual hide. Armed to the
teeth with the newly assembled tools of Generative linguistics, most
particularly the Katz-Fodor brand of discrete, context-free Platonic
categories, structural Anthropologists such as Lounsbury (1965), Schef¬
fler and Lounsbury (1971) or Scheffler (1972) strove to resurrect
Malinowski's biology-based — or nuclear-family-based -- analysis of Tro-
briand kinship. At first glance, the grafting of Malinowski's func¬
tionalism onto an extreme school of linguistic structuralism appears
somewhat bizarre. The secret of their underlying compatibility lies, I
think, in two related points:

(a) Resurrecting Malinowski automatically knocks down Leach's

socio-cultural functionalism; and
(b) Malinowski's functional analysis, being more clearly biol¬
ogy-based, is also more transparently universalist.

Beyond mere logic, a similar blend has been negotiated more recently, in
the grafting of Bickerton's (1981) 'bio-program' onto Chomsky's struc¬
tural innatism (see section 9.6, below).
It was inevitable that a distinct brand of linguistics - either one of the
two successive schools of American structuralism - would be enlisted
in the service of anthropological structuralism. The tools of structuralist
linguistics indeed fit the occasion well:
360 Mind, Code and Context

(a) Katz-Fodor type discrete categorial analysis

(b) accidental homonymy in the Saussurean tradition of
'arbitrary code'
(c) primary vs. extended reference
(d) conjunctive vs. disjunctive definitions
(e) deep underlying abstract rules.

Lounsbury (1965) summarizes as follows the facets of Leach's

analysis he finds most objectionable:

Kinship terms are 'category terms'...

(2) Kinship terms, so-called, do not necessarily have anything to do
with relationships defined by genealogical connections...
(3) ...The nuclear family...is not universal; or at least not universally
relevant as an element in social organization of kinship..." (1965,
p. 143)

Of the three points in contention, (1) has always been a mere semantic
quibble, a red herring. The gist of the argument is indeed (2) and (3). In
challenging Leach's analysis, Lounsbury suggests that:

"...there is a perfectly comprehensible underlying logic to the terms and

their referents as reported by Malinowski... I will take the opportunity
to express my scepticism about this whole view of primitive kinship...
this mystique about a pre-bilateral mentality, as it were, that cannot see
the individual family as a significant legal institution in such societies,
that denies therefore the roles of its unique constituents as furnishing
the basis for calculus of legally and socially significant relationships, and
that consequently cannot see the difference between kinship and social
structure..." (1965, p. 146; emphases are mine; TG)

Lounsbury's apriori preoccupation with the biological nuclear family as

the human-universal basis for all kinship system is openly acknowledged:

"...the real issue, which is that of the precise nature of the jural rules of
society, and the role that family relationships have in formulating
these..." (1965, p. 147)

Finally, in keeping with the structuralist tradition of being more inter¬

ested in formalization than in explanation, Lounsbury observes:

"...I have not given a sociological explanation of the Trobriand kinship

terms as Leach has attempted to do. All that has been given here is a for¬
mal analysis stated in terms of semantic rules. But these semantic rules
 Language, Culture and Translation 361

must have certain clear implications of a sociological nature. Their dis¬

covery and statement, moreover, constitute a necessary preparation for
a proper sociological explanation; for it is these semantic rules that have
to be accounted for. When these fundamental principles are explained,
then the whole terminological structure that results from them is ex¬
plained..." (1965, p. 175; emphases are mine; TG)

There is nothing methodologically outrageous about the insistence that

description precedes explanation. The problem with Lounsbury's ap¬
proach arises, rather, from the conflation of four other pre-empirical
predilections:

(a) The apriori assumption that the core of kin terms must be
found in nuclear-family blood relations;
(b) The apriori insistence that semantic description look like a
calculus, i.e. a deductive logic;
(c) The belief — rather Chomskian -- in deep underlying abstract
rules that govern the behavior of surface forms;
(d) The cleaning up of data that do not fit either assumption (a)
or (b) by invoking the 'linguistic' notions of polysemy or
analogic extension.

These methodological assumptions bear a transparent kinship to deduc¬

tive-rationalist linguistics. The ensuing universalism is a necessary by¬
product of the methodology.

9.5.5. The crux of the matter: Kinship terms and translation

It ought to be obvious why the analysis of kinship terms is so crucial

for a theory of language universals and translatability. If Lounsbury and
Scheffler are right, and if kin terms 'at the core' are based on transparent
bio-functional universals, then the cross-cultural diversity of kinship
terminology is relatively superficial, merely an 'extension by similarity'
from common 'core meanings'. In that case, Translatability Hypothesis
II (24) can then be resurrected; cross-cultural translation is relatively
easy, at least in the most important 'core areas'. In the case of the
Trobriand tama and English 'father', tama's 'extended' connections are
mere superficial embellishments; its 'core' remains identical with that of
'father'. If, on the other hand, one adopts Leach's socio-cultural
functionalism, translatability between tama and 'father' in either direc¬
tion is problematic in the extreme. The overlap of cultural contexts is
much smaller; the connections within the two networks are strikingly
different. Consequently, the core vs. periphery universalist hypothesis
(24) is much less useful as a model for successful cross-cultural transla¬
tion.
362 Mind, Code and Context

9.6. The methodological conundrum:

Navigating between untenable extremes

Cultural universalism has come down to us in two main strands. The

oldest, philosophical variety, dates back the traditional Platonic con¬
cerns with absolute, immutable, context-free categories of The Human
Mind (or Aristotle's equally immutable 'forms' of External Reality).
Whether in the guise of mentalism (a la Von Humboldt, H. Paul or
Sapir), or of structuralism (a la Kroeber, Levy-Strauss, Lounsbury or
Chomsky), this extremist strain of universalism is well entrenched
today in Philosophy, Linguistics and Cognitive Psychology.32
A second, equally extreme variety of more recent universalism, can be
traced down from Darwin through Malinowski's 'naive' functionalism,
all the way to an updated version, so-called Socio-biology. What is
taken to be universal under this version is not Human Mind, but rather
Human Nature.33 This bio-functional universalism tends to assume, in
Geertz's (1984) inimitable words:

"...that culture is [the] icing, biology [the] cake; that we have no choice
as to what we shall hate... that differences are shallow, likenesses [are]
deep..." (1984, p. 269)

The inherent circularity of this raw bio-functional 'explanation' of cul¬

tures, and its cavalier dismissal of cultural diversity, is aptly noted by
Geertz:

"...The issue is not whether human beings are biological organisms

with intrinsic characteristics. Men can't fly and pigeons can't talk. Nor
is it whether they show commonality in mental functioning wherever
we find them. Papuans envy. Aborigines dream. The issue is, what are
we to make of these undisputed facts as we go about explicating rituals,
analyzing ecosystems, interpreting fossil sequences, or comparing lan¬
guages..." (1984, p. 268)

When one rejects both versions of extreme universalism, however,

one must recognize that extreme Whorfian relativism won't do either.
This extremism would mean, in essence, suspending the study of
anthropology -- and thus linguistics -- as a systematic discipline.
Is there a middle ground? Let us consider the words of two scholars

32 In addition to Katz and Fodor's various works, and its 'cognitive' counterparts such as
Miller and Johnson-Laird (1976), see more recent examples in Hollis and Lukes (eds, 1982),
particular Gellner (1982), Horton (1982), or Sperber (1982).
33 See Midgeley (1978), Edgerton (1978), Baggish (1983), Spiro (1978) or Salkever (1983), inter
alia. For other attacks on cultural relativism from a moral perspective, see Johnson (1983)
or Thomas (1983).
 Language, Culture and Translation 363

who have wrestled with this issue. First, Wierzbicka's reflection on

cross-cultural translation:

"...Does language reflect culture? In many ways, it undoubtedly does,

although it is not always easy to determine which aspects of the culture
reflected in a given language pertain to the present and which to the
past... The dangers of subjectivism and arbitrariness involved in a
search for such correlations are no doubt real enough. But to abandon
the search because of these dangers is, to my mind, analogous to saying,
as Bloomfield did, that linguistics should stay clear of meaning because
all attempts to study meaning are fraught with dangers of subjectivism
and arbitrariness.
As I see it, the important thing is to try to sharpen our analytic tools,
and to develop safeguards for the study of "dangerous areas". A semantic
metalanguage for a cross-cultural comparison of meaning seems to me, in
this respect, a requirement of the first priority..." (1986, p. 368; emphases
are mine; TG)

The crux of the matter is -- what exactly does one mean by 'meta¬
language'? If it turns out to be a throwback to rigid, Platonic inventory
of discrete universal categories, then one is bound to admit we have not
advanced far. If, on the other hand, one takes 'metalanguage' to mean a
methodological exhortation to strive gradually toward developing a tenta¬
tive, flexible, phenomenological vocabulary, then Wierzbicka's point seems
rather well taken.
Consider next the following observations made by C. Geertz:

"...The besetting sin of interpretive approaches to anything -- litera¬

ture, dreams, symptoms, culture -- is that they tend to resist, or to be
permitted to resist, conceptual articulation and thus to escape a sys¬
tematic mode of assessment. You either grasp an interpretation or you do
not, see the point of it or you do not, accept it or you do not. Imprisoned
in the immediacy of its own detail, it is presented as self-validating, or,
worse, as validated by the supposedly developed sensitivities of the
person who presents it; any attempt to cast what it says in terms other
than its own is regarded as a travesty - as, the anthropologist's severest
term of moral abuse, ethnocentric.
For a field of study which, however timidly (though I, myself, am not
timid about the matter at all), asserts itself to be a science, this just will
not do. There is no reason why the conceptual structure of a cultural in¬
terpretation should be any less formulable, and thus less susceptible to
explicit canons of appraisal, than that of, say, a biological observation or a
physical experiment..." (1973, p. 24; emphases are mine; TG)
364 Mind, Code and Context

The middle path, as Lao Tse observed, is often murky, dull, unattractive.
It so obviously lacks the captivating brilliance of the clean extremes. But
the siren call of ingenious extremist is seldom good science, however
good rhetoric it may be.

9.7. A pragmatic approach to translation

t • *

"...Translation, as every translator learns quickly, is not just a

matter of imitation, or finding our words to imitate their
words, but it is also the recreation of the context of the foreign
text..."

A.L. Becker, "Communication across

Diversity" (1979, p. 2)

9.7.1. The mirage of reductionism

The first admission one must make concerning cross-cultural transla¬

tion, is that it should have never been formulated as a question of either
or. Such formulation is the sad legacy of our Western reductionist tradi¬
tion. Rather, translation is a matter of more or less. Cultural experience --
hence context -- is always different, however 'trivial' the differences
may be judged. Therefore, perfect Katzian translation of Platonic con¬
cepts and propositions, whether locutionary or per-locutionary, is in
principle an anti-empirical mirage. But equally destructive is the ex¬
treme doctrine of total relativism. Aside from being empirically prema¬
ture, it also deprives us of binding methodology, or in Geertz's words,
of 'explicit canons of appraisal'.

9.7.2. The illusion of perfect communication

Is there anything particularly earthshaking about admitting that

cross-culture translation is a matter of degree? Consider for a minute the
normal process of communication within the same culture, the same
sub-culture, the same sub-dialect: Is it ever anything but a rough ap¬
proximation? Have two minds ever shared an identical point of view,
the very same context? Hardly. Has this ever been considered an insur¬
mountable barrier to communication? Hardly. Communication pro¬
ceeds as a matter of course in a noisy channel, with relative success
and, on occasion, spectacular failures. Indeed one of the main functions
of declarative discourse is to bridge over -- temporarily, contingently,
and for limited tactical goals -- the yawning gaps in 'shared' knowledge
and 'common' point-of-view among members of the 'same' culture, so
 Language, Culture and Translation 365

that a specific communicative transaction may take place. In elaborating

his view of culture as organized diversity, Wallace (1961) notes:

"...Culture...is characterized internally not by uniformity, but by

diversity of both individuals and groups, many of whom are in con¬
tinuous and overt conflict in one sub-system and in active cooperation
in another..." (1961, p. 28)

9.7.3. The negotiation of ’shared’ context

A large portion of grammar as a communicative device serves for the

negotiation of common grounds between speaker and hearer. The
negotiation of context, being a prerequisite to a reasonable degree of
successful communication, is a direct consequence of infra-culture
diversity. The process of context-negotiation never arrives at estab¬
lishing an absolute identity of point of view. A lifetime would not suffice
to achieve that. But the lack of total understanding does not necessarily
entail a total lack of understanding. What the negotiation of context
aims for is a reasonable degree of overlap in points of view, a relative,
temporary, contingent commonality, one that will 'do the job', more or
less. The same goes for cross-cultural communication. The Platonic
quest for 'exact', complete translation remains a mirage. But partial shar¬
ing of experience -- and the enlargement of one's context through con¬
tact with other points of view — has always been the norm, within as
well as across cultures. In fact, the ability to learn — and learn from -- the
experience of other members of the social unit is probably at the very
core of the notion 'culture'.

9.7.4. Relative overlap of contexts

The degree of overlap between two points of view, be they of two per¬
sons or of two cultures, is an empirical matter and a matter of degree. One
can well imagine an extreme condition whereby the contexts -- for a
concept to be translated -- overlap so little, that only a few connections
in the network are shared. Under such conditions, translation is a hope¬
less task. One may render this Diagrammatically as:
366 Mind, Code and Context

(35) Minimal overlap of contexts

One may also imagine a situation whereby two points of view may
overlap considerably, share many connections within the network of
surrounding the concept, not only the immediate connections but also
their respective connections, and their respective connections, and so
on. Diagrammatically this may be rendered as:

(36) Large overlap of contexts

 Language, Culture and Translation 367

The more deeply — and richly -- the two contexts overlap, the more suc¬
cessful is a translation likely to be.
However difficult the task is of determining what cultures share and
how exactly they differ, it is nonetheless an empirical issue, to be faced,
agonized, wrestled with. Practicing either anthropology or linguistics as
a science is often frustrating. But that is not a cogent reason for giving it
up. Or, as Geertz puts it:

"...The objection to anti-relativism is not that it rejects an it's-all-how-

you-look-at-it approach to knowledge or a when-in-Rome approach to
morality, but that it imagines that they can only be defeated by placing
morality beyond culture and knowledge beyond both. This, speaking of
things which must needs be so, is no longer possible. If we wanted
home truths, we should have stayed at home..." (1984, p. 276)
 ‘

.
CHAPTER 10

ADAPTIVE BEHAVIOR,
GROUP VARIABILITY AND
THE GENETIC CODE:
THE PRAGMATICS OF
EVOLUTIONARY CHANGE

10.1. Prospectus*

This chapter constitutes an attempt to resurrect, in a modified form

and hopefully with some lasting impact, one of the most persistent
themes in the history of biology: Lamarck's phoenix-like Second Law of
biological evolution. To be more precise, I will attempt to vindicate
some of the underlying intuitions of the Lamarckian position. In under¬
taking this, one must concede from the start that the actual mechanism
proposed by Lamarck to account for -- as he saw it -- the inheritance of
acquired traits, has been shown to be empirically untenable. Here is
how Lamarck (1809) expressed his Second Law:

"...All the acquisitions or losses wrought by nature on individuals...

are preserved by reproduction to the new individuals which arise,
provided that the acquired modifications are common to both sexes, or
at least to the individuals which produce the young..." (1809, p. 113; see
also Mayr, 1972)

With many other astute biologists (Darwin included), Lamarck could

not help observing the correlation between the use and disuse of organs,
on the one hand, and their evolutionary fate, on the other. The Western
propensity for reductionism, however, has backed us into the usual
choice between two uncomfortable extremes: A rigid mutationist stance,
or equally rigid Lamarckist rendition of the interplay between individual
behavior and genetic substance in determining the course of evolution.
Mayr (1982) sums up this reductionism as follows:

* I am greatly indebted to Prof. Ernst Mayr for many valuable suggestions and comments on
an earlier draft of the manuscript.
370 Mind, Code and Context

"...Behavior was for Lamarck an important evolutionary mechanism.

The physiological process initiated by behavioral activity ("use versus
disuse"), combined with an inheritance of acquired characters, were for
him the cause of evolutionary change...the mutationists went to the
other extreme. According to them, major mutations generate new struc¬
tures, and these "go in search of appropriate function"..." (1982, p. 611)

In the last one-hundred years, noh-reductionist reassessment of the

role of the individual's lifetime adaptive behavior in shaping the evolu¬
tion of gene-coded traits has indeed taken place. Mayr's articulation of
such a non-extremist position closely parallel's Kant's interactionist —
pragmatic -- position in epistemology:

"...The modem evolutionist rejects both interpretations. To him, chan¬

ges in behavior are indeed considered important pacemakers in evolu¬
tionary change..." (ibid; emphases are mine; TG)

In a similar non-reductionist formulation of the interaction between

'habit' and 'structure', Mayr (1960) observes:

"...It is now quite evident that every habit has some structural basis
but that the evolutionary changes that results from adaptive shifts are
often initiated by a change in behavior, to be followed secondarily by a
change in structure...The new habit often serves as the pacemaker that
sets up selection pressures that shifts the mean of the curve of struc¬
tural variation [within the population; TG]..." (1960, p. 106; emphases
are mine; TG)

In combining the role of genetic variability within the population, on

the one hand, and natural selection, on the other, the mechanism by
which behavior affects evolution has been de-Lamarckized, with the
original intuition preserved intact:

"...Many if not most acquisitions of new structures in the course of

evolution can be ascribed to selection forces exerted by newly acquired
behavior..." (Mayr, 1982, p. 612).

The mechanism I will discuss below, of how adaptive (thus to my

mind also purposive) behavior guides evolution, is in a clear sense a fur¬
ther elaboration of Mayr's non-reductionist position. It involves two
closely related features of biological information processing, discussed
in two earlier chapters. The first pertains to the organization of knowl¬
edge structure; the second to the mode of information processing. Ex¬
pressed as developmental processes, the two are:
 Adaptive Behavior, Group Variability, and the Genetic Code 371

(a) The rise of categorization schemata (ch. 2)

(b) The rise of automated processing (ch. 7)

While the argument itself is relatively simple, the background required

for appreciating it is indeed complex. It spans the protracted struggle
between the Platonic and pragmatic interpretation of the biological
universe, and the position of human life and mind within it.

10.2. Pre-evolutionary biology

The nature of biological organisms, their uniqueness vis-a-vis the inor¬

ganic universe, and why they happen to display their particular structural
design, have been recurrent themes in the history of Western thought.
With the advent of evolutionary thinking, three related questions joined -
and eventually almost superseded -- the old biological agenda:

(a) How did biological organisms get to be -- over time — the

way they are?
(b) How does biological evolution connect to the phenomenon
of mind?
(c) How do biological evolution, and the evolution of mind,
connect to the phenomenon of culture?

As noted earlier above,1 Western biology got an early start as a

functionalist discipline when Aristotle disposed of the early Greek struc¬
turalist schools. In so doing, Aristotle inadvertently also set the stage for
the major cleavage in the Western approach to biological design. On the
one hand stood the Platonic -- or philosophical Rationalist -- tradition,
treating biological design the way it treated mental categories (or, for that
matter, the structure of the universe itself): They were all there, perfect
and immutable, since the very beginning and for all times, pre-ordained
by some higher intelligence.2

1 See the discussion of Aristotle's biological functionalism in Ch. 1, as well the general
discussion of functionalism as a pragmatic approach to the bio-behavioral sciences in Ch.
8.
2 Mayr (1959a), following Popper (1950), characterizes the Platonic position variably as
"essentialist" and "typological", in the sense that members of a Platonic 'type' exhibit the
same identical 'essence' (eidos) and no intra-population variability. Mayr contrasts this with
"population" or "variationist" thinking (essentially similar to our pragmatic alternative to
Platonic categorization; see Ch. 2, above), whereby 'types' show a distributional curve of
their characteristic ('prototypical') properties. For extensive discussion of the inherent
variability of biological species, its origins and its role in evolution, see Futuyma (1979, ch.
8,9,10,11 and 16). It is a somewhat amusing historical curiosity that 'typological' in modem
linguistics has come to mean more-or-less the opposite of what it means for Mayr, i.e. the
recognition of the inherent variability within the type 'human language'.
372 Mind, Code and Context

Perfection of design, however, may be interpreted differently for the

biological organisms than for the inorganic (or mental) universe. Aris¬
totle, while sharing Plato's belief in a stable, perfect universe, nonethe¬
less pointed out that the perfection of biological design must be inter¬
preted in functional -- i.e. pragmatic -- terms:

(a) The design is there to serve some function.

(b) Biological structure is motivated by considerations that are
not merely formal or aesthetic.
(c) The motivation involves an interaction between the or¬
ganism and its environment — i.e. its context.

From here, the road toward an adaptive, behavioral — and eventually

evolutionary — brand of functionalism is in principle short:

(d) That interaction between organism and environment invol¬

ves the organism's adaptive behavior.
(e) Adaptive behavior is governed by the organism's perceived
adaptive tasks (or 'purpose'), given perceived environmen¬
tal challenge.
(f) Biological adaptation occurs over time, it involves the
organism's modification of its structures in order to better
perform various functions.3

It would be misleading to suggest that an evolutionary, bio-adaptive

interpretation was the only logical conclusion that could be drawn from
the transparent function-adapted nature of biological structure. Indeed,
Aristotle himself interpreted the instrument's seemingly perfect fitness
to its task in strictly Platonic terms: Biological structure has been so
created at the beginning, it is part and parcel of the immutable Grand
Design.
Biological science under the Church and through the Middle Ages
perpetuated, as a matter of theological necessity, this Platonic version of
Aristotelian functionalism. Thus, for example, St. Thomas Aquinas' fifth
teleological argument for the existence of God, in Summa Theologicae,
hinges in part on the perfect, task-adapted, functional design of bio-or¬
ganisms. As we shall see below, a similar Platonic stance was retained
by the early exponents of both embryonic ('ontogenetic') and speciated

3 Anti-functionalist linguists often cite the existence of excess structure, some of which has no
obvious functional motivation, as an argument against functionalism. The same argument
was raised in biology against adaptive natural selection. As Mayr (1962) points out, the
argument is against naive functionalism, rather than against adaptive selection per se. The
extremely complex nature of genetic structure, the complex interactive relationship among
genes — and between genotype and phenotype — are largely responsible for 'excess structure'
(see also Futuyma, 1979, ch. 3).
 Adaptive Behavior, Group Variability, and the Genetic Code 373

(phylogenetic') evolution. Indeed, the conversion of Aristotelian

functionalism into modern bio-adaptive evolutionary thinking waited
upon the eventual conflation of three ranges of facts that were either not
available, or in one case available only scantily, to pre-scientific
biologists.

10.3. The rise of evolutionary thinking: Biological diversity,

embryology and the fossil record4

10.3.1. Preamble

Perhaps one of the best arguments for the fact-driven nature of

hypothesis formation5 concerns the three ranges of facts that eventually
motivated — i.e. allowed the abduction of — an evolutionary interpreta¬
tion of biological design. The three are:

(a) Biological diversity

(b) Embryonic development
(c) The fossil record

We will discuss the three in order.

10.3.2. Biological diversity

Biological diversity had been well documented since Aristotle.6 How¬

ever, extensive geographic exploration and the flowering of medical in¬
vestigation from the Renaissance onward, contributed to an exponential
growth in the available information on extant flora and fauna species.
With the advent of the microscope, van Leeuwenhoek's discovery of
mono-cellular micro-organisms, on the one hand, and Malpigi's
pioneering of histology, on the other, added a vital new dimension to
the study of biological diversity: Micro-diversity, the diversity of cell
and tissue types. The explosion of documented diversity soon created
the need for the first level of processing the facts, i.e. classification.
Linnaeus' taxonomy,7 a natural response to that pressing need, was ac¬
complished -- like all taxonomies -- through the contemplation of
similarities and their gradation. Similarities, and in particular fine

4 In my perusal through the history of evolutionary thinking, as well as in other matters

concerning biology, I have relied repeatedly on the works of two profound scholars, Ernst
Mayr (in particular Mayr, 1976,1982) andStephen J. Gould (in particular Gould 1977,1980).
5 SeeCh. 8.
6 Aristotle, De Generatione Animalium.
7 Linnaeus (1753,1758)
374 Mind, Code and Context

gradations, in turn demand some cogent interpretation. One possible in¬

terpretation of graded similarities among species was evolution.

10.3.3. Embryonic development

The history of embryology — the study of ontogenetic evolution --

goes all the way back to Aristotle. Its importance for the eventual emer¬
gence of evolutionary thinking springs from two sources:

(a) The development of embryos — ontogenesis -- is itself an

evolutionary process.
(b) From the very start, ontogenesis was viewed as proceeding
by analogy with the external ('macrocosmic') diversity of the
biological world.

Thus Aristotle, in De Generatione Animalium, already viewed ontogenesis

as a clear recapitulation of the ranking of the extant speciated diver¬
sity:8

"...We must observe how rightly nature orders generation in regular

gradation. The more perfect and hotter animals [mammals; TG] produce
their young perfect... The third class [birds; TG] do not produce a perfect
animal, but an egg and this egg is perfect. Those whose nature is still
colder than these produce an egg, but an imperfect one, which is per¬
fected outside the body..." (De Generatione Animalium, 733b, lines 1-10)

Aristotle's ranked taxa were not evolutionarily ordered, but rather or¬
dered in terms of increased perfection. Ontogenesis was thought to
recapitulate this static ranking in its temporal sequence, through which
the embryo gradually acquired the three types of 'souls' used to rank
the biological taxa:

"...For nobody would put down the unfertilized embryo as soulless or

in every sense bereft of life... That they possess the nutritive soul is plain.
As they develop they also acquire the sensitive soul in virtue of which an
animal is an animal..." (ibid., 736a, lines 33-38)

The last — rational -- soul was reserved, as a matter of course, for man¬
kind. Similarly:

8 Aristotle's 5 classes, in order of perfection, were: (1) mammals, (2) ovoviparous sharks; (3)
birds and reptiles; (4) fish, cephalopods, and crustaceans; and (5) insects.
 Adaptive Behavior, Group Variability, and the Genetic Code 375

"...At first all such embryos seem to live the life of a plant..." (ibid.,
736b, lines 13-14) "...Since the embryo is already potentially an animal
but an imperfect one, it must obtain its nourishment from elsewhere: ac¬
cordingly it makes use of the uterus and the mother, as a plant does of
the earth, to get nourishment, until it is perfected to the point of being
now an animal potentially locomotive..." (ibid., 740a, lines 24-28)

The great embryologists of the 17th, 18th and early 19th Centuries
simply reaffirmed this Aristotelian recapitulationist creed as their data¬
base increased in size and diversity.9 But again, their interpretation was
not initially evolutionary. Indeed, two pre-evolutionary schools en¬
deavored to interpret the embryological sequence. The first believed in
epigenesis, whereby some 'vital force' powered ontogenesis. The
second, more literally inclined, believed in preformation, whereby the
entire sequence of ontogenetic development was germinally present,
'encapsulated' in (to render Bonnet's emboitment), or 'rolled into' the
embryo from the start, then merely 'un-rolled' during ontogeny.10 Both
schools thus endeavored to preserve a measure of Platonic stability in
the biological universe. And all pre-evolutionary schools tended toward
a liberal measure of what Mayr (1982) calls Natural Theology, whereby
the perfect progression observed in both the ranked species and on¬
togenesis reflected God's perfect scheme of Creation.11 Only the even¬
tual evolutionary interpretation of Aristotle's ranking dispatched this
persistent view.

10.3.4. The fossil record

The third range of facts that contributed crucially to the rise of evolu¬
tionary thinking was the -- relatively late - discovery of the fossil record

9 See for example. Sir Thomas Browne, in Religio Medici (1642): "...For first we are rude mass,
in the rank of creatures which onely are, and have a dull kind of being, not yet privileged
with life... next we live the life of Plants, the life of Animals, the life of Men...” (quoted from
Gould, 1977, p. 16). Or, consider the following from the writings of John Hunter (published
posthumously by Richard Owen, 1841): "...If we were capable of following the progress of
increase of the number of the parts of the most perfect animal, as they first formed in
succession, from the very first to its state of full perfection, we should probably be able to
compare it with some one of the incomplete animals themselves, of every order of animals
in Creation, being at no stage different from some of the inferior orders..." (1841, p. 14)
10 The term 'evolution' was first used to characterize this ontogenetic process of 'unrolling'.
Both postulated causes of ontogenesis — the epigenesist 'vital force and the preformationist
'encapsulation' - found their eventual translation as the genetic code. See extensive
discussion in Gould (1977, pp. 16-28).
11 John Ray's (1691) The Wisdom of God Manifested in the Works of Creation may be dted as a
typical example. But as Mayr (1982) points out, eminences such as Leibniz, Linnaeus and
Herder were inclined toward similar views.
376 Mind, Code and Context

in the late 18th Century.12 The intellectual impact of this discovery on

biological thinking could not be over-estimated. To begin with, it con¬
fronted contemporary thinking — static, Platonic, creationist -- with in¬
controvertible evidence of hundreds of perfectly well-formed species
that used to exists but didn't any more. In that connection, it destroyed
the prevailing notion of a single scala naturae of species, gradually as¬
cending toward perfection: Many of the extinct fossils represented
separate, unique branches that somehow petered out. What is more, in
the profile of rock strata, the graduated similarities between the fossil
inhabitants of adjacent strata are often striking. The temptation to inter¬
preted such gradation developmentally, thus in analogy with on¬
togenesis, is indeed strong.
With the fossil record in place, an evolutionary interpretation of the
three ranges of fact became feasible, indeed almost necessary. Such new
interpretation was an abduction that would relate all three factual
domains within a single wider context. It required two crucial intellec¬
tual adjustments:

(a) Extending the developmental, temporal concept of

'evolution' from ontogenesis (where is originated) to
phylogenesis, which then displaced the abstract scalar rank¬
ing of extant species.

(b) Reinterpreting the old Aristotelian recapitulation not as an

analogical relation between ontogenesis and the abstract
scala of the species, but as a much more powerful analogy
between two developmental process, ontogenesis and
phylogenesis.

10.4. Rear-guard defense of Platonism

10.4.1. Preamble

As noted above, it was always possible for devout Platonists,

religious and otherwise, to interpret both the implicit scale of biological
diversity and the developmental course of embryology as controlled by
a resident universal principle of divine guidance toward perfection. Charles
Bonnet expresses this most succintly in terms of the familiar Great
Chain of Being:

12 Assodated initially with Cuvier (1769-1832) in Paris.

 Adaptive Behavior, Group Variability, and the Genetic Code 377

"...Between the lowest and the highest degree of spiritual and corporal
perfection, there is an almost infinite number of intermediate degrees.
The succession of degrees comprises the Universal Chain. It unites all
beings, ties together all worlds, embraces all the spheres. One single being
is outside this chain, and this is He who made it..." (1764, p. 27)

While evolutionary thinking increasingly permeated the intellectual

climate of the late 18th and early 19th Centuries, Platonic interpreta¬
tions of the biological order persisted. These interpretations retained
their non-pragmatic character even -- rather paradoxically — when prag¬
matic notions such as context and function were invoked. In the follow¬
ing two sections we will consider briefly two historical high-points that
constituted the immediate precursors to Darwin.

10.4.2. Naturphilosophie
Associated initially and most strongly with German Romanticism,
and with illustrious names such as Goethe, Herder and Schelling, this
school endeavored to describe one fundamental universal order, of
metaphysics, the physical world, biology and Man. The Romantic agen¬
da sought to demonstrate:13

a. The ever-present universal order

b. The superiority of Man
c. The scalar contiguity of all life; and, thus,
d. Mankind's affinity to the other divisions of Nature

The biological agenda of Naturphilosophie was profoundly develop-

mentalist, evolutionist and — because of its overriding commitment to a
single underlying principle -- also profoundly recapitulationist:

(i) All developmental processes in nature abide by a single ten¬

dency and a single developmental course.
(ii) The same laws govern embryonic development (on¬
togenesis) and the evolution of species (phylogenesis).
(iii) The animal kingdom must itself be considered a single or¬
ganism, evolving from lower to higher stages.

In Tiedemann's words:

"...Even as each individual organism transforms itself, so the whole

animal kingdom is to be thought of as an organism in the course of
metamorphosis..." (1808, cited in Russell, 1916, p. 215)

13 For the discussion in this section I relied heavily on Gould (1977, pp. 35-50). For the general
agenda of German Romanticism see also Goede von Aesch (1941).
378 Mind, Code and Context

Or in the words of the French transcendentalist Serres:

"...The entire animal kingdom can, in some measure, be considered

ideally as a single animal which, in the course of formation and
metamorphosis of its diverse organisms, stops its development, here
earlier and there later..." (1860, p. 834)

What had continued to elude this profoundly developmentalist vision

were the two inter-dependent pragmatic, contextual ingredients, whose
interaction would eventually extract evolutionary thinking from its
residual Platonism: The ecological context of the environment, on the
one hand; and the functional context of the organism's adaptive be¬
havior, on the other.

10.4.3. Lamarck’s adaptive functionalism

In the transformation of biological thinking — and of evolutionary
theory - from a Platonic to a pragmatic approach, and thus in the
protracted transition from Aristotle to Darwin, Jean-Baptiste Lamarck
(1744-1829) occupies a pivotal, if paradoxical, position. Following Cuvier,
and at a relatively advanced age, Lamarck was quick to recognize the
devastating implications of the fossil record for the single, idealized Aris¬
totelian Scala Naturae. Both the details of separate branches and the extinc¬
tion of terminal species within branches forced a hierarchic, tree-like re¬
formulation of the old scale. Further, by analogy with the fossil record, the
scale could at last be re-interpreted in evolutionary terms.
The fossil record also brought with it the second — perhaps most
revolutionary — element in Lamarck's evolutionary thinking: The recog¬
nition of environmental change as the context that motivated biological
change. The nascent study of geology increasingly pointed to a recur¬
rent coincidence of the two developmental processes: Biological species
residing in the contiguous geological strata seemed to change together
with their environment. The impact of this recognition on Lamarck's
thinking — and through it on the history of Biology -- could not be over¬
emphasized.
Still, Lamarck, in keeping with the religious climate of his time, strove
to retain as large a residue of Platonism as could be compatible with the
newly discovered facts. He accomplished this by recognizing two
separate, indeed conflicting, mechanisms that may causally drive -- i.e.
explain - evolution. The first one is transparently Platonic:14

14 For the discussion of Lamarck, I relied heavily on Mayr (1982, pp. 343-362), as well as on
Mayr (1972).
 Adaptive Behavior, Group Variability, and the Genetic Code 379

(a) An innate, God-given capacity -- or drive -- for progress

toward perfection (with the latter defined in purely formal
terms of structural complexity):

"...Nature, in successively producing all species of animals, beginning

with the most imperfect or the simplest, and ending her work with the
most perfect, has caused their organization gradually to become more
complex... [And, the cause of this progress toward increasing perfection
derives] from powers conferred by the Supreme Author of all things..."
(1809, pp. 60,130, resp.)

Lamarck's second mechanism, in spite of a Platonic residue ('perfect

harmony'), is context-dependent, thus pragmatic:

(b) An interactive process of constant adaptation to the ever-

changing environment (so as to maintain as much as pos¬
sible perfect harmony with the environment).

A third element in Lamarck approach is even more transparently prag¬

matic. It pertains to the putative role the adaptive behavior of individuals
plays, given perceived environmental-adaptive tasks, in the evolution of
species. Lamarck's position on this issue has remained deservedly con¬
troversial in evolutionary biology. In essence, he proposed a mechanism
by which the individual's life-time adaptive change is incremented into
the species' genetic pool, Lamarck's Second Law:

"...Everything which nature has caused individuals to acquire or lose

as a result of the influence of environmental conditions to which their
race has been exposed over a long period of time -- and consequently, as
a result of the effects caused by either the extended use (or disuse) of a
particular organ — is conveyed by generation to new individuals des¬
cending therefrom..." (1809, p. 113)

We shall return to discuss this position of Lamarck's further below.

10.4.4. Darwin

The transformation of evolutionary thinking was completed by Dar¬

win, whose position -- with important refinements -- has remained
dominant to this day. Darwin's re-interpretation of the three bodies of
biological data — diversity, embryology and the fossil record -- hinged
upon two processes, one in-built into the organism and thus incipiently
Platonic, the other contextual-interactive and thus pragmatic:
380 Mind, Code and Context

(a) Species have a certain characteristic range of variability in

both their genotype and phenotype. That variability is now
known to be the product of two forces:
(i) The incidence of random mutation; and
(ii) The massive reshuffling of genetic material due to
recombination through sexual reproduction.15

(b) The interaction with the environment weeds out the less
well-adapted variants by the process of natural selection.

The translation of (a) into molecular-genetic terms was only a natural

step, once Mendel's pioneering work had been re-discovered. The
modem field of population genetics has also filled in the blanks about
the genetic variability within populations and its interaction with
natural selection.
As pragmatic as (b) seems, it nonetheless misses one crucial com¬
ponent of evolution — the role played by the adaptive behavior, indeed
adaptive experimentation, of individuals. The missing ingredient is
Mayr's "pacemaker of evolution". Stripped of the actual erroneous
mechanism, this was the essence of Lamarck's grand vision.

10.5. The siren-call of Lamarckism

There has been, from the very start, one seductive aspect to in
Lamarck's Second Law. And while we know now that the actual transfer
mechanism proposed by Lamarck has no factual support, the attraction
of that one element persists. It is worth noting, for example, that Darwin
himself was far from opposed to Lamarck's position on the inheritance of
acquired traits. Thus, we find the following words in The Descent of Man,
under the heading of The effect of the increased use and disuse of parts:16

"...it is well known that use strengthens the muscles in the individual,
and complete disuse, or the destruction of the proper nerve, weakens them.
When the eye is destroyed, the optic nerve often becomes atrophied. When
an artery is tied, the lateral channels increase not only in diameter but in

15 This element in Darwinian thinking, given the current emphasis on populations genetics,
intra-species variability and non-Platonic distribution of genetic traits within the species,
is as dose as Darwinians ever get to the old Platonic (Mayr's "essentialist") view of fixed
genetic types.
16 Mayr (1972) notes that Darwin referred to use and disuse, as a mechanism for transfer of
acquired traits, on 13 separate pages in The origin of Species, the same number of references
found in Lamarck's Philosophic Zoologiqne. Mayr (in personal communication) also notes
that the general thesis of Lamarck's Second Law, the inheritance of acquired traits, had
been the received wisdom in biology since the Greeks.
 Adaptive Behavior, Group Variability, and the Genetic Code 381

thickness and strength of their coats. When one of the kidneys ceases to act,
the other increases in size, and does double work....Whether the several
foregoing modifications would become hereditary, if the same habits of life
were followed during many generations, is not known, but it is probable..."
(1861, pp. 24-25; emphasis is mine; TG)

Darwin follows this observation with many examples in support of

Lamarck's Second Law, concluding with the following remarks concern¬
ing its applicability to human evolution:

"...Although man may not have been much modified during the latter
stages of his existence through the increase or decrease use of parts, the
facts now given show that his liability in this respect has not been lost;
and we positively know that the same law holds good with the lower
animals. Consequently we may infer that when at a remote epoch17 the
progenitors of man were in a transitional state, and were changing from
quadrupeds into bipeds, natural selection would probably have been greatly
aided by the inherited effects of increased or diminished use of different parts of
the body..." (1861, p. 27; emphasis is mine; TG)

What is so attractive about Lamarck's interactive view of evolution, a

process whereby the efforts invested by individuals in survival strategies
during a life-time of learning, of searching for adaptive solutions to en¬
vironmental dilemmas, somehow contributes to the long-term adaptive
history of the species? What has continued to beckon to thinkers as
diverse as the sedately empirical Charles Darwin, the Stalinist stooge
Trofim Lysenko,18 the mercurial dilettante Arthur Koestler19 and the
prophetic Gregory Bateson?20 One may speculate on this briefly.
For Darwin, the attraction was simple: The facts seem to point in that
direction; increased use, increased development and eventual in¬
heritance of the increased trait seemed to correlate, go hand in hand.
Lysenko's motives were the most suspect, if not necessarily the most
transparent. Scientific conviction and naked ambition aside,21 the

17 Darwin here displays, inadvertently and to be sure only by the weakest of implications,
the fatal tendency of separating the early - 'physical' - epoch of human evolution from
the more recent - 'cognitive' - phase. As we shall see shortly, he is in illustrious company
there.
18 Lysenko is considered such an unworthy scientist that neither Mayr (1982) nor Gould (1977,
1980) dignify his work with a bibliographic citation.
19 As in Koestler (1967,1972). For the former, see in particular Ch. XI: "Acting before reacting
— Once more Darwin and Lamarck".
20 As in Bateson (1979).
21 Gould (1980) is surprisingly charitable in referring to Lysenko: "...Lysenko's debate with
the Russian Mendelians was, at the outset, a legitimate scientific argument. Later, he held
on through fraud, deception, manipulation, and murder..." (p. 67)
382 Mind, Code and Context

ideological context whereby an anti-Mendelian dogma became de

rigueur in Soviet biology owes much to Marxist resentment of inherited
social privilege. The transmutation of acquired protoplasmic ex¬
perience into inherited traits was viewed by the Stalinists as a clear
analog to the proletariat bettering its social lot through life-long struggle.
The analogy concerns the relationship between the cell's relatively
large, hard-working protoplasm, saddled with all the day-to-day meta¬
bolic chores, governed by the small, metabolically inactive nucleus that
controls and transmits inheritance.
Koestler arrived at the necessity of Lamarckian evolution with a dis¬
tinct spiritual agenda. In adopting a neo-Lamarckian view, he seems to
have not been bound by the need to elucidate a coherent alternative to
Lamarck's defunct transfer mechanism. Rather, Koestler's Lamarckian
forays seem to spring from his Romantic preoccupation with the crea¬
tive spirit of all sentient beings.22 His actual argument boils down to
roughly this: If organisms seem to behave as if they creatively seek an
adaptive solution, they must therefore in fact be doing just that. However
sympathetic one may be to Koestler's spiritual presuppositions, the em¬
pirical merit of his argument — except perhaps as a heuristic -- is ques¬
tionable.
Bateson, whatever his initial motivation,23 came a considerable dis-

22 See discussion of the German Romantic movement in section 10.4.2., above.

23 The title of Bateson (1979), "Mind and Nature, A Necessary Unity", is of course suggestive
of Bateson's intellectual agenda, which — among other things — demanded the abolition of
the old Cartesian dualism, i.e. the cleavage between the evolution of mind and the
evolution of body. As we shall see further below, this dualism has persisted in 20th Century
philosophy and linguistics. It was also taken for granted by many of Darwin's great
evolutionist contemporaries. In his Introduction, Bateson takes pains to distinguish his
proposal from the traditional Great Chain of Being. His meta-observations on his own
agenda are a rare model of empirical accountability: "...In what is offered in this book, the
hierarchic structure of thought, which Bertrand Russell called logical typing, will take place
of the hierarchic structure of the Great Chain of Being and an attempt will be made to
propose a sacred unity of the biosphere [emphasis is mine; TG] that will contain fewer
epistemological errors than the versions of that sacred unity which the various religions of
history have offered. What is important is that, right or wrong, the epistemology shall be
explicit. Equally explicit criticism will be then possible..." (1979, p. 21)
 Adaptive Behavior, Group Variability, and the Genetic Code 383

tance toward resolving the puzzle of the transfer mechanism. He seems

to have identified a number of the necessary ingredients: (a) The role of
a hierarchic categorization of the phenomenological domain ('per¬
ceived task environment');24 and (b) the process of practiced, skilled
learning.25 In rejecting the old Lamarckian mechanism, Bateson also
seems to have adopted a version of modern population thinking:

"...The reader is reminded here of what was said about the fallacy of
Lamarckism...Lamarck proposed that environmental impact could
directly affect the genes of the single individual (emphasis is mine; TG).
That is untrue. What is true is a proposition of next-higher logical type:
that the environment does have direct impact on the gene pool of the
population..." (1979, p. 131),

Lamarck's Second Law may now be re-considered in light of the interac¬

tion among three factors:26

(a) The adaptive behavior of individuals

(b) The genetic and behavioral variability within the population
(c) The genetic evolution of the species

24 Here Bateson (1979, pp. 21,127-139, and elsewhere) credits Russell's Theory of Types (see
discussion in Ch. 1). Bateson makes use of Russell's idea of 'types' to argue for the hierarchic
relation between individuals and populations. While the argument is important in its own
realm, it is not connected to the second ingredient of the mechanism, automated processing.
25 See Bateson (1979, pp. 216-217). Bateson did not tie the two together as necessary com¬
ponents of a unified mechanism by which life-long learning is transferred into the
population's genetic pool.
26 Bateson has rightly viewed the progression from (a) to (b) as a progression up the hierarchy
of types. In some yet-to-be elucidated sense, perhaps the progression from (b) to (c) is also
an instance of meta-level shifting. Bateson's position on this second possibility is not clear.
The phenomenology of human language is replete with examples of close interaction
between the group and the type. Consider, for example, the use of plurality in pronominal
reference, as in: 'If anybody comes late, they better...' (type reference) 'I talked to the team
and they... '(group reference) In English (as well as in Spanish, Hebrew and many other
languages), impersonal-'passive' semi-referring or non-referrmg subjects are coded as
plurals, as in: 'They found him drunk on the beach'. For a recent discussion, see Gernsbacher
(1987). The process of generalization of experience also displays a progression from
plurality to generality. Thus, the progression: T saw a mule there yesterday' (single instance)
'I saw mules there regularly' (plural instances) 'One sees mules there regularly' (generalized
experience). For further discussion of this, see Givon (1984f, section 8). While the actual
solution to the Lamarckian puzzle seems to have eduded him, Bateson came close. His last
book (1979) was written as he waged a losing battle with cancer.
384 Mind, Code and Context

10.6. The transfer mechanism: Adaptive behavior,

population dynamics and species evolution

"...Learning alters the shape of the search space in which

evolution operates and thereby provides good evolutionary
paths toward sets of co-adapted alleles. We demonstrate that
this effect allows learning organisms to evolve much faster
than their non-learning equivalents, even though the
characteristics acquired by the phenotype are not
communicated to the genotype..."

Hinton and Nowlan, "How learning

can guide evolution" (1987)

10.6.1. Preamble

In the years since Darwin, the Lamarckian intuition of "behavior

guiding evolution" did not die out. Rather, it persisted as the middle
ground between the two reductionist extreme, naive Lamarckism and
mechanistic mutationism. The interaction between learning behavior
and evolution was suggested by both Baldwin (1896) and Lloyd Morgan
(1896). It was pursued by Waddington (1942, 1953) under the label of
canalization or genetic assimilation, and has by now become a respect¬
able mainstream idea among Darwinians.27 The element of this
'guidance' that has been left under-elaborated will be the focus of the
discussion below.

10.6.2. The software of adaptive behavior: Skilled user strategies

In this section I will sketch out an explicit mechanism by which the

acquired life-time adaptive learning by individuals may be transmitted
into the species' genetic pool. This mechanism does not attempt to repli¬
cate Lamarck's naive notion of direct, mechanistic causal link between
increased use, somatic ('protoplasmic') growth and inheritance. Further,
the mechanism I propose takes for granted both random mutation and
recombination, thus incorporating Darwin's idea of 'spontaneous'
variation as interpreted by current molecular and population genetics.
Lastly, the mechanism also takes for granted Darwin's notion of natural
selection. I will propose, however, a crucial shift of emphasis concern¬
ing the idea of adaptive traits. Both Lamarck and Darwin, and indeed

27 See e.g. Futuyma (1979, pp. 376-378) and Mayr (1982, pp. 611-612). Elsewhere Mayr (in
personal communication) notes: "...As the Darwinians say: "Behavior is the pacemaker of
evolution"...”
 Adaptive Behavior, Group Variability, and the Genetic Code 385

many contemporary evolutionary biologists, persist in interpreting

'trait' to mean purely physical trait -- i.e. a particular organ, its detailed
shape and physical design. In other words, they interpret evolution to
mean primarily the evolution of adaptive hardware. The hardware,
however, like all instruments and organs, is totally useless -- indeed
meaningless -- without its concomitant software: The cognitive-be¬
havioral strategies and habits that govern the hardware's use.
It has been convenient for biologists, particularly in the study of so-
called lower organisms, to downgrade the role of adaptive 'user
strategies' and 'behavioral habits' in task-driven evolution. This has
been done, quite routinely, by assuming that together with physical
modification of an organ, new genetic combinations also bring with
them — automatically -- the appropriate behavioral habits, the neces¬
sary strategies of use. And, further, that from the very start these
software concomitants of structural-genetic modification are instinctual
and wired-in.28 But hardware use is not always 100% predicted from
hardware structure. More often than not, the hardware entails — leaves
open -- a range of behavioral options, among which individuals make
deliberate adaptive choices.29 This is, presumably, what Mayr (1974)
had in mind in coining the term 'open behavioral program'.
Persisting in a mechanistic, deterministic approach to the relationship
between hardware and software is both empirically risky and theoreti¬
cally impoverished. This is certainly the case in the study of evolution¬
ary change in 'higher species',30 those that possess:

(a) a complex neurology,

(b) complex cognitive skills, and
(c) complex socio-cultural behavior and communication.

A complex neurology, and its entailed range of complex cognitive skills,

evolves interactively, with the neurology often modified during the

28 Concerning such behavioral, strategic software, Mayr (in personal communication) com¬
ments: "...But this is a non-inheritable product of the hardware..." However see discussion
directly below of Mayr's (1974) distinction between 'open' and'closed' behavior programs.
His 'open' program is an evolutionary strategy that leaves important adaptive choices open
to the individual, thus not fully determined by gene-coded structure.
29 Gregory Bateson's view of evolution involves a similar cyclic interaction between the
hardware and software of adaptive behavior. He writes: "...the hierarchy is not only a list
of classes, classes of classes, and classes of classes of classes, but has also become a zigzag
ladder of dialectics between form and process..." (1979, p. 215; emphases are mine; TG) Bateson
applies this interactive/dialectic - pragmatic - model to a wide range of phenomena. His
approach is strikingly similar, in general outline, to Kant s pragmatic middle ground in
epistemology, i.e. the interaction between percept and concept (see Ch. 1).
30 There is, to my mind, no compelling reason for assuming that the same argument does not
hold true for our study of the evolution of lower' organisms. It is our recalcitrant Cartesian
dualism that leads us to persist in assuming otherwise. See discussion further below.
386 Mind, Code and Context

life-time of the individual. That interactive modification has been

studied in great detail at both the behavioral-cognitive and neurological
levels. It comes under the heading of automated processing, or skilled
learning, as discussed in Chapter 7. In the next section I will sketch out
the relevance of skilled, automated learning to the way behavior guides
evolution.

10.6.3. The adaptive advantage of skilled learning

and automated processing
Within each population ('social group'), all members of the same
species, biologists acknowledge a range of genetic variation that
governs the distribution of phenotype traits, both physical and
neurological. But the use individuals make of their genetically-based
traits is not completely predictable. Rather, in many important areas, in
particular the areas of the most complex adaptive behavior,31 the
hardware leaves individuals a range of options. As a result, each in¬
dividual, however genetically defined, has a certain range of lifetime
learning ('experience') that is unpredictable and potentially unique.
The life experience of an individual includes preferred strategies or
routinized behavior patterns; these are the individual's particular solu¬
tions to adaptive tasks, the way that individual perceives them. Let us
consider a simple example. Gould (1980) described the evolution of the
Panda's prehensile "thumb", a 'sixth toe' that in other five-toed mam¬
mals has nothing to do with toes or their prehensile use, from the radial
sesamoid bone. The concomitant musculature evolved, in parallel
fashion, from the accompanying abductor muscle. The development —
now genetically coded — was presumably gradual. At the very begin¬
ning, presumably, the tool -- bone & muscle -- was not obviously suited
to its new use. A newly-recombined genetic pattern, resulting in a
slightly divergent phenotype, may have facilitated recognition of the
potential for a new, prehensile, use of the radial sesamoid. But is it really
necessary to assume that a new genetic pattern must have preceded the
discovery of the new use? As Mayr (1982) notes, such extreme
mutationism is unnecessary.32 Suppose instead:

31 See discussion in Ch. 7, concerning the retention of non-automated processing in higher

monitoring functions. This is another instance where the life-time behavior of the organism,
in partly automating those higher functions through skilled learning, pioneers the eventual
transformation of those function into genetically-determined "closed programs".
32 Recall Mayr's (1982) rejection of extreme mutationism: "...Many if not most acquisitions of
new structures in the course of evolution can be ascribed to selection forces exerted by
newly acquired behavior..." (p. 612; see Also Mayr, 1960).
 Adaptive Behavior, Group Variability, and the Genetic Code 387

(i) Some member(s) of the Panda population discovered, through

their own process of adaptive trial-and-error and analysis, the
potential of the radial sesamoid for prehensile use;

(ii) Those individuals perfect their behavioral pattern of use of

the new instrument. In the best tradition of routinized, auto¬
mated skilled learning, they become experts in the use of the
instrument. Their behavior becomes automated, skilled, a
habit.

(iii) As is well established, routinized processing involves a cer¬

tain rigidified patterning of the neurological pathways --
'connections' -- that govern the behavior.

(iv) Presumably, the physical instrument itself also develops in

size and appropriate design during this protracted, expert
life-time use.

So far, none of the developmental changes catalogued above could be

transmitted to subsequent generations. Nothing has penetrated the
genotype, yet. All we have yet, within the population, is one or more,
certainly a small minority of individuals who answer to the following
description:

(a) They have adapted the old organ to an unconventional use,

as a new tool;
(b) Their behavior pattern -- including attended neurology —
has become highly automated vis-a-vis the new use.
(c) The organ itself has adopted physically, at least to some de¬
gree, the shape appropriate for the new use.

Presumably, other individuals within the population have made other

behavioral adaptation during their lifetime. Some of those adaptations
pertains to either:

(a) Different physical-behavioral solutions to the same task; or

(b) Different task interpretations of the same organ.

In the case of the Panda's thumb, adaptation type (a) may be, for ex¬
ample, the use of the 'true' thumb as a prehensile instrument. Adapta¬
tion type (b) may be, fort example, a weapon-like use of the radial
sesamoid. All these individuals, with their routinized behavior patterns,
now live together within the same population, interacting and inter¬
breeding.
388 Mind, Code and Context

Enter classical Darwinian new genetic combination. Among the many

types of new genetic variants, only a few are relevant either to the task
('prehensile control') or the organ (radial sesamoid or the true thumb). Let
us consider specifically what may happen when the new genetic variant
that enters the population makes the radial sesamoid more amenable to
prehensile use. Via interbreeding it soon begins to distribute gradually
through a wider range of the population. Suppose, further, that four in¬
dividuals -- A,B,C,D ~ who possess, this new gene now proceed to
develop the following life-long strategic experience:

A routinizes the prehensile use of the radial sesamoid;

B routinizes the prehensile use of the true thumb;
C routinizes the defensive use of the radial sesamoid;
D routinizes none of the above.

The new genetic pattern favors them -- in classical Darwinian selection

terms -- differentially. It favors most individual A who has routinized —
on his/her own — the behavior pattern (i.e. 'software') that goes with
the use of the new physical equipment (i.e, 'hardware'). That individual
will be more successful in spreading the new genetic trait in to the gene
pool.
Consider next a slightly modified scenario: Suppose that two new
variations entered the genetic pool at the same time: variation (i) with
an enlarged real thumb, and variation (ii) with an enlarged radial
sesamoid. Suppose also that individual A above was the unaware but
lucky repository of variation (ii), and individual D above was also a
receptor of variation (ii). Further, no such individuals as B or C above
existed within the population; so that whatever individual received
variation (i) resembles C in being behaviorally unskilled vis-a-vis either
(ii) or (i). In terms of the endowed genetic hardware, individuals A and
D are equally competitive. However, in terms of the use of the hardware
— the life-time acquired, routinized behavioral software — individual A
has a clear adaptive advantage.
Consider, finally, the survival chances of the two genetic variations.
Variation (ii) has a higher probability of spreading in the gene pool,
since an individual exists who has the habituated skills to take ad¬
vantage of it, who can appreciate its adaptive edge. In contrast, varia¬
tion (i) has a lower probability of spreading: No individual has bothered
to acquire and automate the behavioral software that would take ad¬
vantage of its potential. Behavior, especially routinized skilled learning,
is indeed the pacemaker of evolution.33

33 In a recent computer simulation study, Hinton and Nowlan (1987) have demonstrated the
viability of the mechanism outlined above. An earlier oral report of their finding was given
in Hinton (1986). While my own thinking on the subject goes back at least to Givon (1982a),
I am delighted to see this convergence.
 Adaptive Behavior, Group Variability, and the Genetic Code 389

10.7. Socio-cultural, transmitted learning

and evolutionary continuity

10.7.1. Background

In advocating rejection of Lamarck's old transfer mechanism in favor

of a neo-Darwinian combination of genetic variation and natural selec¬
tion, Gould (1980) makes the following observation concerning one
developmental domain, uniquely human, where Lamarckism seems to
have at long last triumphed:

"...Lamarckism, so far as we can judge, is false in the domain it has al¬

ways occupied -- as a biological theory of genetic inheritance. Yet, by
analogy only, it is the model of the "inheritance" for another and very
different kind of "evolution" -- human cultural evolution. Homo sapiens
arose at least 50,000 years ago, and we have not a shred of evidence for
any genetic improvement since then...All that we have accomplished
[since then; TG], for better or worse, is a result of cultural evolution.
And we have done it at rates unmatched by orders of magnitude in all
the previous history of life....Cultural evolution has progressed at rates
that Darwinian processes cannot begin to approach. Darwinian evolu¬
tion continues in Homo sapiens, but at rates so slow that it no longer has
much impact on our history. This crux in the earth's history has been
reached because Lamarckian processes have finally been unleashed
upon it. Human cultural evolution, in strong opposition to our biologi¬
cal history, is Lamarckian in character. What we learn in one generation,
we transmit directly by teaching and writing. Acquired characters are
inherited in technology and culture..." (1980, pp. 70-71)

There's much that one would like to agree with, indeed applaud, in
Gould's formulation. Our cultural evolution is indeed supremely
Lamarckian. But there are also some disturbingly sweeping attitudes
that cry for careful challenge.

(a) The lack of genetic change in Homo sapiens:

The reason why there is no evidence for significant genetic modifica¬

tion since 50,000 years ago is due to the nature of the evidence and the
likely nature of more recent adaptive change: The evidence for change
has hinged traditionally on skeletal structure, the only surviving
remains of earlier man. Profound progressive changes in neurological
organizations that may have been responsible for the evolution of
higher cognitive capacities, reasoning, planning, strategic judgement
and language may have taken place since then without any trace on
390 Mind, Code and Context

cranial remains, and in particular on cranial shape and overall capacity.

But such changes may have been quite crucial for our cognitive and cul¬
tural development during that period. Homo sapiens, as Gould himself
suggests, has evolved toward increasing adaptive dependency on the
software of cognitive strategies and socio-cultural behavior patterns.
But the genetic modification supporting those may go completely un¬
detected from cranial hardware alone.

(b) The uniqueness and non-contiguity of human culture

In making culture to be a totally novel phenomenon, a human-

specific evolutionary departure, Gould departs --1 confess somewhat to
my own disappointment -- from his exemplary attempts elsewhere34 to
resist, indeed debunk, the sharp demarkation between pre-human and
human evolution. The truth of the matter is that socio-cultural organiza¬
tion, and the dependence for survival on socio-behavioral strategies,
precedes Homo sapiens by a wide evolutionary margin. It is of course
true that the pace has markedly accelerated in our most recent evolu¬
tion. But the precursors for that process had evolved long before the ad¬
vent of man. Some of our close relatives, such as the Chimpanzee or the
Japanese Macaque, are known for transmitting to the social group be¬
havioral traits that were discovered by solution-seeking, task-driven in¬
dividual members.35 But many perfectly serious biologists have traced
the evolution of socio-culture much further back.36 And the difference
between 'simple imitation' and 'deliberate teaching' -- the latter the
hallmark of cultural transmission, turns out to be a matter of gradation
(Bonner, 1980, ch. 6). So that there is no reason to assume that other so¬
cial mammals (such as wild dogs, wolves, or horses), as well as lower
social species, are incapable of acquiring adaptive behavior by observ¬
ing the behavior of an enterprising, inventive, adventuresome member
of their social group.37
It is of course perfectly true that the rate of transmission of informa¬
tion is vastly faster in cultural than in 'purely genetic' transmission
(Bonner, 1980, ch. 2). Further, the rate of cultural transmission itself has
been vastly accelerated with the advent of human language, and again
accelerated exponentially with the development of writing. Nonethe¬
less, socio-cultural transmission of the individual's adaptive skilled
learning predate Homo sapiens by a wide margin.

34 See in particular Part 2, "Darwiniana" in Gould (1980), where the theme recurs both in
discussing Wallace and in discussing the extremist positions of socio-altruism and socio-biol¬
ogy. We will return to this subject further below.
35 See, for example, discussion in Goodall (1965), Simonds (1974) or dc Waal (1982) inter alia.
36 See Bonner (1980) for an extensive discussion and literature review.
37 See, for example, discussion of wild dog social behavior in van Lawick-Goodall and van
La wick (1971). For a general discussion, see again Bonner (1980).
 Adaptive Behavior, Group Variability, and the Genetic Code 391

10.7.2. The interaction between individual learning

and group transmission in evolution

It is easy to see now why socio-cultural development did not intro¬

duce the Lamarckian pattern into evolution. Rather, it served to vastly
accelerate an already established evolutionary pattern. To argue this, we
will first recapitulate the discussion in section 10.6.3. above in a
simplified form.

10.7.2.1. Role of the individual’s routinized behavior in gaining

adaptive-selectional advantage

(i) Suppose one random mutation "M" is introduced into a

population, a mutation favorable to gaining a certain adap¬
tive advantage;

(ii) Next, suppose there are two groups of individuals in the

population:
Group A = routinizes an adaptive strategy compatible
with mutation M;
Group B = does not routinize that strategy;

(iii) Suppose also that the distribution of individuals displaying

mutation M and those that don't is equal within groups A and B.

(iv) In consequence, the following four sub-groups are now

present in the population:
Group A-M = has both mutation M and its appropriate,
routinized use-strategy;
Group A-0 = doesn't have mutation M but has the
appropriate routinized use-strategy;
Group B-M = has mutation M but doesn't have the
appropriate routinized use-strategy;
Group B-0 = has neither mutation M nor the
appropriate routinized use-strategy;

(v) In terms of chances for survival during their life-time, the

four groups are ranked as follows:
A-M = highest survival chances
A-0 = intermediate survival chances
B-M and B-0 = lowest survival chances

(vi) Consequently, the gene carrying mutation M has the highest

probability of spreading through the population through the
progeny of sub-group A-M.
392 Mind, Code and Context

So far we have reckoned without any socio-cultural transmission of

behavior strategies. That is, within our putative population, individuals
experiment strictly on their own and do not imitate the behavior of
proximate members of their social group. But even before the rise of
communication and explicit teaching,38 imitation of successful adaptive
behavior is likely. Socio-cultural transmission must therefore predate
deliberate, explicit communication.
\ ■ »

10.7.2.2. The role of socio-cultural transmission in accelerating the

spread of favorable genetic traits

It is easy to see now how socio-cultural transmission of acquired be¬

havior patterns, whether by implicit imitation or explicit instruction,
could vastly accelerate the adaptive advantage that the individuals of
Group A, above, possess. Many individuals of group B-M — i.e. those who
had the appropriate genetic trait but did not develop the appropriate use-
strategy on their own -- may acquired that behavior strategy by contact
with members of group A-M. In a population with little socio-cultural
communication, their survival chances remain low. Cultural transmission
now increases those chances in the direction of the most favored group, A-
M. Indeed, the ranks of group A-M will now swell by those erstwhile B-M
members. And as a result, the spread of mutation-M within the popula¬
tion will increase more rapidly, since more survivors carry it.39

10.7.3. Summary

To paraphrase Hinton (1986) on the course of directed evolution: In¬

dividuals within a population act as advanced scouts for evolution.
Their life-time experience, in experimenting and testing the range of
likely survival strategies available to the species, and the resultant
skilled learning and routinized neuro-behavioral patterns, have
profound consequences for the eventual course of the genetic evolution
of the species. These genetic consequences can be postulated even
without the introduction of socio-cultural transmission. The latter vast¬
ly amplifies them; it gives a new advantageous genetic variant a tremen¬
dous boost, by accelerating the rate of its spread within the population.

38 Explicit instruction is well-documented in primates and carnivores (see footnotes 35, 36,
37, above). Implicit transmission via imitation undoubtedly pre-dates it. And, as Bonner
(1980) has noted, the difference between imitation and explicit teaching is not sharp, but
rather is a matter of degree.
39 Hinton and Nowlan's (1987) simulation study disregard this possibility, thus supplying a
formal elucidation of our earlier - pre-cultural - scenario (section 10.6.3, above).
 Adaptive Behavior, Group Variability, and the Genetic Code 393

10.8. The difference of mind and the difference it makes40

10.8.1. The difference of Man

Our historical survey (see sections 10.2. and 10.3., above) makes it
clear that the idea of the uniqueness of Man — the apex of perfection¬
seeking biological evolution -- had permeated biological thinking all the
way to Darwin and even beyond. Thus Aristotle, in setting up his three-
step progression of 'souT up scala naturae — from nutritive (metabolic) to
sensible (sensory-motor) to rational (cognitive) -- explicitly singled out
the rational mind as the criterial trait by which to differentiate human
from pre-human. This tradition has persisted, via Descartes and others,
into 20th Century philosophy and linguistics. It is a tradition that insists
on a sharp cleavage in the continuity of the evolutionary ladder, a
cleavage that is invariably placed between mankind and its immediate
ancestors.
One of the hallmarks of the sharp cleavage tradition is the Platonic
rigidity of its criteria. It strives to define various cognitive faculties and
cultural traits from a discrete, anthropo-centric perspective. Pre-human
precursors of any alleged human-specific trait are thus excluded from
consideration by being pronounced 'different in kind', rather than
'different in degree'. Chomsky's (1968) view of the sharp break between
pre-human and human communication is an example of this conceptual
trick.41 The updated Cartesian checklist of Man's unique attributes is
given by Mortimer Adler as:42

40 The title of this section owes an obvious debt to Mortimer Adler's (1967) The Difference of
Man and the Difference it Makes; the position for which I argue is the exact opposite of Adler's.
41 See extensive discussion of this issue in Ch. 3, in particularly as related to the distinction
between symbolic vs. iconic signs. In addition, by branding the inventory of animal signs
"closed", Chomsky implicitly likens it to an instinctive behavioral program, reminiscent of
Mayr's (1974) 'closed program'.
42 Summarized from Adler (1967, p. 91).
394 Mind, Code and Context

1. Language: Propositions, sentences, verbal symbols

2. Tools: Useful artifacts, technology
3. Organization: Social, political and legal structure
4. Culture: Transmitted traditions, history
5. Religion: Magic, ritual, spiritual belief
6. Ethics: Morality, conscience, sense of right and wrong
7. Aesthetics: Adornment, entertainment, art

With the exception of the last three traits, which Adler himself concedes
to be 'interpreted behavior', the sharp break between human and pre¬
human behavior hinges primarily on definitional games. In each one of
traits (1) through (4) above, extensive pre-adaptations can be shown to
have existed prior to the evolution of man.43 We will touch upon the
neurological basis for such pre-adaptations further below.

10.8.2. Darwin, Wallace and evolutionary contiguity

One of the great signs of intellectual courage that characterized

Darwin's work was that he did go ahead and write The Descent of Man,
where he insisted on straight-forward evolution continuity and no ex¬
emption for 'the top rung of the ladder'. As Gould44 points out, Alfred
Wallace and Charles Darwin arrived at the explanatory principle of
natural selection independently, and roughly at the same time. Unlike
Darwin, Wallace chose to exempt the evolution of mind from the normal
adaptive, selectional constraints on physical bio-evolution. A religious

43 See extensive discussion in Lamendella (1975, 1976), de Waal (1982), Simonds (1974) or
Lieberman (1984), among many others. A wonderful example of the transformation of an
iconic into symbolic sign is described in de Waal (1982), where the Chimpanzee's gesture
of extending the arm-and-hand for the purpose of grasping has transformed gradually into
the conventionalized signs for: (a) 'I want it'; (b) 'give it to me'; (c) 'I need your support
against my enemy'; (d) 'I need love and comfort'. The chain of gradual symbolic transfor¬
mation bears striking resemblance to normal metaphoric extension in human language (see
Ch. 2). Serious main-stream evolutionary biologists have also recognized the early pre¬
human roots of both consciousness and culture. For representative examples, see Bonner
(1980) or Griffin (1976,1984).
44 Gould (1980, part 2, chapter 4).
 Adaptive Behavior, Group Variability, and the Genetic Code 395

motivation for this exemption may be taken for granted. But Wallace
also produced cogent theoretical arguments in support of retaining the
Cartesian dichotomy. His argument turns out to have been an extreme
super-selectionist extension of Darwin's position: All evolving traits
must have some selectional advantage. Darwin never espoused such ex¬
tremism, writing:

"...I am convinced that natural selection has been the main but not the
exclusive means of modification..." (The Origin of Species, 1872 edition)45

Rather, Darwin believed that "adaptive change in one part can lead to
non-adaptive modifications of other features".46
Wallace, in surveying the brain-size of non-Western 'savages', which
together with Darwin and all their contemporaries he considered both
intellectually inferior and evolutionary prior,*7 observed:

"...In the brain of the lowest savages, and, as far as we know, of the
prehistoric races, we have an organ...little inferior in size and com¬
plexity to that of the higher type..." (1890)48

Thus concluding that the brain may be vastly over-designed, i.e. an in¬
stance of excess structure, Wallace argues that the evolution of mind is
not governed by the same principle -- adaptive selection -- as the evolu¬
tion of the body:

45 Quoted in Gould (1980, p. 45)

46 See Gould (1980, p. 45)
47 A pervasive sense of racism is virtually taken for granted in the discussion of 'savage
human races' by Darwin and his contemporaries, for whom these races exhibited clear
signs of lower evolutionary characteristic, somewhat intermediate between ape and man.
Indeed, part of Darwin's argument for an evolutionary descent of man hinged, at least by
implication, on the intermediate evolutionary status of the non-Caucasian races. Thus, in
discussing the importance of the olfactory sense in pre-human mammals, Darwin writes:
"... The sense of smell is of the highest importance to the greater number of mammals...But
[it is] of extremely slight service, if any, even to the dark-coloured races of man, in whom
it is much more highly developed than in the white and civilized races..." (The Descent of
Man, 1871, p. 12) And further: "...It appears as if the posterior molar or wisdom teeth were
tending to become rudimentary in the more civilized races of man...In the Melanian races,
on the other hand, the wisdom teeth are usually furnished with three separate fangs, and
are generally sound; they also differ from the other molars in size, less than in the Caucasian
races...” (ibid., p. 14) And further: "...In man, the canine teeth are perfectly efficient instru¬
ments for mastication. But their true canine character, as Owen remarks, "is indicated by
the conical form of the crown...The conical form is best expressed in the Melanian races,
especially the Australian..."" [ibid., pp. 30-31)
48 Quoted from Gould (1980, p. 48).
396 Mind, Code and Context

"...Natural selection could only have endowed savage man with a

brain a few degrees superior to that of an ape, whereas he actually pos¬
sesses one very little inferior to that of a philosopher...The habits of
savages give no indication of how this faculty could have been
developed by natural selection, because it is never required or used by
them. The singing of savages is more or less monotonous howling, and
the female seldom sings at all. Savages certainly never chose their wives
for fine voices, but for rude health, and strength, and physical beauty..."
(1890)49

We will return to the question of evolutionary contiguity directly

below.

10.8.3. More rear guard skirmishes of Platonism:

The abduction of Gould’s punctuated equilibrium

"...Both Darwin and Agassiz, then, denied the "biological

reality" of species, but for diametrically opposed reasons.
Darwin denied it because he considered species an arbitrary
segment in a continuous stream of individuals. Agassiz
denied it because to him not the physical species as such had
reality but only the category of thought that we call
"species"..."

E. Mayr, "Agassiz, Darwin and

Evolution" (1959c, p. 261)

As the idea of evolution becomes more widely accepted in the social

and human sciences, Platonic-Cartesian biases for the 'uniqueness of
man' undergo strange metamorphoses. Often, the ensuing conceptual
stance is eerily reminiscent Wallace's. Thus Chomsky, for example, has
more recently taken to citing Gould's position on gapped evolution --
punctuated equilibrium — 50 in support of the uniqueness and evolution¬
ary non-contiguity of the human language faculty. Chomsky's agenda is
two-fold, and Gould is pressed into service in support of its first firma¬
ment: The process of evolution seems to alternate between periods of
gradual ('quantitative') and abrupt ('qualitative') change, with the latter
most characteristic of rapid speciation. The evolution of human lan¬
guage, Chomsky asserts, must have been of that second type.
The idea of gradual change has been anathema to Platonists from the
very start, running smack against their notion of discrete ideal types. We

49 Quoted from Gould (1980, pp. 49-50).

50 See Eldredge and Gould (1972) and Gould and Eldredge (1977).
 Adaptive Behavior, Group Variability, and the Genetic Code 397

have already noted (Ch. 1) how Aristotle wrestled with the implications
of gradual change by positing an added level of reality, the sinolon,
where change could register in the graduated space between the forms,
while the forms themselves remain discrete and immutable. By a coin¬
cidental inverse analogy, early evolutionists used the evident graduated
similarities among biological species as an argument for evolutionary
change and against abrupt creationism. Thus, Lamarck (1809) observed:

"...With regard to living bodies, it is no longer possible to doubt that

nature has done everything little by little and successively..." (1809, p.
11) and "...These changes only take place with an extreme slowness,
which makes them always imperceptible..." (ibid., p. 30)

Gould's notion of punctuated equilibria,51 and in particular the way

Chomskians have made use of it to support the alleged evolutionary
discontinuity (thus Cartesian uniqueness) of man, mind and language,
is open to a number of factual and theoretical objections.

(a) Generality:
Gould himself cautions against too general applicability of
'punctuated' change:

"...I emphatically do not assert the general "truth" of this philosophy

of punctuational change. Any attempt to support the exclusive validity
of such a grandiose notion would border on the nonsensical.
Gradualism sometimes works well..." (1980, p. 153).

(b) Speciation and accelerated change:

It is generally accepted now (see e.g. Mayr, 1959b, 1970) that abrupt
evolutionary gaps are associated most typically with speciation — the
emergence of new species. And further, that the primary impetus for
this abruptness is the combination of:

(i) The isolation of a small population;

(ii) Consequently, lack of dilution ('neutralization') of new
genetic traits; and thus
(iii) The accelerated spread of innovation across the isolated
population.

Futuyma (1979) sums up the seeming contradiction between gapped

speciation and gradual genetic change as follows:

51 See Gould (1980, part 5, ch. 17) as well as Odredge and Gould (1972) and Gould and
Eldredge (1977).
398 M ind. Code and Context

"...Thus the theory of punctuated equilibria suggests that major chan¬

ges can happen very rapidly, but it still does not tell us whether they
occur through intermediate morphological steps..." (1979, p. 163; emphasis
is mine; TG)

Gould himself puts it as follows:

"...A new species can arise when 3 small segment of the ancestral
population is isolated at the periphery of the ancestral range..." (1980, p.
152)

And:

"...Large, stable central populations exert a strong homogenizing in¬

fluence. New and favorable mutations are diluted by the sheer bulk of
the population through which they spread. They may build slowly in
frequency, but changing environments usually cancel their selective
value long before they reach fixation..." (ibid.)

But the bulk of the non-variant population could only dilute the new
genetic variant if it is inter-breedable with it. Or, in other words, if the
new variant has not diverged across an inter-species gap. The process of
speciation must therefore be a graduated — though possibly quite rapid
-- cumulation of successive small steps. It is only its ultimate product
that is guaranteed to seem gapped. This is so because of the interaction
of several factors, all already noted above:

(i) The geographic isolation of the small sub-population,

within which selectionally-favored innovations spread
without dilution;
(ii) The acceleration of the rate of spread of the mutation due to
automated, skilled-learning by individual members. That
is, the "guidance of evolution by behavior";
(iii) The further acceleration of spread of the new favorable
variation due to socio-cultural transmission.52

None of these factors militates for the evolutionary non-contiguity of

higher cognitive faculties. Rather, they suggest that pre-human biologi¬
cal evolution must have already been guided by purposive behavior, de-

52 One must note, lastly, that pre-adaptation via automated behavior and skilled learning
favors not only the first mutation in the appropriate adaptive direction, but also the entire
chain of compatible mutations. Thus, the first panda who routinized the prehensile use of
the radial sesamoid created a favorable survival chance not only for the first genetic variant
leading to the re-shaping of that bone as a 'thumb', but for the entire chain of convergent
variations leading in the same direction.
 Adaptive Behavior, Group Variability, and the Genetic Code 399

pendent on population dynamics, and quick to take advantage of socio¬

cultural transmission.

10.8.4. Modularity and the Platonization of mental faculties

The Chomskians' rear-guard struggle to salvage the Cartesian agenda

apparently also requires that human language be discontinuous with
so-called 'general cognitive capacities',53 perhaps because such capac¬
ities -- or their unmistakable pre-adaptive precursors -- can be shown to
exist in pre-human mammals. To this end, latter-day Platonists press
into service the notion of neurological modularity,54 whereby the human
language-processing capacity is alleged to be a separate neurological
module, quite independent of the neurological regions that support the
other, 'lower' cognitive functions.
The neurological evidence concerning the left-hemisphere specializa¬
tion -- lateralization -- of language, as well as the seeming grammar-
specificity of Broca's region, has often been used to prop the argument
for a human-specific language-module. But as Lieberman (1984) points
out, a number of other complex faculties are also controlled from the
same brain region. In particular, that region seems to specialize in
automated rhythmic-hierarchic skills, including complex motion,
music and hierarchically-organized visual patterns.55
The claims of separate neurological module for language also fly in
the face of what is known in biological evolution in general as the
process of homoplasy, whereby old organs are adapted to perform
novel (but initially similar and thus compatible) functions (Futuyma,
1979, p. 140). As Gould (1980) points out, "nature is a tinkerer", taking
advantage of existing structures whenever they are extendable, rather
than inventing new ones. The extension of bio-design thus proceeds in a
Rube Goldberg fashion. And the extension of both bio-design and bio¬
function is in a sense rather analogous to metaphoric ('Wittgen-
steinean') extension of meaning:56 Small steps of novel contextual adap¬
tation proceed along gradients of similarity of both structure and
function.

53 See Chomsky (1968), as well as more recent formulations in Chomsky's various contribu¬
tion in Piattelli-Palmarini (ed., 1980).
54 See Fodor (1983).
55 See some discussion in Ch. 7. The neurological literature supporting the alternative
approach, that of shared modules, is quite extensive (e.g. Bradshaw and Nettleton, 1981;
Greenfield and Schneider 1977; Grossman, 1980; Ibbotson and Morton, 1981; Lieberman,
1984; Kimura, 1979; Martin 1972; Poeck and Huber 1977; Robinson, 1977; Robinson and
Solomon, 1974; Whitaker, 1983; inter alia. More recently the same approached to module
sharing has been demonstrated in the study of rhythmic motor skills (Keel, 1986; Keel and
Ivry, 1986; Keel et al, 1985,1986).
56 See discussion in Ch. 2.
400 Mind, Code and Context

10.9. Closure: The language parallels

10.9.1. Variation and change

"...Every characteristic of organisms has a frequency

distribution...Evolution in a broad sense may be considered
any change in the form of the frequency distribution of one
or more characteristics..."

D. Futuyma, Evolutionary Biology

(1979, pp. 19-21)

As noted above, biological evolution proceeds through the interaction

of three factors:

(a) The inherent variability within the species ('type')

(b) The adaptive-purposive behavior of individuals
(c) The adaptive selection pressure of the environment

The parallels between biology and linguistics here are indeed strik¬
ing. In Ch. 2 we noted how the only approach to mental categories that
is compatible with behavioral facts is a variationist approach, akin to
Mayr's population thinking. Bill Labov's pioneering work on variation
(see e.g. Labov, 1975b) has shown the applicability of this thinking to
populations of speakers, as well as to population of utterances. Further,
Labov's work has also pointed out another pervasive parallelism be¬
tween linguistics and biology: The intimate connection between
synchronic variation and diachronic change. So that " today's variants
are tomorrow's main-stream behavior". Much as in bio-evolution, lin¬
guistic diachrony can be characterized as a shift in the frequency dis¬
tributions that characterizes the population.
In its much shorter history as a separate empirical discipline, linguis¬
tics has experienced the same damaging pressures from Platonic
idealization -- Mayr's "essentialism" — as had biology. In both dis¬
ciplines, the advent of contextual-pragmatic thinking is marked by the
very same features:

(i) Population-variationist thinking

(ii) Graduated-scalar definition of 'types'
(iii) Functionalist-adaptive approach to both behavior and structure
(iv) Ecological-environmental framing
(v) Historical-evolutionary explanation of stasis

The fact that Aristotle turns out to have been the spiritual father of prag-
 Adaptive Behavior, Group Variability, and the Genetic Code 401

matic thinking in both disciplines may of course be an accident, but

surely a happy one.

10.9.2. History, diachrony and recapitulation

"...Every biological phenomenon, every structure, every

function, indeed everything in biology, has a history. And the
study of this history, the reconstruction of the selection pres¬
sures that have been responsible for the biological world of
today, is as much a part of the causal explanation of the world
of organisms as physiological or embryological explanations.
A purely physiological-ontogenetic explanation that omits
the historical side is only half an explanation..."

E. Mayr, Evolution and the Diversity of

Life (1976, p. 16)

In arguing against Dawkins' non-behavioral, non-functional ap¬

proach to biological evolution, an approach that reduces 'adaptation' to
a mere lottery of genotypes, and would thus effectively ignore the inter¬
action between the phenotype — task-oriented organs -- and the environ¬
ment, Gould (1980) makes a plea for a historical interpretation of bio-or¬
ganisms:

"...[Dawkins' reductionist] ideas have been successful in our study of

simple objects, made of few components, and uninfluenced by prior his¬
tory...But organisms are much more than amalgamations of genes. They
have a history that matters..." (1980, p. 77; emphasis is mine; TG)

One of the most salient features of bio-evolution as a historical process

is the parallelism, noted in one guise or another ever since Aristotle, be¬
tween the evolutionary sequence ('phylogenesis') and the embryonic se¬
quence ('ontogenesis'). How this parallelism -- usually referred to as
recapitulation — actually comes into being is a complex, absorbing story
that is not within the immediate scope of our discussion.57 As noted in
Chapter 8, both language and socio-culture are adaptive, functional-
pragmatic, historically-derived structures. They are historical in very
much the same sense as Mayr and Gould consider bio-organization to
be historical:

57 See Gould (1977). The most vigorous exponent of classical recapitulationism was Haeckel
(1866). A comprehensive overview may be also found in de Beer (1930).
402 Mind, Code and Context

(a) They bear the traces of their evolutionary history; and

(b) that evolution is itself task-driven.

The possibility that in language -- as in biology — ontogenesis

recapitulates phylogenetic evolution has been discussed in some detail
elsewhere.58 What human language and socio-culture add to the
recapitulation paradigm is a third component — socially-transmitted
history. As Gould (1980, Part 2, ch. 7) notes, the historical transmission
of language and socio-culture is a clear instance of Lamarckian evolu¬
tion. The fact that Lamarckian principles may also be involved in pre¬
human bio-evolution detracts little from Gould's observation.
The striking parallels between child language development (on¬
togenesis) and language history (phylogenesis) have been noted ear¬
lier.59 As suggested in chapter 7, the rise of grammatical morphology
and syntactic constructions involves the -- at least partial — automat¬
ization of language processing. The course of child language acquisition
shows striking parallels to the historical rise of grammar. Mayr's notion
of behavior-guided evolution applies well to the three developmental
domains of language: Ontogenesis, phylogenesis and diachronic ('his¬
torical') change. In the phylogenetic evolution of language, the three in¬
teract as follows:

(a) Ontogenesis: The adaptive behavioral experimentation by

the young of the species constitutes an important com¬
ponent of 'directed' evolution. The young are characterized -
neurologically and otherwise — by more flexible, less
differentiated structures. Such 'immature' structures are still
capable of life-time skilled, automated learning in new adap¬
tive directions.

(b) History: Linguistic and socio-cultural history constitute an

intermediate selection process, through which the lifetime
behavioral innovation of both young and adult is subjected
to the test of adaptive, task-driven socio-cultural selection,
within the variant pool of the population.

(c) Phylogenesis: Finally, the adaptive experimentation by in¬

dividuals, transmitted via the filter of socio-cultural history,
is the behavioral precursor of eventual genetic evolution.

58 See Lamendella (1976), Givon (1979a, chapters 5 & 7) and Lieberman (1984, ch. 11)
59 See Slobin (1977) or Givon (1979a, chapters 5,6,7), inter alia.
 Adaptive Behavior, Group Variability, and the Genetic Code 403

10.9.3. The evolution of communicative context

As noted in a number of chapter above (see in particular Ch. 3),

human language at its present stage of evolution makes systematic com¬
municative use of three main divisions of context:

(a) The generic culturally-shared context

(b) The deictic situation-shared context
(c) The textual shared-discourse context

Given the reasonable certainly that culture organization and its con¬
comitant, shared cognitive world-view, predate homo sapiens (Bonner,
1980; Griffin, 1976,1984); and given the clear evidence of animal deictic
'pointing' behavior (see Ch. 3, above), it is most likely that the use of the
first two categories of context had already become well-established in
pre-human communication. The evolution of human communication
can be viewed, from one perspective, as a systematic shifting of the
balance among the three, toward increased reliance on the third division
of context: the stored shared text.
That the notion 'shared text' is itself contiguous to both 'generic' and
'situational' shared knowledge is also likely. Higher mammals' pointing
behavior suggests just that. Its amplitude can escalate gradually if:60

(i) the interlocutor fails to respond; or

(ii) the pointed object is more surprising; or
(iii) The communication is more urgent

Human anaphoric reference shows a similar strategic escalation under

the very same conditions. But in pre-human communication, the 'text'
that one relies on (or 'refers to') remains largely what one mind assumes
the other knows. The rise of propositional-declarative language may be
viewed as an attempt to make 'shared information', a sine qua non for
biologically-based communication, more explicit.

60 See discussion in Chapters 3 and 6.

* .
CHAPTER 77

THE MYSTIC AS PRAGMATIST:

LAO TSE AND TAOISM

11.1. Introduction*

In this chapter we survey one pre-Westem pragmatic tradition. We

could have just as easily selected one of the pre-Socratic Greeks, such as
Anaximander, Heraclitus or Protagoras. However, their teachings sur¬
vive mostly in fragments or incidental references in the works of others.
In electing to go instead with a non-Western dialectician, Lao Tse, one
has the advantage of at least one coherent, if short and sometimes enig¬
matic, text of Philosophical Taoism (the Tao Teh Ching), together with the
more earthy writings of its second major exponent, Chuang Tse.
I must confess to another motivation for favoring Lao Tse's prag¬
matism over the pre-Socratic Greeks. Many pragmatic traditions seem to
reveal a stubborn streak of mysticism. In rational, post-Socratic Western
philosophy, this streak got stamped out vigorously -- together with the
pragmatic middle ground. I have often wondered (i) why the mystic
streak was such a persistent, natural concomitant of the pragmatic tradi¬
tion; and (ii) why reductionist philosophy found it so necessary to stamp
out both. My tentative answer to (i) is that the open-endedness that is
tolerated by the pragmatic method, and in particular the non-closure of
knowledge and description, leaves ample room for the possibility that
some metaphysical entity may lurk beyond the reach of our limited point
of view. My tentative answer to (ii) is that open-endedness and non¬
closure have been anathema to Platonism from the very start.
The Western rational tradition is reasonably consistent in its attempt
to deal separately with the three major branches of philosophy:

(i) Metaphysics (cosmology; theory of origins)

(ii) Epistemology (theory of knowledge or method)
(iii) Ethics (theory of human conduct)

Whether by logical necessity or by cultural drift, the separation has eventual¬

ly lead to a massive narrowing of the horizons in 'respectable' 20th Century

*1 am indebted to T.K. Bikson, Pete Becker, Haj Ross and Martin Tweedale for comments on
an early precursor of this chapter.
406 Mind, Code and Context

philosophy, where a brand of inquiry has evolved that preoccupies itself al¬
most exclusively with the method itself, or worse, with the meta-language.
The Tao Teh Ching, in its own peculiar way, displays the converse of
the narrow preoccupation of major contemporary figures such as Rus¬
sell, Carnap or Wittgenstein. It deals with metaphysics, epistemology
and ethics not only within the same slim volume or the same short
sutra, but often in the same breath. One may of course ascribe this to the
mystic's fabled lack of analytic rigor. But one then risks missing an im¬
portant point: This seeming disregard for sub-disciplinary boundaries is
the product of deliberate choice; it reflects the unitary, integrative vision
of Taoism, whereby all three sub-fields of inquiry flow from a single
source, the metaphysics of Tao. When tracing out the separate philo¬
sophical strands running through Lao Tse's intellectual tapestry, one
must bear in mind that the endeavor is to some extent futile, it does
violence to the strong undercurrent of oneness that is so pervasive in
Taoist thinking. Thus, in dissecting Taoist dialectics with Western anal¬
ytic tools, a measure of pragmatic — context-mediated — double vision is
sometimes useful and occasionally indispensible.

11.2. Taoist metaphysics: Unity in diversity

11.2.1. The One

"...The world was so brand new, that many things had not yet
been named, so that to refer to them one had to point with
the finger..."
G. Garcia-Marquez, Cien Ahos de
Soledad (1975, p. 9)
At the core of Taoist metaphysics rests the notion of Tao. Variously
rendered as 'The Way', 'The Power', 'Nature', 'God', or 'Entropy7, it
defies translation. From a synchronic perspective, it is the unifying
principle, as well as the ultimate reality behind the phenomenological
universe. It is the law of nature, the flow of gravity, the inherent direc¬
tional bias of the universe. From a diachronic perspective, it is the
mystery before the beginning, the primal source. It is pre-sensory, pre-
cognitive, non-verbalizable. Or, as the Tao Teh Ching puts it:1
The Tao that can be told of in not the real Tao,
Names that can be given are not real names
(1)

1 All citations from the Tao Teh Ching are taken from my own unpublished translation,
(Copyright © 1976 by the Shaolin-West Foundation) for which much thank is due to Jia-Hsi
Wu (in personal communication). Numbers refer to traditionally-numbered sutras in the
book. The attribution of authorship of the book to Lao Tse is traditional, with the most likely
codification occurring in the 5th or 6th Century B.C. (see Welch, 1957).
 The Mystic as Pragmatist 407

The first fundamental dichotomy set up by Lao Tse is betweenTao, the

unnameable, and the phenomenological world of forms. In the latter,
entities are individuated, distinctions can be made, names may be
given. The world of forms flows from Tao ('The Father'), once a Prin¬
ciple of Differentiation ('The Mother') is introduced:
Nameless is the Father of heaven and earth.
Named is the Mother of all things
(1)
The undifferentiated, pre-formed Tao is likened to a piece of uncarved
wood (Pm), in which the world of forms is latent but not yet manifest:
Tao is absolute, nameless.
Like a piece of uncarved wood
(32)
The dichotomy between the pre-formed world of Tao and the
phenomenological universe of forms is made relevant, from the very
start, to the realm of human conduct. This is done through depicting Tao
as equally irrelevant to human emotions as it is to human cognition:
Seek the hidden core without passion.
Revel in its myriad forms with passion
(1)
In spite of the dichotomy, Taoist mysticism insists that Tao ('the hid¬
den core'), while impenetrable to human distinction making, is nonethe¬
less the source of the phenomenological world ('the myriad forms'):
The two flow from the same source
But get different names2
(1)
This theme, of the inaccessibility of Tao, is indeed recurrent:
Looked at, it cannot be seen.
It is the invisible.
Listened to, it cannot be heard.
It is the inaudible.
Reached for, it cannot be touched.
It is the intangible...
...Without beginning, without end.
It cannot be defined.
And so it reverts to the realm of no thing
(14)

2 In Taoist epistemology, 'names' stand equally for 'sensory distinction' and 'cognitive
categories'. As noted in Chapter 1, the historical cleavage between the Aristotelian and
Platonic, about the primacy of concepts or percepts, was resolved in the Western tradition
by Kant. This chicken-egg quibble never arose in Taoism, which in a sense has always been
Kantian.
408 Mind, Code and Context

From a diachronic perspective, within Taoist cosmology, Tao is the

primal source:

Before heaven and earth

There was a mystery.
Silent, boundless, alone.
Changing yet always the same.
Mother of all things
(25)

All things in the formed world arose from that primal 'no thing':

All things arise from being.

Being arises from non-being
(40)

11.2.2. The Two

"...What keeps the Yin and the Yang together.
The circle around them
Or the line inbet ween?..."
S. Buker

In the Western rational inquiry, the existence of discrete categories is

founded upon the Law of the Excluded Middle. Taoist dialectics remains
squarely outside this tradition. Within the unifying principle of Tao,
contraries do not clash, but rather find their resolution:

In it sharp edges are blunted.

Tangles resolve.
Bright lights are tempered.
Conflicts submerge
(4)

What makes this contradictory doctrine coherent is its belief in the non¬
contradictory nature of apparent dichotomies. In the more analytic
terms of Western pragmatics, the Taoist position exploits two com¬
plementary features of pragmatics:

(a) The Wittgensteinean scalarity of conceptual space; and

(b) The context-dependent nature of perceived opposites; i.e.
the fact that they are irreconcilable only from a certain
perspective.

The major symbolic representation of the Taoist doctrine of com¬

plementarity of opposites is the emblem of the Supreme Ultimate (Tai
Chi), made out of the Yin and Yang:
 The Mystic as Pragmatist 409

(1)

Traditionally, Yin stands for the dark, female, curving, passive; while
Yang stands for the light, male, angular, active. But as the emblem sug¬
gests, there is a speck of Yin within the Yang and a speck of Yang within
the Yin. Their contrariness is thus only relative, a matter of degree or
perspective.
The mystic union of Yin and Yang within the circle of Tao represents
the very essence of Taoist metaphysics -- unity in diversity. This
metaphysics may be interpreted both diachronically and synchronically:

From Tao one is born.

From one, two.
From two, three.
From three, all.

Yin is the back of all.

And Yang its face.
From the union of the two
The world attains its balance
(42)

The latency of one opposite within its contrary is of course familiar to

students of the semantics of antonyms3 and the pragmatics of negation.4
The Tao Teh Ching expounds this dialectics as follows:

What shrinks must first be large.

What weakens must first be strong.
What falls must first be high.
What loses must first possess

3 See discussion of the semantics of paired adjectives in Bierwisch (1967) or Givon (1970),
inter alia.
4 See discussion in Chapter 4.
410 Mind, Code and Context

The same dialectics characterizes, by analogy, the relationship between

life and death:

Thirteen organs accompany life.

Thirteen organs usher in death.
Thirteen organs move a man
Through life to death.
Why is this so?
Because living tips the scale toward death
(50)

While Yin-Yang is per se a binary distinction, it is reasonably clear from

the Tao Teh Ching context that duality is a mere stand-in for diversity or
multiplicity. That is, once the first dichotomy has been provided for, in
departing from the pre-formed Tao, one has opened the Pandora Box of
the world of forms. The Taoist approach to this post-Tao universe is
reminiscent of the Buddhist notion of Samsara — the world of illusion.
Yet a subtle difference persists, and is indeed the hallmark of Taoist
pragmatism. Lao Tse never denies the reality ('usefulness') of the world
of forms. While exhorting us to not lose sight of the forest (Ultimate
Reality) for the trees (worldly forms), the latter is conceded its rightful
context:

Thirteen spokes unite at the hub.

But the wheel hinges on its empty hole.
Clay is molded into a cup.
But the space within is what is filled.
Walls and a roof make a house.
But the hollow inside is where you live.
Thus, while the tangibles have their place.
It is the intangible that is used
(11)

As we shall see further below, the dialectic unity-in-diversity of Yin and

Yang is equally central to Taoist epistemology and ethics.
 The Mystic as Pragmatist 411

Lastly, Lao Tse sensibly emphasizes the pre-rational, indeed mind-

boggling nature of Tao as ultimate reality.5 Human attempts to rational¬
ize it are bound to yield logically contradictory results:
Tao is an empty bowl.
Drawn from, it remains full.
Fathomless, it is the source of all things
(4)
And further:
The thing called Tao
Is elusive, evasive.
Evasive, elusive.
Yet full of latent forms.
Elusive, evasive.
Yet full of latent objects
(21)
And again:
Thus, it is said of Tao:
Whoever understands it seems duller.
Whoever follows it seems to retreat.
Its even path seems crooked
(41)

5 One may wish to compare Lao Tse's terseness and common sense with Heidegger's (1959)
belabored discussion of 'being' and 'non-being', and heavy-handed rationalization of the
irrational: "...The question "How is it with being?" is included as a preliminary question in
our central question "Why are there essents rather than nothing?" If we now begin to look
into that which is questioned in our preliminary question, namely being, the full truth of
Nietzsche's dictum is at once apparent. For if we look closely, what more is "being" to us
than a mere word?..." (1959, p. 32) It has been noted (see e.g. Popper, 1950) that the marriage
of Platonic reductionism and obfuscatory mysticism tends to produce virulent results. Here
is how Heidegger motivates his metaphysical quest: "...We ask the questions "How does it
stand with being?" 'What is the meaning of being?" not in order to set up an ontology of
the traditional style, much less to criticize the past mistakes of ontology. We are concerned
with something totally different: to resolve man's historical being-there — and that always
includes our own future being-there in the totality of the history allotted us — to the domain
of being, which it was originally incumbent upon man to open up for himself..." (ibid., p.
34) Heidegger's grand design slowly unfolds with a Nitzschean Jeremiad: "...The spiritual
decline of the earth is so far advanced that the nations are in danger of losing the last bit of
spiritual energy..." (ibid., p. 31; emphasis is mine; TG) Now the juxtaposition begins to sound
familiar: "...What philosophy essentially can and must do is this: a thinking that breaks the
path and opens the perspectives of knowledge that sets the norms and hierarchies, of the
knowledge in which and by which a people fulfills itself historically and culturally...the
challenge is one of the essential prerequisites for the birth of all greatness, and in speaking
of greatness we are referring primarily to the works and destinies of nations..." (ibid., p. 9;
emphases are mine; TG) The original version of this text was delivered as Heidegger's
inaugural lecture, on the occasion of his elevation to the post of Rector of the University of
Freiburg in 1935.
412 Mind, Code and Context

11.2.3. The circle: Too as drift

The Supreme Ultimate, by whatever name, is taken by all mystic

traditions to hold sway over the world of forms. The capricious pique of
Zeus, the jealous fury of Jehovah, are both reflections — somewhat
crude, to be sure -- of this power. In contrast, Tao is said to be a gentle,
non-coercive force. It prevails not by brute exertion, but through its
non-arbitrariness. This aspect of Tao is an essential ingredient in Taoist
metaphysics (as well as epistemology and ethics):

Tao acts by returning.

Its function is soft, gentle
(40)

It is this non-coercive yet ultimately prevailing force that is likened to

the uncarved wood (Pu):

Tao is absolute, nameless.

Like a piece of uncarved wood.
Useless, harmless
(32)
And again:

Tao is everywhere.
Left, right, source of all.
Nothing is rejected.
It gets involved but lets be.
Provides but lays no claim.
Mild and even-tempered.
It seems small
(43)

And, in a metaphor reminiscent of the proverbial fisherman:

Tao's net is loose, wide.

Its mesh is large.
Yet nothing slips through
(73)

The concept of 'power' that emerges here is akin to gravity or

entropy. Tao does not prevail by force, nor by arbitrary choice, nor by
the exercise of will, but through being the inherent directionality of the
universe. The symbolic expression of this mode of power, the circle in
which the Yin and the Yang both clash and unite, is again a metaphor for
Taoist cosmology as well as epistemology. In the latter, it stands for the
two core pragmatic features:
 The Mystic as Pragmatist 413

(i) The graduality of the transition along semantic space;6 and

(ii) the contextual relativity of polar distinctions.

The Yin-Yang emblem assumes, in Taoism, the broad scope and sig¬
nificance that similar pictorial metaphors carry in other great mystic
traditions.7

11.3. Taoist epistemology

11.3.1. Two realities, two modes of knowledge

Western epistemology, whether empiricist or rationalist, shares the

fundamental assumption concerning the separateness of mind and
world. Taoist epistemology concedes such separateness as one half of
the truth -- in the context of the world of forms. In this world, percepts
and concepts are relevant and necessary. Lao Tse's intuition in this mat¬
ter is surprisingly Kantian, taking it for granted that perceptual-concep¬
tual knowledge will never penetrate the Ding am Sich:

The Tao that can be told of is not the real Tao,

Names that can be given are not real name
(1)

Behind the reality of the world of forms lies another reality, that of Tao.
Since the two are not apprehended by the same mode, human
knowledge remains fundamentally paradoxical:

In accumulating knowledge.
Can you renounce the mind?
(1)

Scholarly knowledge gets the short end of the stick in Taoist writings,
a fact that is perhaps motivated by the contemporary context of Con-
fucian scholarship:

Banish wisdom, discard knowledge.

And people will know a hundred-fold more
(19)

6 This aspect of the circle serves, in Taoist pragmatics, the same role as Aristotle's synolon,
which made possible the change from one discrete 'form' to another.
7 The Buddhist Mandala to name one, the Plains Indians' Medicine Shield to name another,
the Navajo-Hopi Four-Comers emblem, or the Indo-European Wheel of Fortune.
414 Mind, Code and Context

Nature has only few words.

How many fewer should man have?
(23)

The distinctions that the intellect can make are relevant only within
their narrow domain:

You break up nature to make tools and artifacts.

In the hands of the sage, tools are means to an end.
The master carver never carves
(28)

As Lao Tse suggests, perceptual-conceptual knowledge can easily be¬

come a trap, enticing one to carve up the ultimate unity into irrelevant
detail:

Civilization breeds names.

Names have their limit
(32)

Scholars aim to know.

The sage aims to loose knowledge
(48)

Taoist epistemology thus takes for granted two types of reality, each with its
own — contextually appropriate -- mode of knowledge. Mundane reality can
be apprehended through perceptual and conceptual distinctions. In the ul¬
timate reality, all such distinctions dissolve. Or, in the words of Chuang Tse:8

Great knowledge sees all in One,

Small knowledge breaks One into many
(ii, 2)

Having thus dispensed with rationalism, Lao Tse proceeds to dismiss

empiricism with equal vigor:

The five colors blind the eye.

The five notes deafen the ear.
The five tastes dull the palate
(12)

8 From Merton's (1965) translation.

 The Mystic as Pragmatist 415

The supreme ultimate, Tao, is in principle inaccessible to distinction¬

making:

Looked for, it cannot be seen.

Listened to, it cannot be heard
(35)

Is Tao then knowable? Lao Tse, like other mystics before and since,
proffers somewhat contradictory advice. First, he suggests that through
knowing the ultimate one may know the mundane:

The source of the universe

May be likened to its mother,9
From knowing the mother
One may know the sons
(52)

Second, he suggests that the ultimate can be known through self-con¬

templation:

How do I know this of the world?

Through myself
(54)
He who knows others is wise.
He who knows himself is wisest
(33)

11.3.2. Yin and Yang: The context-dependence of categories

As we have noted on several occasions above, post-Socratic Western

epistemology is founded upon the Law of the Excluded Middle, thus
upon categories that are discrete and mutually exclusive. The Taoist ap¬
proach to categories, on the other hand, recognizes the relativity of bi¬
nary distinctions, the fact that they are polar only within a limited con¬
text. The best expression of this doctrine, below, is indeed reminiscent of
Socrates' pragmatic line in Hippias Major (itself attributed to Heraclitus;
see Chapter 1):

9 'Mother' in this particular sutra stands for the pre-formed world of Tao. In another instance
(cf. sutra 1 of the Tao Teh Ching), the same term is used for the first stage of the post-Tao
universe, i.e. the yet-to-be-divided-but-already-divisible universe.
416 Mind, Code and Context

When the world sees beauty, it knows ugliness.

When it perceives good, it recognizes evil.
Thus,
The dark and the light reflect each other.
The hard and the soft explain each other.
The long and the short reveal each other.
The high and the low define each other.
The loud and the silent expose each other.
The front and the back outline each other

Taoist epistemology may be summed up as follows:

(a) Discrete categories and binary distinctions are arbitrary

divisions of the continuum;
(b) Reality is context-bound and multiple; therefore
(c) Paradoxes are inherent; and
(d) Each mode of knowledge is applicable only within its right¬
ful domain.

11.4. Taoist ethics

11.4.1. The ethics of entropy

Ethics, the branch of philosophy dealing with social conduct and
motivation, has remained a problematic link in the chain of argument in
most philosophical traditions. East and West. Rational Western phi¬
losophers have striven, for the most part in vain, to make ethics a logical
derivative of their epistemology or metaphysics.10 Western religions have
derived ethics, somewhat arbitrarily, from the Supreme Deity's com¬
mandments, with an elaborate reward-and-punishment scheme to secure
compliance.11 In this tradition, the absolute, discrete -- Platonic --
categories good and evil somehow contrive to emerge — Deus ex machina —

10 One of the least satisfactory aspects of Kant's philosophy is his muddled attempt to
integrate his ethics into the grand schema of his metaphysics and epistemology.
11 Modem utilitarians, such has Mill, have not fared much better, substituting the vagaries of
the social covenant for the wrathful Deity.
 The Mystic as Pragmatist 417

as the product of the deity's arbitrary will and infinite power to mete
reward and punishment.
Taoist ethics, like all else, springs from Tao as the natural flow, the in¬
herent directionality, the grain against which the carver finds the going
rough and unrewarding. The doctrine of Wu-Wei ('no do'), the cor¬
nerstone of Taoist ethics, is rooted in this aspect of Tao. Taoist ethics, un¬
like the Western right/wrong, the Buddhist Karma12 or the Confucian
Canon (in opposition to which Lao Tse fashioned his approach), makes
no reference to absolute 'right' or 'wrong'. And although Lao Tse extols
the utilitarian virtues of Wu-Wei, utility and the common good do not
play a central role in motivating Taoist ethics. The core dichotomy is
rather that of natural vs. unnatural, that which conforms with Tao vs.
that which goes against the flow. The latter is not evil or immoral, but
only unwise and self-destructive.

11.4.2. Wu-Wei and Tao: The water metaphor

Wu-Wei is variously translated as 'no do', 'inaction', 'passivity7,

'serenity', 'gentleness' or even 'peace'. The most persistent metaphor
used by Lao Tse to expose this principle of human conduct is that of
gravity-bound water:

Be like water.
It quenches all thirst yet lays no claim.
It flows to the lowest place
And thus rejoins Tao
(8)
Nothing is weaker than water.
Yet nothing like it wears down the strong.
Nothing is quite like it.
The weak wears down the strong.
The soft overcomes the hard
(78)

The interplay between Wu-Wei and Tao is made explicit from the very
start:

12 One may argue that Karma entails no moral imperatives, but only the factual observation
concerning the cause-and-effect relationship between one's personal conduct and eventual
destiny. It is nevertheless true that in both main-line Mahayana and Theravada Buddhism,
the long path of Karmic accretion has become identified, at least in practice, with a rigid
system of good-and-evil, reward-and-punishment ethics. Both branches of Vajrayana
(Tibetan Tantrism and Zen) assert this to be the case. For some more discussion, see Blofeld
(1958).
418 Mind, Code and Context

Tao never acts.

Yet through it all is done.
If rulers did likewise.
The world will turn of its own accord
(37)

And further:

Those who aim to win the world

And bend it in their image
Will fail.
The world is God's vessel.
It cannot be re-made.
To tamper is to spoil.
To grab is to lose hold.
Some things go fast while others lag.
Some things glow hot while others freeze.
Some things grow strong while others wilt.
Some break, some mend.
The sage never interferes
(29)

The follower of Tao, the would-be sage, is exhorted to abide by the

paradoxical 'no do':

Yield and you shall become whole.

Bend and you shall remain strait.
Be empty and you shall be filled.
Wear out and you shall be renewed.
Have nothing and you shall have all
(22)
Attain the utmost by being passive.
Follow the peaceful road.
When things burst around,
I contemplate their immobility.
Like lush growth
Forever returning to earth.
To return is to regain peace.
Peace is regaining one's destiny.
To regain one's destiny
Is to know the Eternal One
(16)
 The Mystic as Pragmatist 419

The most common misinterpretation of Wu-Wei is to equate it with in¬

action. Rather, Wu-Wei may be viewed as an exhortation to shun ar¬
bitrary, unmotivated action, shun going against the grain, swimming
up-river, opposing Tao. In the Western ethical tradition, action is mo¬
tivated through will, that hallmark of the purposive agent. Taoist ethics
elects to emphasize instead recognition or perception of the natural
drift. Action, when undertaken, best accommodate to such perception.
Much like knowledge, action remains a paradox:

The master carver never carves

(28)

But never carving does not mean 'not carving', much like Wu-VJei does
not mean 'inaction'. Rather, one is exhorted to not carve against the grain,
not act against the drift.

1 1.4.3. Wu-Wei as a utilitarian creed

Above all, Lao Tse says, going with the flow is practical:

To strive is to oppose Tao,

Whoever opposes Tao dies young
(55)

As noted earlier above, Taoist ethics is not founded upon utilitarian con¬
siderations. Nonetheless, its practicality is never ignored. The follower
of Tao is said to possess the quality of Teh, another term that defies trans¬
lation, standing for the 'essence of the sage'.13 Thus:

Whoever has Teh is like a child.

Snakes won't bite him.
Beasts won't attack him.
Birds won't prey on him.
His bones are soft, his flesh tender.
Yet his grasp is firm
(55)

The dangers of over-doing receive particular attention in the prag¬

matic ethics of Lao Tse:

13 The common rendition of Teh as either 'virtue' or 'power' (cf. Waley, 1934) is both restricted
and misleading, pulling Teh toward the Confucian Li.
420 Mind, Code and Context

Pull the bow-string too hard

And you wish you'd stopped in time.
Temper a sword too sharp
And the edge will soon wear out.
Fill your house with jade
And it cannot be guarded
(9)

Arbitrary striving is disparaged in the following words:

He who stands on his tip-toes will soon tire.

He who hurries his step will not walk far.
He who burns bright will not shine for long
(24)

In contrast:

A good traveller leaves no tracks,

A good speaker is seldom heckled,
A good merchant needs no scales,
A well-shut door requires no bolt
(27)

11.4.4. The Code: Li vs. natural kinship

The natural ethics of Wu-VJei evolved in the context of the prevailing
Confucian Canon. Rigid and prescriptive, the Canon rested upon the
notion of Li, variously translated as 'morals', 'standards', 'etiquette',
'propriety', 'ritual' or 'ceremony'. To this social-based standard, Lao Tse
contrasted his natural morality, in which the only imperative — if impera¬
tive it indeed is -- is to recognize the flow of Tao and refrain from oppos¬
ing it. Like Rousseau, Lao Tse was a firm believer in the inherent decen¬
cy of unspoiled humanity au naturel. That natural state is often likened
to the innocence of newborn babes. The Tao Teh Ching is replete with
passages describing the sage as gentle, considerate, self-effacing and
compassionate. On this background, many commentators have puzzled
over certain sutras that appear to clash with the pervading humane tone
of the Tao Teh Ching. For example:

When Tao is lost, 'compassion' and 'justice' arise.

Then 'knowledge' and 'wisdom' appear
With hypocrisy in their wake
(18)
 The Mystic as Pragmatist 421

To understand this seeming contradiction, one must consider again the

historical context of Lao Tse's writings, i.e. the Confucian Canon with its
extensive inventory of dos and don'ts and assorted utilitarian incentives.
Lao Tse considered such behavior-modifying ethics worthless. For him,
ethical behavior must spring from a deeper, natural source. As Lao Tse
saw it, Tao obviates unsatisfactory arbitrary concepts such as Justice or
Compassion.
Equally, the Tao Teh Ching disparages the kinship-based, reward-laden
Confucian code:

When the Six Relations are lost,

'Kind fathers' and 'devoted sons' appear
(18)

The 'Six Relations' represent natural kinship, deep loyalty that springs
from Tao and requires neither justification nor exhortation nor reward.
The imperatives of the Code, Lao Tse suggests, are only necessary when
natural kinship is lost or disregarded. Here is what he says of the Con¬
fucian Li:

When Tao is lost, 'compassion' arises.

When 'compassion' is lost, Li arises,
Li is the death of loyalty and honesty
And the beginning of chaos
(38)

In this context:

Drop 'compassion', abandon 'justice'.

And people will find their true loyalties
(19)

11.4.5. The doctrine of Straw Dogs

One passage of the Tao Teh Ching has baffled critics and admirers
alike, prompting some to interpret Taoist ethics as a-moral, relativistic,
harsh. The passage invokes the metaphor of Straw Dogs, the expendable
sacrificial figurines burned as offering at ancestral altars en lieu of blood
sacrifice:

The world is not 'good'.

To it all things are Straw Dogs.
The sage is not 'good'.
To him all people are Straw Dogs
(5)
422 Mind, Code and Context

The impression of indifference, of moral neutrality, is further strength¬

ened by the following description of the sage:

He treats the good kindly.

He treats the bad kindly...
...He believes the honest.
He also believes the liar
• ' (49)

Again, echoes of early Christianity seem to reverberate here,14 with the

age-old confrontation between the mystic's paradoxical, natural a-
morality, and organized religion's institutional, codified ethics. Lao Tse's
seeming nihilism is grounded in his vision of the all-embracing Tao:

Tao is a mystery, source of all.

Treasure to the good.
Refuge to the bad.
Fine words may come at a price.
Fine manners may be contrived.
Though the bad be bad,
They need not be rejected...
...Seek the guilty and then forgive
(62)

And further:

The executioner is often also killed.

To presume to be one
Is to presume to wield the Master Carver's axe:15
Whoever wields the Master Carver's axe
May chop his own hands
(74)

Taoist ethics emerges here as non-judgemental as Christ's ('Let he

who is without sin...'), and just as compassionate:

I have three treasures:

The first is love.
The second moderation.
The third humility
(67)

14 'I have not come to call the righteous, but the sinners to repent'.
15 The Master Carver is used by both Lao Tse and Chuang Tse as metaphor for the sage. In
the present context it is just as likely a metaphor for Tao.
 The Mystic as Pragmatist 423

Lastly, Lao Tse's seeming permissiveness and moral relativism are

firmly grounded in his figure-ground, context-dependent, dialectic epis¬
temology:

How different is 'aye' from 'nay'?

How different is 'good' from 'evil'?
(20)
Who knows the end?
Who knows the norm?
Norms soon become abnormal.
Good turns into evil.
So long have humans been misguided!
(58)

11.5. Closure

The history of pragmatics has often yielded a curious mix of compell¬

ing intellectual triumph and spectacular media failure. This has to do, I
believe, with an in-built hazard of all middle-ground positions: They
never seem quite as attractive, quite as clear-cut, striking or salient, as
the well-articulated, aggressive extremes.16 The middle ground often is,
as Lao Tse observed, socially murky, dull, unattractive:

All the people have plenty,

I alone have nothing,
Like a fool, muddled, befogged....
...All the people are smug and clever,
I alone am dim, dull.
Aimless as the sea, forever drifting
(20)

The way 'great philosophical questions' have always been formulated, in

reductionist terms, invites extreme responses. When one asks: "Is it the
chicken or is it the egg that came first?", one presupposes that either one or
the other must have come first, that complex interactions can and must
be reduced to neat, linear, cause-and-effect scripts.17 It is an old story.
In epistemology, traditional rationalism and empiricism have both in¬
sisted on an either-or solution to the ontology of percepts and concepts,
Kant's middle ground notwithstanding. In semantics, extreme
Platonists and avid Wittgensteineans have both sought pure answers to

16 The fact that the middle also catches the flak from both extremes is an added hazard.
17 As Alfred North Whitehead observed: "...philosophical truth is to be sought in the presup¬
positions of language rather than in its expressed statements..." (1938, p. vii)
424 M ind. Code and Context

the puzzle of natural categories, rejecting a compromise through which

both categoriality and gradation are preserved — in the appropriate con¬
text. In the empirical study of cognition, both Chomsky and Skinner
have insisted on a reductionist rendition of the interaction between
mind and input, whereas the facts point toward a context-mediated
mix.
In linguistics likewise, one is told that only law-like generalizations
are worthy of the name; that grammar's that are not context-free will bog
us in contextual mire; that failure to demonstrate that all features of
grammar are functionally motivated must entail failure to argue for
functional motivation for some or most.
In the philosophy of science, deductivists and inductivists share their
abiding faith in extreme reduction, never mind the facts of doing real
science, never mind abduction. In cultural anthropology, extreme uni-
versalists and extreme relativists struggle for exclusive dominion. In
evolutionary biology, extreme Lamarckians and extreme mutatio-selec-
tionists still act as each other's deadly foils. In artificial intelligence, top-
down (deductive, linear) and bottom-up (inductive, parallel) processing
are presented as exclusive alternatives rather than natural comple¬
ments. And so it goes.
What falls by the wayside, in each case, is the vast empirical evidence
suggesting that living organisms have long ago, perhaps from the very
start, opted for interactive, hybrid, pragmatic solutions. That they are
capable of a wide range of behavioral responses. That their adaptive-
context demands -- indeed rewards -- selective access to what Mayr has
called 'open programs'.
The position of pragmatics in the intellectual arena is just as frustrat¬
ing today as it must have been in Lao Tse's or Aristotle's time. As then,
the pragmatist must put up with a certain measure of murk that comes
with the territory, a whiff of Wittgenstein's 'friction'. As then, the prag¬
matist must resist, against considerable odds, the seductive clarity of
reductionist answers; or rather, of reductionist questions.
BIBLIOGRAPHY

Achinstein, P. (1965) "The problem of theoretical terms", in B. Brody (ed.,

1970)
Adams, M.J. (1979) "Models of word recognition", J.V.L.V.B., 16: 277-304
Adler, M. (1967) The Difference of Man and the Difference it Makes, NY:
World Publ. Co.
Ackrill, J.L. (tr. & ed., 1963) Aristotle's Categories and De Interpretatione Ox¬
ford: Clarendon Press
Allen, R.E. (ed., 1966) Greek Philosophy: Thales to Aristotle NY: Macmil¬
lan/The Free Press
Andersen, H. (1983) "A map of the territory", paper read at the Symposium on
Iconicity in Syntax, Stanford University, June, 1983 (ms)
Anderson, A. S.C. Garrod and A.J. Sanford (1983) "The accessibility of
pronominal antecedents as a function of episodic shift in narrative text".
Quarterly J. of Experimental Psychology, 35A: 427-440
Anderson, J.R. and G.H. Bower (1973) Human Associative Memory,
Washington, DC: Winston
Anderson, J.R. and R. Paulson (1978) "Interference in memory for pictorial infor¬
mation", Cognitive Psychology, 10:178-202
Anderson, R.C. and J.W. Pichert (1978) "Recall of previously unrecallable infor¬
mation following a shift in perspective", J.V.L.V.B., 17:1-12
Andrade, E.N. (1954) Sir Isaac Newton: His Life and Work, Garden City, NY:
Doubleday
Anttila, R. (1977) Analogy; State of the Art Report no. 10, The Hague: Mouton
Aristotle, Categories, in J.L. Ackrill (tr. & ed., 1963)
Aristotle, De Anima, in R. McKeon (ed., 1941)
Aristotle, De Generatione Animalium, in R. McKeon (ed., 1941)
Aristotle, De Interpretatione, in J.L. Ackrill (tr. & ed., 1963)
Aristotle, De Partibus Animalium, in R. McKeon (ed., 1941)
Aristotle, De Sophisticis Elenchis, in R. McKeon (ed., 1941)
Aristotle, Ethics (Ethica Nichomanchea), in R. McKeon (ed., 1941)
Aristotle, Metaphysics, tr. by R. Hope, Ann Arbor: University of Michigan Press
(1952)
Aristotle, Posterior Analytic, in R. McKeon (ed., 1941)
Aristotle, Prior Analytic, tr. by Jenkinson, ed. by Ross, Oxford: Oxford Univer¬
sity Press
426 Mind, Code and Context

Aronson, D. and S. Ferres (1983) "Lexical categories and reading tasks", J. of Ex¬
perimental Psychology: Human Perception and Performance, 9
Aronson, D. and H.S. Scarborough (1976) "Performance theories for sentence en¬
coding: Some quantitative evidence", J. of Experimental Psychology: Human
Perception and Performance, 2
Ascher, J.J. and R. Garda (1969) "The optimal age to learn a foreign language".
Modern Language Journal, 8
Attneave, F. (1957) "Transfer of experience with a class-schemata to identification
learning of patterns and shapes", J. of Experimental Psychology, 54
Attneave, F. (1959) Application of Information Theory to Psychology, NY: Holt
Austin, J. (1962) How to Do Things with Words, Cambridge: Cambridge
University Press
Bach, E. (1965) "Structural linguistics and the philosophy of science", Diogenes,
51
Baggish, H. (1983) "Confessions of a former cultural relativist". Anthropology,
83/84
Baldwin, J.M.A. (1896) "A new factor in evolution", American Naturalist, 30
Bartlett, F. (1932) Remembering: A Study in Experimental and Social Psychol¬
ogy, Cambridge: Cambridge University Press
Bartsch, R. (1984) "Norms, tolerance, lexical change, and context-dependence of
meaning", J. of Pragmatics, 8.3
Bates, E. (1976) Language in Context: The Acquisition of Pragmatics, NY:
Academic Press
Bateson, G. (1972) Steps to an Ecology of Mind, NY: Ballantine Books
Bateson, G. (1979) Mind and Nature: A Necessary Unity, NY: Bantam
Becker, A.L. (1979) "Communication across diversity", in A.L. Becker and A.A.
Yengoyan (eds) The Imagination of Reality, Norwood, NJ: Ablex
Bentivoglio, P. (1983) "Topic continuity and discontinuity in discourse: A study
of spoken Latin-American Spanish", in T. Givon (ed., 1983a)
Berlin, B. and P. Kay (1969) Basic Color Terms: Their Universality and Evolu¬
tion, Berkeley: U.C. Press
Bever, T.G., J.A. Fodor and W. Weksel (1965) "Is linguistics empirical?".
Psychological Review, 72
Bickerton, D. (1975) "Creolization, linguistic universals, natural semantax and
the brain". University of Hawaii Working Papers in Linguistics, Honolulu (ms)
Bickerton, D. (1981) Roots of Language, Ann Arbor: Karoma
Bickerton, D. (forthcoming) Language Evolution
Bierwisch, M. (1967) "Some semantic universals of German adjectives". Founda¬
tions of Language, 3.1
Blofeld, J. (1958) The Zen Teachings of Hunag-Po, NY: Grove Press
Bloomfield, L. (1922) "Review of Sapir's Language", The Classical Weekly, 18
 Bibliography 427

Bloomfield, L. (1926) "A set of postulates for the science of language". Language, 2
Bloomfield, L. (1933) Language, NY: Holt, Rinehart and Winston
Bolinger, D. (1952) "Linear modification", in his Forms of English, Cambridge:
Harvard University Press (1965)

Bolinger, D. (1954) "Meaningful word-order in Spanish", Boletfn de Filologfa,

Universidad de Chile, vol. 8
Bolinger, D. (1977) The Forms of Language, London: Longmans
Bolinger, D. (1978a) "Intonation across languages", in J. Greenberg et al (eds)
Universals of Human Language, vol. 2, Phonology, Stanford: Stanford U. Press
Bolinger, D. (1978b) "Yes and no questions are not alternative questions", in H.
Hiz (ed.) Questions, Synthese Language Library, #1
Bolinger, D. (1984) "The inherent iconicity of intonation", in J. Haiman (ed., 1984)
Bonner, J.T. (1980) The Evolution of Culture in Animals, Princeton: Princeton
University Press
Bonnet, C. (1764) Contemplations de la Nature, vol. 2, Amsterdam: Marc-
Michel Rey
Bradshaw, J.L. and N.C. Nettleton (1981) "The nature of hemispheric specializa¬
tion in man". Behavior and Brain Science, 4
Bransford, J. and Franks, J.J. (1971) "Attraction of linguistic ideas". Cognitive
Psychology, 2
Broadbent, D.E. (1958) Perception and Communication, London: Pergamon
Press
Broadbent, D.E. (1977) "Levels, hierarchies, and the locus of control". Quarterly
J. of Experimental Psychology, 29
Brody, B. (ed., 1970) Readings in the Philosophy of Science, Englewood Cliffs,
NJ: Prentice-Hall
Bromberger, S. (1966) "Why questions", in B. Brody (ed., 1970)
Brown, P. and S. Levinson (1978) "Universals of language usage: Politeness
Phenomena", in E. Goody (ed., 1978)
Brown, P. and S. Levinson (1979) "Social structure, group and interaction", in K.
Schere and H. Giles (eds) Social Markers in Speech, Cambridge: Cambridge
University Press
Buchler, J. (1939) Charles Peirce's Empiricism, London: Kegan Paul
Bybee, J. (1985) Morphology: A Study of the Relation Between Meaning and
Form, TSL vol. 9, Amsterdam: J. Benjamins
Cairns, H.S. and J. Kamerman (1975) "Lexical information processing during
sentence comprehension", J.V.L.V.B., 14
Carnap, R. (1923) "Uber die Aufgabe der Physik", Kantstudien, 28
Carnap, R. (1934) Der Logische Aufbau der Welt, Tr. by R.A. George as The
logical structure of the world, Berkeley: UC Press (1967)
Carnap, R. (1947) Meaning and Necessity, Chicago: University of Chicago Press
428 Mind, Code and Context

Carnap, R. (1956) "The methodological character of theoretical concepts", in H.

Feigl and M. Scriven (eds) Minnesota Studies in the Philosophy of Science, vol. I
Carnap, R. (1958) Introduction to Symbolic Logic and its Applications, NY:
Dover
Carnap, R. (1959) The Logical Syntax of Language, Patterson, NJ: Littlefield,
Adams & Co.
Carr, T. (1985) "Perceiving visual language", in L. Kaufman, J. Thomas and K.
Boff (eds) Handbook of Perception and Human Performance, NY: J. Wiley
Carroll, J.B. (1953) The Study of Language, Cambridge: Harvard University
Press
Cassierer, E. (1933) "La langue et la construction du monde des objets", J. de
Psychologie Normale et Pathologique, 30
Chafe, W. (1970) Meaning and the Structure of Language, Chicago: University
of Chicago Press
Chafe. W. (1976) "Givenness, contrastiveness, definiteness, subjects, topics and
point of view", in C. Li (ed., 1976)
Chafe, W. (1979) "The flow of thought and the flow of language", in T. Givon (ed.,
1979c)
Chafe, W. (ed., 1980) The Pear Stories, Norwood, NJ: Ablex
Chafe, W. (1987) "Cognitive constraints on information flow", in R. Tomlin (ed.,
1987)
Chafe, W. and J. Nichols (eds, 1986) Evidentiality: The Coding of Epistemology
in Language, Norwood, NJ: Ablex
Chase, W.G. and K.A. Ericsson (1981) "Skilled memory", in J. Anderson (ed.)
Cognitive Skills and their Acquisition, Hillsdale, NJ: Erlbaum
Chase, W.G. and K.A. Ericsson (1982) "Skill and working memory", in G. Bower
(ed.) The Psychology of Learning and Motivation, vol. 16, NY: Academic Press
Chase, W.G. and H.A. Simon (1973) "Perception in chess". Cognitive Psychology, 4
Chang, F. (1980) "Active memory processes in visual sentence comprehension:
Clause effect and pronominal reference". Memory and Cognition, 8
Chisholm, W. (ed.) Interrogativity, TSLvol.4, Amsterdam: J. Benjamins
Chomsky, N. (1957) Syntactic Structures, The Hague: Mouton
Chomsky, N. (1959) "Review of Skinner's Verbal Behavior", Language, 35
Chomsky, N. (1964) "Current issues in Linguistics", in J. Fodor and J. Katz (eds)
The Structure of Language, Englewood Cliffs, NJ: Prentice-Hall
Chomsky, N. (1966) Cartesian Linguistics, NY: Harper and Row
Chomsky, N. (1968) Language and Mind, revised edition [1972], NY: Harcourt,
Brace and World
Chomsky, N. (1981) "On the representation of form and function". The Linguis¬
tic Review, 1:3-40
 Bibliography 429

Cirilo, R.K. and D.J. Foss (1980) "Text structure and reading time for sentences",
J.V.L.V.B., 19

Clark, E. (1971) "What's in a word", in T. Moore (ed.) Cognitive Development

and the Acquisition of Language, NY: Academic Press
Clark, H. (1970) "Linguistic processes in deductive logic". Psychological
Review, 76.4

Clark, H. and B. Malt (1984) "Psychological constraints on language", in W.

Kintsch et al. (eds.) Methods and Tactics in Cognitive Science, Hillsdale, NJ:
Erlbaum
Claudi, U. and B. Heine (1984) "From metaphor to grammar: Some examples
from Ewe", Koln: Institut fur Linguistik, Universitat zu Koln (ms)
Cocchiarella, N.B. (1965) Tense and Modal Logic: A Study in the Typology of
Temporal Reference, PhD dissertation, U.C.L.A. (ms)
Cohen, A. and S. Nooteboom (eds, 1975) Structure and Process in Speech Per¬
ception, Heidelberg: Springer-Verlag
Cohen, N.J. and L.R. Squire (1980) "Preserved learning and retention of pattern
analyzing skill in amnesia: Dissociation of knowing how and knowing that".
Science, 210
Cole, P. (ed., 1978) Pragmatics, Syntax and Semantics, vol. 9, NY: Academic
Press
Cole, P. (ed., 1981) Radical Pragmatics, NY: Academic Press
Cole, P. and J. Morgan (eds, 1975) Speech Acts, Syntax and Semantics, vol. 3,
NY: Academic Press
Cole, R.A. and A.I. Rudnicky (1983) "What's new in speech perception? The re¬
search ideas of William Chandler Bagley, 1874-1946", Psychological Review, 90:
94-101
Coleman, L. and P. Kay (1981) "Prototype semantics and the English word lie",
Language, 57.1
Collins, and M.R. Quillian (1969) "Retrieval time for semantic memory",
J.V.L.V.B., 8:240-247
Cooreman, A. (1982a) "Transitivity, ergativity and topicality in Chamorro", BLS
#8, Berkeley: University of California, Berkeley Linguistics Society
Cooreman, A. (1982b) "Topicality, ergativity and transitivity in narrative dis¬
course: Evidence from Chamorro", Studies in Language, 6.3
Cooreman, A. (1985) Studies in Chamorro Grammar and Discourse, University
of Oregon, Eugene, PhD dissertation (ms)
Cooreman, A., B. Fox and T. Givon (1984) "The discourse definition of er¬
gativity", Studies in Language, 8.1
Corbett, A.T. and F.R. Chang (1983) "Pronoun disambiguation: Accessing poten¬
tial antecedents". Memory and Cognition, 11
Cresswell, M.J. (1972) "The world is everything that is the case", Australian Jour¬
nal of Philosophy, 50.1
430 Mind, Code and Context

Crouch, J.E. (1978) Functional Human Anatomy, 3rd edition, Philadelphia: Lea
and Fabiger
Curtiss, S. (1977) Genie: A Psycholinguistic Study of a Modern-Day
"Wildchild", NY: Academic Press
Cutler, A. and D.J. Foss (1977) "On the role of sentence stress in sentence proces¬
sing", Language and Speech, 21
Cuvier, G. (1828) Le regne animal distribue d'apres son organization, Paris:
Fortin
Dahl, D.A. and J. K. Gundel (1982) "Identifying referents for two kinds of
pronouns", Minnesota Papers in Linguistics and Philosophy, 7
Darwin, C. (1859) The Origin of Species, London: J. Murray
Darwin, C. (1871) The Descent of Man, and Selection in Relation to Sex,
(Heritage Edition, 1971), Norwalk, Conn.: Heritage Press
Davison, A. (1975) "Indirect speech acts and what to do with them", in P. Cole
and J. Morgan (eds, 1975)
Dawkins, R. (1976) The Selfish Gene, NY: Oxford University Press
de Beer, G.R. (1930) Embryology and Evolution, Oxford: Clarendon Press
DeLancey, S. (1984) "The analysis-synthesis-lexis cycle in Tibeto-Burman: A case
study in motivated change", in J. Haiman (ed, 1984)
Dell, G., G. McKoon and R. Ratcliff (1983) "The activation of antecedent information
during the processing of anaphoric reference in reading", J.V.L.V.B., 22:121-132
Denier van der Gon, J.J. and J.P. Thuring (1965) "The guiding of human writing
movements", Kyb emetic, 2
Derbyshire, D. (1985) "Topic continuity and OVS order in Hixkaryana", in J.
Sherzer and G. Urban (eds) Native South American Discourse (ms, SIL-Dallas)
Derbyshire, D. and G. Pullum (1981) "Object initial languages", I.J.A.L., 47.3
Derwing, B. (1973) Transformational Grammar as a Theory of Language Ac¬
quisition, Cambridge: Cambridge University Press
de Waal, F. (1982) Chimpanzee Politics: Power and Sex among the Apes, Lon¬
don: Unwin Paperbacks/Counterpoint
Diels, H. (1969) Die Fragmenten der Vorsokratiker, herausgegeben von Walter
Kranz, 3 vols, Dublin and Zurich: Weidman
Dik, S. (1978) Functional Grammar, Amsterdam: North Holland
Dijk, T. van and W. Kintsch (1983) Strategies of Discourse Comprehension, NY:
Academic Press

Donellan, K. (1966) "Reference and definite description". The Philosophical

Review, 75.3

Dreizin, F. (1980) "The flavor of Russian negation: Some notes on and around".
University of Haifa, Israel (ms)

DuBois, J. (1980) "Beyond definiteness", in W. Chafe (ed., 1980)

DuBois, J. (1985) "Self evident", in W. Chafe and J. Nichols (eds, 1985)
 Bibliography 431

Eco, U. The Name of the Rose, tr. by W. Weaver, NY: Harcourt, Brace and World
Eckert, R. and D. Randall (1978) Animal Physiology: Mechanisms and Adapta¬
tions, NY: Freeman

Edgerton, R. (1978) The study of deviance: Marginal man or Everyman", in G.

Spindler (ed.) The Making of Psychological Anthropology, Berkeley: U.C. Press
Einstein, A. (1934) The world as I see it, tr. by A. Harris, NY: Covici and Friede
Eldredge, N. and S.J. Gould (1972) "Punctuated equilibria: An alternative to
physical gradualism", in T.J.M. Schoopf (ed.) Models in Paleobiology, San Fran¬
cisco: Freeman, Cooper & Co.

Elster, J. (ed., 1985) The Multiple Self, Cambridge: Cambridge University Press
Ericsson, K.A. (1985) "Memory skill", Canadian J. of Psychology, 39.2
Ervin-Tripp, S. (1974) "Is second language learning like the first?", TESOL
Quarterly, 8.2

Feyerabend, P. (1970) "How to be a good empiricist — a plea for tolerance in mat¬

ters epistemological", in B. Brody (ed., 1970)
Feyerabend, P. (1975) Against Method: An Outline of an Anarchic Theory of
Knowledge, NY: The Humanities Press
Fillmore, C. (1963) "The position of embedding transformations in grammar".
Word, 19
Fillmore, C. (1968) "The case for the case", in E. Bach and R.T. Harms (eds)
Universals in Linguistic Theory, NY: Holt, Rinehart and Winston
Fillmore, C. (1971) "How to know whether you're coming or going", in K. Hyl-
gaard-Jensen (ed.) Linguistik 1971, Athenaum-Verlag
Firbas, J. (1966) "On defining the theme in functional sentence perspective",
Traveaux Linguistiques de Prague, 2
Firbas, J. (1974) "Some aspects of the Czechoslovak approach to problems of
functional sentence perspective", in F. Dane§ (ed.) Papers on Functional Sen¬
tence Perspective, Janua Linguorum, sm 147, The Hague: Mouton
Fletcher, C.R. (1982) "Markedness and topic continuity in discourse processing",
J.V.L.V.B., 21
Fodor, J. (1983) The Modularity of Mind, Cambridge: MIT Press
Foley, W. and R. van Valin (1985) Functional Syntax and Universal Grammar,
Cambridge: Cambridge University Press
Fossberg, H. (1985) "Ontogeny of human locomotive control, I: Infant stepping,
supported locomotion and transition to independent locomotion", Experimental
Brain Research, 57
Fox, A. (1983) "Topic continuity in Biblical Hebrew", in T. Givon (ed., 1983a)
Fox, B. (1985) "Word order inversion and discourse continuity in Tagalog", in T.
Givon (ed. 1985b)
Frost, N.A.H. (1971a) "Clustering of visual shape in the free recall of visual
stimuli", J. of Experimental Psychology, 88
432 Mind, Code and Context

Frost, N.A.H. (1971b) Interaction of Visual and Semantic Codes in Long-Term

Memory, PhD dissertation, U. of Oregon, Eugene (ms)
Futuyma, D.J. (1979) Evolutionary Biology, Sunderland, Mass.: Sinauer
Galambov, R. and C.T. Morgan (1960) "The neural basis of learning", in J. Field,
H.W. Magoun and V.E. Hall (eds) Handbook of Physiology, vol. 3, section 1:
Neurophysiology, Washington, DC: American Physiological Society
Ganong, W.F. (1980) "Phonetic categorization in auditory word perception”, J. of
Experimental Psychology: Human Perception and Performance, 6:110-125
Garcfa-Marquez, G. (1975) Cien Anos de Soledad, Buenos Aires: Editorial
Sudamericana
Garfinkel, H. (1972) "Remarks on ethnomethodology", in J. Gumpertz and D.
Hymes (eds) Directions in Sociolinguistics, NY: Holt, Rinehart and Winston
Garrod, S. and A. Sanford (1977) "Interpreting anaphoric relations: The integra¬
tion of semantic information while reading", J.V.L.V.B., 16: 77-90
Gazdar, J. (1979) Pragmatics: Implicature, Presupposition and Logical Form,
NY: Academic Press
Geertz, C. (1972) "Linguistic etiquette", in J.B. Pride and J. Holmes (eds)
Sociolinguistics, Harmondsworth: Penguin
Geertz, C. (1973) The Interpretation of Cultures, NY: Basic Books
Geertz, C. (1984) "Anti anti-relativism", American Anthropologist, 86
Gellner, E. (1982) "Relativism and universals", in M. Hollis and S. Lukes (eds,
1982)

Gernsbacher, M.A. (1985) "Surface information loss in comprehension". Cogni¬

tive Psychology, 17

Gernsbacher, M.A. (1987) "Conceptual anaphora in English: An experimental ap¬

proach", University of Oregon, Eugene (ms)

Gernsbacher, M.A. and D. Hargreaves (1986) "Cognitive accessibility of sentence

participants". University of Oregon, Eugene (ms)

Givon, T. (1966) "Leonard Bloomfield as a philosopher of language". University

of California, Los Angeles (ms)

Givon, T. (1970) "Notes on the semantic structure of English Adjectives", Lan¬

guage, 46.4

Givon, T. (1971) "Historical syntax and synchronic morphology: An

archaeologist's field trip", CLS 7, Chicago: University of Chicago, Chicago Lin¬
guistics Society

Givon, T. (1972) Studies in ChiBemba and Bantu Grammar, Studies in African

Linguistics, Supplement 3

Civon, T. (1973a) "The time-axis phenomenon". Language, 49.4

Givon, T. (1973b) "Opacity and reference in language: An Inquiry into the role of
modalities", in J. Kimball (ed.) Syntax and Semantics, vol. 2, NY: Academic
Press
 Bibliography 433

Givon, T. (1975a) "Serial verbs and syntactic change: Niger-Congo", in C. Li (ed.)

Word Order and Word Order Change, Austin: University of Texas Press
Givon, T. (1975b) "Cause and control: On the semantics of interpersonal
manipulation", in J. Kimball (ed.) Syntax and Semantics, vol. 4, NY: Academic
Press

Givon, T. (1975c) "Focus and the scope of assertion: Some Bantu evidence".
Studies in African Linguistics, 6.2

Givon, T. (1976a) "Topic, pronoun and grammatical agreement", in C. Li (ed.,

1976)

Givon, T. (1976b) "On the VS word-order in Israeli Hebrew: Pragmatics and

typological change", in P. Cole (ed.) Studies in Modern Hebrew Syntax and
Semantics, Amsterdam: North Holland
Givon, T. (1977) "The drift from VSO to SVO in Biblical Hebrew: The pragmatics
of tense-aspect", in C. Li (ed.) Mechanisms for Syntactic Change, Austin: U. of
Texas Press

Givon, T. (1978a) "Definiteness and referentiality", in J. Greenberg et al (eds)

Universals of Human Language, vol. 4: Syntax, Stanford: Stanford University
Press
Givon, T. (1978b) "Universal grammar, lexical structure and translatability", in F.
Guenthner and M. Guenthner-Reutter (eds) Meaning and Translation, London:
Duckworth
Givon, T. (1979a) On Understanding Grammar, NY: Academic Press
Givon, T. (1979b) Ute Dictionary, Ignacio, Colorado: Ute Press
Givon, T. (ed., 1979c) Discourse and Syntax, Syntax and Semantics, vol. 12, NY:
Academic Press
Givon, T. (1980a), "The binding hierarchy and the typology of complements".
Studies in Language, 4.3
Givon, T. (1980b) Ute Reference Grammar, Ignacio, Colorado: Ute Press
Givon, T. (1981a) "Typology and functional domains". Studies in Language, 5.2
Givon, T. (1981b) "On the development of the numeral 'one' as an indefinite
marker". Folia Linguistica Historica, 2.1
Givon, T. (1982a) "Logic vs. pragmatics, with human language as the referee:
Toward an empirically viable epistemology". Journal of Pragmatics, 6.1
Givon, T. (1982b) "Tense-aspect-modality: The Creole prototype and beyond", in
P. Hopper (ed.) Tense and Aspect: Between Semantics and Pragmatics, TSL vol.
1, Amsterdam:]. Benjamins
Givon, T. (1982c) "Evidentiality and epistemic space". Studies in Language, 6.1
Givon, T. (ed., 1983a) Topic Continuity in Discourse: Quantified Cross-Lan¬
guage Studies, TSL #3, Amsterdam:]. Benjamins
Givon, T. (1983b) "Topic continuity in spoken English", in T. Givon (ed., 1983a)
Givon, T. (1983c) "Topic continuity in discourse: An introduction", in T. Givon
(ed., 1983a)
434 Mind, Code and Context

Givon, T. (1984a) Syntax: A Functional-Typological Introduction, Vol. 1,

Amsterdam:]. Benjamins
Givon, T. (1984b) "Universals of discourse structure and second language ac¬
quisition", in W. Rutherford (ed.) Language Universals and Second Language
Acquisition, TSL vol. 5, Amsterdam: J. Benjamins
Givon, T. (1984c) "Iconicity, isomorphism and non-arbitrary coding in syntax", in
J. Haiman (ed., 1984)
Givon, T. (I984d) "Direct object and dative shifting: Semantic and pragmatic
case", in F. Plank (ed.) Objects, NY: Academic Press
Givon, T. (1984e) "The speech-act continuum", in W. Chisholm (ed.) Interrog-
ativity, TSL vol.4, Amsterdam: J. Benjamins
Givon, T. (1984f) "Prolegomena to discourse pragmatics", in C. Caffi (ed.)
Metapragmatics, special issue of J. of Pragmatics, 8.4
Givon, T. (1985a) "The pragmatics of referentiality", in D. Schiffrin (ed.) Mean¬
ing, Form and Use in Context, Washington, DC: Georgetown University Press
Givon, T. (ed., 1985b) Quantified Studies in Discourse, special issue of Text,
5.1/2

Givon, T. (1987a) "Beyond foreground and background", in R. Tomlin (ed., 1987)

Givon, T. (1987b) "The pragmatics of word-order: Predictability, importance and
attention", in E. Moravcsik et al (eds) Typology and Language Universals, TSL
vol.16, Amsterdam:]. Benjamins
Givon, T. (1987c) "Serial verbs and the mental reality of 'event'". Final N.E.H.
Project Report, University of Oregon, Eugene (ms)
Givon, T., W. Kellogg, M. Posner and P. Yee (1985) "The identification of referents
in connected discourse: Automatic vs. attended processing". Cognitive Science
Working Papers, University of Oregon, Eugene (ms)
Givon, T. and A. Kimenyi (1974) "Truth, belief and doubt in KinyaRwanda",
Studies in African Linguistics, Supplement #5
Glazner, D., B. Fischer and D. Dorfman (1984) "Short term storage in reading",
J.V.L.V.B., 23

Goede von Aesch, A. (1941) Natural Science in German Romanticism, NY:

Columbia University Press

Goffman, H. (1974) Frame Analysis, Cambridge: Harvard University Press

Goffman, H. (1976) "replies and responses". Language and Society, 5
Goguen, J. (1969) "The logic of inexact concepts", Synthese, 19
Goodall, ]. (1965) "Chimpanzees of the Gombe reserve", in E. deVore (ed.)
Primate Behavior, NY: Holt, Rinehart and Winston

Goody, E. (ed., 1978) Questions and Politeness: Strategies in Social Interaction,

Cambridge: Cambridge University Press

Gordon, D. and G. Lakoff (1971) "Conversational postulates", CLS 7, Chicago:

University of Chicago, Chicago Linguistics Society
 Bibliography 435

Gould, S.J. (1977) Ontogeny and Phylogeny, Cambridge: Harvard University

Press

Gould, S.J. (1980) The Panda's Thumb, NY: Pelican

Gould, S.J. and N. Eldredge (1977) "Punctuated equilibria: The tempo and mode
of evolution reconsidered". Paleobiology, 3

Grace, G. (1986) "The intertranslatability postulate and its consequences", in his

Ethnolinguistic Notes, 3.25, Honolulu: University of Hawaii at Manoa (ms)
Green, G. (1970) "Whimperatives: Schizophrenic speech acts". University of
Michigan, Ann Arbor (ms)

Green, G. (1975) "How to get people to do things with words: The whimperative
question", in P. Cole and J. Morgan (eds, 1975)
Greenfield, P. and L. Schneider (1977) "Building a tree structure: The develop¬
ment of hierarchic complexity and interrupted strategies in children's construc¬
tion activity". Developmental Psychology, 13.4
Grice, H. P. (1968/1975) "Logic and conversation", in P. Cole and J. Morgan (eds,
1975)

Griffin, D.R. (1976) The Question of Animal Awareness, NY: Rockefeller

University Press
Griffin, D.R. (1984) Animal Thinking, Cambridge: Harvard University Press
Grillner, S. (1975) "Locomotion in vertebrates -- central mechanisms and reflex
interaction". Physiological Review, 55
Grossman, M. (1980) "A central processor for hierarchically-structured material:
Evidence from Broca's aphasia", Neuropsychologia, 18
Gumperz, J. (1977) "Sociocultural knowledge in conversational inference", in M.
Saville-Troike (ed.) Linguistics and Anthropology, Washington, DC: George¬
town University Press
Gumperz, J. (1982) Discourse Strategies, Cambridge: Cambridge University
Press
Gurfinkel, V.S. and M.L. Shik (1977) "The control of posture and locomotion", in
A.A. Gyikov, N.T. Tanov and D.S. Kosarov (eds) Motor Control, NY: Plenum
Press
Haberland, H. (1985) "Review of Klaus Heinrich's Dahlemere Vorlesungen I, J.
of Pragmatics, 9
Haberlandt, K. (1980) "Story grammar and reading time of story constituents".
Poetics, 9
Haeckel, E. (1866) Generelle Morphologie der Organismen: Algemeine
Gerundziige der Organischen Formen-Wissenschaft, Mechanisch begrundet
durch die von Charles Darwin reformierte Descendenz-Theorie, (2 vol.), Ber¬
lin: Georg Reimer
Haiman, J. (1978) "Conditionals are topics". Language, 54
Haiman, J. (1979) Hua: A Papuan Language of the Eastern Highlands of New
Guinea, Amsterdam: J. Benjamins
436 Mind, Code and Context

Haiman, J. (1980) "The iconicity of grammar: Isomorphism and motivation".

Language, 56.3
Haiman, J. (1983) "Iconic and economic motivation". Language, 59
Haiman, J. (ed., 1984) Iconicity in Syntax, TSL vol.6, Amsterdam: J. Benjamins
Haiman, J. (1985) Natural Syntax, Cambridge: Cambridge University Press
Halliday, M.A.K. (1967) "Notes on transitivity and theme in English", J. of Lin¬
guistics, 3
Hallett, G. (1977) A Companion to Wittgenstein's "Philosophical Investiga¬
tions", Ithaca: Cornell University Press
Hanson, R.N. (1958) Patterns of Discovery, Cambridge: Cambridge University Press
Hanson, R.N. (1961) "Is there a logic of scientific discovery?", in H. Feigl and G.
Maxwell (eds) Current Issues in the Philosophy of Science, NY: Holt, Rinehart
and Winston
Hargreaves, D. (1983) Evidentiality in Newari, MA Thesis, University of
Oregon, Eugene (ms)
Haviland, J.B. (1986) "Complex referential gestures in Guugu Yimidhirr", Reed
College, Portland, Oregon (ms)
Hawkins, H.L. (1969) "Parallel processing in complex visual discrimination".
Perception & Psychophysics, 5

Hawkinson, A. and L. Hyman (1974) "Natural topic hierarchies in Shona",

Studies in African Linguistics, 5

Heidegger, M. (1959) Introduction to Metaphysics, NY: Doubleday (1961 ppbk

edition)

Heine, B. and M. Reh (1984) Grammaticalization and Reanalysis in African

Languages, Hamburg: Helmut Buske Verlag
Heller, J. (1962) Catch-22, London: Transworld Corgi (1970 ppbk edition)
Hempel, C. (1945) "Studies in the logic of confirmation". Mind, LIV.l
Hempel, C. (1959) "The logic of functional analysis", in B. Brody (ed., 1970)
Hempel, C. (1966) Philosophy of Natural Science, Englewood Cliffs, NJ: Pren¬
tice Hall

Hempel, C. and P. Oppenheim (1948) "Studies in the logic of explanation".

Philosophy of Science, XV reprinted in B. Brody (ed., 1970)

Herman, R., S. Grillner, P. Stein and D.G. Stuart (eds, 1976) Neural Control of
Locomotion, vol. 18, NY: Plenum Press

Hertzberger, H. (1971) "Setting Russell free", Toronto, University of Toronto (ms)

Hetzron, R. (1971) "Presentative function and presentative movement". Studies
in African Linguistics, Supplement #2

Hinton, G. (1986) "Learning in massive parallel networks", paper read at the Of¬
fice of Naval Research Conference, Eugene, Oregon, (November 1986; ms)
Hinton, G. and S.J. Nowlan (1987) "How learning can guide evolution". Com¬
puter Science Dept., Carnegie Mellon University (ms)
 Bibliography 437

Hinttika, J. (1967) "Individuals, possible worlds and epistemic logic". Nous, 1

Hintzman, D.L. (1970) "Effects of repetition and exposure duration on memory",
J. of Experimental Psychology, 83

Hintzman, D. (1983) "Schema abstraction in multiple trace memory model".

University of Oregon, Eugene (ms)

Hintzman, D. and D.L. Ludlam (1980) "Differential forgetting of prototypes and

old instances". Memory and Cognition 8
Hocart, A.M. (1937) "Kinship systems", in P. Bohanan and J. Middleton (eds)
Kinship and Social Organization, NY: Natural History Press
Hollis, M. and S. Lukes (eds, 1982) Rationality and Relativism, Cambridge: MIT
Press
Hopper, P. and S. Thompson (1980) "Transitivity in grammar and discourse".
Language, 56.3
Hopper, P. and S. Thompson (eds, 1982) Studies in Transitivity, Syntax and
Semantics, vol. 15, NY: Academic Press
Hopper, P. and S. Thompson (1984) "The communicative basis for lexical
categories". Language, 60
Horn, L. (1972) On the Semantic Properties of Logical Operators in English,
PhD dissertation, U.C.L.A. (ms)
Horton, R. (1982) "Tradition and modernity revisited", in M. Hollis and S. Lukes
(eds, 1982)
Huang, F. (1985) "The numeral 'one' as an indefinite marker in Mandarin
Chinese”, University of Oregon, Eugene (ms)
Hume, D. (1739/1740) A Treatise on Human Nature, Oxford: Clarendon Press (1978)
Hyman, L. (1973) "How do natural rules become unnatural?" paper read at the
LSA Winter Conference, San Diego, Dec. 1973 (ms)
Hyman, R. and Frost, N.A.H. (1975) "Gradients and Schema in pattern recogni¬
tion", in P.M.A. Rabbitt and S. Dornic (eds) Attention and Performance, V, New
York: Academic Press
Ibbotson, N.R. and J. Morton (1981) "Rhythm and dominance". Cognition, 9
Jackendoff, R. (1971) "Modal structure in semantic representation". Linguistic
Inquiiy, 2.4
Jelinek, E. (1986) "Yaqui minus control", in S. DeLancey and R. Tomlin (eds)
Pacific Linguistics Conference, II, Eugene, Oregon: University of Oregon
Jespersen, O. (1924) The Philosophy of Grammar, NY: Norton [1965 edition]
Johnson, M. (ed., 1981) Philosophical perspectives on metaphor, Minneapolis:
University of Minnesota Press
Johnson, P. (1983) Modern Times: The World from the Twenties to the Eighties,
NY: Harper and Row
Johnson-Laird, P. (1983) Mental Models, Cambridge: Harvard University Press
438 Mind, Code and Context

Jorgensen, J. (1972) The Sundance Religion: Power for the Powerless, Chicago:
University of Chicago Press
Karttunen, L. (1974) "Presupposition and linguistic context". Theoretical Lin¬
guistics, 1.2
Katz, J.J. (1972) Semantic Theory, NY: Harper and Row
Katz, J.J. (1976) "A hypothesis about the uniqueness of natural language", in S.R.
Harnad, H.D. Steklis and J. Lancaster (eds) Origin and Evolution of Language
and Speech, Annals of the N.Y. Academy of Science, 280
Katz, J.J. (1978) "Effability and translation", in F. Guenthner and M. Guentner-
Reutter (eds) Meaning and Translation, London: Duckworth
Katz, J.J. and J. Fodor (1964) "The structure of a semantic theory", in J. Fodor
and J.J. Katz (eds) The Structure of Language, Englewood Cliffs, NJ: Prentice-
Hall
Katz, J.J. and P. Postal (1964) An Integrated Theory of Linguistic Descriptions,
Cambridge: MIT Press
Kay, P. and C. K. McDaniel (1978) "The linguistic significance of the meaning of
basic color terms". Language, 54.4
Keele, S.W. (1969) "The repetition effect: A memory dependent process", J. of Ex¬
perimental Psychology, 80
Keele, S.W. (1972) "Attention demands on memory retrieval", J. of Experimental
Psychology, 81
Keele, S.W. (1986) "Sequencing and timing in skilled action: An overview". Tech¬
nical Report 86-1, Cognitive Science Program, University of Oregon, Eugene
Keele, S.W. and R. Ivry (1986) "Components of the motor program: The cerebel¬
lum as an internal clock". Technical Report 86-7, Cognitive Science Program,
University of Oregon, Eugene
Keele, S.W., R. Ivry and R.A. Pokorny (1986) "Force control and its relation to
timing". Technical Report 86-4, Cognitive Science Program, University of
Oregon, Eugene
Keele, S.W., R.A. Pokorny, D.M. Corcos and R. Ivry (1985) "Do perception and
production share common time mechanisms?", Acta Psychologica, 60
Keenan, E.L. (1969) A Logical Base for a Transformational Grammar of
English, PhD dissertation. University of Pennsylvania (ms)
Keenan, E.L. (1971) "Two types of presupposition in natural language", in C.
Fillmore and T. Langendoen (eds) Studies in Linguistic Semantics, NY: Holt,
Rinehart and Winston
Keenan, E.L. (1975) "Some universals of passive in Relational Grammar", CLS 11,
University of Chicago, Chicago Linguistics Society
Keenan, E.L. (1976) "Remarkable subjects in Malagsy", in C. Li (ed., 1976)
Keenan, E.L. (1978) "Some logical problems in translation", in F. Guenthner and
M. Guenthner-Reutter (eds) Meaning and Translation, London: Duckworth
 Bibliography 439

Keenan, E.O. and B. Schieffelin (1976) "Topic as a discourse notion: A study of

topic in the conversation of children and adults", in C. Li (ed., 1976)
Kemp, J. The Philosophy of Kant, Oxford: Oxford University Press
Kim, A. H. (1986) "Semi-clausal modals in Korean verb morphology". University
of Oregon, Eugene, Linguistics Colloquium, November, 1986 (ms)
Kim, A.H. (1987) "Korean honorification system and relevance theory", in S. De-
Lancey and R. Tomlin (eds) Pacific Linguistics Conference II, Eugene, Oregon:
University of Oregon
Kimura, D. (1979) "Neuromotor mechanisms in the evolution of human com¬
munication", in D.H. Stelkis and M.J. Raleigh (eds) The Neurology of Social
Communication in Primates, NY: Academic Press
Kintsch, W, (1974) The Representation of Meaning in Memory, Hillsdale, NJ:
Erlbaum
Kintsch, W. (1977) Memory and Cognition, NY: Wiley
Kintsch, W. and E. Green (1978) "The role of culture-specific schemata in the
comprehension and recall of stories". Discourse Processes, 1:1-13
Kintsch, W. and J.M. Keenan (1973) "Reading rate and retention as a function of
the number of the propositions in the base structure of sentences". Cognitive
Psychology, 5: 257-274
Kintsch, W. and T. van Dijk (1978) "Toward a model of text comprehension and
production". Psychological Review, 85:363-394
Kiparski, P. and C. Kiparski (1968) "Fact", in M. Bierwisch and K.E. Heidolph
(eds) Progress in Linguistics, The Hague: Mouton
Klatt, D.H. (1979), "Speech perception: A model of acoustic-phonetic analysis
and lexical access", J. of Phonetics, 7: 279-312
Koestler, A. (1967) The Ghost in the Machine, NY: Macmillan
Koestler, A. (1972) The Case of the Midwife Toad, NY: Random House
Krashen, S.(1972) "Lateralization, language learning and the critical period".
Language Learning, 23.1
Krashen, S., S. Long and R. Scarcella (1979) "Age, rate and eventual attainment in
second language acquisition", TESOL Quarterly, 13.4
Kripke, S. (1963) "Semantic considerations on modal logic", Acta Philosophica
Fennica, 16
Kripke, S. (1972) "Naming and necessity", in D. Davidson and G. Harman (eds)
Semantics of Natural Language, Dordrecht: Reidel
Kroeber, A.L. (1909) "Classificatory systems of relationships", reprinted in P.
Bohannon and J. Middleton (eds, 1968) Kinship and Social Organization, NY:
Natural History Press
Kroon, F. (1981) "Kant and Kripke on the identifiability of modal and epistemic
notions". Southern Journal of Philosophy, 19.1
Kuhn, T. (1962) The Structure of Scientific Revolutions, Chicago: University of
Chicago Press
440 Mind, Code and Context

Labov, W. (1972a) Sociolinguistic Patterns, Philadelphia: University of Pennsyl¬

vania Press
Labov, W. (1972b) "Rules for ritual insults", in D. Sudnow (ed.) Studies in Social
Interaction, NY: The Free Press
Labov, W. (1972c) "Where do grammars stop?", in R. Shuy (ed.) Georgetown
Monograph on Language and Linguistics, Washington, DC: Georgetown
University Press
Labov, W. (1972d) "For an end to the uncontrolled use of linguistic intuitions".
Paper read at the LSA meeting, Atlanta (ms)
Labov, W. (1975a) "The quantitative study of linguistic structure", in K.-H.
Dahlstedt (ed.) Proceedings of the Second International Conference on Nordic
and General Linguistics, Umea
Labov, W. (1975b) "Empirical foundations of linguistic theory", in R. Austerlitz
(ed.) The Scope of American Linguistics, Lisse: Peter de Ridder Press
Labov, W. and J. Waletzky (1967) "Narrative analysis: Oral versions of personal
experience", in J. Helm (ed.) Essays on the Verbal and Visual Arts, Seattle: U. of
Washington Press
Laertius, Diogenes, Lives of Eminent Philosophers, vols 1,2, tr. by R.D. Hicks
(1925), Cambridge: Harvard University Press
Lakoff, G. (1973) "Hedges: A study in the meaning criteria and the logic of fuzzy
concepts", J. of Philosophical Logic, 2
Lakoff, G. (1977) "Linguistic gestalt", CLS 13, University of Chicago: Chicago
Linguistics Society

Lakoff, G. (1982) "Categories and cognitive models", Berkeley Cognitive

Science Report, no. 2, Berkeley: Institute for Human Learning, University of
California

Lakoff, G. and M. Johnson (1980) Metaphors We Live By, Chicago: University of

Chicago Press

Lamarck, J.-B. (1809) Philosophic zoologique, ou exposition des considerations

relatives a l'histoire naturelle des animaux, Paris; English tr. Hugh Elliot (1914),
The Zoological Philosophy, London: Macmillan

Lamendella, J. (1975) The Early Growth of Cognition and Language: A Neuro¬

psychological Approach, Cal. State University, San Jose (ms)
Lamendella, J. (1976) "Relations between the ontogeny and phylogeny of lan¬
guage: A neo-recapitulationist approach", in S.R. Harnad, H.D. Stelkis and J.
Lancaster (eds) The Origins and Evolution of Language and Speech, NY: New
York Academy of Science

Langacker, R. (1987) Cognitive Grammar, vol. I, Stanford: Stanford University

Press

Leach, E.R. (1958) "Concerning Trobriand clans and the kinship category 'tabu'",
in J. Goody (ed.) The Development Cycle in Domestic Groups, Cambridge
,
Papers in Social Anthropology, 1 Cambridge: Cambridge University Press
Lees, R.B. (1957) "Review of Chomsky's Syntactic Structures", Language, 33
 Bibliography 441

Levinson, S. (1983) Pragmatics, Cambridge: Cambridge U. Press

Levy, D. (1979) "Communicative goals and strategies: Between discourse and
syntax”, in T. Givon (ed., 1979c)

Levy, W.B. and O. Steward (1983) "Temporal continuity requirements for long¬
term associative potentiation/depression in the hippocampus". Neuroscience, 8
Lewis, D. (1972) "General semantics", in D. Davidson and G. Harman (eds)
Semantics for Natural Language, Dordrecht: Reidel

Lewis, D. (1979) "Score keeping in language games", J. of Philosophical Logic, 8

Li, C.N. (ed., 1976) Subject and Topic, NY: Academic Press
Li, C.N. and S. Thompson (1976) "Subject and topic: A new typology for lan¬
guage", in C.Li (ed., 1976)

Lieberman, P. (1984) The Biology and Evolution of Language, Cambridge: Har¬

vard University Press

Lindlahr, V. (1940) You Are What You Eat, NY: Lancer Books [1972 ppbk edition]
Linnaeus, C. (1753) Species Plantarum, Halmiae
Linnaeus, C. (1758) Systema Naturae, Stockholm
Lloyd Morgan, C. (1896) "On modification and variation". Science, 4
Loftus, E.F. (1980) Eyewitness Testimony, Cambridge: Harvard University Press
Longacre, R. (1964) Grammar Discovery Procedures: A Field Manual, The
Hague: Mouton
Lounsbury, F.G. (1965) "Another view of the Trobriand kinship categories", in For¬
mal Semantic Analysis, special issue of American Anthropologist, 67.5, part 2
Luce, P.A. and D.B. Pisoni (1984) "Speech perception: Recent trends in research,
theory and applications", in H. Winitz (ed.). Human Communication and its
Disorders, Norwood, NJ: Ablex
Lyons, J. (1977) Semantics, vol. 1,2, Cambridge: Cambridge University Press
Malinowski, B. (1926) "Anthropology", Encyclopaedia Britannica, London and
NY: The Encyclopaedia Britannica, Inc.
Malinowski, B. (1932) The Sexual Life of Savages, 3rd edition, London
Malinowski, B. (1935) Coral Gardens and Their Magic, London
Malinowski, B. (1954) Magic, Science and Religion, and Other Essays, Garden
City, NY: Doubleday Anchor Books
Mandler, J.M. (1978) "A node in the code: The use of a story schema in retrieval".
Discourse Processes, 1:14-35
Mandler, J.M. and N. Johnson (1977) "Remembrance of things parsed: Story
Structure and recall". Cognitive Psychology, 9:111-151
March, J. and H.A. Simon (1958) Organizations, NY: Wiley
Marslen-Wilson, W., L.K. Tyler and M. Seidenberg (1978) "Sentence processing
and the clause boundary", in W.J.M. Levelt and G.B. Flores (eds) Studies in the
Perception of Language, London: Wiley
442 Mind, Code and Context

Martin, J.G. (1972) "Rhythmic (hierarchic) versus serial structure in speech and
other behavior". Psychological Review, 79.6
Mayr, E. (1959a) "Typological vs. population thinking", in Mayr (1976)
Mayr, E. (1959b) "Isolation as an evolutionary factor", in Mayr (1976)
Mayr, E. (1959c) "Agassiz, Darwin and evolution", in Mayr (1976)
Mayr, E. (1960) "The emergence of evolutionary novelties", in Mayr (1976)
Mayr, E. (1962) "Accident or design: The paradox of evolution", in Mayr (1976)
Mayr, E. (1970) Populations, Species and Evolution, Cambridge: Harvard
University Press
Mayr, E. (1972) "Lamarck revisited", in Mayr (1976)
Mayr, E. (1974) "Behavior programs and evolutionary strategies", in Mayr (1976)
Mayr, E. (1976) Evolution and the Diversity of Life, Cambridge: Harvard
University Press
Mayr, E. (1982) The Growth of Biological Thought, Cambridge: Harvard
University Press
McKeon, R. (ed., 1941) The Basic Works of Aristotle, NY: Random House [22nd
printing, 1970]
McKoon, G. and R. Ratcliff (1980) "The comprehension processes and memory
structures involved in anaphoric reference", J.V.L.V.B., 19:668-682
McLeod, P., C. McLaughlin and I. Nimmo-Smith (1985) "Information encapsula¬
tion and automaticity: Evidence from visual control of finely timed actions", in
M. Posner and O. Marin (eds, 1985)
Medin, D.L. and M.M. Schaffer (1978) "Context theory of classification and learn¬
ing", Psychological Review, 85.3
Merton, T. (1965) The Way of Chuang-Tse, NY: New Directions
Midgeley, M. (1978) Beast and Man: Roots of Human Nature, Ithaca: Cornell
University Press
Miller, G.A. and P. Johnson-Laird (1976) Language and Perception, Cambridge:
Cambridge University Press

Mishkin, M., B. Malamut and J. Bachevalier (1984) "Memories and habits: Two
neural systems", in G. Lynch, J.L. McGaugh and N.M. Weinberger (eds) The
Neurobiology of Learning and Memory, NY/London: Guilford Press
Mithun, M. (1985) "Is basic word order universal?", in R. Tomlin (ed., 1987)
Montague, R. (1970) "Pragmatics and intensional logic", Synthese, 22; reprinted
in D. Davidson and G. Harman (eds, 1972) Semantics of Natural Language,
Dordrecht: Reidel

Morris, C. (1938) Foundations of the Theory of Signs, Chicago: University of

Chicago Press

Morris, W. (ed., 1969) The American Heritage Dictionary, Boston: Houton and
Mifflin
 Bibliography 443

Neely, ].H. (1976) "Semantic priming and retrieval from lexical memory:
Evidence for facilitatory and inhibitory processes". Memory and Cognition, 4:
648-654

Neely, J.H. (1977) "Semantic priming and retrieval from lexical memory: Roles of
inhibitionless spreading activation and limited-capacity attention", J. of Ex¬
perimental Psychology: General, 106: 226-254
Nissen, M.J. (1976) Semantic Activation and Levels of Processing, University of
Oregon, Eugene: Unpublished PhD Dissertation (ms)
Nissen, M.J. and P. Bulleme (1986) "Attention requirements of learning: Evidence
from performance measures". University of Minnesota, Minneapolis (ms)
Nola, R. (1986) "The status of Popper's theory of scientific method", Auckland
University, New Zealand (ms)
Nooteboom, S.G. (1981) "Speech rate and segmental perception: The role of
words in phoneme identification", in T. Myers, J. Laver and J. Anderson (eds).
The Cognitive Representation of Speech, Amsterdam: North Holland
Ochs, E. (1979a) "Social foundations of language", in R. Freedle (ed.) New Direc¬
tions in Discourse Processing, vol. 3, Norwood, NJ: Ablex
Ochs, E. (1979b) "Introduction: What a child can contribute to pragmatics", in E.
Ochs and B. Schieffelin (eds, 1979)
Ochs, E. (1979c) "Planned vs. unplanned conversation", in T. Givon (ed., 1979c)
Ochs, E. and B. Schieffelin (eds, 1979) Developmental Pragmatics, NY: Academic
Press
Oh, C.-K. and D. Dinnen (eds, 1979) Presupposition, Syntax and Semantics, 11,
NY: Academic Press
Ortonyi, A. (ed., 1979) Metaphor and Thought, Cambridge: Cambridge Univer¬
sity Press
Osgood, C. (1971) "Where do sentences come from?", in D. Steinberg and L.
Jacobovitz (eds) Semantics: An Interdisciplinary Reader in Philosophy, Lin¬
guistics and Psychology, London: Cambridge University Press
Osgood, C. and T. Sebeok (eds, 1954) Psycholinguistics: A Survey of Theory
and Research Problems, I.J.A.L. Memoir #10
Osgood, C., G. Suci and P. Tannenbaum (1957) The Measurement of Meaning,
Urbana: U. of Illinois Press
Owen, R. (1841) John Hunter's Observations on Animal Development, Edited,
and his Illustrations of that Process in the Bird Described, London: R. and J.E.
Taylor
Paillard, J. (1960) "The patterning of skilled movement", in J. Field, H.M.
Magoun and V.E. Hall (eds) Handbook of Physiology, vol. 3, section 1,
Neurophysiology, Washington, DC: American Physiological Society
Paivio, A. (1971) Imagery and Verbal Processes, NY: Holt, Rinehart and Winston
Patterson, K.E. and A.J. Marcel (1977) "Aphasia, dislexia, and the phonological
coding of written words". Quarterly J. of Experimental Psychology, 29:307-318
444 Mind, Code and Context

Pawley, A. (1966) The Structure of Karam: A Grammar of a New Guinea High¬

lands Language, ph.D. dissertation, Auckland University, Auckland, NZ (ms)
Pawley, A. (1976) "On meeting a language that defies description by ordinary
means", Auckland University, Auckland, NZ (ms)
Pawley, A. (1987) "Encoding events in Kalam and English: Different Logics for
reporting experience", in R. Tomlin (ed., 1987)
Payne, D. (1983) "Basic constituent order in Yagua clauses: Implications for
word-order universals", in D. Derbyshire and G. Pullum (eds) Handbook of
Amazonian Languages, vol. 1, Berlin: Mouton (ms)
Payne, D. (1985) "Information structuring in Papago", Language
Pearson, T.R. (1985) A Short History of a Small Place, NY: Ballantine
Peirce, C.S. (1931) Collected Writings, vol. 1, Cambridge: Harvard University
Press
Peirce, C.S. (1934) Collected Writings, vol. 5, Cambridge: Harvard University
Press
Peirce, C.S. (1940) The Philosophy of Peirce, J. Buchler (ed.) NY: Harcourt, Brace
Perfetti, C.A. and S.R. Goldman (1974) "Thematization and sentence retrieval",
J.V.L.V.B., 13: 70-79

Piattelli-Palmarini, M. (ed.,1980) Language and Learning: The Debate Between

Jean Piaget and Noam Chomsky, Cambridge: Harvard University Press
Pirozzolo, F.H. and K. Rayner (1977) "Hemispheric specialization in reading and
word recognition". Brain and Language, 4: 248-261
Poeck, K. and W. Huber (1977) "To what extent is language a sequential ac¬
tivity?", Neuropsychologia, 15

Pollatsek, A. and T.H. Carr (1979) "Rule-governed and wholistic encoding

processes in word perceptions", in P.A. Kolers, M.E. Wrolstad and H. Bonma
(eds) Processing Visible Language, NY: Plenum
Popper, K. (1950) The Open Society and its Enemies, vol. 1: In the Spell of
Plato, London: Rutledge and Kegan Paul

Popper, K. (1959) The Logic of Scientific Discovery, revised edition (1968), NY:
Harper and Row

Popper, K. (1974) The Philosophy of Karl Popper, ed. by P.A. Schlipp, La Salle,
Ill.: Open Court

Popper, K. and J. Eccles (1977) The Self and Its Brain, London: Rutledge &
Kegan Paul

Posner, M.I. (1969) "Abstraction and the process of recognition", in G.H. Bowers
and J.T. Spence (eds) The Psychology of Learning and Motivation, vol. 3, NY:
Academic Press

Posner, M.I. (1985) Hierarchically distributed networks in neuropsychology of

selective attention". Cognitive Science Working Papers, TR-85-I, University of
Oregon, Eugene
 Bibliography 445

Posner, M. (1986) "Empirical studies of prototypes", in C. Craig (ed.). Categoriza¬

tion and Noun Classification, TSL vol. 7, Amsterdam: J. Benjamins
Posner, M.I. and S. W. Boies (1971) "Components of attention". Psychological
Review, 78

Posner, M. and S. Keele (1968) "On the genesis of abstract ideas", J. of Ex¬
perimental Psychology, 77

Posner, M. and S. Keele (1970) "Retention of abstract ideas", J. of Experimental

Psychology, 83
Posner, M.I. and R. Klein (1971) "On the function of consciousness", paper read
at the Fourth Conference on Attention and Performance, University of
Colorado, Boulder (ms)
Posner, M.I. and O.S.M. Marin (eds, 1985) Attention and Performance, XI,
Hillsdale, NJ: Erlbaum
Posner, M.I. and C.R.R. Snyder (1974) "Attention and cognitive control", in R.L.
Solso (ed.) Information Processing and Cognition: The Loyola Symposium,
Hillsdale, NJ: Erlbaum
Posner, M. and R.E. Warren (1972) "Traces, concepts and conscious construc¬
tions", in A.W. Melton and E. Martin (Eds) Coding Processes in Human
Memory, Washington, DC: V.H. Winston & Sons
Potter, M.C., V.V. Valian and B.A. Faulconer (1977) "Representation of a sentence
and its pragmatic implications: Verbal, imagistic, or abstract?", J.V.L.V.B., 16
Powell, A. (1971) Temporary Kings, NY: Popular Library (1986 paperback
reprint)
Price, H.H. (1953) Thinking and experience NY: Hutchinson's University Press
Prince, E. (1979) "On the given/new distinction", CLS 15, Chicago: University of
Chicago, Chicago Linguistics Society
Prince, E. (1981) "On the inferencing of indefinite-this NPs", in A. Joshi, B. Web¬
ber and I. Sag (eds) Elements of Discourse Understanding, Cambridge:
Cambridge University Press
Purtill, R.L. (1968) "About identity through possible worlds". Nous, 2
Pylyshyn, Z.W. (1973) "What the mind's eye tells the mind's brain: A critique of
mental imagery". Psychological Bulletin, 80
Quillian, M.R. (1968) "Semantic memory", in M. Minsky (ed.) Semantic Informa¬
tion Processing, Cambridge: MIT Press
Quine, W. van O. (1953) "Reference and modality", Ch. 8 of his From a Logical
Point of View, Cambridge: Harvard University Press
Rafferty, E. (1987) "Word order in intransitive clauses in High and Low Malay of
the Late 19th Century", in R. Tomlin (ed., 1987)
Ramsay, V. (1987) "Preposed and postposed 'if' and 'when' clauses in English
discourse", in R. Tomlin (ed., 1987)
Rapoport, A, (ed., 1968) Clausewitz's On War, NY: Penguin Classics
446 Mind, Code and Context

Ratcliff, R. and G. McCoon (1980) "Automatic and strategic priming in recogni¬

tion", J.V.L.V.B., 20:204-215
Reyburn, W.D. (1958) "Kaka kinship, sex and adultery". Practical Anthropology,
5.1., reprinted in W.A. Smalley (ed.) Readings in Missionary Anthropology, So.
Pasadena, California: William Carey Library
Robinson, G.M. (1977) "Rhythmic organization in speech processing", J. of Ex¬
perimental Psychology: Human Perception and Performance, 3.1
Robinson, G.M. and D.J. Solomon (1974) "Rhythm is processed by the speech
hemisphere", J. of Experimental Psychology, 102.3
Rosch, E. (1973a) "On the internal structure of perceptual and semantic
categories", in T. Moore (ed.) Cognitive Development and the Acquisition of
Language, NY: Academic Press
Rosch, E. (1973b) "Natural categories". Cognitive Psychology, 4
Rosch, E. (1975) "Human categorization", in N. Warren (ed.) Advances in Cross-
Cultural Psychology, London: Academic Press [1977]
Rosch, E. and B.B. Lloyd (eds, 1978), Cognition and Categorization, Hillsdale, NJ:
Erlbaum

Rosch, E. and C.B. Mervis (1975) "Family resemblances: Studies in the internal
structure of categories". Cognitive Psychology, 7: 573-605
Rosch, E., C.B. Mervis and W. Gray (1976) "Basic objects in natural categories".
Cognitive Psychology, 8: 382-439
Ross, J.R. (1972a) "The category squish: Endstation Hauptwort", CLS 8, Chicago:
University of Chicago, Chicago Linguistics Society

Ross, J.R. (1972b) "Act", in G. Harman and D. Davidson (eds) Semantics of

Natural Language, Dordrecht: Reidel

Ross, J.R. (1973) "Nouniness", in D. Fujimura (ed.) Three Dimensions of Lin¬

guistics, Tokyo: TEC Corporation

Ross, J.R. (1974) "Clausematiness", in E.L. Keenan (ed.) Semantics for Natural
Languages, Cambridge: Cambridge University Press
Rude, N. (1983) "Transitivity, topicality and direct object in Nez Perce discourse"
I.J.A.L., 52.2 (1987)

Rude, N. (1985) Studies in Nez Perce Grammar and Discourse, University of

Oregon, Eugene, PhD dissertation (ms)

Rumelhart, D. (1977) "Toward an interactive model of reading", in S. Dorni£ (ed.)

Attention and Performance VI, Hillsdale, NJ: Erlbaum

Rumelhart, D. and A. Ortony (1977) "The representation of knowledge in

memory", in R. Anderson, R. Spiro and W. Montague (eds). Schooling and the
acquisition of knowledge, Hillsdale, NJ: Erlbaum
Russell, B. (1905) "On defining". Mind, 14

Russell, B. (1919) Introduction to Mathematical Philosophy, London: Allen and

Unwin
 Bibliography 447

Russell, B. (1948) Human knowledge, its scope and limits, NY: Simon and
Schuster

Russell, E.S. (1916) Form and Function: A Contribution to the History of

Animal Morphology, London: J. Murray

Rybarkiewicz, W. (1984) "Word-order flexibility in Polish", University of Oregon,

Eugene (ms)

Sacks, H., E. Schegloff and G. Jefferson (1974) "A simplest systematic for the or¬
ganization of turn-taking in conversation". Language, 50.4
Sadalla, E. and W.J. Burroughs (1981) Food Preferences and Social Identity, U.S.
Science and Education Administration, Technical Report, Phoenix: Arizona State
University
Sadock, J. (1970), "Whimperatives", in J. Sadock and A. Vanek (eds) Studies
Presented to R.B. Lees, Edmonton: Linguistic Research
Sadock, J. and A. Zwicky (1985) "Speech act distinctions in syntax", in T. Shopen
(ed.) Language Typology and Syntactic Description, vol. 1, Cambridge: Cam¬
bridge University Press
Saffran, E. and O.S.M. Marin (1977) "Reading without phonology: Evidence from
aphasia". Quarterly J. of Experimental Psychology, 29:515-525
Salisbury, M. (1986) "Negation and irrealis in Pukapukan", Auckland University,
Auckland, N.Z. (ms)
Salkever, S. (1983) "Beyond interpretation: Human agency and the slovenly
wilderness", in N. Haan, R.M. Bellah, P. Rabinow and W.M. Sullivan (eds) Social
Science as Moral Inquiry, NY: Columbia University Press
Sapir, E. (1949) Selected Writings of Edward Sapir, ed. by D.G. Mandelbaum,
Berkeley: U.C. Press
Saussure, F. de (1915) Course in General Linguistics, edited by C. Bally and A.
Sechehaye, tr. by A. Reidlinger, NY: Philosophical Library (1959)
Schank, R. and R. Abelson (1977) Scripts, Goals, Plans and Understanding,
Hillsdale, NJ: Erlbaum
Scheffler, H.W. (1972) "Kinship semantics", in B. J. Siegel (ed.) Annual Review of
Anthropology
Scheffler, H.W. and F.G. Lounsbury (1971) A Study in Structural Semantics: The
Siriono Kinship System, Englewood Cliffs, NJ: Prentice-Hall
Schieffelin, E. (1976) The Sorrow of the Lonely and the Burning of the Dancer,
NY: St. Martin's Press
Schneider, W. (1985) "Toward a model of attention and the development of auto¬
matic processing", in M.I. Posner and O. Marin (eds, 1985)
Schneider, W. and A.S. Fisk (1983) "Attention theory and the mechanisms for
skilled performance", in R.A. McGill (ed.) Memory and Control of Action, NY.
North Holland
Schneider, W. and R.M. Shiffrin (1977) "Controlled and automatic human infor¬
mation processing, I: Detection, search and attention , Psychological Review, 84
448 Mind, Code and Context

Schmidt, R.A. (1975) "A schema theory of discrete motor skill learning".
Psychological Review, 82
Schmidt, R.A. (1980) "Past and future issues in motor programming". Research
Quarterly for Exercise Sports, 51.1
Scollon, R. (1974) Conversations with a One-Year Old Child, Honolulu: Univer¬
sity of Hawaii Press
Scott, D. (1970) "Advice on modal logic"; in K. Lambert (ed.) Philosophical
Problems in Logic, Dordrecht: Reidel
Scriven, M. (1962) "Explanations, predictions and laws", in B. Brody (ed., 1970)
Searle, J. (1970) Speech Acts, Cambridge: Cambridge University Press
Searle, J. (1975) "Indirect speech acts", in P. Cole and J. Morgan (eds, 1975)
Searleman, A. (1977) "A review of right hemisphere linguistic capabilities".
Psychological Bulletin, 84: 530-528
Serres, E. (1860) Principles d'embryologie, de zoogenie, et de teratogenie,
Mem. Acad. Sci., 25:1-943
Shallice, T. and E.K. Warrington (1975) "Word recognition in a phonemic dys¬
lexic patient". Quarterly J. of Experimental Psychology, 27:187-199
Shapiro, D.C.A. (1977) "A preliminary attempt to determine the duration of a
motor program", in D.M. Landers and R.W. Christina (eds) Psychology of Motor
Behavior and Sports, vol. 3, Champaign, Ill.: Human Kinetics
Shapiro, D.C.A. (1978) The Learning of Generalized Motor Programs, PhD dis¬
sertation, University of Southern California, Los Angeles (ms)
Shapiro, D.C.A. and R.A. Schmidt (1980) "The schema theory: Recent evidence
and developmental implications", in J.A.S. Kelso and J.E. Clark (eds) The
Development of Movement Control and Coordination, NY: Wiley
Shapiro, D.C.A., R.F. Zernicke, R.J. Gregor and J.D. Diestel (1980) "Evidence for
generalized motor programs using gait pattern analysis", J. of Motor Behavior
Shapiro, M. (1983) The Sense of Grammar, Bloomington: Indiana University
Press
Shir, N. (1979) "Discourse constraints on dative movement", in T. Givon (ed.,
1979c)
Short, T.L. (1981) "Semeiosis and intentionality". Transactions of the Charles S.
Peirce Society, 17
Shroyer, S. (1985) "The demonstrative 'this' as an indefinite article in spoken
English", in S. DeLancey and R. Tomlin (eds) Proceedings of the First Pacific
Linguistics Conference, Eugene: University of Oregon
Silva-Corvalan, C. (1977) A Discourse Analysis of Word Order in the Spanish
Spoken by Mexican-Americans in West Los Angeles, MA Thesis, UCLA (ms)
Simonds, P. (1974) The Social Primates, NY: Harper & Row
Skinner, B.F. (1938) The Behavior of Organisms, NY: Applton-Century-Croft
Skinner, B.F. (1957) Verbal Behavior, NY: Applton-Century-Croft
 Bibliography 449

Slobin, D. (1973) "Cognitive prerequisites for the development of grammar", in

C. Ferguson and D. Slobin (eds) Studies in Child Language Development, NY:
Holt, Rinehart and Winston

Slobin, D. (1977) "Language change in childhood and history", in J. MacNamara

(ed.) Language Learning and Thought, NY: Academic Press
Slobin, D. and A. Aksu (1982) "Tense, aspect and modality in the use of Turkish
evidentials", in P. Hopper (ed.) Tense and Aspect: Between Semantics and Prag¬
matics, TSL vol.l, Amsterdam: J. Benjamins
Smalley, W.A. (1958) "Vocabulary and the preaching of the gospel". Practical
Anthropology, 6.4., reprinted in W.A. Smalley (ed.) Readings in Missionary
Anthropology, So. Pasadena: William Carey Library
Smith, E.E. and D. Medin (1981) Categories and Concepts, Cambridge: Harvard
U. Press
Smith, J.L. (1980) "Programming of stereotyped limb movements by spinal gen¬
erators", in G.E. Stelmach and J. Requin (eds) Tutorials in Motor Behavior, NY:
North Holland
Smith, N.K. (tr., 1929) Emanuel Kant's Critique of Pure Reason, London/NY:
Macmillan [1973 edition]
Sperber, D. (1982) "Apparently irrational beliefs", in M. Hollis and S. Lukes (eds,
1982)
Sperber, D. and D. Wilson (1988) Relevance: Communication and Cognition,
Cambridge: Harvard University press
Spiro, M. (1978) "Culture and human nature", in G. Spindler (ed.) The Making
of Psychological Anthropology, Berkeley: U.C. Press
Squire, L.R. and N.J. Cohen (1984) "Human memory and amnesia", in G. Lynch,
J.L. McGaugh and N.M. Weinberger (eds) Neurobiology of Learning and
Memory, NY/London: Guilford Press
Sridhar, S.N. (1980) Cognitive Determination of Linguistic Structure: A Cross-
Linguistic Experimental Study in Sentence Production, PhD dissertation.
University of Illinois, Urbana (ms)
Stein, P.S.G. (1978) "Motor systems, with specific reference to the control of
locomotion". Annual Review of Neuroscience, 1

Strawson, P. (1950) "On referring". Mind, 59

Sun, C.-F. and T. Givon (1985) "On the so-called SOV word-order in Mandarin
Chinese: A quantified text study and its implications". Language, 61.2
Sundberg, K. (1985) "Word-order in Klamath: The effect of topic continuity in
determining pre- and post-verbal position". University of Oregon, Eugene (ms)
Syder, F. and A. Pawley (1974) "The reduction principle in conversation", Auck¬
land University, Auckland, NZ (ms)
Szentagothai, J. (1977) "The neuron network of the cerebral cortex: A functional
interpretation". Proceedings of the Royal Society of London, B, 201
Tesniere, L. (1959) Elements de Syntaxe Structurale, Paris: Klinksack
450 Mind, Code and Context

Thelen, E. (1984) "Learning to walk: Ecological demands and phylogenetic con¬

straints", in P. Lipsitt and C. Rovee-Collier (eds) Advances in Infancy Research,
vol. 3, Norwood, NJ: Ablex
Thomas, H. (1983) "The inferno of relativism". Times Literary Supplement, July
8,1983
Thompson, S. (1973) "Resultative verb compounds in Mandarin Chinese: A case
for lexical rules". Language, 49.2
Thompson, S. (1985) "Grammar and written discourse: Initial vs. final purpose
clauses in English", in T. Givon (ed., 1985)
Thompson, S. (1987) "That deletion in English from a discourse perspective",
BLS, Berkeley: University of California (ms)
Thurman, R. (1978) Interclausal Relations in Chuave, MA Thesis, University of
California, Los Angeles, (ms)
Tomlin, R. (1987) "Linguistic reflections of cognitive episodes", in R. Tomlin (ed.,
1987)
Tomlin, R. (ed., 1987) Coherence and Grounding in Discourse, TSL vol. 11,
Amsterdam: J. Benjamins
Traugott, E. (1974) "Spatial expressions of tense and temporal sequencing: A con¬
tribution to the study of semantic fields", Stanford: Stanford University (ms)
Tsuchihashi, M. (1983) "The speech-act continuum: An investigation of Japanese
sentence-final particles", J. of Pragmatics, 7.4
Tversky, B. (1969) "Pictorial and verbal encoding in a short-term memory task".
Perception and Psychophysics, 6
Tversky, B. (1975) "Pictorial encoding of sentences in sentence-picture com¬
parison", Quarterly J. of Experimental Psychology, 27
Tweedale, M. (1986) "How to handle problems about forms and universals in
Aristotle's work", Auckland University, Auckland, NZ (ms)
van der Muelen, E.C. (1977) "A survey of multi-way channels in information
theory: 1961-1976", IEEE Transactions in Information Theory, IT-23
van Lawick-Goodall, J. and H. van Lawick (1971) Innocent Killers, Boston:
Houghton and Mifflin

Verhaar, J. (1985) "On iconicity and hierarchy". Studies in Language, 9.1

Verhaar, J. (1984) "Topic continuity in Indonesian discourse", St. Michael's In¬
stitute, Gonzaga University, Spokane, Wash, (ms)
von Wright, G.H. (1971) Explanations and Understanding, Ithaca: Cornell
University Press

Waddington, C.H. (1942) "Canalization of development and the inheritance of

acquired characters". Nature, 150

Waddington, C.H. (1953) "Genetic assimilation of an acquired character". Evolu¬

tion, 7

Wald, B. (1973) Tense Variation in Mombasa Swahili, PhD dissertation, Colom¬

bia University (ms)
 Bibliography 451

Wald, B. (1983) "Referents and topics within and across discourse units: Obser¬
vations from contemporary English", in F. Klein-Andreu (ed.) Discourse
Perspectives on Syntax, NY: Academic Press
Waley, A. (1934) The Way and its Power, London: Macmillan (reprinted 1958,
NY: Grove Press)
Wallace, A.F.C. (1961) Culture and Personality, NY: Random House
Wallace, A.R. (1890) Darwinism, London: Macmillan
Welch, H. (1957) Taoism: The Parting of the Way, Boston: Beacon Press
Whitaker, H. (1983) "Toward a brain model of automatization: A short essay", in
R.A. Magill (ed.) Memory and Control of Action, Amsterdam: North Holland
Whitehead, A.N. (1938) Modes of Thought, NY: Capricon Books (1958 ppbk edi¬
tion)
Whiting, W.C., R.F. Zernicke, T.M. McLaughlin and R.J. Gregor (1980) "The
recognition and correlation of human movement patterns", J. of Biomechanics,
13
Whorf, B.L. (1950) "An American Indian model of the universe", I.J.A.L., 16
Whorf, B.J. (1956) Language, Thought & Reality: Collected Writings, ed. by J.B.
Carroll, Cambridge: MIT Press
Wierzbicka, A. (1986) "Does language reflect culture: Evidence from Australian
English", Language and Society, 15
Winograd, T. (1970) A Computer Program for Understanding Natural Lan¬
guage, PhD Thesis, Dept, of Mathematics, MIT (ms)
Wittgenstein, L. (1918) Tractatus Logico Philosophicus, tr. by D.F. Pears and B.F.
McGuinness, NY: The Humanities Press (1966)
Wittgenstein, L. (1953) Philosophical Investigations, tr. by G.E.M. Anscombe,
NY: Macmillan
Woodruff, P. (1982) Plato, Hippias Major, Indianapolis: Hackett Publishing Co.
Wright, S. and T. Givon (1987) "The pragmatics of indefinite reference". Studies
in Language, 11.1
’

'
 INDEX

abduction 13, 242, 275, 276 Categorization 36, 37, 38, 40

abductive induction 244 causal explanation 304
abductive inference 7,13,20,242 causation, abduction and 244
abductive reasoning 286 certainty 127,139,164
Achinstein, P. 289 Chomsky 77, 94, 107, 122, 283, 321,
adaptive behavior 369, 372 393, 396
Adler, M. 393 ChuangTse 405
analogical reasoning 7, 286 code 69, 76
analogy 6, 45 coding 81
anaphoric predictability 214 cognitive science 33
anaphoric reference 205 communicative context, evolution of
Anaximander 9 403
apagoge13 communicative contract 129,130
Aristotelian functionalism 372 completeness 3
Aristotle 10, 70,128, 238, 371, 374 Confucian Code 420
Artificial Intelligence 31 consciousness, levels of 252
attended processing 250 consistence 3
attention 235 context 2, 73, 74, 75
Austin 27,151 context, overlap of 365, 366
automated processing 248, 250,386 context-dependence 44
automated processing, evolution of contextualization 301
259 continuum 5
automated processing, function of contradiction 24,167, 240
256 coreference 198
automaticity, levels of 252 Cresswell 30
backgroundedness 146 Culture 323,339
Baldwin 384 culture, grammar and 354
Bateson 381,382, 383 Darwin 32, 379,380, 381, 384, 394
biological diversity 373 De Interpretatione 128
Bloomfield 27,277, 278,326 deduction 239
Boaz 26 deductive explanation 301, 302, 303
Bolinger 106, 223 deductive logic 24
Bonnet 375 deductivism 269, 279, 283, 287, 301
Carnap 173, 273 deference 164
Cassierer 79 definiteness 205, 208
cataphoric importance 214 definiteness, sources of 206, 207
categories 35 degree 7
454 INDEX

Descartes 95 Great Chain of Being 376

designatum 76, 77, 78, 84 Gumperz 30
dialecticians 9 Haberland, H. 151
Difference of Man, the 393 Haiman, J. 77, 99,104, 111, 333
discourse pragmatics 83,91 Hanson, R.N. 286, 288, 289, 290, 320
discovery 284 hardware 388
effability 329 Haviland, J. 345,346
embryonic development 374 Hempel 309, 317
empiricism 14 Heraclitus 9
empiricists 14 Herder 377
epigenesis 375 Hinton, G. 392
epistemic modalities 128,131,148 history 401, 402
events 88 Hockart 359
evidentiality 130, 131, 134, 135, 136, Humboldt 26
137 hypothesis formation 281, 286, 288
evocation 198 icon 125
evolution 31,369, 371,373, 391 iconicity 69, 94, 96,104,121
excess structure 395 importance 6
Excluded Middle, law of 5,15 indefinite reference 175,179
exemplars 51 indirect speech-acts 152
existence 173 induction 239, 243, 274, 276
explanation 300 inductivism 273, 277
explanation, generalization as 301 inference, modes of 241
falsification 279, 280 information 127
family resemblance 25,49 innate ideas 95
fossil record 375 intent 127
frame 2 internal reconstruction 294
frequency distribution 40,41, 42 interpretant 21, 70, 71, 72, 73
function 8,123 irrealis 133,137
function, in biology 310 isimorphism 94, 96,117
functional explanation 308, 317 Jespersen 27
functional realms 82 Kant 14,16,17,18,19, 67,321
functionalism 12,26 Katz, J. 329
Futuyma, D. 397 Keenan, E. 28
fuzzy logic 31 kinesiology 253
Geertz 90, 362, 363, 367 kinship 355
genetic code 369 knowledge, modes of 237
Gestalt 31 Koestler 382
Goethe 377 Kroeber 26,355, 356
Goffman 30 Labov, W. 400
Golden Mean 11 Lakoff, G. 38
Gougen, J. 31 Lamarck 369, 370, 378, 379, 381, 384
Gould, S.J. 314,386,389,390,396,397, Lamarck's Second Law 381
398, 401 LaoTse 405
Grace, G. 329 Leach 353, 357,358, 359
gradation 5 Levinson, S. 30
grammatical mode 246 Lewis, D. 29,170
grammaticalization 57, 295 lexical meaning 83
 INDEX 455

Lieberman, P. 399 preformation 375

Lloyd Morgan 384 presupposition 28,132,135,136,142,
logic 269 205
Lounsbury, F. 359, 360 processing, modes of 237
Lysenko 381 propositional code 86
Malinowski 353, 357,359 propositional frame 127
manipulation 148,149 propositional information 83
Mayr 369, 370, 380, 385, 400 propositional modalities 127, 141,
Mean, doctrine of The 11 177, 241
meta-iconicity 120 prototype 38,39
meta-levels 3 prototypes 34,54
metaphoric extension 54 punctuated equilibrium 396, 397
method 269 purpose 8
middle ground, the 423 Quine 175
modal logic 29 Rationalism 14
modularity 399 rationalists 14, 329
Montague, R. 29 realis 133,137
Morris, C. 70 recapitulation 401
mystic 405 reductionism 14, 270, 271, 364
Naturphilosophie 377 reference 173
negation 156 referential confusion 218
negotiation 170 referential import 181
non-contiguity 390 referential intent 174,196
non-discreteness 22, 48,147,193 referential opacity 175,176
norm 8 relevance 6
normative action 192 relevance, abduction and 244
norms 51 Rosch, E. 38
object 220, 221 Ross, J.R. 38
object, direct and indirect 228 routinization 237
ontogenesis 374, 376, 402 routinized behavior 386,391
ontogeny 124 routinized learning 254
Ordinary Language Philosophy 27 routinized processing 249
other minds 205 Russell 3,4, 28,173,273
Pawley, A. 332 Sadalla, E. 349
Peirce 20, 70, 71,72, 99, 242, 243, 288 saliency 8
Philosophical Investigations 43 Sapir 27, 326,330, 341
phylogenesis 376, 402 Sapir-Whorf Hypothesis 327
pidgin 246 Saussure 95
Plato 35,151 Scala Naturae 378
Platonism 376,378, 399 Schank, R. 31
point of view 1, 89 Scheffler, I. 359
politeness 165 Schegloff, E. 30
Popper 21, 270, 273, 279, 281, 289 Schelling 377
Posner, M. 38, 64, 251, 252 Schneider, W. 250, 254
Possible Worlds 29,174 science, philosophy of 269
pragmatics of reference 181 scientific method 286
Pragmatism 20 Scriven, M. 300
pre-grammatical mode 246 selection 384
456 INDEX

semeiotic 8 universal core 345

Serres 378 universalist 329
similarity 6,45 universality 324
similarity, abduction and 244 universe of discourse 173
skilled learning 386, 390 urigubu 353
socio-cultural transmission 392 variability 369
Socrates 10 Verhaar, J. 78, 79, 93
software 388 verification 280
speech-act 27,148 Wallace, A.R. 394,395
speech-act continuum 151 Whorf 326, 327, 328
St. Augustine 372 Wierzbicka, A. 363
stereotypicality 117 Winograd, T. 31
Straw Dogs, doctrine of 421 Wittgenstein 23, 27,35,37,43,80,167,
structuralism 27, 320,355 238, 274, 325
subject, grammatical 220 word-order, pragmatics of 222
symbol 125 world view 344, 349
synolon 11 Wu-Wei 417, 419
Tao 406,411,412 Yin-Yang 408, 409, 415
Tao Teh Ching 406
Taoism 405
Taoist epistemology 413
Taoist ethics 416
Taoist metaphysics 406
tautology 24,123,167, 240, 319
teleology, abduction and 244
thematic predictability 219, 233
theoretical terms 289
theory-laden facts 289, 290, 291
Thompson, S. 333, 337
topic 209
topic accessibility 215
topic importance 226
topic predictability 215
topic, scalarity of 211, 212
topic, subject and 211
topicality 198,205, 208
topicality, measurement of 214
Tractatus Logico-Philosophicus 238
Transcendental Schema 18
transitivity 59,112,113,115
translation 323, 328, 339
translation, cross-cultural 339
translation, grammar and 330
translation, kinship and 361
translation, pragmatics of 364
Trobriand 353
truth 127,144
Types, Theory of 3, 202