Indefinite pronouns

Martin Haspelmath

1997

Manuscript version submitted to OUP

 1

Overview

The subject matter of this book are the formal and semantic properties of
INDEFINITE PRONOUNS, expressions such as those highlighted in (1)-(4), in the
languages of the world.

(1)English
Someone once said that anything goes.
(2)Italian
Nessuno ha mai detto questo.
nobody has ever said that
'Nobody has ever said that.'
(3)Russian
Kto ugodno mož et prijti.
who INDEF can come
'Anyone can come.'
(4)Japanese
Dare-ka ni ki-ite mi-masyoo.
who-INDEF DAT ask-CONV try-POL:HORT
'Let's ask somebody.'

Indefinite pronouns have traditionally played only a minor role in descriptive

linguistics, but the theoretical work in semantics, pragmatics and syntax of the
last few decades has shown that the distribution of indefinite pronouns is
highly complex and interesting in many ways. These theoretical discussions
have often used the narrower terms QUANTIFIERS or NEGATIVE POLARITY ITEMS
for certain sub-classes of indefinite pronouns.
This book gives a comprehensive overview of the main theoretical debates
around the semantic and syntactic properties of indefinite pronouns. Its major
original contributions are a large-scale cross-linguistic study of indefinite
pronouns (see especially chapter 4 and Appendix A) and a detailed
investigation of the diachronic sources of the markers of indefinite pronouns
(chapters 6-8). This rich factual material is brought to bear on the issues that
have been present in the literature. The new generalizations that emerge from
the typological and diachronic study are discussed and explanations are
provided. Throughout the book, particular emphasis is put on links between
the formal properties of indefinite pronouns and their functional (semantic and
syntactic) properties.
In chapter 2 after this overview, I give a brief introduction to the goals and
methods of the typological approach that is applied in this book. Typological
work presupposes cross-linguistically applicable concepts, so a definition of
'indefinite pronoun' that is independent of language-particular properties is
provided. Another important precondition for typology is the availability of
data from a wide range of languages, so the issue of sampling is also discussed
in chapter 2. This study is based on quite detailed data from 40 languages, and
on very limited data from another sample of 100 languages which comes close
to being representative of the world's languages. The 40-language sample is
 2

biased toward European languages, as is inevitable due to limitations on the

availability of data.
Chapter 3 maps out the space of formal and functional variation that is
found in indefinite pronouns. Formally, indefinite pronouns typically occur in
several series, with each series comprising a set of indefinites referring to the
major ontological categories (person, thing, place, time, manner,...). For
instance, English has the some-series (somebody, something, somewhere,...), the any-
series (anybody, anything,...), and the no-series (nobody, nothing,...); Russian has
the -to-series (kto-to 'somebody', č to-to 'something', gde-to 'somewhere',...), the -
nibud'-series (kto-nibud' 'anybody', č to-nibud' 'anything',...), and others. In terms
of synchronic formal structure, languages are remarkably similar typologically,
but extremely interesting cross-linguistic variation is observed in the different
meanings or functions that these different series can express. In chapter 3, I
identify nine core functions (i.e. meanings and/or contexts) that must be
distinguished for the purposes of cross-linguistic comparion. Not all of these
functions are distinguished formally in every particular language, but each
function is justified by attested differences between the functional ranges of
different indefinite pronoun series. These nine functions are as follows:

specific, known to the speaker

(5) Somebody called while you were away: guess who!

specific, unknown to the speaker

(6) I heard something, but I couldn't tell what kind of sound it was.

non-specific, irrealis
(7) Please try somewhere else.

polar question
(8) Did anybody tell you anything about it?

conditional protasis
(9) If you see anything, tell me immediately.

standard of comparison
(10) In Freiburg the weather is nicer than anywhere in Germany.

direct negation
(11) Nobody knows the answer.

indirect negation
(12) I don't think that anybody knows the answer.

free choice
(13) Anybody can solve this simple problem.

Some of these functions are well known and well established (e.g. 'free choice',
'negation'), while others are less known (e.g. 'known vs. unknown to the
speaker', 'irrealis'). There is one distinction that is very often made in the
literature but that does not appear in the above list: NEGATIVE POLARITY.
'Negative polarity items' are often said to be restricted to the functions in (8)-
(12), but the cross-linguistic data show that the notion of negative polarity is too
 3

crude. In addition to indefinite pronoun series that are restricted to precisely

these five functions, there are other indefinites that are restricted to a subset of
these five functions, or to a set of functions that comprises some non-negative
polarity functions in addition to some negative-polarity functions. The final
section of chapter 3 then discusses alternative ways of expressing what many
languages express by indefinite pronouns, including the rare but apparently
attested case of a language that completely lacks indefinite pronouns.
Chapter 4 presents the principal typological generalizations that emerge
from the data of the 40-language sample. These take the form of implicational
universals concerning the distribution of indefinite pronoun series over the
nine functions distinguished in chapter 3. While there are many differences in
the distribution of indefinite pronoun series of different languages, the
variation is not without limits. As elsewhere in typology, we find many
regularities in the bewildering diversity that strikes us at first sight. This can be
illustrated with examples from Russian, which differs in an interesting way
from English. The sentences in (14)-(17) are the Russian equivalents of (6), (7),
(9), and (10).

Russian
(14) Ja uslyš ala č to-to, no ja ne ponjala, kakoj zvuk èto bylo. (=6)
I heard what-INDEF but I not understood what sound it was

(15) Popytajtes', pož alujsta, gde-nibud' v drugom meste. (=7)

try:IMPV:PL please where-INDEF in other place

(16) Esli ty uvidiš č to-nibud', skaž i mne srazu. (=9)

if you see what-INDEF tell:IMPV me immediately

(17) Vo Frejburge pogoda lučš e, č em gde-libo v Germanii. (=10)

in Freiburg weather better than where-INDEF in Germany

Although above I glossed Russian kto-to, č to-to as 'somebody, something', and

kto-nibud', č to-nibud' as 'anybody, anything', these examples are sufficient to
demonstrate that the English some-series and the Russian to-series are by no
means equivalent, and neither are the any-series and the -nibud'-series. The -to-
series cannot be used in (15), and the any-series cannot be used in the English
equivalent (7). Furthermore, the -nibud'-series cannot be used in (17), where
English has the any-series, and Russian employs a third series, the -libo-series.
But cross-linguistic variation of this kind is strongly constrained by a number of
implicational universals which are summarized by the IMPLICATIONAL MAP in
Figure 1.
 4

Figure 1. The implicational map for indefinite pronoun functions

(7)
direct
(4) (6) negation
question indirect
negation
(1) (2) (3)
specific specific irrealis
known unknown non-specific
(5) (8)
conditional comparative
(9)
free choice

The map is to be read as follows: An indefinite pronoun series will always

express a set of functions that are adjacent on the map. For instance, the English
any-series covers functions (4)-(9), and the Russian -nibud'-series covers
functions (3)-(5). Since adjacency on such an implicational map can only be
explained by similarity of function or meaning, the implicational map may be
considered as a geometric representation of its semantic or cognitive domain.
In chapter 5, I discuss a variety of theoretical approaches to the functions of
indefinite pronouns, beginning with the tradition of structural semantics.
Probably the best-known approach is that of logical semantics, where
indefinites are seen as the counterparts of the existential or the universal
quantifiers. While many important insights have been gained within this
approach, there is one crucial problem that has remained unresolved: the
existence of indefinite pronouns which appear to correspond to the existential
quantifier in some contexts, and to the universal quantifier in others (for
instance, English any- and Italian nessuno 'anybody; nobody'). Another
interesting approach (Progovac 1994) tries to subsume the distribution of
certain indefinites under the syntactic binding theory that was originally
developed for reflexive and reciprocal pronouns. Next I present the mental-
space-approach (Fauconnier 1985), which offers important insights into the
nature of non-specificity and related phenomena. Finally Fauconnier's (1975a,
1975b) work on pragmatic scales and scalar implicatures is discussed, which
provides a perspective for solving the puzzle of the double nature of any- and
similar indefinites in other languages. On the basis of the various theoretical
proposals I then attempt an explanation of the implicational map.
Chapter 6 begins the diachronic investigation of indefinite pronouns. Since
very few in-depth studies of diachronic changes in indefinite pronouns are
available, most of the typological generalizations in the diachronic parts of this
book are based on etymological information. In chapter 6 I look at those
indefinites that are best analyzed as having arisen by grammaticalization from
a complex syntactic construction involving an interrogative pronoun (these
constructions thus explain in part why indefinites are so often based on
interrogatives). I identify four such constructions: 'I don't know wh-', 'wh- you
want', 'wh- ever it may be', 'no matter wh-'. The main effects of
grammaticalization, such as phonological erosion, desemanticization, reduced
syntactic scope etc. can all be shown to be present in the attested
grammaticalized indefinites. The semantic aspect of grammaticalization can be
seen in the extension of an indefinite's range of functions from its original
function to other functions along the implicational map.
 5

Chapter 7 deals with a number of further sources of indefinite pronouns that

cannot be easily subsumed under grammaticalization. These are indefinites
marked by scalar additive focus particles ('also', 'even', 'at least'), reduplicated
indefinite pronouns, indefinite pronouns derived from generic nouns ('person',
'thing', etc.), and others. A very important type of indefinite pronoun across
languages is represented by the use of interrogative pronouns as indefinites
without any additional markers. I argue that the indefinite use is always
secondary in such instances. As in chapter 6, for each of the diachronic sources
of indefinite pronouns I identify the original motivation and thereby contribute
to explaining their formal properties.
In chapter 8, I single out one of the nine functions, the function of direct
negation (as in (11) above), for further detailed discussion. I show that the
cross-linguistic patterning of negative indefinite pronouns cannot be
understood without placing them in the context of other indefinites and that
the implicational map captures important aspects of their behavior. Negative
indefinite pronouns are particularly interesting because they may or may not
cooccur with the normal sentential negation marker on the verb. The situation
in which negative indefinites and verbal negation cooccur has sometimes been
called 'double negation', but in fact this situation is much more common cross-
linguistically than negative indefinites of the standard European type which
simultaneously act as sentential negators. I formulate a number of typological
generalizations concerning the cooccurrence of negative indefinites and verbal
negation and propose explanations for them. The chapter ends with an
examination of the diachronic sources of negative evidence, which are quite
parallel to those of other indefinites.
The concluding chapter 9 summarizes the typological generalizations of this
book and briefly looks at wider typological and areal connections.
The book is concluded with two appendices: Appendix A gives detailed data
on indefinite pronoun series and their distribution in the languages of the 40-
language-sample, providing many examples of indefinite pronouns. Appendix
B gives the data from the world-wide 100-language-sample.
With this book I hope to contribute to Wilhelm von Humboldt's program of
investigating universal grammar through broad cross-linguistic comparison,
which he formulated in his 1827 study of the dual:

'Unter den mannigfaltigen Wegen, welche das vergleichende Sprachstudium

einzuschlagen hat, um die Aufgabe zu lösen, wie sich die allgemeine menschliche
Sprache in den besonderen Sprachen der verschiedenen Nationen offenbart? ist
einer der am richtigsten zum Ziele führenden unstreitig der, die Betrachtung eines
einzelnen Sprachtheils durch alle bekannten Sprachen des Erdbodens hindurch zu
i
verfolgen.'
 6

i
'Among the manifold ways that comparative linguistics must pursue towards a solution
of the question of how the general human language is manifested in the particular
languages of the different nations, indisputably one of the most successful ones is
examining a single part of language throughout all the known languages of the earth.'
 9

A Typological Perspective on Indefinite Pronouns

2.1. Language Typology

This book approaches the topic of indefinite pronouns with the goals and
methods of language typology as founded by Wilhelm von Humboldt (e.g.
1827), revived by Joseph Greenberg (1963) and summarized most recently in
Croft (1990). A few general remarks on the typological approach are thus in
order before we turn to a definition of indefinite pronouns in the next section.
Language typology, as understood here, is the scientific study of variation and
the limits to variation in the structure of languages. There are several reasons
why typological research is of central importance to our understanding of
human language.
Despite the bewildering diversity of their structure, different languages have
always been regarded as roughly equivalent instantiations of the more abstract
notion of 'human language', a unique and universal endowment of human
beings. However, the question arises how language could be universal and at
the same time be manifested in forms that show seemingly unlimited variation.
Systematic cross-linguistic investigations of language structure, as conducted
by typologists at least since Humboldt (1827), have made possible significant
advances toward a resolution of the apparent contradiction between
universality and diversity. Above all, typological studies have demonstrated
that cross-linguistic variation is by no means random. On the contrary, the
grammatical systems of languages around the world show such striking
similarities that it seems not inconceivable to extract a common core, a
'universal grammar', out of the individual grammars.
On another level, linguistic typology is indispensable for our goal of
explaining particular grammatical phenomena and of detecting significant
generalizations. The fundamental problem that we are faced with is to state
generalizations at the right level of generality. Linguists often fall into the trap
of explaining a language-particular phenomenon by a high-level generalization
(making reference e.g. to innate language structure or to general cognitive
capacities), thus wrongly predicting that the phenomenon in question should
be universal. And conversely, linguists also often explain a very general
phenomenon by a low-level explanation (making reference e.g. to historical or
other accidents in a particular language), thus wrongly predicting that the
phenomenon in question should be restricted to the particular language they
are dealing with. Cross-linguistic studies can help steer us through the Scylla of
overgeneralization and the Charybdis of undergeneralization. If in a cross-
linguistic study a phenomenon turns out to be universal, we know that it must
be explained with reference to universal factors; if on the other hand a
phenomenon turns out to be restricted to a single language or a small set of
isolated cases, then a particular explanation is called for.
Now it turns out that many, if not most, interesting grammatical phenomena
fall between these two extremes: They are not completely universal, but they
are not restricted to a small random set of languages either. However, more
sophisticated techniques have been developed by typologists that allow us to
find universal generalizations even for non-universal phenomena. The central
 10

notion is that of an IMPLICATIONAL UNIVERSAL, i.e. a property that holds in all

languages that have some other property. In a way implicational universals are
even more important than absolute universals because they give us some
additional insight into the nature of grammar: If a property A is implied by a
property B, then it is quite likely that the two are causally related. The major
cross-linguistic generalizations of this work will take the form of implicational
universals.
An important practical problem with typology is that typological studies can
never reach the same depth of analysis as studies on individual languages. The
more languages a typologist investigates, the less attention he or she can give to
each individual language, so typological studies are of necessity more
superficial than single-language studies, and this book is no exception. It is true
that there is a certain danger that data from individual languages are isolated
from their larger context in broad cross-linguistic studies. But on the other
hand, single-language studies that do not consider the typological variation
also isolate their data from the cross-linguistic context and thereby are in
danger of misinterpreting the facts, however deep and fine-grained their
analyses are. The obvious solution to the limitations of typological and single-
language work is to combine the two to the extent that this is possible. Both ap-
proaches to language are complementary, because necessarily one loses depth
as one gains in breadth, and conversely, one loses breadth as one gains in
depth. But neither depth nor breadth are inherently more important – what
counts is the volume of the new knowledge which results from multiplying
breadth and depth.
On the empirical side, the present work focuses more on breadth, because
there are already a substantial number of in-depth studies of indefinite
pronouns in individual languages which I have been able to make use of,
whereas the typological perspective has not been elaborated sufficiently in
previous research. This empirical breadth is used as a measure for theoretical
approaches and thus hopefully leads to greater depth of explanation.
A typological study logically consists of four steps. First, the domain of
phenomena to be compared across languages is delimited by formulating a
definition that is cross-linguistically applicable. Second, the space of typological
variation is mapped out by providing a complete taxonomy of the various
means by which the phenomenon under discussion is expressed in different
languages. Third, correlations between individual structural options and other
parts of the grammar are identified and formulated as implicational universals.
Fourth, explanations for these universals are sought. In this book, the first step
(definition) is taken in the following section (§ 2.2), the second step (taxonomy)
follows in chapter 3, the third step (correlations) is the topic of chapter 4, and
the fourth step is the explanation of the implications in semantic terms that is
proposed in § 5.6.

2.2. Indefinite Pronouns:

Definition and Delimitation of the Domain of Inquiry
2.2.1. Mixed functional-formal definitions

This work applies the cross-linguistic perspective to INDEFINITE PRONOUNS like

English somebody, something, somewhere, anything, anytime, or Japanese dareka
‘somebody, anybody’, dokoka ‘somewhere, anywhere’, daremo ‘anybody,
 11

nobody’, nanimo ‘anything, nothing’. Such indefinite pronouns generally occur

in SERIES, e.g. the English some-series and the any-series, or the Japanese -ka-
series and the -mo-series. The functional distinctions exhibited by such
contrasting series will be the main object of my interest.
There is both a functional and a formal component to the definition of the
subject matter of this study: From a functional point of view, the expressions
investigated here can be characterized as INDEFINITE (in the sense discussed
below), and from a formal point of view they can be characterized as
PRONOUNS, i.e. grammatical elements.
This double or hybrid definition is as it should be: The subject matter of a
typological investigation cannot be defined either on an exclusively formal
basis or on an exclusively functional basis. Purely formal definitions are
impractical because there are extremely few if any structural (or formal)
properties that can be identified and compared across languages. It is possible,
for example, to study the Instrumental case in various Slavic languages on the
basis of a purely formal definition (case forms which are the reflexes of Proto-
Slavic *-mĭ , *-jo4, *-mi, etc.), but a definition of this kind cannot serve as the
basis for a typological study, simply because most languages do not have
suffixes of these shapes. On the other hand, purely functional definitions have
the disadvantage that they tend to pick out quite heterogeneous expressions.
For example, a typological study of temporal expressions that is not formally
delimited would have to consider such diverse phenomena as verbal tense
inflections, tense iconicity in coordinate structures (like I came, I saw, I
conquered), and temporal adverbs and nouns like tomorrow and hour. Although it
is rarely made explicit, typological investigations are usually based on a
combined functional-formal definition of their domain of inquiry, e.g. Dahl’s
(1985) study of tense-aspect systems, Svorou’s (1994) study of spatial
adpositions, Koptjevskaja-Tamm’s (1993) study of action nominal constructions,
Nedjalkov’s (ed.) (1988) study of resultative constructions, Stassen’s (1985)
study of comparative constructions, and many others.

2.2.2. The formal criterion: 'pronoun'

The formal criterion used here is the pronominal, i.e. grammatical na-ture of the
expressions. In particular, I will not deal with lexical indefinite expressions like
a woman, a tree, a meadow, or with general nouns which are not pronouns by any
criterion, like a person, a thing, a place, etc. Since semantically there is almost no
difference between an indefinite pronoun like someone and a lexical noun
phrase like a person, this is a formal criterion.
It is true that it is not always immediately obvious whether an expression is a
pronoun or not. Often pronouns are easily recognizable as such because of
highly specific phonological and morphological properties, e.g. the Genitive
singular suffix -go in Polish pronouns (je-go ‘his’, ko-go-ś ‘someone’s’) vs. -y, -u,
-a in nouns (osoby ‘a person’s’). But in many other cases, formal pronominal
features are harder to detect. The English indefinite pronoun something differs
from the lexical phrase some thing only in its prosodic ('– – vs. ª– '–),
morphological (*somethings vs. some things) and syntactic properties (something
important vs. some important thing). The French pronoun quelque chose
‘something’ differs from the phrase quelque chose ‘some thing’ in its agreement
properties: the pronoun takes masculine agreement (quelque chose s’est passé
‘something happened’), whereas the noun chose ‘thing’ is feminine. In Turkish,
bir s3ey ‘something’ formally resembles a phrase (bir ‘one’, s3ey ‘thing’), but the
 12

process of m-reduplication shows that we are dealing with a single word. This
process applies to words, so kitap-mitap ‘books and so on’ is possible, while *bir
kitap mir kitap is impossible. However, we do have bir s3ey mir s3ey ‘something
and so on’, so bir s3ey must be a single word, and hence a pronoun.
But even if it were possible to obtain all the relevant data also for little-
known languages, it would still be futile to try to draw a clear-cut boundary
between lexical expressions and grammatical, pronominal expressions that
applies to all cases. When new pronouns are created in a language, they are
often grammaticalized from nouns that have a very general meaning (cf.
Lehmann 1982a: 38-56, Sugamoto 1989), and since grammaticalization is a
gradual process, there must be plenty of cases in many languages that are
somewhere in the middle on their way from the lexicon to the grammar.
It should furthermore be emphasized that the term pronoun is used here in its
broad sense where it comprises not only PRO-NOUNS (i.e grammatical items that
can replace nouns or noun phrases), but also PRO-ADVERBS like somewhere,
sometime (which can replace adverbs or adverbial phrases), PRO-ADJECTIVES like
Polish jaki-ś ‘some kind of’, and even PRO-VERBS. Thus the term pronoun
(without a hyphen) should not be understood in its etymological sense
(‘replacing a noun’), but in its traditional sense.i Since I am mainly interested in
the differences between different series of indefinite pronouns, such as the
English some-series (somebody, something, somewhere, etc.) and the any-series
(anybody, anything, anywhere, etc.), it would not make any sense to exclude the
non-nominal members of these series from my considerations.
Also included are indefinite determiners that clearly belong to a series of
indefinite pronouns, such as English some and any, or German irgend-ein (which
belongs to the irgend-series, cf. irgend jemand ‘somebody’, irgend etwas
‘something’, irgend-wo ‘somewhere’, etc.). Determiners are not pronouns be-
cause they do not replace anything, but in traditional grammar determiners are
often treated as pronouns or at least together with pronouns, so including them
here does not do too much violence to the term pronoun.

2.2.3. The functional criterion: 'indefinite'

We now turn to the functional part of the definition of INDEFINITE PRONOUN.

Again, the definition is quite straightforward: I will regard as indefinite
pronouns such pronouns whose main function is to express indefinite ref-
erence. While this terminological usage is highly transparent, it represents a
certain departure from traditional terminology. In traditional Western
grammar, the following types of pronouns are usually distinguished (e.g.
Grevisse 1986: § 625-736, Braune 1961: § 282-300):

(18) (i) personal pronouns

(ii) demonstrative pronouns
(iii) relative pronouns
(iv) interrogative pronouns
(v) indefinite pronouns

Whereas the first four of these form relatively coherent sets, the category of
indefinite pronouns seems to function as a sort of waste-basket category in
many traditional descriptive grammars. In addition to indefinite pronouns in
the narrow sense (in which the term is used throughout this work), the
following four types of expressions are commonly put in this category:
 13

(a) MID-SCALAR QUANTIFIERS like few, several, [Link] These express quantity
and have nothing to do with indefiniteness—many of them combine freely with
the definite article (the few, the many). However, some mid-scalar quantifiers are
incompatible with the definite article in various languages, e.g. English some
(*the some), German einige ‘several, some’ (*die einigen), French maints ‘many’
(*les maints). And more importantly, some mid-scalar quantifiers are formally
very similar to indefinite pronouns (and diachronically derived from them), e.g.
English some [sm] (mid-scalar quantifier) vs. some [sUm] (indefinite pronoun),
Russian ne-kotorye ‘some’ (mid-scalar quantifier, plural) vs. ne-kto ‘someone’
(indefinite pronoun). These facts may explain why mid-scalar quantifiers have
often been lumped together with indefinite pronouns.
(b) GENERIC PRONOUNS like French on ‘one’, German man ‘one’, English one
are also often called 'indefinite pronouns'. It is true that they are both
pronominal and indefinite, so they would fall under the definition as stated
above, but since generic pronouns have very different properties from words
like someone, they are not further considered here. Moreover, formally distinct
generic pronouns do not seem to be very widespread: most languages make do
without them, whereas the large majority of languages seem to have indefinite
pronouns like someone.
(c) UNIVERSAL QUANTIFIERS like all and every. The collective universal
quantifier ‘all’ is quite compatible with the definite article in many languages
(e.g. all the children, French tous les enfants, Hebrew kol ha-yeladim), and even the
distributive universal quantifier ‘every’ can be combined with the definite
article in some languages, e.g. Modern Greek i káthe ghlósa ‘every language’. As
a rule, universally quantified noun phrases are semantically definite, even
when not marked as such in languages that have a definite article.
That said, it must be acknowledged that there are close connections between
distributive universal quantifiers (‘every’) and indefinite pronouns that express
irrelevance of choice (‘any’). Although universal quantifiers are not considered
indefinite pronouns and are not the focus of my attention in the present study,
they will be mentioned and discussed in connection with free-choice indefinites
later (§ 6.5). (Cf. Gil 1991 for an excellent typological study of universal
quantifiers.)
(d) IDENTITY PRONOUNS/DETERMINERS like other and same. These express
identity and non-identity and do not show an affinity to indefiniteness at all.
After these negative statements, what kinds of pronouns are left for our
investigation? Essentially, indefinite pronouns like those in (19a-c).

(19) a. someone, something, somewhere, somehow,...

b. anyone, anything, anywhere, anyhow,...
c. no one, nothing, nowhere, never,...

It could be objected that the pronouns of the any-series mainly express free
choice rather than indefiniteness (as in You may take anything), and that the
pronouns of the no-series express non-existence rather than indefiniteness. This
may be true, and perhaps the definition 'pronouns whose main function is to
express indefinite reference' should be made more specific to include all of
them.
However, the crucial point to note is that there are very close connections
between different series of indefinite pronouns as in (19a-c), not just in English,
but in many other languages as well, as I will show in the course of this work.
In English, there are contexts where either the some- or the any-series can be
 14

used, without a great meaning difference (If you see someone/anyone, tell me), and
there are contexts where either the no-series or the any-series can be used
without any meaning difference (I saw nobody/I didn’t see anybody). Such close
connections, which will be the focus of our attention in the main part of this
work, justify the inclusion of all of these pronoun types in the typological
investigation. The points of contact and overlap with other grammatical
phenomena are minor compared to the internal coherence of indefinite
pronouns like those in (19a-c).

2.3. Earlier Work

The in-depth study of indefinite pronouns does not have a long history in
linguistics. To be sure, indefinite pronouns have generally been described in
reference grammars. In the European tradition of grammaticography that goes
back several centuries, there is always a separate section on pronouns, divided
into sub-sections on different kinds of pronouns, usually in the following
sequence: personal pronouns, demonstrative pronouns, relative pronouns,
interrogative pronouns, indefinite pronouns. But the description of indefinites
is very often restricted to the formal side in such works, i.e. to their derivational
structure and their inflectional properties. Only reference grammars with a very
complete syntactic component, such as Kühner and Stegmann (1914) (Latin),
Paul (1916-20) (German), Schwyzer (1953) (Classical Greek) have had anything
to say on the functions of indefinite pronouns. In these older grammars, much
of the syntax component consists of the description of syntactic and semantic
properties of grammatical categories and grammatical words, and the functions
of indefinite pronouns are then described in the section on pronoun syntax. In
more modern grammars, where the syntax is generally given more autonomy,
the situation sometimes becomes worse: The functional properties of indefinite
pronouns do not fit into any of the major parts of the syntax, and as a result of
this, even voluminous grammars may be completely silent on the functions of
indefinite pronouns, e.g. Š vedova (ed.) (1980) (Russian), Engel (1988) (German).
In smaller reference grammars, especially grammars of little-known languages,
indefinite pronouns are often completely ignored.
The first comparative studies of indefinite pronouns were conducted within
the framework of historical-comparative linguistics of the older Indo-European
languages (Brugmann 1911, Delbrück 1893), the Romance languages (Lombard
1938-39), and the Finno-Ugrian languages (Beke 1913-14). Since indefinite
pronouns even in closely related languages often show quite different forms,
these comparative studies are not very successful in meeting their own goals,
but instead they tend to take on a typological character. Some later comparative
studies by Indo-Europeanists such as Frei (1940), Gonda (1954-55) explicitly
take into account typologically similar phenomena from non-Indo-European
languages.
The first theoretically-oriented in-depth studies of the functions of indefinite
pronouns of individual languages appear in the 1960s,iii motivated by the
structuralist movement in semantics (Greimas 1963, Veyrenc 1964, Manoliu-
Manea 1966, Kř íž ková 1971, Topolinska 1972), by logical semantics
(Seliverstova 1964, Dahl 1970, Horn 1972, Grzegorczykowa 1972b, Levin 1973,
Paduč eva 1974), or by transformational grammar (Lees 1960: 123-125, Klima
1964, Labov 1972, Ali 1970, 1972). Since the 1970s, there is also a steady flow of
papers that deal contrastively with indefinite pronouns in two languages
 15

(Rybák 1965, Donč eva 1970, Pálfy 1982, Stephanides 1983, 1985, Kirova 1986,
Mostovska 1988).
While the present study is unique in the breadth of its scope, it is not the first
typological study of indefinite pronouns. Valuable typological generalizations
concerning the formal properties of indefinite pronouns have been made in
Majtinskaja (1969), Coyaud and Aït Hamou (1972, 1976), and von Bremen
(1983). Schmid (1980) is the only previous study that attempts to combine a
world-wide typological comparison (involving a dozen languages) with a closer
look at the environments (mainly negative, interrogative and conditional
clauses) where different indefinite pronouns appear. Gil (1991), in his
typological study of 'universal quantifiers', also looks at free-choice indefinites
in some detail. Kahrel (1989) and Bernini and Ramat (1992) are large-scale
typological studies of negative pronouns.
In the 1980s and the early 1990s, negative-polarity indefinites have again
become the topic of studies within the Chomskyan syntactic framework (e.g.
Aoun 1985, 1986, Mahajan 1990, Progovac 1992, Zanuttini 1991, Li 1992, Laka
1994), among them also an ambitious typologically-oriented proposal (Progovac
1994). Other theoretical approaches that have dealt with indefinite pronouns are
Antoine Culioli’s 'enunciative' approach (Culioli (ed.) 1980, 1984, Fisher and
Franckel (eds.) 1983, Dhelverúdhi 1989, Le Goffic 1994, Mellet 1994), the
Guillaumean approach (Hirtle 1982, 1988), Gilles Fauconnier’s theories of
pragmatic scales (Fauconnier 1975a, 1975b, 1977, 1979, 1980, Ladusaw 1980) and
mental spaces (Fauconnier 1985, 1987), and Ronald Langacker’s Cognitive
Grammar (Langacker 1992: 103-41).
It is obviously far beyond the scope of this work to discuss all the previous
work on indefinite pronouns. However, the major theoretical approaches will
be discussed in chapter 5, and many other smaller contributions to the
theoretical debate will be mentioned throughout this study. Many of the more
descriptively oriented works on individual languages have served as useful
sources of data (cf. Appendix A, where many of these studies are cited),
without which the typological generalizations of this study would have been
impossible.
Thus, I see the present work as firmly rooted in the unbroken tradition of
previous generations of linguists. I hope that by collecting and comparing the
insights and data accumulated in these earlier works, and by highlighting the
cross-linguistic generalizations that emerge, this work will constitute a step
forward in our knowledge and understanding of indefinite pronouns in human
language.

2.4. The Language Samples

2.4.1. General problems of typological sampling

Even though its ultimate goal is generalizing over all the languages of the
world, typological work can in practice only be based on a relatively small
subset of languages. In selecting the languages to be compared, the three most
important considerations are the size and the representativeness of the sample
and the availability of data.
The choice of the best SIZE of the sample is easiest: One simply has to select
the relation between breadth and depth (cf. § 2.1.1) that best suits one’s goals. If
the questions asked in a typological study are fairly superficial, a relatively
 16

large sample of languages can be covered. The deeper the questions are, the
fewer languages can be handled.
A more difficult problem is to find a compromise between
representativeness and availability of data. A sample is REPRESENTATIVE of the
world’s languages to the extent that its internal diversity corresponds to the
diversity of the whole set of languages. The most important factor limiting
linguistic diversity is, of course, genetic relationship. Thus, the languages
included in a representative sample should be as genetically distant from each
other as possible. This in turn presupposes that we can determine genetic
closeness and distance, and to the extent that our knowledge of the genetic
relationships between languages is incomplete, a sample based on it is also
problematic. Various ways of dealing with the problem of sampling genetic
groupings have been proposed, each with their own advantages and
disadvantages (Dryer 1989, Perkins 1980, Nichols 1992, Rijkhoff et al. 1993).
Another potential problem is areal relatedness (Sprachbund phenomena), which
can be mitigated by selecting languages that are areally as distant from each
other as possible.
However, a factor that restricts typological research much more seriously
than sampling problems is the AVAILABILITY OF DATA. Data from well-studied
and widely spoken languages are usually much more easily accessible than
data from little-studied languages spoken by small groups in 'remote' areas,
and the widely spoken languages come from just a few genetic groups (10 of
the world’s 20 largest languages are Indo-European, 4 are Sinitic). This problem
of potential genetic bias becomes worse the deeper the questions are that one
wishes to investigate, because in-depth data are available for much fewer
languages than superficial descriptions. But even if one limits oneself to
questions that can be answered by an average reference grammar, it is not easy
to obtain data of equal quality and depth from a balanced, unbiased sample
that is truly representative of the genetic diversity of the world’s languages.
Thus, any typological study has to make a conscious compromise between the
limiting factors of potential genetic bias and availability of data.
It is important to recognize that this choice is not entirely independent of the
kind of question under investigation. The general goal of maximal genetic
diversity of a typological sample is based on the assumption that genetic
relatedness is by far the most important factor of bias. In general, this is
probably true: For example, the fact that most Dravidian languages have
retroflex consonants and postpositions (rather than prepositions) is certainly
not an accident or evidence for a typological connection between the two, but is
simply due to common inheritance from Proto-Dravidian, which also had these
features. But not all structural features of a language have the same degree of
DIACHRONIC STABILITY (cf. the discussion in Nichols 1992: ch. 5). Some features
are very stable and change only very slowly, whereas others tend to change
quite rapidly. This phenomenon has long been known in the domain of lexical
change: There is no doubt that words belonging to the basic vocabulary (e.g.
body part terms, kinship terms, numerals) are diachronically more stable than
peripheral words. But in the domain of grammatical features, differential rates
of change have only just begun to be investigated systematically (Nichols 1992).
Nevertheless, the rate of change of the feature under study has important
consequences for the selection of the sample (as noted in Croft 1990: 23). Highly
unstable phenomena are much less likely to be common retentions from the
parent language, and hence the problem of genetic bias is less severe when one
studies grammatical features with low diachronic stability.
 17

It turns out that the feature studied in this work, indefinite pronouns and
their functions, shows a remarkably low degree of diachronic stability.
Indefinite pronoun systems may differ substantially even in languages that are
closely related, such as Polish and Russian, Dutch and German, Catalan and
Portuguese, Latvian and Lithuanian (cf. the relevant data in Appendix A,
especially the distribution of indefiniteness series). This impressionistic
statement could easily be made more precise by introducing some measure of
diversity and comparing indefinite pronouns to other, more stable features such
as marking type (head marking vs. dependent marking), alignment type
(ergative, accusative, etc.), or pronominal-argument type (independent
pronouns vs. pronouns as verbal inflection). However, I trust that the
impression is sufficiently robust to convince readers. This means that the
problem of genetic bias is probably much less severe for this study than one
might first have thought.

2.4.2. Selection of the samples for this study

After the general remarks of the preceding section, let me now explain the
choices made in this work. To deal with the problem of sample size and the
relation between breadth and depth, I have chosen a two-level-approach, i.e. I
use two different samples showing different combinations of breadth and
depth: First, a large sample of 100 languages is investigated with respect to very
few superficial parameters that can even be found in a sketchy description (cf.
Appendix B). Second, a smaller sample of forty languages is studied in some
depth with respect to questions that can only be answered by excellent
reference grammars or dictionaries, by specialized studies of indefinite
pronouns in these languages, or by information obtained from native speakers
(cf. the data in Appendix A).
The problem of representativeness has been solved satisfactorily only for the
100-language-sample, which is evenly distributed among the genetic groups of
the world (for minor reservations, cf. the discussion in § [Link]). The 40-
language-sample is very biased genetically toward Indo-European languages
(22 of the 40 languages), and even more areally toward European languages (30
if the Caucasus is included in Europe, 27 if it is excluded). The reason for this
bias is of course the limited availability of data for other languages. However,
since a diachronically unstable phenomenon like indefinite pronouns is less
affected by genetic bias (as discussed in the previous subsection), I hope that
the 40-language-sample still gives a fairly good picture of the variation found in
human languages in general.
It should perhaps also be noted here explicitly that this study is limited to
spoken languages, i.e. that no sign languages have been taken into account. The
cross-linguistic study of sign languages is still in its infancy, and my own
competence does not allow me to say anything even about an individual sign
language. But clearly, eventually linguistic theory will have to account for
spoken languages and for signed languages in a unified way.

2.4.3. The two samples

[Link]. The 40-language sample. The 40 languages are listed in Table 1. See
Appendix A for the data on these languages, as well as the sources for these
data.
 18

Table 1. The languages of the 40-language sample

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
Indo-European
Germanic
West 1. German
2. Dutch
3. English
North 4. Swedish
5. Icelandic
Italic
Latin-Romance 6. Latin
Western 7. Portuguese
8. Catalan
9. French
10. Italian
Eastern 11. Romanian
Greek 12. Modern Greek
Slavic
South 13. Bulgarian
14. Serbian/Croatian
West 15. Polish
East 16. Russian
Baltic 17. Lithuanian
18. Latvian
Celtic 19. Irish
Indo-Iranian
Iranian
Eastern 20. Ossetic
Western 21. Persian
Indic 22. Hindi-Urdu
Turkic
Southern 23. Turkish
Central 24. Kazakh
Northern 25. Yakut
Finno-Ugrian
Ugrian 26. Hungarian
Finnic 27. Finnish
Manchu-Tungusic 28. Nanay
Nakh-Daghestanian 29. Lezgian
Afro-Asiatic
Semitic 30. Maltese
31. Hebrew
Chadic 32. Hausa
Niger-Congo 33. Swahili
Kartvelian 34. Georgian
Dravidian 35. Kannada
Sino-Tibetan 36. Chinese
Andean 37. Ancash Quechua
affiliation unclear 38. Japanese
39. Korean
40. Basque
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

[Link]. The 100-language sample. This sample was selected according to the
sampling method proposed in Rijkhoff et al. (1993), which measures the internal
diversity of genetic groupings by means of a simple calculation based on the
algebraic structure of a genetic language tree. This method is adopted here
because it has the advantage that no information in addition to a complete
classification of the world’s languages (as offered in Voegelin and Voegelin
(1977), Ruhlen (1987), Bright (ed.) (1992)) is required, in contrast to the methods
employed in Dryer (1989) and Nichols (1992), which presuppose a lot of
 19

additional knowledge about the time depth of genetic groupings. As in the

example samples in Rijkhoff et al. (1993), my sample is based on the
classification in Ruhlen (1987) because it is the most readily available
classification.
The actual procedure of selecting the languages is described in detail in
Rijkhoff et al. (1993) and will not be repeated here. Note, however, that I deviate
from their model in one respect: Whenever no data were available for any
language of a grouping that should have been represented in the sample, rather
than reducing the sample size I added another language elsewhere in the next
higher grouping. In this way, my 100-language sample really contains 100
languages, not 76 as would have been the case if Rijkhoff et al.’s proposal had
been followed. Thus, my 100-language sample may show some distortions, but
since imprecision is present elsewhere in the procedure (Ruhlen’s (1987)
classification is far from perfect, and Rijkhoff et al.’s sampling method is only an
approximation), I do not think that this is a serious problem. In any event, not
too much hinges on the quality of the 100-language sample because the data
that I have for these languages is quite restricted anyway. The main point of
this sample is to provide an impression of the world-wide situation, making it
plausible that the conclusions based on the heavily biased 40-language sample
can be extrapolated to the world’s [Link]
For genetic groupings which contain a greater number of languages for
which information is available (Indo-European, Altaic, Semitic, etc.), I selected
the languages randomly. In many other cases, I had to be content if I found any
data at all.
The languages of the 100-language sample are listed in Table 2. A more
detailed description of the sample, including the genetic classification of the
languages is given in Appendix B, together with the data that I gathered for
these languages.
 20

Table 2. The languages of the 100-language sample

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
Abkhaz Ibibio Ntomba
Ainu Babungo Harar Oromo
Akkadian Indonesian Paez
Amele Ingrian Paiwan
Amharic Jaqaru Palauan
Atayal Kabyle Panare
Bambara Kaingang Pashto
Bashkir Kalkatungu Eastern Pomo
Basque Kanuri Romantsch
Bobo Ket Samoan
Bugis Khmer Santali
Burmese Kilivila Seychelles Creole
Burushaski Koasati Slave
Cayuga Kobon Songhai
Chinese Koromfe Upper Sorbian
Djaru Koryak Sumerian
Dogon Kwamera Sumu
Dongolawi Lak Tagalog
Dyirbal Lakhota Takelma
Ancient Egyptian Lango Takia
Ewe Logbara Telugu
Fula Malagasy Thai
Gbeya Mangarayi Tok Pisin
Goajiro Manipuri Warndarang
Gooniyandi Mapuche Welsh
W Greenlandic Margi Yagua
Guarani Maricopa Yapese
Haruai Masalit Yoruba
Hausa Mixtec Yukaghir
Hittite Khalkha Mongolian Yuwaalaraay
Hmong Njua Nahali Xinh Mul
Hua Classical Nahuatl Xhosa
Huastec Newari
Iatê Nivkh
———————————————————————————————————
————————————
 21

i
If one wants to use pronoun only in its etymological sense (‘noun-substitute’), as
suggested e.g. by Jespersen (1924: 83), one has to come up with a new cover term
comprising pro-nouns, pro-adverbs, pro-adjectives, etc. The term pro-form is sometimes
used in this sense, e.g. by Vater (1975). I go along with the tradition in ignoring the
etymology of pronoun and regarding it as a term with no internal structure, so that com-
bined terms like pronominal adverb, pronominal verb, etc., are not contradictory.
ii
I call such words mid-scalar quantifiers because they can be arranged on a scale from
maximal to minimal quantity (all – most – many – several – few – none, cf. Horn (1972:
61)), where they occupy the middle.
iii
Descriptively and/or diachronically oriented studies had appeared sporadically earlier,
e.g. Gessner (1895) (Spanish), Einenkel (1903) (English), Etzrodt (1909), Ohlhoff
(1912), Foulet (1919) (French).
iv
Note also that a typological approach that seeks to discover universal properties (like
this book) is really interested in the notion of a POSSIBLE HUMAN LANGUAGE (in
contrast to what Nichols (1992) calls 'population typology', which deals with existing
languages), and we have no way of being sure that the existing diversity at the end of
the second millennium CE is represenatative of the possible diversity. Inevitably, thus,
our 'representative samples' are only gross approximations.
 27

Formal and Functional Types of

Indefinite Pronouns

3.1. The Main Formal Types of Indefinite Pronouns

In the first section of this chapter, I give a survey of the formal types of
indefinite pronouns. We first look at the types of markers that characterize
morphologically complex indefinite pronouns, and then examine the kinds of
bases to which these markers are attached.

3.1.1. Types of indefiniteness markers

Indefinite pronouns normally occur in SERIESi which have one member for each
of the major ONTOLOGICAL CATEGORIESii such as person, thing, property, place,
time, manner, amount, plus a few others. Some examples of different indefinite
pronoun series in different languages are given in (20).

(20) a. English some-series any-series no-series

person: somebody anybody nobody
thing: something anything nothing
place: somewhere anywhere nowhere
time: sometime anytime never
manner: somehow anyhow no way
determiner: some any no

b. Polish nie-series ś -series -kolwiek-series ni-series

(‘somebody’, etc.) (‘somebody’, etc.) (‘anybody’, etc.) (‘nobody’)
person: niekto ktoś ktokolwiek nikt
thing: nieco coś cokolwiek nic
property: niejaki jakiś jakikolwiek nijaki
place: — gdzieś gdziekolwiek nigdzie
time: niekiedy kiedyś kiedykolwiek nigdy
manner: niejako jakoś jakkolwiek nijak
amount: — ileś ilekolwiek —
determiner: niektóry któryś którykolwiek z%aden

c. Lezgian sa ... jata’ni-series x̂ ajit’ani-series sa ...-ni-series

(‘somebody’, etc.) (‘anybody’, etc.) (‘nobody’, etc.)
person: sa wuž jat’ani wuž x̂ ajit’ani sa kas-ni
thing: sa wuč jat’ani wuč x̂ ajit’ani sa zat’ni
property: sa hix̂ tin jat’ani hix̂ tin x̂ ajit’ani —
place: sa hina jat’ani hina x̂ ajit’ani sana-ni
time: sa mus jat’ani mus x̂ ajit’ani sadra-ni
manner: sa hik’ jat’ani hik’ x̂ ajit’ni sak’-ni
amount: sa hiq’wan jat’ani hiq’wan x̂ ajit’ani —
 28

In the most common case, indefinite pronouns consist of (i) a stem indicating
the ontological category, plus (ii) a formal element shared by all members of an
indefinite pronoun series, such as some- and any- in English, nie-, -ś , -kolwiek
and ni- in Polish, and sa ... jat’ani, x̂ ajit’ani and sa ...-ni in Lezgian. This element
will be called INDEFINITENESS MARKERiii in the present work.
The indefiniteness marker is an affix or a particle which stands next to the
pronoun stem. It may be a prefix, like English some-, any- no-, Polish nie-, ni-, or
it may be a suffix, like Polish -ś , -kolwiek. The indefiniteness marker may consist
of two or more parts, e.g. Lezgian sa ... jat’ani, where the first part is placed
before and the second part is placed after the stem. Even indefiniteness markers
consisting of three or four parts are not unheard of, e.g. French que ce soit (in qui
que ce soit ‘anybody’, quoi que ce soit ‘anything’, etc.) or Russian by to ni bylo (in
gde by to ni bylo ‘anywhere’, kak by to ni bylo ‘anyhow’).
When the indefiniteness marker is an affix, it is usually the outermost affix of
an indefinite pronoun, i.e. it is an [Link] Case inflections are normally
inside the indefiniteness marker. This is best illustrated with a postfixal
indefiniteness marker like Latin -dam in qui-dam, or Georgian -me in ra-me.

(21) Latin (‘somebody’) (22) Georgian (‘anything’)

Nom. qui-dam Nom. ra-me
Gen. cujus-dam Dat. ras-me
Dat. cuj-dam Gen. ris(a)-me
Acc. quem-dam Instr. rit(i)-me
Abl. quo-dam Adv. rad-me

Case prefixes are rarer, but if one considers Russian 'prepositions' like s ‘with’,
u ‘at’, as case prefixes, then the extrafixal (more specifically, antefixal) nature of
the Russian prefixal indefiniteness markers ni- and koe- can be illustrated:

(23) Russian
Nom. koe-kto ‘somebody’ ni-kto ‘nobody’
u-case koe u kogo ni u kogo
s-case koe s kem ni s kem
ot-case koe ot kogo ni ot kogo

However, like all extrafixes, extrafixal indefiniteness markers show a tendency

to be rearranged. Thus, alongside forms like Georgian ra-d-me and Russian koe u
kogo, there also exist forms like ra-me-d and u koe-kogo. See further § 6.3.3 and
Haspelmath (1993b).
When the indefiniteness marker is not an affix but a particle (or a sequence
of particles), it may stand next to its derivational basis or at the margin of a
phrase. These two positions are different only with indefinite determiners,
which do not form a phrase of their own but combine with a noun phrase.
Examples (24a-c) show such discontinuous indefinites, where the basis (which
is always an interrogative determiner, ‘which?’) precedes the noun and the
indefiniteness marker follows it.

(24) a. Japanese
Yosikoo-ga dono tegami mo yom-ana-katta.
Yoshiko-NOM which letter INDEF read-NEG-PAST
‘Yoshiko didn’t read any letter.’
b. Dutch (Paardekooper 1978: 577)
Ik lees dat liever als wat voor boeken ook.
 29

I read that preferably than what kind books INDEF

‘I read that with more pleasure than any books.’
c. Lezgian (Nakh-Daghestanian; Haspelmath 1993a: 196)
Hi kac x̂ ajit’ani hajwan ja.
which cat INDEF animal is
‘Any cat is an animal.’
d. French
Tu n’es pas capable d’agir par quelque impulsion personnelle que ce soit.
‘You are not able to act by any personal impulse.’

While indefiniteness markers are sequences of segments in most cases, this is

not the only possibility. Like other grammatical morphemes, they can also be
expressed by reduplication and by stem modification.
The signaling of indefinite pronouns by means of REDUPLICATION (mainly
full reduplication) is rather common. Some examples are shown in (25) (see
further § 7.4):

(25) a. Latin
quis-quis ‘anyone’
ubi-ubi ‘anywhere’
b. Malay
apa-apa ‘something’
c. Chinese Pidgin Russian (Nichols 1986)
č ego-č ego ‘something, anything’
kakoj-kakoj ‘some, any’

Finally, the indefiniteness marker may consist in a STEM MODIFICATION. This

type is extremely rare, but I have found an example in Chechen. In this
language indefinite pronouns are formed from interrogative pronouns by
lengthening the final consonant of the stem and adding the particle a.v

(26) Chechen (Nakh-Daghestanian; Maciev 1961)

interrogative indefinite
person: Abs. mila milla a ‘someone’
Gen. Ìē nan Ìennan a ‘someone’s’
Loc. Ìänga Ìängga a ‘at someone’
thing: Abs. hun hu?a a ‘something’
Gen. stē nan stennan a ‘of something’
Loc. stenga stengga a ‘at something’
place: mič Ìa mičč aÌa a ‘somewhere’
time: maca macca a ‘sometime’
manner muxa muxxa a ‘somehow’
determiner mülxa mülxxa a ‘some’

Thus the first important generalization that we can make is that indefinite
pronouns are as a rule DERIVED forms. That this is not a trivial observation can
be seen from the artificial language Esperanto, designed by Ludwik Zamenhof
in 1887. As a rule, Esperanto grammar follows the typological design of
Standard Average European, eliminating irregularities and 'useless' features
like gender and agreement. There are two indefinite series in Esperanto, which
are clearly related to interrogative pronouns, much as in many natural
languages:
 30

(27) Esperanto interrogative indefinite negative

(‘wh-’) (‘some-’) (‘no-’)
person: kiu iu neniu
thing: kio io nenio
property: kia ia nenia
place: kie ie nenie
time: kiam iam neniam
manner: kiel iel neniel

A priori, it seems logical that the indefinite meaning should be expressed by

a maximally unmarked form, as in the Esperanto series iu ‘someone’, io
‘something’, etc. After all, indefiniteness is not a concrete, positive meaning that
can be easily described. Instead, indefiniteness seems to be very similar to the
absence of any meaning at all, and from this point of view, it makes sense to
give indefinite pronouns a maximally unmarked form.
However, natural languages that are structured like Esperanto in this respect
are virtually unattested. I am not aware of a clear case in which an indefinite
pronoun is formally unmarked with respect to a marked interrogative pronoun.
With respect to its indefinite pronoun system, Esperanto is thus probably not a
possible human language. The derived nature of indefinite pronouns is one of
the cross-linguistic observations that call for an explanation.
Let us briefly examine possible exceptions to the rule that if there is a basic-
derived relation between interrogatives and indefinites, it is always the
indefinites which are derived. In a large number of languages, indefinites and
interrogatives are formally identical (see §[Link] and §7.3). Now some
languages require a sentential interrogative marker in question sentences, and
when this interrogative marker attaches to the interrogative pronoun, it might
look as if it were derived from the unmarked indefinite pronoun.5a In the
Australian language Ngiyambaa (Donaldson 1980: 148-50),
interrogative/indefinite roots must be combined with special sentential
markers both when used as interrogatives and when used as indefinites. Thus,
minja- ‘what/something’ appears as minja-wa: ‘what?’ or as minjaN-ga:
‘something’. Thus Ngiyambaa has a system where indefinites are based not
directly on interrogatives, but on the same base as interrogatives.
Exceptions to the general rule that indefinite pronouns are derived are rare,
but occasionally we find indefinite pronouns or even whole series that are not
synchronically related to anything else, as exemplified by the cases in (28).

(28) a. German
nichts ‘nothing’
nirgends ‘nowhere’
b. Catalan
ningú ‘anybody’
res ‘anything’
enlloc ‘anywhere’
mai ‘ever’
cap ‘no’
c. Hindi/Urdu
koii ‘someone’
kuch ‘something’
kahi$ ‘somewhere’
 31

kabhii ‘sometime’

Such cases are not uncommon in the Indo-European languages of Europe, but
outside of Europe they are rare. In my 100-language sample, I have found such
indefinites only in five languages. Diachronically all the above indefinites are
known to go back to interrogatives or generic nouns plus an indefiniteness
marker (or negation).

3.1.2. The derivational bases of indefinite pronouns

In the previous subsection we saw that indefinite pronouns are generally

marked as such by different kinds of indefiniteness markers. Thus, indefinite
pronouns are generally derived forms. But what are they derived from?
There are two main types of derivational bases from which indefinite
pronouns are derived in the world’s languages: (i) INTERROGATIVE PRONOUNS,
and (ii) GENERIC ONTOLOGICAL-CATEGORY NOUNS such as ‘person’, ‘thing’,
‘place’, ‘time’, ‘manner’, etc. In addition, indefinite pronouns may be based on
the numeral 'one'.

[Link]. Interrogative-based indefinites. Most of the examples of indefinite

pronouns given in the preceding subsection illustrate interrogative-based
indefinites: all four Polish series in (20b), two of the Lezgian series in (20c), as
well as the Latin, Georgian, Russian, Japanese, Dutch, French, Malay and
Chechen cases in (21-26) above. In descriptions of indefinites in individual
languages or language families, we often find statements pointing to the
seemingly remarkable fact that indefinite pronouns are based on interrogative
pronouns (e.g. Frei (1940) on Indo-European, Beke (1913-14: 130) on Finno-
Ugrian). This fact may be surprising from the point of view of speakers of
English or French (in which languages the words for ‘somebody’ and
‘something’ are not based on interrogative pronouns), but it is in fact an ex-
tremely common phenomenon in languages across the world, as has been
shown by two studies of the Stanford universals project. According to Ultan
(1978: 230), 'indefinite words are at least in part either formally identical with or
related to question words... The only possible exceptions to this statement in
the present sample appear to be Samoan and Rotuman...' (Ultan’s sample
consists of 79 languages, distributed world-wide.) And Moravcsik (1969) states:
'The indefinite pronoun has been found to be identical or similar, in most cases,
to some question pronoun.' Moravcsik lists 26 languages (representing all
continents) which exhibit this similarity. In my sample of 100 languages, 63
languages show this similarity. It is thus not as ubiquitous as Ultan’s results
suggested, but clearly the majority of the world’s languages have interrogative-
based indefinite pronouns.
As Ultan and Moravcsik observe, indefinite pronouns are sometimes not
overtly derived from interrogative pronouns, but are identical to them. For
example, in Khmer, the shapes of indefinite pronouns (ne´q-naa ‘somebody’,
qw´y ‘something’, etc.) are identical to those of interrogative pronouns (ne´q-naa?
‘who?’, qw´y? ‘what?’, etc.), as shown by the following examples:

(29) Khmer (Huffman 1967: 153, 155)

a. Look cON tIñ qw´y?
you want buy what
‘What do you want to buy?’
 32

b. Kñom tIñ qw´ y ha´y, trOlOp tow pte´h.

I buy what:INDEF PRF return go house
‘After I buy something, I’m going back home.’

In such cases we can say that indefinite pronouns are zero-derived (derived by
conversion) from interrogative pronouns. Here one might object that the
direction of derivation could just as well be the opposite: Interrogative pro-
nouns could be zero-derived from indefinite pronouns. However, this is
unlikely because indefinite pronouns are usually expressed by more
morphemes than, and usually derived from, interrogative pronouns. Hence,
typologically indefinites are a marked category relative to interrogatives.
Moravcsik (1969: 77) states this in even stronger terms: 'If [the indefinite
pronoun and the question pronoun] are partially similar it is the indefinite
pronoun whose shape includes morphologically that of the question pronoun.'
Thus, there is a universal asymmetric markedness relation such that indefinite
pronouns are usually more marked than and derived from interrogative
pronouns. This asymmetry is sometimes neutralized (in languages like Khmer),
but it is virtually never reversed. The semantic and diachronic relationship
between indefinites and interrogatives will be further discussed below (§ 7.3).

[Link]. Generic-noun-based indefinites. The other main type of derivational

basis, general ontological-category nouns, can be exemplified by the Persian
indefinite series in (30).

(30) Persian
kas-i ‘someone’ lit. ‘a person’
č iz-i ‘something’ ‘a thing’
(dær) yek ǰ â-yi ‘somewhere’ ‘(at) one place’
yek vaVt-i ‘sometime’ ‘one time’
yek towr-i ‘somehow’ ‘one manner’

These indefinite pronouns are based on the general nouns kas ‘person’, č iz
‘thing’, ǰ â ‘place’, etc. The indefiniteness marker -(y)i also occurs in ordinary
indefinite noun phrases (e.g. ketâb-i ‘a book’), so it seems that there is nothing
special about the expressions in (30): They are perhaps not indefinite pronouns
at all, but ordinary indefinite noun phrases that Persian uses to express what
many other languages express by means of indefinite pronouns.
It may well be that this is the correct analysis for some languages (cf. the
discussion in § 3.3). However, there are many cases where combinations of a
generic ontological-category noun plus an indefiniteness marker have been
grammaticalized sufficiently to qualify as real pronouns. A good example is
provided by English somebody, something, sometime, someplace, anybody, etc.
These forms are clearly distinct from the indefinite NPs some body, some thing,
some time, etc., both semantically and formally, and thus show that indefinite
pronouns can be derived from general ontological-category [Link]
Another example is provided by Hebrew iš , discussed in Glinert (1982: 460-
61). This is both a noun (‘man’) and an indefinite pronoun (‘anyone’), so one
might ask whether these are really two different items (with iš ‘anyone, no one’
derived from iš ‘man’) The answer is yes: That iš ‘anyone, no one’ is a separate
lexical item is shown by three syntactic facts: (i) iš ‘anyone, no one’ is restricted
to negative-polarity contexts (cf. § 3.2.2), as shown in (31a-b); (ii) iš ‘anyone, no
one’ allows a partitive phrase, like other pronouns, cf. (32); and (iii) iš ‘anyone,
 33

no one’ cannot be modified by an adjective, cf. (33). In all these respects, iš
‘man’ behaves differently.

Hebrew (Glinert 1982: 461)

(31) a. Im yesapru zot le-iš ,...
if they:tell that to-man
‘If they tell it to anyone/to a man,...’
b. Sipru zot le-iš .
they:tell that to-man
‘They will tell it to a man <*to anyone>.’
(32) Iš mehem lo heš iv.
anyone of:them NEG replied
‘None of them replied.’ <*A man of them didn't reply.>
(33) Lo haya š am iš meyuÌad.
NEG was there man special
‘A special man was not there.’ <*No one special was there.>

These syntactic differences show that the pronoun iš is a different lexical item
and can be regarded as derived (at least diachronically) from the general noun
iš ‘man’.
In my 100-language sample, I registered 42 languages with generic-noun-
based indefinites. It may be that many of these cases are not really indefinite
pronouns, but represent languages that lack indefinite pronouns (cf. § 3.3). But
my suspicion is that subtle differences like those cited above can be found in
many of these languages. See § 7.5.1 below for further discussion of the use of
generic nouns as indefinite pronouns.

[Link]. ‘One’-based indefinites. The two types of derivational bases discussed

above are by far the most important ones for the overwhelming majority of
indefinite pronouns in the languages that I have surveyed. However, a minor
third type should also be mentioned: the numeral ‘one’.
Most often this numeral is used in the sense ‘somebody’, and it is not un-
common in languages where indefinite pronouns are not based on inter-
rogatives. For example, in Egyptian Arabic, ‘someone’ is waaÌid, and English
someone is of course also a case in point. Since the numeral ‘one’ is usually the
source for the grammaticalization of indefinite articles (cf. e.g. Givón 1981), it is
hardly surprising that it should be used as the basis of an indefinite pronoun as
well. Why it should consistently have the meaning ‘someone’ rather than
‘something’ is less obvious, but this must have to do with the unique
importance of individuality for people. Very rare are cases like Lezgian (cf.
(20c) above) where the numeral ‘one’ (sa) is also the basis of adverbial indefinite
pronouns such as sana ‘somewhere’, sak’(a) ‘somehow’. See below § 7.5.2 for
further discussion of the role of ‘one’ in indefinite pronouns.

3.1.3. The ontological categories

It is an interesting question why indefinite pronoun series should consist

precisely of words for ontological categories such as ‘person’, ‘thing’,
‘property’, ‘place’, ‘time’, ‘manner’, ‘amount’, ‘reason’. For example, why are
more concrete features like color and size or more abstract notions like con-
cession systematically absent from lists of indefinite pronouns? Jackendoff
1983:48-56 argues that such ontological categories reflect the major constituents
of conceptual structure. Still, one might want to know why conceptual structure
 34

is made up precisely of these constituents (see Heine et al. (1991: 55-58) for some
recent discussion).
Fascinating though this question is, I will not pursue it further here because
it is not specific to indefinite pronouns. Independently of genetic and areal
affiliation, languages express roughly the same ontological categories as
interrogative and demonstrative pronouns, as well as several more specific
pronoun types (e.g. relative pronouns) that are not universal. Since the different
types of pronouns generally show parallels in the way they express the various
ontological categories, traditional grammars of Indo-European languages often
contain a table of 'correlative pronouns' (including pronominal adverbs), e.g.
Table 3 for Classical Greek.

Table 3. 'Correlative pronouns' in Classical Greek (somewhat simplified)

————————————————————————————————
ontological interrogative demonstrative relative indefinite
category pronouns pronouns pronouns pronouns

person tís hou$tos hós tis

‘who?’ ‘this’ ‘which’ ‘someone’
thing tí – – ti
‘what?’ ‘something’
property poi$os toiósde hoi$os poiós
‘what kind?’ ‘this kind’ ‘which kind’ ‘some kind’
place pou$ ekei$ hou$ pou
‘where?’ ‘there’ ‘where’ ‘somewhere’
time póte tóte hóte poté
‘when?’ ‘then’ ‘when’ ‘sometime’
manner põs hoútō s hō s pō s
‘how?’ ‘like this’ ‘as’ ‘somehow’
amount pósos tosósde hósos posós
‘how much?’ ‘this much’ ‘how much’ ‘some amount’
————————————————————————————————
Such an organization of the various types of pronouns is not only typical of
Indo-European languages but can be observed in languages of different types
(cf. the analogous table for Lezgian in Haspelmath (1993a: 188)). The pronoun
types are characterized by their stems (Greek ti-/po-: interrogative/indefinite,
(hou-)to-: demonstrative, ho-: relative) and show similar affixes for the various
ontological categories (in Greek, -ou for place, -ote for time, -ō s for manner,
etc.). There are various idiosyncrasies, as we expect for derivational
morphology.
The seven ontological categories represented in Table 3 are the categories
that are most often expressed by simple means in the languages of the world.
Of course, it is always possible to create more specific complex expressions, e.g
for color (interrogative what color?, indefinite of some color, etc.) or concession
(interrogative: despite what?, indefinite despite something, etc.). This is what
languages do which have fewer than these seven most common categories, e.g.
English, where amount and property have to be expressed by a circumlocution
(how much, what kind).
The distinction between human and non-human referents is made prac-
tically everywhere (‘who’ vs. ‘what’, ‘somebody’ vs. ‘something’), even in
languages where humanness is not very prominent elsewhere in the grammar.
Languages that assign a prominent role to other gender distinctions in their
grammar (e.g. sex-based gender) nevertheless usually make the human/non-
 35

human distinction here, e.g. Classical Greek. Languages like Gothic which have
a gender distinction here (hwas ‘who (masc.)’, hwo ‘who (fem.)’, hwat ‘what’) are
rare, as are languages like Lithuanian which lack a human/non-human
distinction (kas ‘who, what’).
The standard series of seven categories is most often enriched by the addi-
tion of further distinctions in the place category. Many languages have simple
expressions for the directional notions ‘whither’ and ‘whence’, and sometimes
also for the path notion ‘along where’ (e.g Polish te4dy ‘this way’). Furthermore,
many languages have a determiner (‘which’) that is different from both the
substantival ‘who’/‘what’ and the adjectival ‘what kind’. Another notion that is
sometimes expressed by simple means is cause or reason (‘why’).
In general, there do not seem to be great differences between different types
of pronouns with respect to which ontological categories they have. True, there
are sometimes specific gaps for indefinite pronouns: For example, while
English and German have simple cause interrogatives, why and warum, both
languages lack the corresponding indefinites (*somewhy, *irgendwarum).
However, Russian does have a regularly formed pronoun that expresses
precisely this idea: poč emu-to ‘for some reason’, from poč emu ‘why’. Thus, the
English and German gaps may well be accidental.
These few remarks on ontological categories should suffice. We now turn to
a first overview of the functional distinctions expressed by indefinite pronouns,
a topic of central concern to this work.

3.2. The Main Functional Types of Indefinite Pronouns

This section gives an informal description and exemplification of the most
important functional distinctions that are expressed by different series of in-
definite pronouns. Most of these distinctions are discussed again in later parts
of this work.

3.2.1. Negation: negative indefinite pronouns

Many languages have special indefinite pronouns that are only used in
negative sentences where the scope of negation extends over the indefinite. For
example, Latin nemo, English nothing, Bulgarian nikoj, and Lezgian sada-ni al-
ways express negation. Some of these indefinites contain a formally negative
formative synchronically (Bulgarian ni-) or at least diachronically (English no-,
Latin ne-), but many do not.

(34) a. Latin
Nemo venit.
nobody came
‘Nobody came.’
b. English
I noticed nothing.
c. Bulgarian
Ni-koj ne se obadi.
NEG-who NEG REFL phoned
‘Nobody phoned.’
d. Lezgian
Sada-ni za-z am žǧ a-na-č .
[Link]-even I-DAT it:ABS find-AOR-NEG
 36

‘I did not find it anywhere.’

Notice that in some languages, an indefinite pronoun by itself is sufficient to

express sentential negation, e.g. in Latin and English, while in others the
negative pronouns cooccur with verbal negation, e.g. in Bulgarian and Lezgian.
Other languages do not have any special negative indefinites, but employ in-
definites that are also used for other, non-negative functions. For instance,
Swedish någon (ex. 35) and Bambara fEn (ex. 36) are not interpreted negatively
when used in questions. In some cases, a sentence may even be ambiguous,
interpretable as negative or non-negative, as exemplified by Hindi/Urdu (37).

(35) Swedish
a. Fick du någon?
got you some
‘Did you get any?’
b. Jag har inte sett någon av dem.
I have not seen any of them
‘I have not seen any of them.’
(36) Bambara (V.F. Vydrin, p.c.)
a. í yé fE n yé wà?
you PERF thing see Q
‘Did you see anything?’
b. n tE fE n fO
I NEG:FUT thing say
‘I will not say anything.’
(37) Hindi/Urdu
a. Aaj koii aayaa.
[Link] someone came
‘Today somebody came.’
b. Aaj koii nahi$ aayaa.
today someone NEG came
‘Today somebody didn’t come.’ Or: ‘Today nobody came.’

An important subclassification of negation concerns the syntactic position of

the negative element. In the cases that we have seen so far in this section, the
negated indefinite pronoun is an argument of the negated clause. But the
negated indefinite pronoun may also be in a subordinate clause, in particular in
situations of negative raising (i.e. when the negation in the superordinate clause
logically belongs to the subordinate clause). In such situations of
SUPERORDINATE NEGATION, some languages use the same indefinite series as in
direct negation (cf. 38-39), but others require a different series (40-41).

(38) Italian
a. Non è venuto nessuno.
not has come anybody
‘Nobody has come.’
b. Non è necessario che venga nessuno.
not is necessary that come anybody
‘It is not necessary that anybody come.’
(39) Basque
a. Ez naiz i-noiz etorri-ko.
NEG I:AUX INDEF-when come-FUT
‘I will never come.’
 37

b. Ez dut uste gizon hori i-noiz ikusi dud-anik.

NEG I:it:AUX believe man that INDEF-when seen I:him:AUX-COMP
‘I don’t believe that I’ve ever seen that man.’
(40) German
a. Niemand ist gekommen.
nobody is come
‘Nobody came.’
b. Es ist nicht nötig, dass jemand kommt.
it is not necessary that someone comes
‘It is not necessary that anybody come.’
(41) Hungarian
a. Nem lát-t-am sem-mi-t.
NEGsee-PAST-1SG NEG-what-ACC
‘I did not see anything.’
b. Nem hisz-em, hogy vala-ki lát-t-a volna.
NEGthink-1SG that INDEF-who see-PAST-3SG SBJV
‘I don’t think that anybody has seen it.’

Many languages also require special indefinites with implicitly negative ex-
pressions like the preposition ‘without’ or the verbs 'deny' and 'refuse', e.g.
Russian nikakaja pomošč ‘no help’, but bez kakoj-libo pomošč i ‘without any
help’ (*bez nikakoj pomošč i). Indefinites in such contexts of implicitly negative
expressions and indefinites in the context of superordinate negation are usually
the same, so they are treated together as INDIRECT NEGATION here. Van der
Wouden (1994: ch. 1) shows that superordinate negation and implicit negation
share the semantic property of anti-additivity (a Boolean property). Thus, the
function of indirect negation as defined here is ultimately based on meaning.
Topics related to indefinite pronouns and negation will be treated in detail in
chapter 8.

3.2.2. Negative polarity (or scale reversal)

Some series of indefinite pronouns in some languages are also associated with
negative environments (like the negative pronouns in the previous section), but
are not restricted to the expression of non-existence. In addition to negative
clauses, they can also be used in conditional and interrogative clauses, in the
standard of comparison, and in some further environments. Expressions that
show this distribution are generally called NEGATIVE POLARITY [Link]
One type of negative polarity items are lexical minimal-unit expressions of
the type (lift) a finger, (see) a living soul, which probably exist in all languages.
But some languages also have indefinite series that show the same properties.
The best-known negative-polarity indefinite is of course the English any-series,
but similar indefinites can be found in many other languages (cf. Schmid 1980:
ch. 3).
In the following, I will exemplify the various conditions under which
negative polarity items are licensed mainly on the basis of examples from
English (see Linebarger (1981, 1987), von Bergen and von Bergen (1993) for
detailed studies of negative polarity in English). The (a) examples show
minimal-unit negative-polarity items, and the (b) examples show indefinite
pronouns.
In declarative affirmative sentences, negative-polarity items are ill-formed:

(42) a. *He lifted a finger to help her.

 38

b. *He did anything to help her.

The central context where negative polarity items are licensed is direct (or
clausemate) negation.

(43) a. He didn’t lift a finger to help her.

b. He didn’t do anything to help her.

Next, negative-polarity items are licensed in (especially polar) questions (i.e.

interrogative sentences),

(44) a. Would she lift a finger for you?

b. Have you heard anything new about the ozone hole?

and in the protasis of a conditional sentence:

(45) a. If you budge an inch, I’ll hit you.

b. If you tell anybody, we’ll punish you.

Both questions and conditional protases are compatible with the truth of the
corresponding negative clause, so might conceivably be interpreted as natural
extensions of the prototypical licensing context (cf. Progovac 1994).
Negative-polarity items may also occur in quite a few non-negative contexts
that only carry a negative implication od some kind. These include negative
quantifiers like ‘few’,

(46) a. Few people give a damn about the rain forest.

b. Few people show any interest in global issues.

the restrictive focus particle ‘only’,

(47) a. Only Mira has a hope in hell of passing.

b. Only Luozhu has anything substantial to report.

the purposive complement of an excessive comparative (‘too’),

(48) a. Khadija is too tired to give a damn.

b. Ayse is too occupied to invite anyone else.

complements of implicitly negative verbs like ‘doubt’,

(49) a. I doubt that Ronald gave a red cent to charity.

b. We doubt that Juan applied anywhere.

and arguments of implicitly negative adpositions like ‘without’:

(50) a. Noriko is getting by without lifting a finger.

b. Hannu is getting by without doing anything.

A context where the connection to negation is less obvious is the standard of

comparison of inequality or equality:
 39

(51) a. Cows fly more often than John lifts a finger to help Louise.
b. The boy can run faster than anyone in his class.

(52) The boy runs as fast as anyone in his class.

Finally, negative-polarity items are licensed in relative clauses headed by a

universal quantifier:

(53) a. Everyone who knows a damn thing about Dyirbal knows that it’s
syntactically ergative.
b. Everyone who likes any kind of seafood will like our new seafood sticks.

To give some examples of negative-polarity indefinites from a less well-

known language, consider the Catalan indefinite series ningú ‘anybody’, res
‘anything’, cap ‘any’, mai ‘ever’ etc.:

(54) Catalan (Lleó 1983)

a. (direct negation) No et demano res.
not you I:ask anything
‘I do not ask you anything.’
b. (interrogative) Hi ha res de nou?
there has anything of new
‘Is there anything new?’
c. (conditional) Si el veus mai,...
if him you:see ever
‘If you ever see him,...’
d. (implicit negation) Sense dir res, va menjar la sopa.
without say anything did eat the soup
‘Without saying anything, she ate the soup.’

However, in-depth research on negative polarity has shown that the various
contexts that are usually regarded as negative-polarity contexts behave by no
means uniformly (see especially van der Wouden 1994). The most important
subtypes of negative-polarity contexts are questions, conditionals, the standard
of comparison, indirect negation, and direct negation.
We already saw in the preceding section that direct negation and indirect
negation are two functions that must be distinguished. In addition to
indefinites that can only be used in the scope of (direct or indirect) negation,
there are also indefinites that are used in the other negative-polarity contexts
but cannot be used in the contexts of direct negation. Such a case is the Russian
-libo-series:

(55) Ja ne videla *kogo-libo/ ni-kogo.

I not saw whom-INDEF/ INDEF-whom
‘I didn’t see anyone.’
(56) a. Znajut li oni č to-libo ob ozonnoj dyre?
know Q they what-INDEF about ozone hole
‘Do they know anything about the ozone hole?’
b. Esli ty skaž eš ’ komu-libo, my tebja nakaž em.
if you tell whom-INDEF we you punish
‘If you tell anyone, we’ll punish you.’
d. Malo kto pokazyvaet kakoj-libo interes v ètom.
few who shows which-INDEF interest in this
 40

‘Few people show any interest in it.’

Indefinite pronouns very often behave in the same way in questions and
conditionals, but there are also indefinites that are restricted to questions and
cannot be used in conditionals, e.g. the Persian hič -series (cf. 57), and
indefinites that are restricted to conditionals and cannot be used in questions,
e.g. the Modern Greek -dhípote-series (cf. 58).

(57) Persian
a. Âyâ hič č iz-imi-š anav-i?
Q INDEF thing-INDEF IMPF-hear-2SG
‘Can you hear anything?’
b. *Agar hič č iz-i š anid-i, be man begu.
if INDEF thing-INDEF hear-2SG to me tell:IMPV
‘If you hear anything, tell me.’

(58) Modern Greek

a. An dhis oti-dhípote, pés mu.
if you:see what-INDEF tell:IMPV me
‘If you see anything (at all), tell me.’
b. *Ídhes oti-dhípote?
you:saw what-INDEF
‘Did you see anything (at all)?’

The standard of comparison stands apart from questions and conditionals in

that very often an indefinite series that can be used in questions and
conditionals is inadmissible in the standard of comparison, e.g. etwas in
German (cf. 59). Conversely, indefinites that occur in the comparative context
may not be possible in conditionals and questions, e.g. the Romanian ori-series
(cf. 60).

(59) German
a. Hast du etwas gesehen?
Did you see anything?
b. *Das steht dir besser als etwas [Link]
'This suits you better than anything else.'

(60) Romanian (Beatrice Primus, p.c.)

a. Niculina fuge mai repede decît ori-ce fata* din clasa*.
Niculina runs more fast than INDEF-what girl in class
'Niculina runs faster than any girl in her class.'
b. *Daca* auzi ori-ce, trezes3te-ma*.
if hear:2SG INDEF-what wake-me
'If you hear anything, wake me up.'

3.2.3. Specificity and non-specificity

In some languages, different indefinite series are used depending on whether

the indefinite NP is SPECIFIC or NON-SPECIFIC. The concept of specificity is a key
concept in the semantics of reference and has been discussed extensively in the
literature (cf., among many others, Seliverstova 1964, Fodor 1970, Jackendoff
1972a: ch. 7, Ioup 1977, Croft 1983, Galmiche 1983, Fauconnier 1985, Paduč eva
1985, Lavric 1990, Enç 1991). There is no universal agreement on what
 41

phenomena fall under this concept, but there seems to be a broad consensus
that it is best illustrated by a sentence such as (61). In this sentence, the NP a
native speaker of Ainu is ambiguous between a specific reading and a non-
specific reading. On the specific reading, (61) could be continued by (62a), and
on the non-specific reading it could be continued by (62b).

(61) Nobuko wants to marry a native speaker of Ainu.

(62) a. (specific) ...She fell in love with him during fieldwork sessions.
b. (non-specific) ...because she is Ainu herself, and she wants her
children to acquire her ancestors’ language.

In the specific reading, the existence of a uniquely identified Ainu speaker is

presupposed, whereas in the non-specific reading, the identity of the Ainu
speaker has not been established. I will say (preliminarily) that an expression is
specific if the speaker presupposes the existence and unique identifiability of its
referent.
Several overt correlates of the specific/non-specific distinction have been
cited in the literature, especially (i) 'discourse referents', (ii) paraphrasability by
an existential sentence, and (iii) disambiguation by specific determiners.
(i) Only a specific NP can have a 'discourse referent', i.e. can be referred to by
an anaphoric pronoun in a present indicative clause (Karttunen 1976: 366).

(63) a. Cheolbai bought a bicycle [specific]. It is black.

b. Cheolbai wants to buy a bicycle [non-specific]. *It is [Link]

(ii) A paraphrase with an existential sentence is only possible with a specific

NP (Heringer 1969: 90).

(64) a. (specific) There is a native speaker of Ainu who Nobuko wants to

marry. She fell in love with him during fieldwork sessions.
b. (non-specific) *There is a native speaker of Ainu who Nobuko
wants to marry because she is Ainu herself...

(iii) There are determiner-like expressions in some languages which seem to

force a specific reading, especially English a certain and its equivalents in other
languages (e.g. German ein bestimmter, Dutch zeker, Russian opredelennyj, etc.)

(65) English
(only specific) Nobuko wants to marry a certain native speaker of Ainu.
(66) German
(only specific) Murat möchte ein bestimmtes Gemälde kaufen.
Murat wants a certain painting buy
‘Murat wants to buy a certain picture.’

These three tests positively identify only specific NPs. However, once we
bring indefinite pronouns into the picture, we also get a positive criterion for
non-specific NPs. It is not uncommon for languages to have two different
indefinite series for specific and non-specific NPs. I will give some examples
from Russian, Lithuanian, Modern Greek, Georgian, and Kannada. The specific
and non-specific indefiniteness markers are shown in (67). (The interrogative
pronoun on which the indefinites are based is symbolyzed by 'WH'.)
 42

(67) specific non-specific

Russian WH-to WH-nibud’
Lithuanian kaž -WH WH nors
Modern Greek kápjos etc. kanénas etc.
Georgian WH-Vac WH-me
Kannada WH-oo WH-aadaruu

The following sentences illustrate the prototypical context with ‘want’:

(68) Russian (Paduč eva 1985: 211)

a. Ivan xoč et spet’ kakoj-to romans.
Ivan wants sing which-INDEF romance
‘Ivan wants to sing some [specific] romance.’
b. Ivan xoč et spet’ kakoj-nibud’ romans.
Ivan wants sing which-INDEF romance
‘Ivan wants to sing some [non-specific] romance.’
(69) Lithuanian (Pilka 1984: 57)
a. Ji nore%jo i4sigyti kaž -kokia4 preke4 (bet jos negavo).
she wanted acquire INDEF-which thing but it not:got
‘She wanted to acquire some [specific] object (but she didn’t get it).’
b. Ji nore%jo i4sigyti kokia4 nors preke4 (*bet jos negavo).
she wanted acquire which INDEF thing but it not:got
‘She wanted to acquire some [non-specific] object
(*but she didn’t get it).’
(70) Modern Greek
a. Théli na pandreftí me kápja Elinídha.
wants SBJV he:marry with some Greek
‘He wants to marry some [specific] Greek woman.’
b. Théli na pandreftí me kamjá Elinídha.
wants SBJV he:marry with any Greek
‘He wants to marry some [non-specific] Greek woman.’

The specific/non-specific contrast is relevant in many other contexts. In

some contexts, only specific NPs are possible. Such is the case, in particular, in
affirmative declarative sentences in the perfective past or in the ongoing
present. In such prototypical realis sentences, the speaker is committed to the
existence and identifiability of the entity, and indefinites of the non-specific
series are simply unacceptable.

(71) Lithuanian (Pilka 1984: 29, 56)

a. *Kas nors ate%jo. (OK: Kaž -kas ate%jo.)
who INDEF came INDEF-who came
‘Somebody came.’
b. *Ž iū re%-k,kas nors be%ga. (OK: Kaž -kas be%ga.)
look-IMPV who INDEF runs INDEF-who runs
‘Look, somebody is running.’
(72) Modern Greek
*Kanénas írthe. (OK: Ká-pjos írthe.)
anyone came INDEF-who came
‘Somebody came.’
(73) Kannada (Bhat 1981: 7)
*Yaar-aadaruu bandaru. (OK: Yaar-oo bandaru.)
who-INDEF came who-INDEF came
 43

‘Someone came.’

The same ambiguity as in ‘want’ contexts is observed in various types of

other IRREALIS contexts (cf. Croft (1983) for this use of the term irrealis). One
such irrealis context is in future sentences. Since the event is in the future and is
not realized yet, the speaker is not committed to the existence of the referent.

(74) Russian
a. V subbotu oni uedut kuda-to.
in Saturday they go whither-INDEF
‘On Saturday they will go somewhere [specific].’
b. V subbotu oni uedut kuda-nibud’.
in Saturday they go whither-INDEF
‘On Saturday they will go somewhere [non-specific].’

Other irrealis contexts are created by various kinds of non-indicative

modality, e.g. the verb ‘can’.

(75) Lithuanian (Pilka 1984: 29)

a. Gali kaž -kas ateiti.
can INDEF-who come
‘Someone [specific] can come.’
b. Gali kas nors ateiti.
can who INDEF come
‘Someone [non-specific] can come.’
(76) Kannada (Bhat 1981: 4)
Raamu ellig-aadaruu hoodaanu.
Ramu whither-INDEF go:may
‘Ramu may go somewhere [non-specific].’

Non-specific NPs are also admitted in a perfective past and ongoing present
sentence when the sentence is qualified by some indication of epistemic
modality.

(77) Lithuanian (Pilka 1984: 50)

Matyt, kas nors juo buspasiskunde4s.
apparently who INDEF her was complaining
‘Apparently someone [non-specific] complained about her.’
(78) Russian (Kobozeva 1981)
Kaž etsja, kto-nibud’ približ aetsja.
seems who-INDEF approaches
‘Apparently someone [non-specific] is approaching.’

Although these sentences are set in the past and present, respectively, the
speaker’s uncertainty makes them qualify as irrealis, and indefinites may be
non-specific.
In addition to irrealis contexts, ambiguity can also be found in distributive
contexts, for example in the indefinite object NP when the subject is marked as
distributive-key:

(79) Lithuanian (Pilka 1984: 127)

a. Visi kaž -ka4 skaite%.
all INDEF-what read
‘Everybody is reading something [specific].’
 44

b. Visi ka4 nors skaite%.

all what INDEF read
‘Everybody is reading something [non-specific].’

The difference here is that in (79a), everybody reads the same thing, whereas in
(79b), different people may be reading different things. The event in (79b) is
taking place at the moment of speech, so the speaker must presuppose the
existence of the reading material, but since it is distributed over the referents of
the subject NP (‘everybody’), it is not a UNIQUE object and therefore also
qualifies as non-specific.
In another type of distributive context, the indefinite NP is distributed over a
plural event rather than a plural argument, e.g. a habitual event (cf.
Giannakidou 1994c).

(80) Russian
a. Po subbotam sjuda priezž aet kto-to iz Derbenta.
on Saturdays hither comes who-INDEF from Derbent
‘On Saturdays someone [specific] from Derbent comes here.’
b. Po subbotam sjuda priezž aet kto-nibud’ iz Derbenta.
on Saturdays hither comes who-INDEFfrom Derbent
‘On Saturdays someone [non-specific] from Derbent comes here.’
(81) Modern Greek (Veloudis 1982: 181)
Érxete póte-póte kanís ipopsífios.
comes when-when any candidate
‘Now and then a candidate comes.’

So far we have seen contexts where only specific indefinites are allowed
(realis contexts) and contexts where either specific or non-specific indefinites
may occur (irrealis contexts). In addition, there are also contexts where only
non-specific indefinites may occur. These are irrealis contexts where the specific
meaning is excluded for pragmatic reasons.
First of all, let us consider imperatives. Imperatives describe an unrealized
event, hence non-specific indefinites are admissible, as in (82-85) (a). In this
context, specific indefinites do not have another meaning (as in the other
irrealis contexts), but are excluded, as shown by (82-85) (b).

(82) Lithuanian (Pilka 1984: 56)

a. Aplanky-kite mana kada nors.
visit-IMPV.2PL me when INDEF
‘Visit me sometime [non-specific].’
b. *Aplanky-kite mana kaž -kada.
visit-IMPV.2PL me INDEF-when
‘Visit me sometime [specific].’
(83) Russian
a. Kupi mne kakuju-nibud’ gazetu.
buy me which-INDEF paper
‘Buy me some [non-specific] newspaper.’
b. *Kupi mne kakuju-to gazetu.
buy me which-INDEF paper
‘Buy me some [specific] newspaper.’
(84) Modern Greek (Dhelverúdhi 1989: 413)
a. Píjene ke vres kanéna náfti.
go and find any sailor
‘Go and find some [non-specific] sailor.’
 45

b. *Píjene ke vres kápjo náfti.

go and find some sailor
‘Go and find some [specific] sailor.’
(85) Kannada (Bhat 1981: 6)
a. Ellig-aadaruu hoogu.
whither-INDEF go:IMPV
‘Go somewhere [non-specific].’
b. *Ellig-oo hoogu.
whither-INDEF go:IMPV
‘Go somewhere [specific].’

Specific indefinites in imperatives are inadmissible because they would violate

Grice’s cooperative principle (this was observed in Kobozeva (1981) and Croft
(1983)). On the one hand, the speaker asks the hearer to do something, but on
the other hand he or she withholds some crucial information from the hearer.
For instance, by uttering a sentence like (83b), the speaker would make it clear
that he presupposes a unique newspaper that the hearer should buy, but does
not say which one. Thus, the hearer cannot be expected to be able to fulfill the
request. By contrast, if the speaker uses a non-specific NP (kakuju-nibud’ gazetu
in (83a)), he thereby signals to the hearer that the identity of the newspaper is
immaterial. It is left to the hearer to select a newspaper, and he is therefore in a
position to fulfill the request.
A very similar account can be given for questions. In questions, too, in-
definite phrases can only be non-specific. Some examples are given in (86-89).

(86) Russian (Paduč eva 1985: 212)

a. Uvidel li vy kogo-nibud’?
saw Q you whom-INDEF
‘Did you see anyone?’
b. *Uvideli li vy kogo-to?
saw Q you whom-INDEF
(87) Kannada (Bhat 1981: 6)
a. Raamu ellig-aadaruu hoodan-oo?
Ramu whither-INDEF go:he:did-Q
‘Did Ramu go anywhere [non-specific]?’
b. *Raamu ellig-oo hoodan-oo?
Ramu whither-INDEF go:he:did-Q
‘Did Ramu go somewhere [specific]?’
(88) Lithuanian (Pilka 1984: 58)
Tu skaitei ka4 nors apie maju4 kultura4?
you read what INDEF on Mayas’ culture
‘Did you read anything [non-specific] about the culture of the Mayas?’
(89) Modern Greek (Dhelverúdhi 1989: 411)
Sinándises kanénan ekí pu píjes?
[Link] anyone there where [Link]
‘Did you meet anyone [non-specific] at the place where you went?’

Pragmatically, questions are closely related to imperatives: They are requests

by the speaker to the hearer to supply missing information. By using a specific
indefinite phrase in a question, the hearer would again withhold some crucial
information from the hearer, thereby violating Grice’s cooperative principle.
Questions in many ways behave similarly to conditional protases. Like
questions, conditional protases admit only non-specific indefinite phrases, no
specific indefinite phrases (cf. Croft 1983). Examples:
 46

(90) Kannada (Bhat 1981: 7)

a. Yaar-aadaruu bandare heel5utteene.
who-INDEF come:COND tell:FUT:1SG
‘I will tell you if anyone [non-specific] comes.’
b. *Yaar-oo bandare heel5utteene.
who-INDEF come:COND tell:FUT:1SG
‘I will tell you if anyone [specific] comes.’
(91) Modern Greek (Dhelverúdhi 1989: 413)
An me dhí kanís pu dhe me kséri,...
if me sees anyone that not me knows
‘If anyone [non-specific] sees me who doesn’t know me...’

The explanation for this fact is less straightforward, but it seems possible to
extend the account given for imperatives and questions if we recall that a
conditional can be thought of as an instruction to the hearer: 'Suppose someone
sees me who doesn’t know me...'. Again, we have the basic imperative structure
that leads to pragmatic unacceptability when it contains a specific indefinite
phrase.
Now that we have seen questions and conditional protases among the
contexts for non-specific indefinites, I should point out that there is an overlap
with negative-polarity contexts here: Questions and conditionals are also
negative-polarity contexts (see the previous section). This fact has received little
or no attention in the literature, but it is important in the present context. As we
will see in chapter 4, there are quite a few languages in which the same
indefinite pronoun is used both in irrealis non-specific environments (i.e.
imperatives, future, modality, ‘want’, etc.) and in negative-polarity
environments.
The overlap between non-specificity and negative-polarity contexts extends
further to the core cases of negative polarity: Any indefinite phrase that is in the
scope of negation is non-specific. However, unlike questions and conditionals,
negative sentences also allow specific indefinite phrases. Consider the examples
in (92-93).

(92) (non-specific) Niklas couldn’t see anything. (*Everyone else saw it.)
(93) (specific) Niklas didn’t see something. (Everyone else saw it,
only Niklas missed it)

A specific indefinite pronoun in a negative sentence (as in 93) is a rare

phenomenon (cf. Givón 1978), but it is by no means impossible. The context
provided in (93) shows that such sentences are plausible in appropriate cir-
cumstances. In (93), the indefinite phrase is not in the scope of negation, as can
be shown by the paraphrase in (94).

(94) There was something that Niklas didn’t see.

When an indefinite phrase is in the scope of negation, it is necessarily non-

specific. Some languages use their usual non-specific indefinites in negative
contexts as well, e.g. Greek.

(95) Modern Greek

a. Direct negation
Dhen ídha kanéna.
 47

not I:saw anyone

‘I didn’t see anybody.’
b. Implicit negation
Xorís na dhó kanéna...
without SBJV I:see anyone
‘Without seeing anybody...’
c. Superordinate negation
Dhen pistévo pos kanénas írthe.
not I:believe that anyone came
‘I don’t think that anyone came.’

However, many other languages use special indefinites in negative contexts,

either special negative pronouns (§ 3.2.1), or other specialized indefinites. Thus,
Russian -nibud’-indefinites cannot be used in sentences like (95a-c), and
Lithuanian nors-indefinites cannot be used for direct negation as in (95a).
We can now summarize the contexts where specific and non-specific in-
definite phrases are allowed as in Figure 2. In one type of contexts, only specific
indefinites are possible, in another type both are possible but there is a meaning
difference, and in a third type, only non-specific indefinite phrases are allowed.
Negative polarity overlaps only partially with this third type, because
imperatives are not negative-polarity contexts.

Figure 2. Contexts for (non-)specific phrases and negative polarity

perfective ‘want’, imperative question, in the

past, future, conditional scope of
ongoing distributive negation
present
specific possible (specific impossible)

(non-specific non-specific possible

impossible)
no negative polarity negative polarity

Based on these semantic distinctions, we can already identify one important

parameter of cross-linguistic variation: In some languages, different indefinite
series distinguish between negative polarity and no negative polarity (e.g.
English, Catalan, § 3.2.2.), in other languages different indefinite series
distinguish between specificity and non-specificity (e.g. Russian, Modern
Greek, Kannada).

3.2.4. Knowledge of the speaker

Another semantic factor that is sometimes relevant in choosing between

different indefinite series is the knowledge of the speaker: The speaker may or
may not be able to identify the referent of the indefinite pronoun. This
classification applies only to specific phrases, in which the identifiability of
their referents is presupposed. With non-specific phrases, whose referents are
not identifiable in principle, the question of identifiability by the speaker does
not even arise – such expressions are necessarily unknown to the speaker. The
possibilities can be summarized as in Figure 3.
 48

Figure 3. (In-)definiteness, (non-)specificity and knowledge of the speaker

indefinite definite

non-specific specific

unknown known (known to

to the speaker to the speaker speaker and
hearer)

In Russian, for example, the -to-series can only be used if the referent cannot
be identified by the speaker. Thus, in (96a) the speaker does not know who it
was that Maš a met, and in (96b) the speaker has forgotten what she wanted to
tell the hearer.

(96) a. Maš a vstretilas’ s kem-to okolo universiteta.

Maš a met with who-INDEF near university
‘Maš a met with somebody near the university.’
b. Č to-to ja tebe xotela skazat’. (Paduč eva 1985: 211)
what-INDEF I you wanted tell
‘I wanted to tell you something.’

Example (97a) implies that the speaker has already forgotten who she met with,
which makes the sentence somewhat odd, and (97b) is completely deviant,
because it is incoherent to utter a desire about something specific that one
cannot identify.

(97) a. ?Ja vstretilas' s kem-to segodnja v 19 č asov.

I met with who-INDEF today at 19 hours
‘I met with someone [unknown to me] today at 19 hours.’
b. *Ja xoč u spet’ kakoj-to romans. (Paduč eva 1985: 211)
I want sing which-INDEF romance
‘I want to sing some [specific, unknown] romance.’

If the referent can be identified by the speaker but the speaker chooses to
withhold the information about its identity from the hearer, the koe-series or the
determiner odin ‘one’ is used in Russian:

(98) a. Ja vstretilas’ s odnim č elovekom segodnja v 19 č asov.

I met with one person today at 19 hours
‘I met with someone [known to me] at 19 hours.’
b. Maš a vstretilas’ koe s kem okolo universiteta.
Maš a met INDEFwith who near university
‘Masa met with someone [known to me] near the university.’

The semantic distinction of (lack of) knowledge of the speaker is not nearly
as important as the two factors discussed in the previous sections, negative
polarity and (non-)specificity. It has received very little attention in the
theoretical literature. However, it can be found in quite a few languages. In my
40-language sample, I found evidence for such a distinction in ten languages.
In Kannada, indefinites of the WH-oo-series are always specific and imply
that the speaker cannot identify the referent (see Bhat (1981), who calls this
 49

'physical identification'). Thus (99) is unambiguously specific, and the speaker

does not know which book Ramu wants.

(99) Kannada (Bhat 1981: 3)

Raamu-vige yaavud-oo ondu pustaka beekaagide.
Ramu-DAT which-INDEF one book want:having:is
‘Ramu wants a [specific, unknown] book.’

In Lithuanian, the kai-series is restricted to indefinites whose reference is

known to the speaker, while the kaž -series is used when the reference is
unknown to the speaker (this is not surprising since the original meaning of kaž
- is ‘the speaker does not know’, see § 6.2.1):

(100) Lithuanian (Pilka 1984: 97,29)

a. Turiu kai ka4 <?*kaž -ka4 > tiktai tau vienai pasakyti.
I:have INDEF what INDEF-what only to:you alone to:say
‘I’ve got something to say that’s for your ears alone.’
b. Kaž -kas ate%jo.
INDEF-who came
‘Somebody came (I don’t know who).’

In German, the indefinites jemand ‘someone’ and etwas ‘something’ may be

expanded by the prefixed indefiniteness marker irgend only if the referent
cannot be identified by the speaker. Thus, in (101a) the speaker may or may not
know who called, and the hearer could ask who it was. By using irgend in
(101b), the speaker makes it clear that he is ignorant about the caller’s identity,
so the hearer cannot ask who it was.

(101) German
a. Jemand hat angerufen. (—Wer war es?)
‘Someone called. (—Who was it?)’
b. Irgend jemand hat angerufen. (*—Wer war es?)
‘Someone (I don’t know who) called. (*Who was it?)’

In English, too, the (lack of) knowledge of the speaker is relevant, as was
pointed out in Warfel (1972) (cf. also Wierzbicka 1980b: 325, Croft 1983). The
determiner some (pronounced [sUm], stressable), when used with a singular
count noun, implies that the speaker cannot identify the referent. In (102a-c),
the continuations in parentheses presuppose identifiability by the speaker and
are therefore excluded.

(102) English (Warfel 1972: 43-4)

a. John is looking for some book on reserve (*and I know which one).
b. Hortense is watching for some sailor who’s due in port today.
(*He is a friend of mine.)
c. Ralph is worried because he lost some letter he was supposed to mail
(*but I have it right here).

Note that some behaves differently in this respect from the most frequent
members of the some-series: someone and something can be used irrespectively of
whether the speaker can identify the referent or not.
 50

3.2.5. Free-choice indefinite pronouns

Many languages have a special series of indefinite pronouns to express the

meaning of free choice, as illustrated in (103). Such indefinites are often
prosodically prominent in the sentence.

(103) a. English
After the fall of the Wall, East Germans were free to travel anywhere.
b. Russian
Ty mož eš ’ sdelat’ č to ugodno.
you can do what INDEF
'You can do anything.'
c. Lezgian (Haspelmath 1993a: 195)
Mac’a-laj wuč x̂ ajit’ani alaq’-zawa-j-di ja.
Mac’-SUPERDIR what INDEF [Link]-IMPF-PTCP-SBST is
‘Mac’ can do anything.’
d. Hungarian (Stephanides 1985: 80)
A cikk-ek-et meg-ir-hat-om bár-hol, bár-milyen
the article-PL-ACC PREV-write-POT-1SG INDEF-where INDEF-which
körülmény-ek között.
circumstance-PL under
‘I can write the articles anywhere, under any circumstances.’
e. Tagalog (Schachter and Otanes 1972: 532)
Ang kahit na sinu-mang piyanista ay
TOP INDEF LK who-INDEF:LK pianist PT
makakatugtog ng ganyang piyesa.
can:play GEN such:LK piece
‘Any pianist can play a piece like that.’

Free-choice indefinites are semantically similar to universal quantifiers like

‘every’ (‘everyone’, ‘everything’, etc.). In many environments, free-choice
indefinites can be replaced by universal quantifiers without a noticeable change
in the truth conditions.
However, there is always a difference in meaning between free-choice in-
definites and universal quantifiers, and in many contexts there is also a sig-
nificant difference in the truth conditions, as pointed out by Vendler (1967)
(who also coined the now widespread term free choice). For example, (104a) is
quite different from (104b).

(104) a. (free choice) You can take any apple.

b. (universal) You can take every apple.

The precise characterization of the meaning of free-choice indefinites has

proved to be a difficult task. I discuss various theoretical approaches to free-
choice indefinites below in § 5.2.1, § 5.5.
In this section, I will concentrate on presenting and characterizing the
contexts where free-choice indefinites may occur. There seems to be very little
cross-linguistic variation in the conditions under which free-choice indefinites
are possible, and the examples from different languages in the following should
be taken as illustrating the uniformity of free-choice indefinites, not as pointing
to typological differences.
 51

The first thing to note is that free-choice indefinites are semantically non-
specific, so they cannot occur in contexts which do not allow non-specific
indefinites, i.e. in perfective past and ongoing present contexts:

(105) a. English
*Anyone lived in a pretty town.
b. Russian
*Ona kupila č to ugodno.
she bought what INDEF
‘She bought anything.’

However, free-choice indefinites are much more restricted than non-specific

indefinites. The most typical environment for them are sentences that express
possibility, be it objective possibility or permission (as illustrated in (103a-e)
above). In most other typical non-specific contexts, free-choice indefinites are
awkward, although they are somewhat better in these contexts than in specific
contexts.

(106) a. ‘want’
?*Nobuko wants to marry any man who speaks Ainu.
b. imperative
?*When you go to town tomorrow, please buy me anything.
c. future
?*Next week Mahmut will go anywhere.

In particular, free-choice indefinites are bad in contexts of necessity, which is in

striking contrast to their acceptability in contexts of possibility (cf. Horn 1972:
143).

(107) a. You can <*must> marry anybody.

b. I allow <*require> you to marry anybody.

However, free-choice indefinites are allowed in imperatives if the imperative

is interpreted as a permission rather than as a command. As was pointed out by
Strickland (1982: 19-20), out of context an utterance like (108) sounds odd, but
in a context such as (109) or (110), imperatives with any are quite acceptable.

(108) ??Bring me any chair.

(109) A — What chair do you want?
B — Oh, bring me any chair. It doesn’t matter.
(110) A — Bring me a chair.
B — Which one?
A — Oh, any one.

The reason for the acceptability of the free-choice indefinite in (109) is that
although the sentence is structurally an imperative sentence, its communicative
force is not that of a command. In the context it is already clear to A that B
wants A to bring a chair, and B’s utterance in (109) is functionally equivalent to
‘You can bring me any chair’. Similarly, in (110) A’s elliptical utterance any one
can be completed not only by the corresponding parts of A’s previous utterance
(bring me), but one could also say that the non-elliptical version of any one
would be any one is possible (or similar). Compare also the analogous example
from Lithuanian (ka4 nors ‘something’, bet ka4 ‘anything (free choice)’).
 52

(111) Lithuanian (Pilka 1984: 140)

Nupir-k man ka4 nors <*bet ka4 > paskaityti. — O ka4? —
buy-IMPV for:me what INDEF INDEF what to:read and what
Bet ka4 .
INDEF what
‘Buy me something to read. — And what? — Anything.’

A related case are imperatives that function in a way similar to conditional

antecedents. In the sentences in (112), where in each case a free-choice
indefinite is used, the imperative clauses are not commands, but challenges by
the speaker that serve to emphasize the speaker’s argument. Again, a
paraphrase with ‘can’ is always possible.

(112) a. English (Strickland 1982: 33)

Go ask anyone. You’ll see that I’m right.
b. French (Strickland 1982: 33)
Demande à n’importe qui.
‘Ask anyone.’
c. Yakut (Ubrjatova (ed.) 1982: 207)
Kim-ten baV arar yjyt-yN, itinnik etiextere.
who-ABL INDEF ask-iMPV.2PL everybody will:say
‘Ask anyone, everybody will say so.’
d. Italian
Domanda-lo a chi-unque.
ask(IMPV)-it to who-INDEF
‘Ask anybody.’

Thus, imperatives are no exception to the generalization that free-choice

indefinites are allowed in contexts of possibility, but not in other non-specific
contexts.
There is, however, one further important type of contexts in which free-
choice indefinites may occur, which is not easily subsumed under the notion of
possibility: contexts in which a sentence may be interpreted generically.
Consider the examples in (113-114).

(113) English
Any doctor will tell you that Stopsneeze has dangerous side effects.

(114) a. Russian (Paduč eva 1974: 80)

Petja reš it ljubuju zadač u.
Petja solves any problem
‘Petja will solve any problem.’
b. Italian
Qualunque costituente di categoria Xn che viene spostato lascia
any constituent of category Xn that is moved leaves
dietro di sé una categoria vuota dello stesso tipo.
behind of itself a category empty [Link] same type
‘Any constituent of category Xn
that is moved leaves behind an
empty category of the same type.’
c. Lithuanian (Pilka 1984: 141)
Bet kuri kate — ž irduolis.
INDEF which cat mammal
 53

‘Any cat is a mammal.’

d. Lezgian (Haspelmath 1993a: 196)
Hi kac x̂ ajit’ani hajwan ja.
which cat INDEF animal is
‘Any cat is an animal.’

Generic sentences are often in an imperfective present tense, but they may also
be in a perfective future tense, as (113a) and (114a) show.
A further context where free-choice indefinites are licensed are hypothetical
(cf. 115) and counterfactual (cf. 116) sentences:

(115) a. English, French (Strickland 1982: 37)

I would give anything to see that.
Je donnerai n’importe quoi pour voir ça.
b. Italian
Lei parlerebbe dei suoi problemi amorosi con chi-unque.
she would:talk of her problems [Link] with who-INDEF
‘She would talk to anyone about her love problems.’

(116) English, French (Strickland 1982: 36)

a. At that point, I think I would have accepted anything.
b. Je crois qu'à ce moment là, j'aurais accepté n'importe quoi.

And finally, as König (1991: 104) observes, contexts expressing sufficient

conditions license free-choice indefinites, independently of the tense or
modality of the clause:

(117) Any amount is adequate/sufficient.

(*Any amount is required/necessary.)

3.2.6. Summary: the main functional distinctions

We can summarize the main semantic and pragmatic distinctions that recur
cross-linguistically as in Figure 4 (cf. Croft (1983) for a similar list).

Figure 4. The main functional distinctions

specific known to speaker

unknown to speaker
non-specific irrealis contexts
negative polarity conditionals
questions
standard of comparison
indirect negation
direct negation
free choice

These are the functional distinctions that I have investigated systematically in

the 40 languages of the smaller sample. Several further semantic, syntactic and
pragmatic distinctions to which indefinite pronouns may be sensitive are
treated in § 4.7 and § 8.1-2.
 54

3.3. Alternatives to Indefinite Pronouns

Before going on to study systematically the differences between the systems of
indefinite pronouns in the world’s languages, let us ask whether indefinite
pronouns are necessary, or whether a language could do without them, either
for all their functions or for only some of them. It appears that the majority of
languages do have indefinite pronouns, and that all the different functions
discussed in § 3.2 are most often expressed by indefinite pronouns. However,
for all these functions, alternative strategies of expressing them are attested,
and probably quite a few languages lack indefinite pronouns altogether. The
alternative strategies will not be discussed later in this study, but they are noted
in this section.
There are four main ways of expressing what other languages express by
means of indefinite pronouns: (i) generic nouns, (ii) existential sentences, (iii)
non-specific free relative clauses, and (iv) universal quantifiers.

3.3.1. Generic nouns

We saw in § 3.1.2 that indefinite pronouns are sometimes derived from generic
ontological-category nouns such as ‘person’, ‘thing’, ‘place’, ‘time’, ‘manner’.
These generic nouns are very similar in meaning to indefinite pronouns like
‘somebody’, ‘something’, ‘somewhere’, ‘sometime’, ‘somehow’, and languages
may make use of indefinite noun phrases consisting only of such generic nouns
to express what other languages express by means of indefinite pronouns.
It appears that this is the most common way of replacing indefinite pro-
nouns, and it is quite widespread. The following is a small sample of languages
that show this feature, with two examples for each language.

(118) a. Khalkha Mongolian

(neg) xün ‘someone’ = ‘(one) person’
(neg) jum ‘something’ = ‘(one) thing’
b. Irish
duine (éigin) ‘someone’ = ‘a (certain) person’
rud (éigin) ‘something’ = ‘a (certain) thing’
c. Somali (Berchem 1991: 79-84)
nin ‘someone’ = ‘man’
wax ‘something’ = ‘thing’

As I observed in § 3.1.2, it is not always easy to distinguish generic nouns

from indefinite pronouns based on generic nouns. But it does seem that there
are a substantial number of languages that replace indefinite pronouns in this
way. Whenever a grammar that is otherwise quite complete is silent about
indefinite pronouns, it may be assumed that the language does not use gram-
matical means, but the lexical means of general nouns.x
It is also not uncommon for some members of an indefinite series to consist
of a non-grammaticalized expression involving a general noun. Consider the
German and Italian negative series in (129).
 55

(119) German Italian

‘nobody’ niemand nessuno
‘nothing’ nichts niente
‘nowhere’ nirgends in nessun luogo
‘noway’ auf keine Weise in nessun modo
‘no’ kein nessun(o)

The German expression auf keine Weise and the Italian expressions in nessun
modo (lit. ‘in no manner’), in nessun luogo (lit. ‘in no place’) are very weakly
grammaticalized if at all and should probably not be regarded as indefinite
pronouns. Rather, the corresponding indefinite pronoun series are defective in
their places. However, as I made clear earlier (§ 3.1.3), I am not interested here
in which ontological categories are represented in each series, so I will not
always take pains to distinguish grammaticalized generic-noun-derived
indefinites from non-grammaticalized expressions like auf keine Weise. What
counts is that both German and Italian undeniably have an n-series.
A series may even be said to exist when only a determiner exists. For ex-
ample, Russian ljuboj ‘any’ is an indefinite determiner of the type studied in this
work, but all its combinations with generaic nouns are non-grammaticalized
(ljuboj č elovek ‘any person’ = ‘anyone’, v ljubom meste ‘anywhere’ = ‘at any
place’, etc.). Nevertheless these expressions fall under the scope of the present
study.

3.3.2. Existential sentences

In Tagalog (and other Philippine languages, e.g. Cebuano), existential sentences

are used where other languages use specific indefinite pronouns. A sentence
like (120) can perhaps literally be translated as ‘There is (one) having come
yesterday’.

(120) Tagalog (Schachter and Otanes 1972: 276-8)

May dumating kahapon.
exist come:AG yesterday
‘Someone came yesterday.’

Tagalog expresses indefinites in this way not only in agent position. For
indefinites that are patients or places, the patient- and place-oriented verb
forms are used:

(121) May ginawa ang Americano.

exist do:PAT TOP American
‘The American did something.’
(Lit. ‘As for the American, there exists something that was done.’)
(122) Mayroon siya-ng pupuntahan.
exist she:TOP-LK go:LOC
‘She is going somewhere.’
(Lit. ‘As for her, there exists a place where (she) went.’)

The same strategy is also used when other languages use negative indefinite
pronouns, as in (123) (Schachter and Otanes 1972: 521).

(123) Wala-ngdumating kahapon.

[Link]-LK come:AG yesterday
 56

‘No one came yesterday.’

However, the existential strategy does not work when there are two in-
definites in the same clause – unlike the predicate calculus, natural languages
do not allow a sequence of two ‘there is’ expressions in one clause. In such
cases, Tagalog must make use of an indefinite pronoun of its -man-series
(interrogative pronoun plus -man) for one of the two indefinites.

(134) a. At kung ako ’y may nadayang sinu-man,... (NT, Lk 19.8)

any if I PT exist swindle what-INDEF
‘And if I have swindled anyone out of anything,...’
b. Wala sila-ng sinabi kaninu-man. (NT, Mk 16.8)
[Link] they:TOP-LK say who-INDEF
‘They did not say anything to anybody.’

Tagalog also uses its -man-series for the free-choice function (Schachter and
Otanes 1972: 532):

(125) Matatapos namin ito kailan-man.

finish we this when-INDEF
‘We can finish this at any time.’

So not even Tagalog, which uses its existential strategy quite extensively, lacks
indefinite pronouns completely – they are merely restricted in the range of
functions they fulfill.

3.3.3. Non-specific free relative clauses

The free-choice meaning is often expressed most naturally by means of a non-

specific free relative clause whose subject is identical to the subject of the main
clause and whose predicate is an expression of free choice (e.g. ‘want’). In some
languages, this is the only way of expressing the free-choice function. For
example, in Maltese one would normally use a relative clause with ried ‘want’,
as in (126a-b).

(126) Maltese (Haspelmath and Caruana, forthcoming)

a. Tista’tieÌu liema tuffieÌa trid.
you:can:IMPF you:take:IMPF which apple you:want:IMPF
‘You may take any apple.’
(Lit. ‘You may take which(ever) apple you want.’)
b. SetgÌet tmur fejn riedet.
she:can:PERF she:go:IMPF where she:want:PERF
‘She was allowed to go anywhere.’
(Lit. ‘She could go where she wanted.’)

Theoretically one could also use the free-choice determiner kwalunkwe, but this
sounds very artificial because kwalunkwe (from Italian qualunque ‘any’) has not
been assimilated sufficiently yet.

3.3.4. Universal quantifiers

Some languages use universal quantifiers with meanings ‘every’ and ‘all’
instead of indefinite pronouns to translate free-choice indefinites in other
 57

languages. Since the meanings of ‘every’ and ‘any’ are very similar and forms
meaning ‘every’ and ‘any’ are almost interchangeable in some contexts, this is
not unexpected. An example comes again from Maltese, a language lacking
special free-choice indefinites. When the free-choice expression is in subject po-
sition, Maltese uses kull ‘every’ or a related complex expression, e.g.

(127) Maltese (Haspelmath and Caruana, forthcoming)

KulÌadd jista’ jsolvija din il-problema.
everyone he:can:IPF he:solve:IPF this the-problem
‘Anyone/everyone can solve this problem.’

In several languages, speakers are also uncomfortable with indefinites in the

standard of comparison and prefer to use a universal quantifier instead, for
example in Swedish.

(128) Swedish (Kersti Börjars, Martin Tamm, p.c.)

Pojken kan springa fortare än alla i sin klass.
the:boy can run faster than all in his class
‘The boy can run faster than anyone/everyone in his class.’

3.3.5. Languages without indefinite pronouns

While it is easy to find cases in which indefinite pronouns are noz used and
alternative strategies are preferred, it is not easy to find a language that
completely lacks indefinite pronouns. All the languages cited above in this sec-
tion replace indefinites in some functions, but do use them for others. Thus,
Tagalog uses an alternative strategy when specific reference is intended but
employs indefinite pronouns for the free-choice function, and in Maltese the
reverse situation obtains.
When a description of a language does not mention indefinite pronouns, we
cannot of course be sure that they do not exist – the author may simply have
failed to include them for some reason. Grammar books do not often say
explicitly that indefinite pronouns do not exist in the language, and even when
they do, this may only be true for a subset of the functions that indefinite
pronouns may [Link] Unless a grammar is both explicit about the lack of
indefinite pronouns and says what alternative means are used for the functions
that indefinite pronouns typically fulfill, we cannot be sure that they are really
absent.
A good candidate for an example of a language lacking indefinite pronouns
is Lango, a Nilo-Saharan (Western Nilotic) language for which we have the
excellent description of Noonan (1992). For the specific functions, Lango uses
generic nouns (Nàt ‘person’, gìn ‘thing’) combined with the indefinite suffix -
mOrO (Noonan 1992: 161-62), e.g.

(129) Lango (Noonan 1992: 153)

Na~t-tO!rO! ojwato opêo
person-INDEF 3SG:hit:PRF Opio
‘Someone hit Opio.’

The suffix -mOrO is not restricted to a few generic nouns but can be used with
any noun (twòl-lOrO ‘a snake’). Noonan is also quite explicit that negative
pronouns are non-existent in Lango (1992: 242). Instead, an existential
 58

construction (not unlike the Tagalog construction above in § 3.3.2, but involving
generic nouns) is employed:

(130) Lango (Noonan 1992: 242)

a. gên-nOrO amE acamo pe
thing-INDEF REL 1SG:eat:PERF NEG
‘I ate nothing.’ (Lit. ‘A thing I ate doesn’t exist.’)
b. Nat-tOrO amE ocamo dEkko pe
person-INDEF REL 3SG:eat:PERF stew:that NEG
‘No one ate that stew.’
(Lit. ‘A person who ate that stew doesn’t exist.’)

Although Noonan’s grammar is quite comprehensive, he does not say how

the meaning of free-choice indefinites is expressed in Lango, so even in this
well-documented case we cannot be completely sure. But what a case like
Lango does show is that it is quite likely that we will find true cases of
languages lacking indefinite pronouns as we learn more about little-known
languages.
Thus, the subject matter of this study is probably not universal, although it is
very widespread in the world’s languages. Of course, this lack of universality
does not mean that this work is irrelevant to language universals. If universals
research were concerned only with universal features, not much would be left
to investigate: Very few linguistic features that had once been assumed to be
absolute universals have stood the test of time – linguists have discovered
languages lacking subjects, numerals, nasals, and even vowels. All this only
means that the generalizations formulated in this work apply only to languages
that have indefinite pronouns. Technically, they are all implicational universals
whose initial antecedent is ‘If a language has indefinite pronouns, then...’. In
the next chapter, the main implicational generalizations will be presented.
 59

i
The term series was taken from Veyrenc (1964).
ii
Another equivalent term is epistemological category (Durie 1985: ch. 6). Ontological
category is from Jackendoff (1983: 51). Jackendoff’s list of major ontological
categories is: thing, place, direction, action, event, manner, amount. The main
difference between Jackendoff’s list and the lists in (20) is that Jackendoff does not
distinguish between person and thing. It may well be that this distinction is located at a
lower conceptual level, but practically all languages make the person/thing distinction
in their indefinite pronouns at the same level as the other distinctions, so it is best to
regard person and thing as two separate ontological categories.
iii
The term was taken from Veyrenc (1964) (indicateur d’indétermination).
iv
The term extrafix is a cover term for postfixes (suffixes that follow inflectional
suffixes) and antefixes (prefixes that precede inflectional prefixes). The term postfix is
well-established in the literature (e.g. to refer to the Russian reflexive suffix -sja which
follows verbal aspect, tense, and person-number inflections); the neologisms antefix and
extrafix were first used in Haspelmath (1990: 29, 63).
v
A similar phenomenon has been observed for Japanese: The negative indefinite nani-
mo ‘nothing’ (from nani ‘what?’) has the variant nanni-mo. However, this seems to be
an isolated case in Japanese.
5a
This seems to be the situation in Luiseño (Uto-Aztecan), as pointed out to me by
Mark Durie.
vi
English is somewhat peculiar in that it has both interrogative-based and generic-noun-
based indefinite pronouns, even within the same series: somebody, something,
someplace (based on generic nouns) exists alongside somewhere, somehow and
somewhat in the some-series (similarly in the any-series). It is much more common for
all members of an indefinite series to be derived from the same kind of basis.
vii
Since negative polarity items are not restricted to negative contexts, as described in
more detail below, this term (which was coined by Baker 1970) is not particularly
felicitous. As the discussion in § 5.5 will show, a term like scale reversal would be
much more appropriate than negative polarity (and negative polarity items should be
called scale reversal items) However, I continue to use the terms negative polarity,
negative polarity item because they are so entrenched in the linguistic literature.
viii
This sentence is grammatical on the irrelevant reading 'This suits you better than
something else.'
ix
In (63b), a bicycle can also be interpreted specifically, and on this reading the mini-
discourse is fine. A discourse referent is also possible in a sentence with non-indicative
modality or future tense, e.g. Pedro wants to buy a new bicycle [non-specific]. It
must/will be black.
 60

x
Examples of such grammars are Dixon (1988) (Boumaa Fijian) and Foley (1991)
(Yimas). Some grammars also explicitly deny the existence of indefinite pronouns, e.g.
Berchem (1991: 79-84) (Somali), Noonan (1992: 242) (Lango).
xi
For example, Berchem (1991: 79-84) explicitly and emphatically denies the existence
of indefinite pronouns in Somali (cf. the generic nouns in (118c)). However, he also
says that only a limited number of generic nouns may take the negative suffix -na and
function as negative pronouns, as in Nin-na ma iman (person-NEG NEG came) ‘Nobody
came.’ This limitation to a few nouns is evidence that a closed class of negative
pronouns has developed.
 75

An Implicational Map
for Indefinite Pronoun Functions

4.1. Multifunctionality
In the preceding chapter we have seen various syntactic and semantic
conditions under which languages may use different series of indefinite
pronouns. These are the distinctions that were found relevant for languages in
general, but this does not of course mean that all languages reflect all these
functional distinctions in the phonological form of the indefinite pronouns. On
the contrary, there is probably no language that makes all the possible
distinctions, and many languages make only very few distinctions (two series
of indefinite pronouns seems to be the minimum for languages that do not
make extensive use of avoidance strategies as discussed in § 3.3). What we
have, therefore, is MASSIVE MULTIFUNCTIONALITY. The majority of the series of
indefinite pronouns are used to express more than one of the functions
distinguished in § 3.2, as illustrated by the examples in (131-33).

(131) English any-series

a. negative polarity: conditional
If anybody left us, we are lost.
b. free choice
Anybody could have left us.

(132) Hebrew š ehu-series

a. specific (ongoing state)
Balš an ze eÌad š e-mesabex ma-š ehu kal.
linguist that one that-complicates what-INDEF easy
‘A linguist is someone who complicates something easy.’
b. negative polarity: question
Ha?im tafs-u ma-š ehu?
Q catch:PERF-3PL what-INDEF
‘Did they catch anything?’

(133) Yakut da-series

a. negative polarity: standard of comparison (Ubrjatova (ed.) 1982: 209)
Kün annyVar tuox-tan da min bu oloVu taptyy-byn.
this world what-ABL INDEF I this life love-1SG
‘In this world I love life more than anything.’
b. direct negation (Afanas’ev and Xaritonov 1968:335)
Kini kim-i da kör-bö-tö.
he who-ACC INDEF see-NEG-PAST
‘He did not see anyone.’

A more traditional term for such a situation is polysemy, but in many cases
there is no obvious meaning difference between the different functions – these
often seem to be just different CONTEXTS rather than different MEANINGS. In
order not to prejudice the analysis, I will use the neutral term
 76

MULTIFUNCTIONALITY, which unlike the term polysemy does not imply the
presence of multiple meanings, but subsumes polysemy as a special case.
Moreover, even where an indefinite series clearly expresses more than one
semantically distinguishable function, one could maintain that from the point
of view of the individual language, there only is one general meaning
(Gesamtbedeutung) that happens to correpond to several more specific meanings
in other languages. My strategy of investigating multifunctionality allows me
to remain neutral on this issue – I need take no stand on whether
multifunctionality should be analyzed in terms of polysemy or in terms of a
general meaning. All that is necessary is the possibility of identifying the
functions in any language.

4.2. The Semantic Map Methodology

One of the central proposals of this book is to account for cross-linguistic

variation in indefinite pronouns in terms of a semantic map, a strategy that has
been applied successfully to other grammatical categories. In this section I
argue that indefinite pronouns are best understood if they are approached
similarly.

4.2.1. Indefiniteness markers as grammatical categories

The multifunctionality of indefinite pronoun series is hardly a surprising

phenomenon, as we are used to multifunctionality elsewhere in the semantics,
pragmatics and syntax of grammatical items, or GRAMMATICAL CATEGORIES.
For instance, nominal cases are notoriously multifunctional. Thus, the Classical
Greek Dative case may express indirect object, beneficiary, instrument, agent,
cause, and time, and the Russian Instrumental case may express instrument,
agent of passive, time, and predicate nominal.i Or take the example of the
English Future tense, which may express pure future time reference (but not in
conditional and temporal clauses), intention, and generic situations, or the
German Present tense, which may express progressive, habitual, and generic
situations and future time reference. Similarly for voice categories: the Latin
Passive may express passive, anticausative, and reflexive, and the Modern
Greek Middle may express reflexive, reciprocal, anticausative, passive, and
potential passive. With such typical grammatical categories as case, tense and
voice categories, exactly the same problems of description, analysis and cross-
linguistic generalization arise as we noted above for indefinite pronouns. This
means that we can make use of the experience, the methodology and the
theoretical approaches that have been developed for the study of the
prototypical grammatical categories.
One might even go as far as to claim that the different indefiniteness series
ARE grammatical categories. The main difference between prototypical
grammatical categories like case categories or tense categories is that these
apply to an open set of words (nouns, verbs), whereas indefinite markers
combine with a rather small set of words denoting ontological categories
(interrogative pronouns, or generic ontological-category nouns, cf. § 3.1.2).
Even 'normal' grammatical categories like number do not necessarily apply to
an open class of words, e.g. in languages where number is only marked on
pronouns. Anyway, this is ultimately a terminological issue, and my main
reason for mentioning this possibility here is to highlight the similarities
 77

between indefiniteness markers and more prototypical grammatical markers

like case and tense markers.
Several kinds of grammatical categories have been the object of systematic
large-scale cross-linguistic investigations in the last two decades: tense, aspect
and mood markers (e.g. Dahl 1985, Bybee 1985, Bybee et al. 1994), spatial
relation markers (e.g. Svorou 1994), voice categories (e.g. Geniuš iene% 1987,
Haspelmath 1990, Kemmer 1993), case markers (Croft 1991). All these studies
have shown clearly that universal patterns can be found in grammatical
categories. In language after language, independently of genetic affiliation and
areal location, we find similar grammatical meanings (or functions, or uses),
similar patterns of multifunctionality, and similar expression types. Of course,
there are also significant differences between languages of different types, but
this does not detract from the overall universal pattern. These cross-linguistic
studies of grammatical categories have made a considerable contribution to our
knowledge about human languages, and it is in their spirit that the present
work is written.

4.2.2. The use/function-based approach

The typological studies of grammatical categories just cited employ the same
basic research strategy followed in the present work. The strategy was outlined
already in Ferguson’s (1970) programmatic paper. Using the example of
multifunctional case markers, Ferguson showed that the structuralist approach,
which tries to formulate highly abstract general meanings (Gesamtbedeutungen)
for grammatical categories in individual languages (following Jakobson 1936),
is incapable of describing and explaining cross-linguistic similarities and
differences. Consider the dative case in German, Lezgian, and Classical Greek.
In German, the Dative case is mainly used for indirect objects and beneficiaries,
so one might propose a general meaning like ‘recipient’. In Lezgian, the Dative
expresses direction and experiencer in addition to these functions, so one might
try some vague abstract notion of ‘goal’ to encompass all three functions. In
Classical Greek, the functions include indirect object, beneficiary, instrument,
agent, and cause (but not direction), and it seems difficult to come up with
anything more concrete than ‘oblique’. But however ingenious and perhaps
insightful the general meanings may be, they cannot be compared to each
other. What could the relation be between ‘recipient’, ‘goal’ and ‘oblique’?
There is no answer – not surprisingly, there is very little cross-linguistic work
that has come out of the research on general [Link]
As an alternative to this extreme semasiological or bottom-up approach, one
might propose an extreme onomasiological or top-down strategy. This would
involve setting up an exhaustive list of all imaginable functional distinctions in
a domain – probably an impossible task in most cases. It appears that human
cognitive resources are in priniple sufficiently rich to allow for the creation of
an indefinite number of (complex) subtle conceptual distinctions.
Thus, the only possible route left is the middle road between the two ex-
tremes, and this is what Ferguson (1970) proposed and the typologists have
practiced. The functional distinctions to be recognized must be at an inter-
mediate level, the level of USES (or FUNCTIONS). Every grammatical category
can have several uses. Whenever two roughly comparable categories in two
languages turn out to differ in one type of environment or meaning, this is
sufficient for setting up a separate use. Consider the case of the German Dative
and the Lezgian Dative again. These are roughly comparable in that the core
 78

uses of the German Dative, indirect object and beneficiary, are also central uses
of the Lezgian Dative. But the Lezgian Dative can also be used in direction
expressions like Derbent-di-z ‘to (the city of) Derbent’, which is not possible in
German (where a preposition has to be used: nach Derbent). In order to describe
the difference between the German and the Lezgian Datives, we need at least
the function or use 'direction'. By applying this strategy to more and more
languages in the domain of inquiry, the list of different, cross-linguistically
relevant uses is gradually enlarged.
Now it could turn out that this strategy leads to the same unconstrained
proliferation of functional distinctions as the top-down (or onomasiological)
approach, and then it would be of no more use. But experience tells us that
languages do not differ from each other in unlimited ways. Cross-linguistic
variation is tightly constrained also in the area of grammatical meanings. While
the human mind can conceive of an indefinitely large number of ideas, only a
small recurring subset of possible concepts are conventionalized in
grammatical categories. The practice of typology has shown consistently that as
more languages are drawn into the investigation, fewer and fewer new uses
turn up, and the uses that have already been established turn out to be relevant
even for new languages that have never been studied before.
Indefinite pronouns are no exception to this general pattern, as we have seen
in § 3.2. The distinctions described there were found to be of central importance
in a wide variety of languages, independently of genetic and areal connections.

4.2.3. A geometric representation of implicational universals

In order to capture the relationships between different uses within a use-

/function-based framework, the methodology of SEMANTIC MAPS has been
proposed and will be adopted here (cf. Anderson 1982, Bybee 1985: 195-6, Croft
et al. 1987). The highly specific uses or functions that I have identified above in
§ 3.2 are thought of as existing in a semantic or functional space. The relations
among these uses can be studied by determining cross-linguistically which uses
can be covered by the same grammatical marker. These relations are
represented geometrically on a semantic map in such a way that two uses or
functions that are expressed by the same grammatical marker in at least one
language are adjacent on the map. The underlying hypothesis is that it will be
possible to draw up such a map in a way that is consistent with any existing
grammatical system.
Consider first an abstract example:

Figure 5.
c
a b e
d

Figure 5 represents a semantic map of the five uses or functions a, b, c, d, and e.

There are various ways in which these functions can be covered by different
markers in different languages. Three possibilities are given in Figure 6.
 79

Figure 6.
c
Language I: marker X: a, b a b e
marker Y: c, d, e d

c
Language II: marker W: a, b, d, e a b e
marker Z: c d

Language III: marker U: b, c c

marker V: d, e a b e
marker T: a d

However, the semantic map in Figure 5 disallows languages like Language IV

and V in Figure 7, because in these hypothetical languages there are markers
that have several uses that are not adjacent on the semantic map.

Figure 7.
c
*Language IV: marker S: a, e a b e
marker R: b, c, d d

*Language V: marker Q: a, c c
marker O: b, e a b e
marker P: d d

So the semantic-map approach not only helps us to express the relations of

semantic similarity between various uses, but it also makes testable predictions
about what a possible linguistic system is.
Indeed, semantic maps can also be thought of as making statements about
IMPLICATIONAL UNIVERSALS. For any two uses/functions a and b that are ex-
pressed by the same marker, the semantic map predicts that all uses/functions
that are located in the space between a and b are expressed by the same marker
as a and b. Thus, we can also speak of IMPLICATIONAL [Link] The main cross-
linguistic implicational generalizations about the distribution of indefiniteness
markers that I propose in this study will be expressed in the form of an
implicational map.
Another area where semantic maps make predictions is the diachronic
extension of the functions of a marker (cf. especially Croft et al. 1987, Kemmer
1993). The prediction is that when markers gradually acquire new functions,
they will first be extended to those functions that are adjacent to the original
functions on the map, and only later to functions that are further away. The
diachronic extension of indefiniteness markers of indefinite pronouns will be
discussed in chapter 6 (especially § 6.4).
We now turn to the implicational map that I propose.

4.3. The Implicational Map

 80

After examining the evidence from a wide variety of languages, I have arrived
at a two-dimensional implicational map for the uses/functions of indefiniteness
markers. This map is shown in Figure 8.

Figure 8. An implicational map for functions of indefiniteness pronoun series

(7)
direct
(4) (6) negation
question indirect
negation
(1) (2) (3)
specific specific irrealis
known unknown non-specific
(5) (8)
conditional comparative
(9)
free choice

This map consists of nine functions/uses (taken from Figure 4 in § 3.2.6)

arranged in two dimensions. Three of the nine functions are arranged in a one-
dimensional sequence, and a second dimension is introduced for the other six
functions. The functions are numbered to allow us to refer to the various
positions on the map in a simple way.
The evidence for the cross-linguistic generalizations embodied in this
semantic map is found throughout this work. Some evidence has already been
presented in § 3.2. Most of the evidence from the 40-language sample is found
in Appendix A. More evidence, also from other languages that are not in the
sample, is scattered throughout other parts of this work.
In this chapter, I will confine myself to showing how this map works with
three examples: English, Russian, and Nanay (Manchu-Tungusic).

4.3.1. English

The distribution of indefinite pronoun series over the functions on the map in
English is shown in Figure 9. (In this and in subsequent representations of the
map, function numbers and connecting lines between functions are omitted for
the sake of clarity.)

Figure 9. English

direct
question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

English has three main series of indefinite pronouns: the some-series, the any-
series, and the no-series. Each of these corresponds to a continuous area
(graphically delineated by a closed line) on the map.
 81

The some-series covers the five uses ‘specific-known’, ‘specific-unknown’,

‘irrealis’, ‘question', and 'conditional’ – for examples see (98), (102), (74), (151a),
and (154a) respectively. The any-series covers the six uses ‘question',
'conditional’, ‘comparative’, ‘indirect negation’, ‘direct negation’, and ‘free
choice’ – for examples, see (44b), (45b), (51b-52), (49-50), (43), and (103) above.
The no-series covers only the function ‘direct negation’ (e.g. 34b).
Note that two indefinite series may of course overlap in their distribution:
the some-series and the any-series overlap in the uses ‘question' and
'conditional’, and the no-series overlaps with the any-series.
The kind of map exemplified in Figure 6 is simplified in several ways, in
addition to the simplification inherent in the selection of just the nine uses of
Figure 5. First, finer differences between two series that overlap in one function
are not shown, for example the difference between the no- and the any-series
(cf. Bolinger 1977: 37-65), or the difference between the some- and the any-series
in functions where either is possible (cf. § 4.3.7 below). Second, marginal series
are not included. In English, one could perhaps recognize a marginal -ever-
series (whoever, whatever, wherever, etc.), or a no matter-series (no matter who, no
matter what, etc.). The indefinite adverb ever has special properties and could be
thought of as a series of its own. And third, differential behavior of different
members of the same series is ignored. An example of this is the singular
determiner some, which in most ways behaves just like the other members of
the some-series. However, unlike the other members, it does not have the
‘specific-known’ use, as we saw in § 3.2.4. One could easily add this
information to Figure 6 by distinguishing a some1-series and a some2-series, the
latter consisting only of one member, the determiner some. Some2 still conforms
to the implicational generalization, because it occupies the continuous area
from ‘specific-unknown’ to ‘question' and 'conditional’.

4.3.2. Russian

The distribution of indefinite pronoun series in Russian is shown in Figure 10.

(For a full justification of this distribution, see Appendix A, Section 16.)

Figure 10. Russian

direct
question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

Russian is richer than English in its indefinite pronoun system: It has at least
seven core series, as well as several marginal series (see Appendix A). Russian
shows even more overlap between the different indefinite series. For two
functions, three different series of indefinites are possible, again with the
differences that are neglected in Figure 10. In addition to the pragmatic dif-
ferences in terms of speaker expectations (§ 4.7.4), stylistic differences are also
relevant: Where the -nibud’-series and the -libo-series overlap, the -nibud’-series
is more colloquial, while the -libo-series is more formal. But Russian differs
from English not only quantitatively, but also qualitatively. The space of
indefinite pronoun functions is divided up quite differently. There is a special
series for a ‘specific-known’ function (koe-), the two free-choice indefinites
 82

(ugodno/ljuboj) do not have any other functions, and there is a series that spans
the five functions ‘irrealis non-specific’, ‘question', 'conditional’, ‘comparative’,
and ‘indirect negation’ (-libo). On the other hand, there are also some obvious
similarities between Russian and English: a special series for ‘direct negation’
(ni-/no-) as opposed to ‘indirect negation’, and overlap in the question and
conditional functions between a ‘left-hand’ series (-to/some-) and a ‘right-hand’
series (-nibud’/any-). Russian also has two series that overlap completely: the
ugodno-series (kto ugodno ‘anyone’, č to ugodno ‘anything’, etc.), and the ljuboj-
series, which consists only of the determiner ljuboj ‘any’.

4.3.3. Nanay

The distribution of indefinite pronoun series in Nanay (Manchu-Tungusic) is

shown in Figure 11.

Figure 11. Nanay

direct
question indirect negation
specific irrealis negation
non-specific
conditional comparative
free-choice

My information on Nanay (from Onenko 1980, 1986) is less complete than my

information on English and Russian. I lack information on the the difference
between the 'known' and 'unknown' sub-cases of the 'specific' function, and it
may be that Figure 11 requires some changes in details (for instance, it could be
that -nuu and -daa overlap, just like English some-/any-, and Russian -to/-nibud’).
But it is beyond doubt that Nanay again differs significantly from both English
and Russian, while at the same time conforming to the implicational general-
izations embodied in the map. In particular, Nanay has an indefinite series (-
daa) which covers all the functions between ‘direct negation’ and ‘irrealis non-
specific’, as shown in (134-138). Within the irrealis function, -daa may even be
used with past perfective events if an evidential modality is present (138b).

Nanay (Onenko 1980: 441-2, 449)

(134) direct negation
Tuj manga puksin-du xamač aa-daa aliasi-ni goani.
thus strong storm-DAT who-INDEF bear-3SG PT
‘Of course, nobody can bear such a strong storm.’
(135) indirect negation
Duente doowani xaj-daa xaž on anaapulsiu-mi ač aasi.
taiga inside who-INDEF weapon without walk-CONV suit:NEG
‘One should not walk in the taiga without any weapon.’
(136) comparative
Naj xaj-doj-daa murumku, masi, imdar.
human who-than-INDEF smart strong skillful
‘A human being is smarter, stronger, and more skillful than anyone.’
(137) conditional
Xajdu-daa maNga-wa, kamaa-wa baa-xan osini, mimbi-we
where-INDEF difficulty-ACC problem-ACC find-PAST if I-ACC
doolaa xeersigu-xeeri.
quietly call-IMPV
 83

‘If you meet with difficulties anywhere, call me quietly.’

(138) irrealis non-specific: possiblility, evidential
a. Xaj-daa osii-daa osi-ž ara.
what-INDEF [?] happen-FUT.3SG
‘Something may happen.’
b. Sajna, xaj-daa ž aka o-č i-ni biž ere.
probably what-INDEF thing happen-PAST-3SG apparently
‘Probably something has happened.’

The free-choice indefiniteness marker -daa xem aja includes the marker -daa, but
is distinct from it. An example of its use is (139).

(139) free choice (Onenko 1986: 116)

Xaami-daa xem aja, ii-xeri.
when-INDEF come-IMPV
‘Come at any time.’

The indefiniteness marker -nuu is only used with specific expressions:

(140) specific
a. Uj-nuu ž ook-č i laN ž i-č i-ni. (Onenko 1986: 98)
who-INDEF house-DIR near go-PAST-3SG
‘Someone went up to the house.’
b. Ñoambani xajla-nuu bajtalto-j-č i. (Onenko 1980: 443)
they what-INDEF accuse-PRES-3PL
‘They are accusing him of something.’

These three examples should be sufficient to illustrate how the semantic map
captures cross-linguistic generalizations about indefinite pronoun systems and
makes universal predictions. It embodies the hypothesis that no language will
be found where, for instance, an indefinite series expresses both the free-choice
function and the irrealis-non-specific function, but not the question/conditional
and comparative functions, or an indefinite series that expresses both the
specific-known and the irrealis-non-specific functions, but not the specific-
unknown function, etc. In the 40 languages that I looked at, I have not found a
single counterexample to the implicational map. The reader may examine the
data from the 40-language sample in Appendix A. For convenience, the
distribution of the indefinite series in the 40 languages is reproduced in the
following section.

4.4. The Distributional Schemas of 40 Languages

Figures 12.1-40 show the distributional schemas of the 40 languages of the
smaller sample. More details, examples, and sources of the data can be found in
Appendix A for each language. (For some languages, the specific-known
function is missing: This means that I lack the relevant data.)

insert Figures 12.1-40 here

 84

4.5. Further Restrictions on Combinations of Functions

The patterns of the 40 languages shown in § 4.4 exhibit a bewildering diversity
– in fact, every language has its unique pattern, so that there are no two
languages in my sample that have the same system. My data provide evidence
for 37 different combinations of functions in indefinite pronoun series (see
Table 4 below), a much larger number than previous studies suggested. Thus,
with respect to indefinite pronouns, as elsewhere in the grammar, languages
vary widely. However, the variation is not unconstrained: All languages
conform to the adjacency requirement imposed by the implicational map,
which embodies the deeper unity behind the superficial diversity.
It should also be noted that in most languages several indefinites overlap in
their distribution, i.e. some functions may be expressed by several different
indefinite pronouns. Sometimes there are additional subtle meaning
distinctions, but in many cases two (or more) indefinites are equivalent for
most purposes. This massive overlap of indefinites shows that an approach to
grammatical meaning that relies on contrast cannot be correct. If grammatical
meaning consisted in contrasts in a tight system, we would expect to find only
patterns where the space of functions is neatly divided up among the different
indefinite series. But as in the case of tense, aspect, and modality markers (cf.
Bybee 1988), the grammatical meaning of indefiniteness markers must be due
to their semantic substance, not to contrasts in the system.
We have seen that the implicational map puts interesting constraints on the
distribution of indefinite pronoun series. But there are probably further
restrictions on possible combinations of functions in addition to the adjacency
requirement of the map. Given this requirement, there are 95 geometrically
possible combinations of functions that any one indefinite series could express,
but only 37 different combinations are attested in my data. Since several types
are systematically missing, it is unlikely that the other 58 combinations are all
accidental gaps. Table 4 lists the 37 attested combinations and gives an example
of each type (the digits represent the nine functions of Figure 5).

Table 4. Attested combinations of functions

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
 85

1 Russian koe- 45678 Catalan cap

12 Kazakh älde- 456789 English any
123 Serbian/Croatian ne- 4568 German je
12345 English some- 45689 Serbian/Croatian bilo
123456 German etwas 4589 Hungarian akár
1234567 Swedish någon 467 Italian nessuno
2 Kannada -oo 4678 Finnish -kaan
23 Basque bait? 468 Icelandic nokkur
2345 Latin ali-, Greek ka- 568 Russian by to ni bylo
23456789 Portuguese qualquer 5689 Bulgarian -to i da e
2345689 German irgend 589 Modern Greek -dhípote
345 Russian -nibud’ 67 Icelandic n-
3456 Ossetic is- 678 Maltese ebda
34567 Greek típota 6789 Kannada -uu
345678 Nanay -daa 689 German jeder
3456789 Hindi/Urdu bhii 7 German n-
34568 Lithuanian nors 89 Swedish som helst
345689 Dutch dan ook 9 Icelandic sem er
4567 Hebrew iš
––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
Some of the missing combinations are probably true accidental gaps, but many
others are apparently excluded for principled reasons, i.e. they cannot occur in
any language. Two such principles might be formulated along the following
lines:
(i) Principle 1: Combinations of less than three functions are not possible in
the middle of the map, i.e. in the area of functions 4, 5, 6, 8. This excludes the
combinations '4', '5', '6', '8', '45', '46', '58', '68'. (Perhaps the fact that no indefinite
expresses just function 3 is also related to this principle.)
(ii) Principle 2: Functions 9 (free choice) and 8 (comparative) are never
combined with function 1 (specific-known). This excludes combinations such as
'1234568', '1234678', '1234689', '12346789', '12358', '1234589'.
These two principles thus have to be added to the adjacency requirement of
the implicational map. Of course, they are also on need of further explanation.
A speculative explanation for Principle 1 will be suggested in § 5.4.4. Principle 2
can probably be explained by the need to distinguish between meanings that
are too different – the emphatic functions ‘comparative’ and ‘free choice’ are
just too different semantically from ‘specific known’ to be expressible in the
same way (cf. § 5.6.1, § 5.7.2 for more on the nature of these differences).

4.6. Earlier Formulations of Typological Implications

There are two earlier attempts at accounting for the cross-linguistic variation in
terms of implicational constraints: the implicational hierarchy of Edmondson
(1981, 1983) and the map-like schemas of Bernini and Ramat (1992). I became
aware of them only after I had arrived at my implicational map, and I see the
similarities between these two proposals and my own results as confirming the
soundness of the general approach. However, both these attempts are much
smaller in scope and have a number of shortcomings in comparison to my map.
Edmondson (1981, 1983) formulates an implicational hierarchy for various
kinds of negative-polarity items (auxiliaries, intensifying adverbs, indefinite
 86

pronouns). This hierarchy is shown in Figure 13, together with examples of the
four types of negative-polarity items predicted by this hierarchy.

Figure 13. Edmondson's implicational hierarchy for negative-polarity items

comparative ⊃ conditional ⊃ interrogative ⊃ negation

|–––––––––|
Malagasy velively
|––––––––––––––––––––––––––|
Italian mica '(not) at all'
|–––––––––––––––––––––––––––––––––––––––––––|
Latin quis-quam 'anyone'
|–––––––––––––––––––––––––––––––––––––––––––––––––––––––|
English any

Edmondson's implicational hierarchy bears obvious resemblances to my

implicational map, especially with respect to the functions 'negative',
'interrogative' and 'conditional'. Only the position of the comparative function
is incompatible with my map, and the data show that Edmondson's hierarchy
cannot be correct: Many languages have indefinites that express just the
comparative and the free-choice function, but not the conditional, question and
negation functions, e.g. Lithuanian bet-indefinites, Swedish som helst-
indefinites, Basque edo-indefinites and Romanian ori-indefinites. In addition,
the implicational relations cannot be expressed by a hierarchy, because some
indefinites that are used in the conditional and interrogative functions are not
used in the negative functions (e.g. the Russian -nibud'-indefinites).
Bernini and Ramat (1992: ch. 6) distinguish just three main functions of
indefinite pronouns: N (cf. English no-), A (any-), and S (some-). Technically, N
is defined as the indefinite used in elliptical answers, as in (141); A is defined as
the indefinite used in positive or negative questions (141-42); and S is defined
as the indefinite used in positive (and presumably realis) answers, as in (142).

(141) Did you see anyone? — No, nobody.

(142) Didn’t you see anyone? — Yes, (I did see) someone.

Bernini's and Ramat’s map-like schema looks as follows (1992: 126):

(143) negative positive

answer N S
question A A

They do not claim explicitly that this schema expresses an adjacency re-
quirement (and they do not call their schema an implicational map), but in the
seven types that they distinguish, all indefinites (indicated by the letters x, y, z)
occupy contiguous areas, as shown in Figure 14 (Bernini and Ramat 1992: 127-
36).
 87

Figure 14. Types of indefinite pronoun distributions in European languages according

to Bernini and Ramat

neg. pos. neg. pos. neg. pos. neg. pos.

answer x y x y x y x y
question neg z z neg x z (neg) x y (neg) y y
e.g. English e.g. Russian e.g. Romanian e.g. Norwegian
(no/some/any) (ni-/koe-/nibud’) (ni-/-va) (ingen/noen)
N vs. S vs. A N vs. S1 vs. S2 N vs. S N vs. SA

neg. pos. neg. pos. neg. pos.

answer x y neg x y neg x x
question (neg) x x neg x x neg x x
e.g. M. Greek e.g. Basque e.g. Danish
(típota/káti) (i-/-bait) (noget)
NA vs. S A vs. S AS

The similarities between Bernini's and Ramat’s and my approach are evident.
The fact that they distinguish fewer functions results from the scope of their
study, which includes all negation patterns, but includes non-negative
indefinites only peripherally.
However, there are several problems with their approach. First, the use of
the terms N, A and S is confusing. On the one hand, these are said to refer to
functions, but then they are also used to refer to specific distributional patterns.
For example, Romanian is said to have an N vs. S pattern, although the -va-
indefinites are used for the functions S and A. This is mainly a terminological
problem, but the result is that it is not always clear what the authors mean.
Second, I see no reason why negative questions should be given special
prominence. Apparently English is the only language where negative questions
are said to have different indefinites than negative declarative sentences, but
this is not the case (Did you see nobody? is quite acceptable, as is I didn’t see
anybody.) A third problem is that negative elliptical answers are used to define
the function 'N'. As I argue in § 8.1.2, elliptical answers are not a good criterion
for defining negative pronouns.
In conclusion, although the proposals advanced by Edmondson and by
Bernini and Ramat are problematic in several respects, they are interesting
precursors to the implicational map proposed in this work.

4.7. Some Finer Distinctions

The nine functions of indefinite pronoun series that are represented on the
implicational map constitute the core functions with respect to which
indefinites can differ within or across languages. However, they do not exhaust
the possibilities. In this section I mention a couple of further distinctions that
are relevant but for which I could not get systematic cross-linguistic data. Some
of these distinctions can probably be eventually be integrated into the map,
while others seem to belong to a different dimension of variation.

4.7.1. Comparatives
 88

Hoeksema (1983) shows that there are two types of comparative standard
expressions in which indefinites may behave differently, nominal comparatives
and clausal comparatives. Only clausal comparatives allow ook maar-indefinites
in Dutch (cf. 144a), whereas nominal comparatives do not (cf. 144b).

(144) a. Jansen was sneller dan ook maar iemand ooit geweest was.
Jansen was faster than INDEF anyone ever been had
'Jansen was faster than anyone had ever been.'
b. *Jansen was sneller dan ook maar iemand.
Jansen was faster than INDEF anyone
'Jansen was faster than anyone.'

Similarly, in Serbian/Croatian only clausal comparatives (marked by nego

'than') allow i-indefinites (cf. 145), whereas nominal comparatives (marked by
od 'from, than') do not (cf. 146) (Progovac 1994: 71).

(145) Marija je viš a nego i-ko u razredu.

Marija is taller than INDEF-who in class
'Marija is taller than anyone in the class.'
(146) Marija je viš a od *i-koga/bilo koga u razredu.
Marija is taller from INDEF-whom in class
'Marija is taller than anyone in the class.'

The data from Dutch and Serbian/Croatian suggest that the comparative
function should eventually be split up into two functions, the clausal
comparative function being closer to the indirect negation function (Dutch ook
maar-indefinites are used in the indirect-negation and (marginally) in the direct-
negation functions), and the nominal comparative function being closer to the
free-choice function (Serbian/Croatian i-indefinites are not used in the free-
choice function).iv

4.7.2. Indirect negation

The indirect-negation function can be subdivided at least into the two functions
'implicit negation' (i.e. expressions like 'without', 'lack') and 'superordinate
negation' (i.e. sentential negation in a superordinate clause). German jeder and
Latvian jeb WH are examples of indefinites that behave differently in these two
contexts:

(147) German jeder

a. (implicit) ohne jede Vorwarnung
'without any warning'
b. (superordinate) #Es ist nicht nötig, dass jeder kommt.
'It is not necessary that anybody comes.'

(148) Latvian jeb WH

a. (implicit) Tru@ka jeb-ka@ da e@dama@.
lacks INDEF-which food
'Any food is lacking.'
b. (superordinate) *Es nedoma@ju, ka jeb kas atna@ks.
I not:think that INDEF who came
'I don't think that anybody came.'
 89

I have too few examples of such a contrast, but these two cases suggest that the
sub-case 'implicit negation' is closer to the comparative and free-choice
functions, whereas the sub-case 'superordinate negation' is closer to the irrealis-
non-specific function.

4.7.3. Questions

Strictly speaking, the 'question' function on the map refers only to polar
questions. In many cases, different indefinites are used in parametric (or 'WH-')
questions, as for instance in Italian, where niente from the negative-polarity
series is not used in parametric questions, cf. (149a-b).

(149) Italian (Zanuttini 1991: 141)

a. Ha detto niente di nuovo?
has said anything of new
'Has he said anything new?'
b. Quando ha detto qualcosa/*niente di nuovo?
when has said something/anything of new
'When did he say something new?'
c. Quando mai ha detto niente di nuovo?
when ever has said anything of new
'When has he ever said anything new?'

However, two kinds of parametric questions must be distinguished: Ordinary

questions asking for information, and rhetorical questions that presuppose a
negative answer. When a negative-polarity indefinite is used in parametric
questions, usually a rhetorical reading results, as in (149c). Similarly, Progovac
(1994: 98) reports that (150) from Serbian/Croatian, with the negative-polarity
indefinite ikoga, must have rhetorical force. The situation is similar with English
any-indefinites (cf. Lakoff 1969: 609, Borkin 1971, Lawler 1971) and Modern
Greek típota-indefinites (cf. Tsimpli and Roussou 1993: 150-57).

(150) Serbian/Croatian
Kada je Marija i-koga uvredila?
when is Marija INDEF-whom insulted
'When did Marija insult anyone?'

It appears that ordinary parametric questions must be situated to the left of the
'question' function on the map, perhaps just coinciding with the 'irrealis non-
specific' function, whereas rhetorical parametric questions must be close to the
indirect negation function, perhaps just coinciding with the 'question' function.
Positive and negative expectations are also sometimes relevant in polar
questions, as is discussed in the next section.

4.7.4. Expectations of the speaker

In a number of unrelated languages, the pragmatic contrast between positive

and negative or neutral expectations of the speaker is relevant for the choice of
the indefinite pronoun series within the conditional and (polar) question
functions. Invariably the indefinite series that is chosen for positive
expectations is the one that has more functions to the left of the
question/conditional functions on the map, and the series that is chosen for
neutral or negative expectations is the one that has more functions to the right
 90

on the map. It is therefore tempting to propose a further subdivision of the

question and conditional functions and thus to capture this generalization by
making the map more fine-grained. Although such a step would not be against
the spirit of the whole enterprise, it meets with a technical problem: It is not
possible to draw a two-dimensional map in such a way that the adjeacency
requirement is always satisfied – indefinites like the Russian -nibud'-series
cannot be accommodated (see the discussion below). Thus, the parameter of
speaker expectations must be treated as belonging to a different, though
related, dimension. The relevance of speaker expectations has been reported in
the literature for English, Catalan, Russian, and Japanese.
The importance of pragmatic conditions for the English some/any distinction
was originally pointed out by Bolinger (1960) and became widely known
through Lakoff (1969). In questions and conditional protases, the unmarked
case is the occurrence of negative-polarity indefinites of the any-series.
However, if the speaker’s expectations are positive, indefinites of the some-
series are used. Consider Lakoff’s examples:

(151) English (Lakoff 1969: 609-11)

a. Do you think those men want to do some work?
(Because my road needs to be repaired.)
b. Do you think those men want to do any work?
(Because they’ve been standing around all morning telling dirty jokes.)

In (151a), the speaker expects (or hopes) that the answer to his question will be
positive, so some is used. By contrast, in (151b) the speaker expects the answer
to be negative, so only any can be used here.
Two special types of questions that consistently behave according to this
generalization are invitations and requests in the form of a question and
rhetorical questions. In invitations and requests, the speaker expects (or hopes
for) a positive answer, so the some-series must be used (Bolinger 1977: 31):

(152) a. (invitation) Good morning, will you have some <*any> breakfast?
b. (request) May I have something for somebody who is coming by at
four o’ clock to pick it up?

The situation is quite similar with conditional protases. When the speaker
expects that the condition will be fulfilled (or assumes that it has been fulfilled),
some-indefinites are used, but when the expectation is that it will not be
fulfilled, any-indefinites are used. Again, (153a-b) are from Lakoff (1969: 610).

(153) a. Unicorns are mythical beasts: if Joan seens any <??some> unicorns out
there, I’ll eat my hat.
b. If John sees some <??any> goldfish in that tank, it’s not surprising: there
are lots of them in there.

Lakoff also points out special uses of conditionals: conditional promises and
warnings. In conditional promises, the speaker hopes that the hearer will fulfill
the condition, so that (154a) has some, whereas in conditional warnings, the
speaker hopes that the hearer will not fulfill the condition, so that (154b) has
any (Lakoff 1969: 611).

(154) a. I promise you that, if you eat some <*any> spinach, I’ll give you ten
 91

dollars.
b. I warn you that, if you eat any <*some> candy, I’ll whip you.

The relevance of speaker expectations is found in other languages as well. In

Catalan, as in English, negative-polarity indefinites are used when speaker
expectations are negative, whereas positive-polarity indefinites are used when
they are positive. According to Lleó (1983: 326), in (155a) the speaker considers
it more likely that someone came to see her, whereas in (155b) she considers it
more likely that the answer to her question is negative.

(155) Catalan (Lleó 1983: 326)

a. Ha vingut algú a veure’m?
has come someone to see-me
‘Has someone come to see me?’
b. Ha vingut ningú a veure’m?
has come anyone to see-me
‘Has anyone come to see me?’

In questions that are used as requests and therefore expect a positive answer,
the negative-polarity indefinite is impossible (Lleó 1983: 313).

(156) Podríeu tancar alguna <*cap> finestra, si us plau?

could close some any window if you please
‘Could you close some window, please?’

Analogously, in conditional protases a negative-polarity indefinite (157a)

indicates lower likelihood than an affirmative-polarity indefinite (157b) (Lleó
1983: 326):

(157) a. Si en Joan vol res, ja ens ho dirà.

if ART Joan wants anything already us it will:say
‘If Joan wants anything, he will tell us about it.’
b. Si en Joan vol alguna cosa, ja ens ho dirà.
if ART Joan wants some thing already us it say:FUT.3Sg
‘If Joan wants something, he will tell us about it.’

Finally, conditional warnings and promises behave as expected from the

above discussion on English:

(158) a. Si dius res <?alguna cosa>, et castigaran.

if say:2SG anything something you they’ll:punish
‘If you say anything, you’ll be punished.’
b. Si dius alguna cosa <?res>, et donaran un premi.
if say:2SG something anything you give:FUT.3PL a prize
‘If you say something, you’ll get a prize.’

Speaker expectations have been shown to be relevant also in Russian

(Paillard 1984). In (159a), there is no particular expectation on the part of the
speaker, whereas in (159b) the context makes it clear that the speaker expects a
positive answer (cf. also Dahl 1970: 38).

(159) Russian (Paillard 1984: 272-3)

a. Vy kogo-nibud’ ž dete?
you who-INDEF wait
 92

‘Are you waiting for anybody?’

b. Ja celuju Valjuš ku i vsë prisluš ivajus’: ne razdastsja li v
I kiss Valjuš ka and still listen not sounds Q in
koridore zvonok. No zvonka vsë net. – Ty ž deš ’ kogo-to? –

corridor ring but ring still [Link] you wait who-INDEF

spraš ivaet Valjuš ka.
asks Valjuš ka
‘I kiss Valjuš ka and continue to listen if the bell rings in the
corridor. But there is still no ring. – Are you waiting for someone?
– asks Valjuš ka.’

In confirmation questions introduced by pravda, č to...? ‘is it true that...?’, only -

to-indefinites are admissible (Paillard 1984: 272):

(160) Pravda, č to ty kupil kakuju-to <*?kakuju-nibud’>

true that you bought which-INDEF which-INDEF
francuzskuju knigu?
French book
‘Is it true that you bought a French book?’

By contrast, in questions marked with the question particle razve, which

indicates that the speaker would be surprised at a positive answer, only -
nibud’-indefinites are possible (Paillard 1984: 272):

(161) Razve sluč ilos’ č to-nibud’ <*?č to-to>, o č em ja ne znaju?

PRT happened what-INDEF what-INDEF about that I NEG know
‘Did anything happen that I don’t know about?’

Since Russian -nibud'-indefinites are also used in a function to the left of the
question and conditional functions, the irrealis-non-specific function, it is not
possible to express the similarity between English, Catalan and Russian by
making the implicational map more fine-grained. Still, Russian conforms to the
above generalization that the indefinite chosen for positive speaker
expectations is the one that is located further to the left on the map.
A type of question which usually implies positive speaker expectations is the
negated polar question (this seems to be true cross-linguistically). In Japanese,
there is a fairly rigid division between the ka-series of indefinite pronouns,
which occurs in non-negative sentences, and the mo-series, which occurs in
negative sentences.

Japanese (cf. Hasegawa 1991: 271)

(162) a. Dare-ka ki-ta.
who-INDEF come-PAST
‘Someone came.’
b. Dare-mo /*dare-ka ko-nakat-ta.
who-INDEF who-INDEFcome-NEG-PAST
‘Nobody came. /Somebody didn’t come.’

In contrast to English, where the some-series is possible in negative sentences

but is then interpreted as not being in the scope of the negation, the Japanese
ka-series is completely impossible with negation, where the mo-series is
obligatory. However, in negated polar questions this ban is lifted. As McGloin
(1976: 409) points out, ka-indefinites are used in polar questions when the
 93

speaker has a firm basis for believing that the presupposition is true, i.e. when
the speaker expects a positive answer. Thus, (163a) is an information-seeking
question, whereas (163b) is interpreted as an explicit invitation (McGloin 1976:
409).

(163) a. Nani-mo tabe-mas-en ka?

what-INDEF eat-POL-NEG Q
‘Are you not going to eat anything?’
b. Nani-ka tabe-mas-en ka?
what-INDEF eat-POL-NEG Q
‘Aren’t you going to eat something?’

Notice that the same is true for the English some/any contrast (cf. e.g. Bolinger
1977: 24).
And again the situation is quite parallel in conditional antecedents. When the
speaker makes a promise and therefore has positive expectations, the ka-series
is possible with negation, as shown in (164b) (McGloin 1976: 415).

(164) a. Nani-mo tabe-nakat-tara, eiga e ture-te it-te

what-INDEF eat-NEG-COND movie to bring-CONV go-CONV
age-na-i yo.
give-NEG-PRES AFF
‘If you don’t eat anything, I won’t take you to the movie.’
b. Nani-ka tabe-nakat-tara,...
what-INDEF eat-NEG-COND
‘Unless you eat something, I won’t take you to the movie.’

Thus, we have seen four languages (English, Catalan, Russian, Japanese)

where speaker expectations play an important role in selecting an indefinite
pronoun series. This is all the more remarkable as the uses of the two different
series are different in other ways in the four languages: English some/any and
Catalan algú/ningú are non-polarity/negative-polarity items, Japanese ka/mo are
positive/negative, and Russian -to/-nibud’ are specific/non-specific. That
negative expectations are expressed by indefinites that are located further to the
right on the implicational map is of course just what we expect, because the
negation functions are also located on the right. However, the details of an
explanation of these facts remain to be worked out.

4.8. Toward an Explanation of the Implicational Map

At this stage we have completed the third step of typological analysis,
implicational generalizations. The next and final step is the explanation of the
observed patterns. Why are indefinite pronoun series distributed over the
various functions in such a way that the semantic map of § 4.3 results?
I have already alluded to an obvious interpretation of the implicational map:
Closeness on the map means close semantic relatedness. This is of course
intuitively very plausible, but we have to ask further: In what way are the
various functions semantically related? Only once we have an explicit account
of the semantic relationships between the functions of indefinite pronoun series
can we say that we understand the typological pattern.
The next chapter will discuss a number of theoretical approaches to the
interrelations of the various functions of indefinite pronouns, which is a
necessary prerequisite for explanatory success.
 94

i
Note that I follow the convention (used in Comrie (1976), Bybee et al. (1994), and
others) of capitalizing language-particular grammatical categories.
ii
See the works by modern Gesamtbedeutung semanticists like William Diver, Erica
García, Robert Kirsner, Yishai Tobin, and others. None of them has anything to say on
cross-linguistic variation. (An interesting exception to this generalization is David Gil's
work (e.g. 1993), general meanings are investigated cross-linguitically.)
iii
Implicational maps are different from implicational hierarchies in that hierarchies
make predictions for all cases that are higher than an element on the hierarchy, whereas
maps make predictions only about adjacency. Thus, implicational maps represent
weaker generalizations than implicational hierarchies.
iv
Hoeksema (1983) shows that clausal comparatives, but not nominal comparatives, are
semantically 'anti-additive' (in Boolean terms), like indirect-negation contexts, so this
distribution is expected.
 104

Theoretical Approaches to the Functions of

Indefinite Pronouns

This chapter will review and discuss various theoretical approaches to the
semantic and syntactic functions that are found in different indefinite pronoun
series. These approaches generally concern only a small subset of the functions
distinguished in chapter 3. Thus, the approaches dealt with in this chapter are
not necessarily to be seen as alternatives, and some of them could and should
be regarded as complementary (though others are indeed incompatible). Most
of these theoretical approaches are based on a much narrower range of data
than this study, often only data from a single language. My discussion will
highlight the points that are illuminated by the broad cross-linguistic data
assembled in this study. In § 5.6, I make an attempt to show how the various
functions of indefinite pronouns fit together and how the implicational map of
chapter 4 is motivated. Finally, § 5.7 deals with focusing and sentence accent.

5.1. Structuralist Semantics

My discussion begins with structuralist semantics, the study of meaning in the
spirit of Ferdinand de Saussure’s foundational work. (While linguists in the
19th century developed theoretical perspectives in many areas of semantics,
pragmatics, and syntax, the functions of indefinite pronouns were not among
them. The descriptions of indefinite pronouns in pre-structuralist reference
grammars are too sketchy to warrant discussion here.)
The fundamental assumption of structuralist linguistics is that language is a
system whose parts must be defined and described on the basis of their place in
the system and their relation to each other, not on the basis of their own
intrinsic properties (as the opposing view, which might be called 'substantivist',
would maintain). A typical structuralist analysis of a class of linguistic units
classifies them on the basis of several binary distinctions (or 'oppositions').
These binary distinctions are often written in the form '[+α] vs. [–α]' and are
called features ('semantic features' in the case of a semantic analysis).
Structuralist semantics, especially in application to grammatical meaning,
was pioneered by Roman Jakobson (especially 1936, 1957) and was widely
adopted by linguists in the following decades, particularly in Europe. There are
a substantial number of studies on indefinite pronouns that came from this
tradition (Greimas 1963, Veyrenc 1964, Manoliu-Manea 1966, Kř íž ková 1971,
Levin 1973, Dausendschön-Gay 1977, Ponomareff 1978, Pilka 1984,
Gluszkowska 1985). Studies from the generative tradition often preserve much
of the spirit of classical structuralism in their emphasis on abstract formal
systematization (e.g. Labov 1972).
In contrast to the other approaches to be dealt with below, no particular
analysis in this approach has become famous and widely discussed, so an il-
lustrative example should be sufficient here. A typical case of a structuralist
study is provided by Veyrenc (1964). Veyrenc classifies four Russian indefinite
 105

series (-to, -nibud’, koe-, -libo) in the way shown in (165), where kto ‘who’ is taken
as representative of the whole series.

(165) unique non-unique

real kto-to koe-kto
virtual kto-nibud’ kto-libo

The feature 'real/virtual' corresponds closely to my specific/non-specific

distinction (§ 3.2.3), and Veyrenc’s classification makes sense this far: the to-
and koe-series are indeed specific most of the time, while the -nibud’- and the -
libo-series are non-specific. Veyrenc does not mention the possibility of the -to-
series being used non-specifically (cf. Appendix A, Section 16), but this does not
pose a threat to his analysis. He could simply take this as evidence for the un-
marked status of the feature value 'virtual' ('[–real]': When an opposition is
neutralized, neutralization is invariably in favor of the unmarked member of
the opposition.
However, the other feature ('unique/non-unique') is very weakly motivated.
Veyrenc justifies it with reference to the plural-like meaning of the koe-series
(koe-gde ‘somewhere, here and there’, koe-č to ‘something, this and that’). Now
the schema in (165) predicts that the relation between the -nibud’-series and
the -libo-series can be described in the same way, but this is not in fact the case:
The -libo-series does not have a plural-like meaning, but rather it has its most
important function in indirect-negation and comparative functions (cf.
Appendix A, Section 16). Evidently, the binary classification of (165) is a
Procrustean bed into which Veyrenc squeezes a distinction that is of a very
different nature.i While Veyrenc’s analysis is certainly not one of the best
structuralist analyses, it serves well as an illustration of the approach and the
problems associated with it.
First, it is often left unclear what the status of the semantic features is. In
many analyses, it seems that they are created ad hoc and are not interpreted
further, so that they do not have any additional explanatory power and might
as well be replaced by arbitrary numbers. Uninterpreted semantic features
become interesting only if it is hypothesized that there is a universal restricted
(possibly innate) set of them (see Wierzbicka (1980a: ch. 1) for discussion).
Second, the binary privative oppositions that are usually employed,
contrasting an unmarked feature value [–α] with a marked feature value [+α],
often simply do not do justice to the facts. Even if the classification of (165) for
Russian indefinites could be accepted, there are still other indefinite series in
Russian that remain to be accounted for. For instance, the by to ni bylo-series is
in many ways similar to the -libo-series (cf. Appendix A), with which it overlaps
to a certain extent, but it also differs from it. One would have to postulate a
third opposition (say 'emphatic/non-emphatic'), but this would lead to new
possible combinations of features that are not instantiated (e.g. 'real + unique +
emphatic') and would have to be arbitrarily excluded.
Third, the view that grammatical meaning consists in oppositions rather
than in inherent semantic substance leads us to expect no overlap between
contrasting units. But we do find massive overlap of the functions of
grammatical items in many areas, including indefinite pronouns (cf. § 4.5).
Thus, there are certain contexts where either the -nibud’-series or the -libo-series
may be used with only a stylistic difference, and the -to-series and the -nibud’-
series also overlap.
 106

Fourth, structuralist semantics makes wrong predictions about semantic

change. An analysis in terms of binary oppositions would lead us to expect
changes consisting in a feature switching from a plus value to a minus value (or
vice versa), or in adding or replacing a feature. But semantic change seems to
proceed in a gradual fashion, with an expression acquiring only one slightly
different new use at a time. There is good evidence that the -libo-indefinites
originally were free-choice indefinites (§ 6.2.2). The -libo-indefinites lack the
free-choice function now, but they still have the comparative function (very
often expressed by free-choice indefinites in other languages). In addition, they
now have the indirect-negation function and the question/conditional function,
overlapping here with the -nibud’-series. The analysis of (165) does not even
begin to account for these (hypothesized) changes. Quite generally,
grammatical meanings often retain subtle traces of earlier (especially lexical)
meanings which are better accounted for by a substantivist approach that
attributes semantic substance to them (cf. Bybee 1988 for discussion).
Fifth, structuralist semantic analyses of different languages are rarely
comparable and thus are not a suitable basis for typological studies (cf. § 4.2.2).
Of course, if structuralist semanticists operated with a sufficiently restricted
universal set of semantic features, comparability would be ensured (much as
phonemic systems are comparable when based on a restricted universal set of
phonological features). But in the absence of such a restricted set of features,
analysts typically attempt to build a system that best captures the peculiarities
of a given language, with the result that existing analyses of distinct languages
are often radically different.
For all these reasons, I am skeptical about the approach of structuralist
semantics. This does not, of course, mean that there is no value in it – the
studies of indefinite pronouns cited above all contain valuable data and discus-
sion, much of which has been directly useful for this study. But I will base my
conclusions on a substantivist rather than a structuralist view of language.

5.2. Logical Semantics

Since Gottlob Frege’s and Bertrand Russell’s writings, modern linguistics has
been influenced by the thinking and notational devices of mathemati-
cal/philosophical logic. Indefinite pronouns are one of the areas where logical
semanticists have had especially many things to say, based on the predicate
calculus of modern logic. There are two main aspects of indefinites which have
been discussed from a logical-semantic point of view: free-choice indefinites (§
5.2.1) and non-specific indefinites (§ 5.2.2).
As in the preceding section, my evaluation of the logical-semantic approach
will be rather critical, and I will discuss alternative ways of dealing with the
issues raised by logical semantics in § 5.4-5.

5.2.1. Free-choice indefinites and universal or existential quantifiers

There is a huge literature on quantification and corresponding phenomena in

grammar, beginning with the late [Link] Most of this literature is not
immediately relevant to my more narrow concern of exploring the relation
between meaning and form in indefinite pronouns. In the following, I will
concentrate on the logical-semantic literature that deals with indefinite
pronouns in the narrow sense.
 107

In general, logical formulas containing an existential quantifier correspond

to sentences with specific indefinites, and logical formulas containing a
universal quantifier correspond to sentences with universal pronouns,iii as
illustrated in (166-167).

(166) a. (∃x) (Pavlina saw x)

b. Pavlina saw something.
(167) a. (∀x) (Waldemar ate x)
b. Waldemar ate everything.

But what is the logical structure corresponding to free-choice indefinites?

Rather than introduce a new quantifier, logicians and linguists working within
logical semantics have tried to express the meaning of free-choice quantifiers in
terms of the two classical quantifiers, the universal and the existential
quantifiers. Since most of the relevant discussion has taken place in English and
no typological evidence has generally been taken into account, the English any-
series (most often the determiner any) has been at the center of the attention.
This has led to additional problems and challenges, because the English any-
series is also widely used in non-free-choice uses, in particular in negation and
negative-polarity contexts. The prevailing view among linguists seems to be
that two anys have to be distinguished: a 'free-choice any', corresponding to the
universal quantifier, and a 'polarity-sensitive any', corresponding to the
existential quantifier (cf. Carlson 1980, 1981). But a sizable minority has
defended the view that any has a unitary meaning corresponding to the
universal quantifier (e.g. Savin 1974, Gil 1991). The third logical possibility,
unitary meaning corresponding to the existential quantifier, has also found
adherents (Davison 1980, Kadmon and Landman 1993).

[Link]. Univocal universal any. The view that any invariably corresponds to
the universal quantifier was put forward in two famous works of philosophical
logic (Reichenbach 1947: §21, Quine 1960: §29). According to Reichenbach and
Quine, any differs from every in that the universal quantifier to which it
corresponds always has wide scope with respect to other logical operators such
as negation and material implication. In (168), an example of the free-choice use
of any, there is no additional operator, so that any and every are logically
equivalent.

(168) a. (∀x) (x knows this secret)

b. Anybody knows this secret.
c. Everybody knows this secret.

In (169-170), examples of any in negative-polarity contexts, we see how the

difference between any and every is reflected in the different scope relations
between the universal quantifier and the additional operator.

(169) a. (∀: member x) (x contributes ⊃ x gets a poppy)iv

b. If any member contributes, he or she gets a poppy.
c. ((∀: member x) (x contributes)) ⊃ (I’ll be surprised)
d. If every member contributes, I’ll be surprised.
(170) a. (∀: poem x) ¬ (I know x)
b. I do not know any poem.
 108

c. ¬ (∀: poem x) (I know x)

d. I do not know every poem.

If we only consider these examples, the analysis seems satisfactory. Not only
does it allow us to maintain a unitary, non-ambiguous meaning of any, it also
accounts for the fact that an existential meaning seems to be associated with
negative-polarity any. Indeed, (169b) and (170b) can be paraphrased by (171b)
and (172b), which contain the existential expression ‘there is’.

(171) a. ((∃: member x)(x contributes)) ⊃ (x gets a poppy)

b. If there is a member who contributes, he or she gets a poppy.
(172) a. ¬ (∃: poem x) (I know x)
b. There is no poem that I know.

Given the general logical equivalences in (173-174), it follows that (171a) and
(172a) are equivalent to (169a) and (170a), and the (near) synonymy between
(171b) and (169b), and between (172b) and (170b) is explained.

(173) (∀x) (Fx ⊃ p) ≡ ((∃x)(Fx)) ⊃ p

(174) (∀x) ¬ (Fx) ≡ ¬ (∃x)(Fx)

Reichenbach even offers an explanation for the fact that any always has the
widest possible scope: He says that any corresponds to a free variable. As a rule,
logical formulas must be closed in the argument variables, and expressions
containing free variables do not have a determinate truth value. But,
Reichenbach claims (1947: 105), 'when an expression containing a free variable
is true for all values of the variable, it can be asserted.' This idea is not made
precise, but it has a certain intuitive plausibility, and it explains why the scope
of any is always the whole formula (Fx ≡ (∀x)(Fx), ¬ Fx ≡ (∀x) ¬ Fx, Fx ⊃ p ≡
(∀x)(Fx) ⊃ p, etc.).

[Link]. Negative-polarity any as existential. Whereas the univocal analysis of

any has been adopted by some linguists (Bolinger 1960, Bach 1968: 112, Savin
1974, Schmid 1980, Hintikka 1980, LeGrand 1974, Gil 1991), other linguists have
objected that any should correspond to an existential quantifier in negative-
polarity contexts like (45b), (46b). The following five arguments (among others)
have been cited in the literature for the existential nature of negative-polarity
any (Dahl 1970: 38-39, Horn 1972: §2.35, Davison 1980, Carlson 1980, 1981,
Ladusaw 1980: 94-104, Linebarger 1980, 1981: ch. 5, and many others).
First, any-indefinites are ambiguous in negative-polarity environments, as
shown in (175-177). In one of the readings, they are roughly equivalent to every-
pronouns (the 'universal reading'), and in the other reading they are roughly
equivalent to some- or no-pronouns (the 'existential reading'). Note that the two
readings are in fact different suprasegmentally (see Ladd (1980: 160) for some
discussion of the intonational differences).

(175) You can’t do anything here.

a. ‘You can do nothing here.’ (existential reading)
b. ‘You can’t do everything here.’ (universal reading)
 109

(176) If anybody can swim the channel, I can do it.

a. ‘If somebody can swim the channel, I can do it.’ (existential reading)
b. ‘If everybody can swim the channel, I can do it.’ (universal reading)
(177) Can anybody swim the channel?
a. ‘Can somebody swim the channel?’ (existential reading)
b. ‘Can everybody swim the channel?’ (universal reading)

Such cases seem to show that there are two different anys, one corresponding to
a universal quantifier, and one corresponding to an existential quantifier.
Second, in many other languages free-choice indefinites and indefinites in
negative-polarity environments are formally different. Horn (1972: 131)
observes 'the isolation of the situation in English from the usual trend en-
countered in the languages of the world to separate the two cases morpho-
logically'. It is unclear, however, what the basis for Horn’s claim is. My data
show that the situation in English is certainly not 'isolated': 18 of my 40 sample
languages have indefinites which are used both in the free-choice function and
in conditionals. But on the other hand, 19 languages indeed have special free-
choice indefinites which cannot be used in conditionals and questions. And
only 12 languages have indefinites which are used both in the free-choice
function and in the direct-negation function.
Third, only negative-polarity any may occur in existential sentences with
there (cf. 171-172), whereas free-choice any is impossible in such sentences
(*There is any apple that you can take).
Fourth, only free-choice any may be modified by the particles almost/nearly
and absolutely:

(178) a. (free choice) Absolutely/almost anyone can ride a bike.

b. (negative polarity) Has (*absolutely/*almost) anyone seen my bike?

Fifth, the wide-scope universal analysis does not account for all occurrences
of any, e.g. in questions. Sentence (179a) is not correctly paraphrased by (179b).

(179) a. Has wealth made anyone happier?

b. (∀: person x) (has wealth made x happier?)

I might add a sixth argument from the cross-linguistic distribution of in-

definites on the implicational map that we saw in chapter 4: In many
languages, the indefinites that occur in negative-polarity contexts are also used
in irrealis non-specific contexts like imperatives or modal contexts (e.g. Russian
-libo, Lithuanian -nors, Modern Greek típota, Hindi/Urdu bhii, Yakut eme,
Georgian -me, Kannada -aadaruu). In such contexts, the wide-scope universal
analysis is completely ruled out.
For further arguments see Carlson (1980, 1981) (especially concerning more
complicated scope interactions) and Ladusaw (1980) (especially concerning the
formulation of compositional meaning).
It should also be mentioned, however, that there is an additional argument
against separating a free-choice any and a negative-polarity any, in addition to
the general preference for univocal analyses: There are occurrences of any-
indefinites that are, so to speak, neutral between free-choice any and negative-
polarity any. As Carlson (1981: 13) observes, sentences (180a-c) have only one
reading.
 110

(180) a. Bob can run faster than anyone.

b. Bob is unlikely to kick anyone.
c. For Bob to eat anything now would be impossible.

Note that these are precisely the uses (comparative, indirect negation) which
are between the free-choice function and the question and conditional functions
on the implicational map of chapter 4.

[Link]. Problems with the logical-semantic approach to free-choice indefi-

nites. Thus, the logical-semantic approach is in a dilemma: On the one hand,
there is an analysis that accounts for a core set of data and permits the desired
unitary treatment of any, and on the other hand, a closer look seems to show
that any is not only ambiguous, but ambiguous between two meanings
(existential and universal quantification) that are in a sense opposites.v
Another limitation of logical semantics is that it cannot express the
difference between any and every in cases where they are roughly equivalent. In
(181a) (showing free-choice any), we have some kind of universal meaning, but
as Vendler (1967) has pointed out, (181a) and (181b) do not have the same
meaning, and not even the same truth conditions.

(181) a. Any doctor will tell you that Stopsneeze has dangerous side effects.
b. Every doctor will tell you that Stopsneeze has dangerous side effects.

Assuming a meaning element of universal quantification in free-choice in-

definites is particularly suspect in cases like (182a), which is evidently very dif-
ferent from (182b).

(182) a. Take any apple.

b. Take every apple.

Thus, we are justified to be skeptical that universal quantification in the

standard logical sense is involved in free-choice indefinites at all.
It happens that Davison (1980) has claimed that free-choice indefinites are to
be understood as existential quantifiers, too, differing from negative-polarity
indefinites mainly in that they carry a pragmatic implicature of universality (cf.
also Carlson (1981: 16-20) for arguments for this view, and Kadmon and
Landman (1993) for a similar approach). It is not surprising that this option has
also been advocated, because it would again permit a univocal semantic
characterization of any. However, as has been shown in the preceding
subsection, there are many behavioral differences between free-choice any and
negative-polarity any which also argue against a univocal existential analysis.
Thus, although the logical-semantic approach has allowed us to ask many
interesting new questions, it has ultimately remained unsuccessful with respect
to the relation between free-choice indefinites and negative-polarity indefinites.
We will see in § 5.5 that a wider perspective was necessary to make new
progress.

5.2.2. (Non-)specific indefinites and referential opacity

The logical-semantic approach has also been applied to describe the difference
between specific and non-specific noun phrases. However, since English
indefinite pronouns do not make this distinction, (non-)specificity has not often
 111

been discussed with reference to indefinite pronouns in the logical-semantic

literature. The main motivation of these analyses has been to explicate
specific/non-specific ambiguities, not to explicate the meaning of a particular
indefinite expression.
The basic idea is that the two readings of a sentence like (183) can be dis-
tinguished by assigning different scope relations to the existential quantifier
that binds the variable corresponding to the ambiguous NP. In the specific
reading, the existential quantifier has scope over ‘want’ (184a), and in the non-
specific reading, it has narrow scope and is within the scope of ‘want’ (184b).

(183) Bob wants to marry a linguist.

(184) a. (specific) (∃: linguist x) (Bob wants to marry x)
b. (non-specific) Bob wants (∃: linguist x)(Bob to marry x)

The original idea seems to be due to Quine (1956), and it was later taken up by
several linguists, among others C. L. Baker (in an unpublished 1966 paper),
Bach (1968), Dahl (1970), Fodor (1970).
Unfortunately, some confusion has arisen as a result of the different goals of
logicians and linguists. While linguists are mostly interested in explaining
ambiguities, Quine as a logician was primarily occupied with the problem of
REFERENTIAL OPACITY. If the two inference rules of substitutivity of identicals
and existential generalization are not valid in a context, logicians say that the
context is REFERENTIALLY OPAQUE. The ‘want’ context of (183) is such an
opaque context:
(i) Existential generalization is not possible: It could be that there is no
linguist—at least in the non-specific reading; logicians also speak of an OPAQUE
or DE DICTO READING, as opposed to a TRANSPARENT or DE RE READING,
corresponding to the linguists’ specific reading.
(ii) NPs cannot be replaced by other NPs with the same extension without
affecting the truth value. This is best illustrated with a sentence like (185).

(185) Bob wants to marry the linguist he met at the last conference.

Suppose that the linguist he met at the last conference has the same extension as the
mother of Irene, Jonathan and Charlotte. Then one cannot infer from this and (185)
that Bob wants to marry the mother of Irene, Jonathan and Charlotte, because
Bob might not know that the two are identical.
Referential opacity is by no means restricted to contexts like ‘want’ that al-
low NPs to be non-specific. Sentences with non-volitional propositional-at-
titude verbs such as ‘believe’, ‘think’, ‘imagine’ are perhaps even more typical
examples of referential opacity (cf. Jackendoff (1983: ch. 11), Fauconnier (1985:
ch. 1) for convincing treatments of opacity). But a sentence like (186) does not
show the specific/non-specific ambiguity I am interested in [Link]

(186) Mary believes that a friend of mine is a busdriver.

Thus, the specific/non-specific distinction must not be collapsed with the

transparent/opaque distinction, as is sometimes done in the linguistic liter-
ature, even if only terminologically (Fodor 1970, Givón 1973, Lavric 1990: § 3.6).
There are a number of works in the literature in which the analysis of (non-
specificity) in terms of different scopes of an existential quantifier is applied to
indefinite pronouns, most notably Dahl (1970) for Russian -to/-nibud’ (and,
 112

following him, Rufiž ič ka (1973); Pilka (1984: 65) for Lithuanian kaž -/nors). As
we saw in § 3.2.3, (183) has two different translations in Russian and many
other languages, corresponding to the two readings in (184a-b). (In Russian,
these different translations appear only when an indefinite determiner, kakoj-to
or kakoj-nibud’ ‘some’ is used. Thus, (187a-b) correspond more closely to English
Bob wants to marry some linguist. When the determiner is omitted, the Russian
sentence is just as ambiguous as the English one.)

(187) Russian
a. Bob xoč et ž enit’sja na kakoj-to lingvistke.
Bob wants marry on which-INDEF linguist
b. Bob xoč et ž enit’sja na kakoj-nibud’ lingvistke.
Bob wants marry on which-INDEF linguist

Dahl proposes the generalization that 'a variable, bound by an existential

quantifier that stands first in a declarative sentence, will be realized as a to-
pronoun' (1970: 35). This not only accounts for ‘want’ contexts, conditionals,
and questions, but also for the fact that the non-specific -nibud’-series appears
when a distributive meaning is intended. In such cases, the existential
quantifier follows not a non-reality verb like ‘want’, but another quantifier:

(188) a. Vse ljudi č itajut č to-to.

all people read what-INDEF
‘Everybody is reading something.’
(non-distributive: everyone is reading the same thing)
b. (∃: thing x) (∀: people y) (y read x)

(189) a. Vse ljudi č itajut č to-nibud’.

all people read what-INDEF
‘Everybody is reading something.’
(distributive: each person may read something different)
b. (∀: people x) (∃: thing y) (x read y)

Dahl’s rule also accounts for the fact that a specific -to-indefinite may cooccur
with negation only if it is not in the scope of negation. (190-91) are parallel to
(188-89), except that a non-specific indefinite which is in the scope of negation
is obligatorily a ni-indefinite, not a -nibud’-indefinite.

(190) a. Ja ne videl kogo-to.

I not saw whom-INDEF
‘I didn’t see someone.’
b. (∃: person x) ¬ (I saw x)
(191) a. Ja ne videl nikogo <*kogo-nibud’>.
I not saw whom-INDEF whom-INDEF
‘I didn’t see anybody.’
b. ¬ (∃: person x) (I saw x)

But Dahl runs into problems with futures. A non-specific -nibud’-indefinite is

licensed in future sentences as well. Here Dahl is forced to introduce an abstract
performative verb PREDICT. Thus, the two sentences in (192-93) (a) are analyzed
as (192-93) (b).

(192) a. Doroga konč itsja gde-to.

 113

road ends where-INDEF

‘The road will end somewhere [specific].’
b. (∃: place x) I PREDICT (road end at x)
(193) a. Doroga konč itsja gde-nibud’.
road ends where-INDEF
‘The road will end somewhere [non-specific].’
b. I PREDICT (∃: place x) road end at x)

A similar difficulty arises with non-reality verbs that are not followed by a
subordinate clause, as shown in (194a) and its English translation (194b).

(194) a. Ja poprosila ee o kakoj-to/ kakoj-nibud’ knige.

I asked her for which-INDEF which-INDEF book
b. I asked her for some [specific/non-specific] book.

Here we have a difference between Russian -to and -nibud’, and ambiguity in
English, but there are no two different syntactic positions for the existential
quantifier. Again, the only solution for the logical-semantic approach seems to
be to posit an additional abstract predicate that does not appear at the surface
(Bach 1968, Fodor 1970: 17, 1977: 127-9). Thus, the two meanings of (194b) are
analyzed as in (195a-b).

(195) a. (∃: book x) (I asked her (SHE GIVE ME x))

b. I asked her ((∃: book x)(SHE GIVE ME x))

It would be preferable to do without such abstract [Link]

5.2.3. Conclusion

The logical-semantic approach has undoubtedly meant progress in our

understanding of the relevant semantic distinctions. But its limitations are
equally apparent: In both areas considered here, free-choice/negative-polarity
indefinites and specific/non-specific indefinites, some fundamental questions
were left unresolved.
I will try to show in § 5.4-5 that approaches that do not make crucial use of a
logical-semantic representation fare better in coming to grips with the se-
mantics of indefinite pronouns. But first we must consider syntactic approaches
to indefinite pronoun functions.

5.3. Syntactic Approaches

5.3.1. Transformational approaches

The tools developed by the transformational syntacticians in the early 1960s

were soon applied to indefinite pronouns (Lees 1960, Klima 1964). Klima
proposed to account for the alternation between some-, any- and no-indefinites
in English by various transformations, so that the sentences in (196a-b) share
the same deep structure, and (196c) has the same deep structure minus an
abstract negation element.

(196) English
 114

a. There wasn’t any snow falling.

b. There was no snow falling.
c. There was some snow falling.

Some is the spellout of an abstract element QUANT, any is the spellout of the
abstract elements INDEF + QUANT, and no is the spellout of NEG + INDEF +
QUANT. The incorporation of INDEF into QUANT, i.e. in fact the conversion of
some into any, is effected by a transformational rule that inserts INDEF if QUANT
is in the scope of a suitable trigger. Klima noted that in addition to negation,
'INDEF incorporation' is triggered by other expressions such as only, parametric
questions, if, before, lest, and 'adversative predicates' like difficult, afraid, ashamed,
refuse, reluctant, surprised, and he introduced a syntactic feature ‘affective’
shared by all these triggering expressions. This uninterpreted, arbitrary feature
‘affective’ was of course not very illuminating.
Later writers attempted to reduce the triggering environment for negative-
polarity any and other negative-polarity items to negation by proposing that the
core case of ‘affective’ expressions is negation, and that the other ‘affective’
expressions share a negative implicature (Baker 1970, Linebarger 1981, 1987).
But none of these proposals were felt to be totally convincing. The most
problematic case for the negative implicature view are relative clauses
modifying universally quantified NPs, as in (197). It is hard to see where the
negative implicature is here.

(197) Everyone who knew anything shared their knowledge with me.

An additional blow was dealt to syntactic analyses by the demonstration in

Bolinger (1960) and Lakoff (1969) that pragmatic conditions also play a role in
the distribution of some and any (see § 4.7.4). In the 1970s, interests of generative
syntacticians shifted in other directions, and it was not until the late 1980s that
a new attempt at accounting for indefinite pronouns in syntactic terms was
made (but see McCawley (1988: 546-561), where a Klima-style transformation is
still advocated).

5.3.2. A binding approach

A very interesting theory of negative-polarity indefinites in terms of Chomsky’s

(1981) Binding Theory was developed by Progovac (1994) (cf. also Progovac
1988, 1990, 1991, 1992a, 1992b, 1993), elaborating an idea of Milner (1979) and
Aoun (1986). Progovac points out some intriguing similarities between the
syntactic locality restrictions on reflexive pronouns and on negative-polarity
items and formulates a typology of kinds of negative-polarity indefinites in
different languages that is not unlike my own typology formulated in the
implicational map (chapter 4). Although due to her narrow syntactic
perspective Progovac's approach is less successful than it could be, her theory is
original and important and needs to be discussed in greater detail.

[Link]. Principles A and B and raising at LF. Chomsky’s Binding Theory

consists of three principles:

(198) Chomsky’s Binding Theory (1981: 183-222)

Principle A: An anaphor must be bound in its governing category.
Principle B: A pronominal must be free in its governing category.
 115

Principle C: A name must be free.

Binding is defined as co-indexing with a c-commanding antecedent, and the

'governing category' is a local domain of a certain kind, most often a clause.
(The details do not concern us here.) Typical examples of 'anaphors' are re-
flexive pronouns. Principle A explains why (199a) is possible, whereas (199b) is
ungrammatical because the antecedent is outside the clause of the reflexive
pronoun, and thus outside its governing category.

(199) a. Luozhui praised herselfi.

b. *Luozhui was glad [that the teacher praised herselfi]S.

Principle B explains why (200a) is ungrammatical (the 'pronominal' her has its
antecedent within its clause), whereas (200b) is possible (the antecedent is
outside the governing category).

(200) a. *Luozhui praised heri.

b. Luozhui was glad [that the teacher praised heri]S.

Principle C explains why both (201a) and (201b) are ungrammatical.

(201) a. *Luozhui praised Luozhui.

b. *Luozhui was glad [that the teacher praised Luozhui].

Progovac now observes that some indefinites in some languages show a very
similar distribution (her account also extends to other negative-polarity items,
but I will confine the discussion to indefinites, which figure most prominently
in Progovac’s work). In Serbian/Croatian, the language discussed most
extensively by Progovac, the ni-series of indefinite pronouns is reminiscent of
anaphors: It may cooccur with clausemate negation, but not with superordinate
negation.

(202) Serbo-Croatian (Progovac 1994: 40-1)

a. Milan ne vidi ni-š ta.
Milan not sees NEG-what
‘Milan cannot see anything.’
b. *Milan ne tvrdi [da Marija poznaje ni-koga].
Milan not claims that Marija knows NEG-whom
‘Milan does not claim that Marija knows no one/anyone.’

The i-series, on the other hand, is reminiscent of pronominals: It may not occur
with clausemate negation, but it does occur with superordinate negation
(Progovac 1994: 41-3).

(203) a. *Marija ne poznaje i-koga

Marija not knows INDEF-whom
‘Marija does not know anyone.’
b. Milan ne tvrdi [da Marija poznaje i-koga].
Milan not claims that Marija knows INDEF-whom
‘Milan does not claim that Marija knows anyone.’
 116

Progovac proposes to account for these similarities by saying that negative-

polarity items are subject to the same binding principles as reflexive pronouns
and ordinary pronouns. They are bound (and hence licensed) by negation, and
Serbian/Croatian ni-indefinites are subject to Principle A (i.e. they are
'anaphors"), and i-indefinites are subject to Principle B (i.e. they are
'pronominals').viii
But this is not yet sufficient to account for the distribution of i-indefinites.
Ordinary pronouns need not be bound, so (204) is acceptable. By contrast, i-
indefinites must be bound, but not in their governing category. (205) is
ungrammatical because i-koga is not bound by anything at all.

(204) The teacher praised her.

(205) *Marija poznaje i-koga.
Marija knows INDEF-who
‘Marija knows anyone.’

Progovac explains this situation by assuming that i-indefinites raise to a higher

position at the syntactic level of Logical Form (LF), and that they are subject to
Principle A there. In (206), sentence (203b) is shown both at S-structure (=
'surface structure') and at LF.

(206) a. (S-structure) Milan ne tvrdi [da Marija poznaje i-koga].

b. (LF) Milan ne tvrdi i-koga [da Marija poznaje e].

These facts follow from the assumption that i-indefinites are 'anaphoric
pronominals', i.e. that they are subject both to Principle A and to Principle B of
the Binding Theory. Their pronominal requirement (‘must be free in its
governing category’) is satisfied at S-structure, and their anaphoric requirement
(‘must be bound in its governing category’) is satisfied at LF.
A third type of indefinite arises from the possibility of an indefinite series
that is subject only to Principle A, but that raises at LF (like Serbian/Croatian i-
indefinites). Progovac claims that English any-indefinites should be analyzed in
this way. This explains why any-indefinites occur both with clausemate
negation (cf. 4.45), like Serbian/Croatian ni-indefinites, and with superordinate
negation (cf. 4.46), like Serbian/Croatian i-indefinites.

(207) a. (S-structure) Milena did not know anything.

b. (LF) anything [Milena did not know e].
(208) a. (S-structure) Franz did not think [that Milena saw anything].
b. (LF) Franz did not think [anything [that Milena saw e]].

In (207), anything is bound by its clausemate negation and satisfies its anaphoric
requirement at S-structure. In (208), anything is bound by superordinate
negation and satisfies its anaphoric requirement at LF.
Thus, the main three types of negative-polarity indefinites discussed by
Progovac can be represented as in Table 5. The two binary parameters 'non-
raising vs. raising' and 'Principle A only vs. Principle A and B' yield four slots,
of which three are filled by Serbian/Croatian and English indefinites (the
fourth slot cannot be filled because no indefinite can be simultaneously subject
to Principles A and B without raising – these principles contradict each other,
and the only way an item can satisfy both is to raise and satisfy them at dif-
ferent syntactic levels).
 117

Table 5. Three types of negative-polarity indefinites (Progovac 1994)

subject to subject to
Principle A only Principles A and B
non-raising Serbian/Croatian —
ni-indefinites
raising English Serbian/Croatian
any-indefinites i-indefinites

[Link]. Non-negative contexts. Of course, this cannot be the whole story

because any-indefinites and i-indefinites also occur in various non-negative
negative-polarity contexts, e.g. in questions, conditionals, and with 'adversative
predicates', as in (209-11).

(209) a. Da li Milan voli i-koga?

Q Q Milan loves INDEF-whom
b. Does Milan love anyone?
(210) a. Ako Milan voli i-koga, bi-ć e sreć an u ž ivotu.
if Milan loves INDEF-whom be-FUT.3SG happy in life
b. If Milan loves anyone, he will be happy.
(211) a. Sumnja-m da Milan voli i-koga.
doubt-1SG that Milan loves INDEF-whom
b. I doubt that Milan loves anyone.

Progovac (1994: ch. 4) assumes that negative-polarity indefinites in these

contexts are licensed (i.e. bound) by a null 'polarity operator' situated in a high
position in the sentence (specifically, in the COMP position). She further
assumes that the notion of 'governing category' can be defined in such a way
that COMP falls outside of the governing category of the indefinite, so that the
non-raising ni-indefinites cannot be licensed by the polarity operator. Only
raising indefinites can be licensed by it.
Progovac then identifies a further parameter: Some indefinites (e.g. English
any- and Serbian/Croatian i-) are licensed both by superordinate negation and
in non-negative licensing contexts. Other indefinites are licensed only by
superordinate negation, e.g. Turkish hiç-indefinitesix:

(212) Turkish (Progovac 1994: 89; data from Mürvet Enç)

a. (clausemate negation)
Ali hiç kimse-yi gör-me-di.
Ali INDEF anyone-ACC see-NEG-PAST(3SG)
‘Ali did not see anyone.’
b. (superordinate negation)
Ali hiç-bir sey-in bozul-duǧ -u-nu söyle-me-di.
Ali INDEF-one thing-GEN break-REL-3SG-ACC say-NEG-PAST(3SG)
‘Ali didn’t say that anything broke down.’
c. (conditional)
*(Eǧ er) Ali hiç kimse-yi getir-ir-se, televizyon seyred-eceǧ -iz.
if Ali INDEF anyone-ACC bring-AOR-COND television watch-FUT-1PL
‘If Ali brings anyone over, we will watch TV.’

Yet another type, according to Progovac, is represented by Chinese rènhé, which

is licensed only by clausemate negation and non-negative environments, not by
superordinate negation.
 118

(213) Chinese (Progovac 1994: 87; data from Yen-Hui Audrey Li)
a. (clausemate negation)
Tā bù xǐ huan rènhé dō ngxi.
she not like any thing
‘She does not like anything.’
b. (superordinate negation)
*Wǒ méiyǒ u gàosu guo tā nǐ zuò rènhé shìqing.
I NEG:PRF tell ASP he you do any thing
‘I did not tell him that you did anything.’
c. (conditional)
Rúguǒ tā xǐ huan rènhé dō ngxi qing gàosu wǒ .
if he like any thing then tell me
‘If she likes anything, tell me.’

Progovac proposes to account for these types in terms of different raising

possibilities of different negative-polarity indefinites. In the framework of her
assumptions, raising must be to the specifier of COMP in the case of licensing by
superordinate negation, but to a lower position (specifically, IP adjunction) in
the case of non-negative licensing contexts, because the specifier of COMP is
already occupied by the polarity operator. Thus, she can account for the
Turkish data in (212) by hypothesizing that Turkish hiç-indefinites can raise
only to the specifier of COMP, and for the Chinese data in (213) by saying that
Chinese rènhé can only adjoin to IP, whereas English any and Serbian/Croatian
i-indefinites have both options. Thus, by proposing two binary parameters and
subdividing one value into three sub-options, Progovac predicts seven types of
indefinites, five of which are attested. Progovac’s typology can be summarized
as in Table 4.5.

Table 6. Seven possible types of negative-polarity indefinites (Progovac 1994)

subject to subject to
Principle A only Principles A & B
no raising 1. Serbian/Croatian —
ni-
raising specifier of COMP 2. English 5. Serbian/Croatian
& IP adjunction any- i-
specifier of COMP 3. Turkish 6. [?]
only hiç-
IP adjunction 4. Chinese 7. [?]
only rènhé

As far as I can see, nothing excludes types 6 and 7 in Progovac’s theory, but she
does not comment on these types.

[Link]. Comparison with the implicational map. Progovac’s theory distin-

guishes three main types of contexts with respect to which we observe variation
between different indefinites: clausemate negation, superordinate negation,
and non-negative contexts. The distribution of the seven types over these three
contexts can be represented as in Figure 15.

Figure 15. Distribution of Progovac’s indefinite types over contexts

non-negative superordinate clausemate
 119

contexts negation
negation
|——1————-|
|———————————2——————————-|
|———————3————––––––|
|———————5———————–|
|———4———| |———4———–|
|———6————-|
|———7———|

Looking at Progovac’s theory in this way, it can be easily compared to my

approach in terms of the implicational map. The similarities to the upper right
subarea of my map are immediately evident: Progovac’s three contexts are
more or less equivalent to my three functions ‘direct negation’, ‘indirect
negation’, and ‘question/conditional’. Thus, Progovac’s typology and my
typology show some significant resemblances.
There is only one respect in which Progovac’s theory and mine make dif-
ferent predictions: The implicational map excludes the possibility of type 4,
which Progovac allows. Progovac’s data would seem to show that she is right,
because if Chinese rènhé is possible in conditionals and in the direct-negation
function, but not in the indirect-negation function (cf. 213), it would be a
counterexample to my implicational map. However, while (213b) with the
superordinate verb gàosu ‘tell’ sounds odd, rènhé is possible with other
superordinate verbs such as xiangxin ‘think’:

(214) Chinese (Fengxiang Li, p.c.)

Wǒ bù xiangxin rènhé rén lái le.
I not think any person come PFV
‘I don’t think that anybody came.’

The fact that xiangxin ‘think’ allows rènhé, while gàosu ‘tell’ does not, must have
to do with the fact that verbs of thinking in general allow negative raising much
more readily than some other verbs. But whatever the explanation is, a closer
look at the Chinese data shows that it does not contradict my implicational
map.

[Link]. Discussion of Progovac’s theory. Progovac's theory is particularly

successful in bringing out the parallels in the locality restrictions on certain
types of anaphoric pronouns and on certain types of indefinites. Reflexive
pronouns often must be coreferential with a clausemate antecedent, and
similarly indefinite pronouns often must cooccur with clausemate negation.
Other types of pronouns cannot be coreferential with a clausemate antecedent,
and likewise some indefinites cannot occur with clausemate negation. These
parallels would seem to call for a unified account, and Progovac has provided a
concrete proposal of how to achieve it. I will not discuss the details of her
theory here because it presupposes a large number of controversial
assumptions that I do not necessarily share. I will restrict myself to pointng out
a number of general weaknesses of her approach which will have to be
addressed in future work.
First, Progovac assumes that the Binding Theory is a primitive of universal
grammar, although it has been argued (convincingly, to my mind) that the
locality restrictions on reflexive pronouns can be explained with reference to
more general extragrammatical principles (Van Valin 1990, Levinson 1991).
 120

Ultimately we need an extragrammatical account that extends both to reflexive

pronouns and to indefinite pronouns with locality restrictions.
Second, and most importantly, Progovac has made no attempt to incorporate
the undeniable semantic-pragmatic motivation (§ 5.5) into her theory. In her
theory it is not clear what unifies the the two licensers of negative-polarity
items, negation and the empty 'polarity operator' (and of course, the very
existence of this operator, which never surfaces, is doubtful). Furthermore, a
theory like Progovac's, where negative-polarity items must be lexically marked
as subject to certain binding principles and as having certain raising
possibilities, runs into problems with open-class negative-polarity items of the
sort discussed by Schmerling (1971) (I didn’t say a word/do a thing/hear a sound/eat
a bite/move a muscle/...).
Third, Progovac's account in terms of the Binding Theory fails to extend to
related types of indefinites such as irrealis-non-specific indefinites and free-
choice indefinites.x These are clearly semantically related to negative-polarity
indefinites, but they are not susbject to syntactic locality restrictions.
Fourth, as Progovac (1994: 90) notes herself, there is no independent
evidence for the various parametric differences that she assumes. Her system
seems to work, but she has no way of predicting when an indefinite will be
subject to which binding principle, and when it will select which raising option.
It is true that there is probably a lot of arbitrariness, but some things can be ex-
plained. In particular, diachronic explanations are possible – e.g. the fact that
Serbian/Croatian i-indefinites are marked by i, which means ‘and, also, even’.
It is not clear how such connections could be accounted for in Progovac’s
[Link]
Progovac's work has shown clearly that much of the work on binding in
syntax is relevant to an understanding of certain properties of indefinite
pronouns. However, by choosing a narrow syntactic approach, she has missed
the possibility of integrating her insights into a more realistic mixed approach
that recognizes the basic semantic-pragmatic motivation of negative prolarity,
without denying the important role of syntax for an interesting range of facts.

5.4. Mental Spaces

5.4.1. Introduction to the mental space approach

The approach that seems most suitable for describing non-specificity and re-
lated phenomena is Fauconnier’s (1985, 1987) theory of MENTAL SPACES. (A
somewhat similar approach had already been proposed in Jackendoff 1972a: ch.
7). The concept of a mental space is best introduced through a concrete exam-
ple. Consider (215):

(215) Youssouf wants to marry some Frenchwoman.

On the specific reading ('a certain Frenchwoman', say Geneviève) the

Frenchwoman exists quite independently of Youssouf’s desire and his wish
concerns only a particular relation between himself and Geneviève. On the non-
specific reading ('any Frenchwoman, no matter who'), however, the very
possibility of talking about a Frenchwoman arises only because of Youssouf’s
desire. If it makes any sense to speak about existence in this latter case, the
 121

Frenchwoman exists only in the world of Youssouf’s desire, or, in Fauconnier’s

terminology, in the mental space that is set up by ‘Youssouf wants...’.
Mental spaces have a certain similarity to the Leibnitzian concept of POSSIBLE
WORLDS, but in contrast to these they carry no ontological implications – they
are purely cognitive constructs. Furthermore, alternative mental spaces are
concerned only with a small segment of reality (hence, 'space' rather than
'world') and are therefore psychologically plausible constructs, unlike possible
worlds.
We can now describe the ambiguity in (215) in the following way (cf.
Fauconnier 1985: § 1.5): On the non-specific reading, a referent for the phrase
‘some Frenchwoman’ exists only in the mental space of Youssouf’s desire,
whereas on the specific reading a referent for this phrase exists both in the
mental space of the speaker’s reality and in the mental space of Youssouf’s
desire. This can be represented as in Figures 16-17.

Figure 16: Non-specific reading of (215)

Youssouf Youssouf

wants marry

Frenchwoman

speaker’s reality space ‘want’ space

Figure 17: Specific reading of (215)

Youssouf Youssouf

wants marry

Frenchwoman Frenchwoman

speaker’s reality space ‘want’ space

In these diagrams the large circles symbolize mental spaces, the black nodes
symbolize entities, and the white nodes symbolize relations between entities.
Broken lines connect relations with entities, and continuous lines symbolize
pragmatic connectors that link entities in different spaces. (In the examples
cited here, only referentially identical entities are linked by such connectors. See
Fauconnier (1985) for more complicated cases, and detailed justification of the
notion of a pragmatic connector.)
The ‘want’ space is set up by the meaning of the non-reality verb ‘want’
(sometimes called an 'intensional verb' in the philosophical literature). Elements
that set up spaces can be of very different types, to be illustrated below. All
such elements are called SPACE-BUILDERS. Spaces which are set up within other,
'superordinate' spaces are called DEPENDENT SPACES. Thus, the ‘want’ space in
Figure 13-14 is dependent on the speaker’s reality space. Fauconnier applies the
notion of mental spaces to a wide variety of cases, some of which are illustrated
by (216a-e).

(216) a. Lisa, who has been depressed for months, is smiling in this picture.
 122

b. Len believes that the girl with blue eyes has green eyes.
c. In 1929, the lady with white hair was blonde.
d. If he had listened to his mother, this criminal would be a saint.
e. In that movie, Clint Eastwood is a villain.

Thus, mental spaces can be of quite different sorts: for instance picture spaces
(a) in (216), belief-spaces (b), time spaces (c), hypothetical spaces (d), theater
spaces (e).
Fauconnier’s theoretical constructs allow us to make precise the intuition
behind the widespread characterization of (non-)specificity according to which
with a specific phrase, the speaker has a particular entity in mind, whereas with
a non-specific phrase, this is not the case (e.g. Partee 1972: 415, Ioup 1977: 234,
Hawkins 1978: 204). Evidently, what is meant by this informal description is
that in the case of a specific phrase, the referent exists within the speaker’s
reality space, whereas it does not exist there in the case of a non-specific phrase.
We can now also explain two of the three specificity tests mentioned in §
3.2.3, discourse referents and existential paraphrases. A sequence like (217a) is
impossible because the referent of the NP a paper on phonological iconicity is in
Irene’s ‘want’ space, whereas the referent of the anaphoric pronoun ‘it’ is in the
speaker’s reality space. A condition on pronominalization is that the anaphoric
pronoun and its antecedent must be in the same mental space. Thus, (217b),
where ‘it’ is also in the ‘want’ space, is perfectly acceptable.

(217) a. *Irene wants to write a paper about phonological iconicity. It is

brilliant.
b. Irene wants to write a paper about phonological iconicity. She
wants it to be brilliant.

Existential paraphrases are only possible with specific phrases because only
in a reality space can the existence of entities be presupposed by the speaker,
and a presupposition of existence is a necessary prerequisite for an existential
paraphrase.

5.4.2. A definition of (non-)specificity

On the basis of what was said above, I propose the preliminary definitions in
(218), to be revised below.

(218) A SPECIFIC phrase is a phrase that has a referent in the speaker’s

reality space.
A NON-SPECIFIC phrase is a phrase which does not have a referent in
the speaker’s reality space.

Note, incidentally, that this definition does not make reference to the syntactic
category NP (noun phrase) because adverbial phrases may also show this
distinction (e.g. example (74) above). Also, this definition is not restricted to
indefinite phrases, because definite phrases may also be specific or non-specific
(pace Hawkins 1978: 204). This is illustrated in (219). ‘The referee’ is non-spe-
cific in (219a), and non-specific in (219b). (We will see further examples below,
e.g. (225-28).)

(219) a. (non-specific)
 123

We should go to a hockey game tomorrow. [I don’t care which

one.] I want to take a picture of the referee.
b. (specific)
We should go to a hockey game tomorrow. [It’s Lahore against
Rawalpindi.] I want to take a picture of the referee. [It’s Hanif
Abdul Nasir, my brother-in-law.]

Notice further that the definition in (218) does not deny the existence of a
referent of a non-specific phrase, so non-specific is not the same as NON-
REFERENTIAL. (This latter term is sometimes used in the sense in which I use
non-specific, e.g. in Givón (1973, 1978), Bakker (1988), Ladusaw (1993). I would
reserve the term non-referential for predicative noun phrases or dependent
compound members.)
The definition in (218) would also describe ambiguities as in (220) in terms of
the (non-)specificity distinction. The phrase ‘a dragon’ may have a referent only
in the space set up by ‘Alice believes...’.

(220) Alice believes that a dragon ate her petunias.

Such a terminological choice is sometimes made in the literature (e.g. Fodor

1970, Ioup 1977; see footnote 6 of this chapter), but I prefer a definition of the
term non-specific which allows it to be used for characterizing the behavior of
certain indefinite pronoun series in some languages (cf. § 3.2.3, and see Croft
1985). The non-specific indefinites in the languages that I have come across are
not used in sentences like (220), but only in contexts where their unique
identifiability is not presupposed by the speaker. These include, above all, other
dependent spaces which describe unrealized situations like ‘want’-spaces.
Space-builders that set up such IRREALIS SPACES are other modals like ‘can’ and
‘must’, the future tense, and the imperative. Conditionals and questions can
also be regarded as setting up irrealis spaces because of their semantic
similarities to imperatives (§ 3.2.3). Furthermore, distributive quantification can
also license non-specific indefinites, so mental spaces set up by distributive
quantifiers (cf. especially Fauconnier 1987) should also be accounted for in the
definition.
We can now revise the definition in (218) as in (221).

(221) A NON-SPECIFIC phrase is a phrase which has a referent only in a

dependent irrealis or distributive space.
A SPECIFIC phrase is a phrase which also has a referent also in some
other space.

It would of course be interesting to know exactly what it is that unites irrealis

and distributive mental spaces, but this question is beyond the scope of this
book. I refer to Giannakidou (1994a, b, c) for further discussion in different
terms (Giannakidou uses the semantic notion of (non-)veridicality which
basically corresponds to my notion of irrealis spaces).

5.4.3. Advantages of the mental space approach

The mental space approach and the classical logical-semantic approach show
some obvious similarities, and one might even claim that they are largely
notational variants of each other. Existence of a referent in a dependent space
 124

corresponds to narrow scope, dependent spaces correspond to logical operators

and opaque contexts, etc. However, mental spaces have certain advantages
over logical semantics.
First, the terminology and notation of the mental-space approach is much
more appropriate for the cognitively oriented study of language than the
terminology and notation of predicate logic. Logic was developed by
philosophers and mathematicians who were not concerned with language and
cognition, and the use of logic in linguistics has sometimes given rise to
misunderstandings. The quasi-spatial representations of Figures 16-17 are also
much more likely to be realistic than logical formulae because there is
considerable evidence that much of abstract thinking has a spatial dimension
(as shown by the pervasive spatial metaphors in language).
Second, the mental space approach provides a unified account of a large
number of apparently disparate phenomena, as discussed in detail by
Fauconnier (1985). Only a small number of these phenomena are amenable to a
logical-semantic treatment.
Third, and more concretely, there is nothing in Fauconnier’s theory that says
that space builders must be of a particular form. The construction of mental
spaces is a relatively autonomous cognitive process which makes use of
linguistic information of various sorts, but there are few constraints on the
expression type of this information, and even non-linguistic pragmatic
information may be sufficient to set up a mental space. By contrast, the logical
approach always needs some tangible element in the logical formula to
represent scope differences. As we saw in § 5.2.2, this forced logicians and
linguists to introduce abstract elements into the representation. While this is
certainly a possible strategy, it has an unpleasant flavor of artificiality. The
solution (first proposed in Jackendoff (1972a: ch. 7) and adopted by Fauconnier)
to separate the representation of syntactic structure from the representation of
mental-space structure (in Jackendoff’s terms, 'modal structure'), seems more
adequate.
Fourth, there are some semantic phenomena that can only be expressed in
the mental space framework. Thus, Fauconnier points out that a sentence like
(222) has at least four readings, whereas a logic-based description can only
capture three of these readings.

(222) Every child wants to see a movie.

Thus, the mental-space approach is also descriptively the most adequate

approach.

5.5. Pragmatic Scales and Scale Reversal

An approach to the meaning and distribution of free-choice and negative-
polarity indefinites like any that differs radically from the logical-semantic
approach of § 5.2.1 has been advanced in a series of papers by Gilles Fauconnier
(1975a, 1975b, 1977, 1979, 1980) (and see Krifka (1992) for further refinements). I
believe that this is the most successful treatment of free-choice and negative-
polarity indefinites proposed so far, and I will discuss it in some detail.
 125

5.5.1. Pragmatic scales

The key concept of Fauconnier’s theory is that of a PRAGMATIC SCALE, which

may lead to pragmatic SCALAR IMPLICATURES. The notion of a pragmatic scale is
also crucial in the domain of SCALAR FOCUS PARTICLES (cf. König 1991), and
scalar focus particles may be used to illustrate what is meant by a pragmatic
scale. Consider (223), where (223a) shows the non-scalar focus particle ‘also’,
and (223b) shows the scalar focus particle ‘even’.

(223) a. Véronique also speaks Dutch.

b. Natasha even speaks Dutch.

The meaning difference between the two sentences is that whereas (223a) only
implies that Véronique speaks some other language in addition to Dutch, (223b)
also implies that Dutch is a particularly surprising or unlikely language for
Natasha to speak. The focus value ‘Dutch’ and the included implicit
alternatives (i.e. other languages) are taken to constitute a (partially) ordered
scale, e.g. (224).

(224) (1) Russian, (2) English, (3) German, (4) French,... (n) Dutch

‘Even’ characterizes its focus value as ranking high on such an ordered scale, so
(223b) is only appropriate if a scale like (224) is taken for granted.

5.5.2. Scalar implicatures

Semantic and pragmatic scales often give rise to scalar implicatures. Consider
sentence (225), which has two readings: one in which the weakest cow is specific
(a), and one in which it is non-specific (b).xii

(225) The weakest cow can swim through this river.

a. (specific) ‘A certain cow, who is weaker than all others, can swim
through this river.’
b. (non-specific) ‘Any cow can swim through this river.’

Note that the paraphrase of the second reading contains the free-choice in-
definite ‘any’, suggesting that this reading is quantificational in some way.
Fauconnier calls superlatives with readings like (225b) QUANTIFYING SUPERLA-
TIVES. This sense arises by scalar implicature. On the quantifying reading, (225)
is associated with a scale of cow weakness, as illustrated in Figure 18.

Figure 18. Pragmatic scale of cow weakness

|———|———|———|———|———|———|
m x1 x2 x3 x4 x5 M
weakest cow strongest cow
low end high end

Pragmatically, it is assumed that if a cow xn can swim through a river, then any
cow xn+m who is higher on the weakness scale (i.e. is stronger) can also swim
through it. Superlatives specify an extreme value on a scale, so the weakest cow
implicates that all cows can swim through the river, and the resulting meaning
is similar to that of universal quantifiers like ‘every’ and ‘all’. Note that this
 126

universal reading is based on pragmatic implicature, not on logical entailment:

Logically, a weaker cow could of course be able to swim through a river (e.g. by
compensating weakness by special swimming skills), whereas a stronger cow
could be unable (e.g. because of hydrophobia). It is important to note that
not any superlative gives rise to a universal reading, but only superlatives that
express the lowest point on the relevant scale. Thus, in (226-28) only the (a)
cases allow a universal reading (marked by '+U'), whereas the (b) cases, which
express the highest point on the relevant scale, do not (marked by '–U'). Thus,
in the (b) cases only the specific reading (analogous to (225a)) is possible.

(226) a. (+U) The weakest cow can swim through this river.
b. (–U) The strongest cow can swim through this river.
(227) a. (+U) The most popular politician can plummet in the ratings by
raising the gasoline tax.
b. (–U) The least popular politician can plummet in the ratings by
raising the gasoline tax.
(228) a. (+U) My friend will review the cheapest book.
b. (–U) My friend will review the most expensive book.

Similar scalar implicatures arise also in cases like (229), where an individual
that is well-known for some extreme property is taken to represent an extreme
point on the relevant scale.

(229) a. Even Croesus would hesitate to buy this house.

b. Even Kasparov could lose this game.
c. Even Saddam Hussein would have been impressed by this
diplomat.

5.5.3. Scale reversal

Now observe what happens to the quantificational reading if the examples of §

5.5.2 occur in a negated sentence or in a (rhetorical) question:

(230) a. (–U) The weakest cow cannot swim through this river (so we
must leave it behind).
b. (+U) The strongest cow cannot swim through this river (so we
cannot even think about crossing it).
(231) a. (–U) Can the weakest cow swim through this river? (Then we
wouldn’t have to leave it behind when crossing the river.)
b. (+U) Can the strongest cow swim through this river? (And you are
proposing that we should cross it!)
(232) a. ?*Even Croesus would not hesitate to buy this house.
b. OK: Even a beggar would not hesitate to buy this house.

We see that suddenly the superlative that gave rise to the universal implicature
(‘the weakest cow’) does not have this implicature anymore, and the
superlative that denotes the other end of the scale acquires it. The explanation
for this is that in negated and conditional clauses, pragmatic scales are
reversed. Consider the examples in (233-34), where Squeaky and Lowy are two
cows, and Squeaky (e.g. x1 in Figure 18) is weaker than Lowy (e.g. x4 in Figure
18).
 127

(233) a. Squeaky can swim through this river.

⇒ Lowy can swim through this river.
b. Lowy can swim through this river.
/ Squeaky can swim through this river.
(234) a. Lowy cannot swim through this river.
⇒ Squeaky cannot swim through this river.
b. Squeaky cannot swim through this river.
/ Lowy cannot swim through this river.

Similar examples could be given for conditionals or questions. The fact that the
pragmatic implicatures are reversed in negative, conditional and interrogative
sentences, as shown in (233-34), explains the reversal of the pragmatic scales
and the disappearance of universal scalar implicatures in (230-32) (a).
The reversal of pragmatic implicatures in negative and conditional clauses is
closely related to the reversal of entailment relations that has been highlighted
in Ladusaw (1980: ch 6). He observes that a sentence like (235a) entails (235b),
whereas in negative and conditional clauses the opposite entailment holds, cf.
(236-37).

(235) a. Johanna ate brussels sprouts for dinner.

⇒ b. Johanna ate a green vegetable for dinner.
(236) a. Johanna didn’t eat a green vegetable for dinner.
⇒ b. Johanna didn’t eat brussels sprouts for dinner.
(237) a. If Johanna eats a green vegetable for dinner, she will stay healthy.
⇒ b. If Johanna eats brussels sprouts for dinner, she will stay
healthy.

Since the entailment is from hyponyms to hyperonyms in (235), and the other
way round in (236-37), Ladusaw calls contexts like (235) upward-entailing, and
contexts like (236-37) downward-entailing. However, since the scale-reversing
effect of these contexts is their crucial property in the present context, I will call
them SCALE-REVERSING.
As Fauconnier notes (1975a: 195-6, 1977: 24-35; cf. also Ladusaw 1980: 147-
71), the scale-reversing contexts are precisely those contexts that license
negative-polarity items. In addition to overt negation and conditionals, these
include the standard of ordinary comparison (238) and excessive comparison
(‘too’, 237), emotive verbs like ‘be surprised’ (240), implicit negation (241), and
[Link] In the following examples (238-41), the scale-reversing effect of each
context is illustrated by the scale of cow weakness, where the weakest cow gives
rise to universal scalar implicatures in non-reversing contexts (cf. 225), but the
strongest cow results in such universal implicatures in scale-reversing contexts
(e.g. 230-31 (b)).

(238) standard of ordinary comparison

The dog can swim through wider rivers than the strongest cow.
(239) standard of excessive comparison
The river is too wide for the strongest cow to swim through.
(240) emotive verbs
I’m surprised that the strongest cow can swim through this river.
(241) implicit negation
She denies that the strongest cow can swim through this river.
 128

Characterizing the licensing contexts for negative-polarity items as scale-

reversing accounts better for the observed restrictions than all other alternatives
proposed so far (cf. Ladusaw 1980: 147). It avoids the arbitrariness of the
syntactic feature [+affective] of Klima (1964), and it is more adequate than
attempts to reduce all cases of negative-polarity licensing to negation in one
way or another (Baker 1970, Linebarger 1987, 1991, Progovac 1988), because the
context of universal quantifiers (ex. 197), for one, clearly does not involve nega-
tion. And, most importantly for my goals, it allows us to explain many formal
properties of negative-polarity items, in particular negative-polarity indefinites.
Somewhat surprisingly, this point has rarely been addressed (though see
Linebarger (1981: ch. 8) for a subset of negative-polarity items), but it is my ma-
jor concern here.

5.5.4. Negative polarity items express the low point on a scale

As Fauconnier (1975b: 196) notes, the distribution of many negative-polarity

items can be explained in the same way as the distribution of quantifying
superlatives once we recognize that negative polarity items are generally id-
iomatic expressions for a low point on a pragmatic scale. This is particularly
clear with negative polarity items that denote a MINIMAL UNIT, e.g. a red cent, a
jot, a tittle, and equivalents in other languages, as in (242-43) (for more
examples, see Horn 1989: 452-3, von Bergen and von Bergen 1993: 139-54).

(242) a. She didn’t give a red cent to him.

b. If she gives a red cent to him, I’ll be surprised.
c. Everybody who gives a red cent to him must be pretty stupid.
(243) a. *She gave a red cent to him.
b. *She would give a red cent to him.

Since they express the low point on a pragmatic scale, these minimal-unit
expressions also have universally quantified readings, quite analogously to
quantifying superlatives (she didn’t give a red cent = she gave nothing at all). The
perfect parallelism is best illustrated by examples (244-45), corresponding to
(242-43).

(244) a. (+U) She didn’t give the smallest sum to him.

b. (+U) If she gives the smallest sum to him, I’ll be surprised.
c. (+U) Everybody who gives the smallest sum to him must be pretty
stupid.
(245) a. (–U) She gave the smallest sum to him.
b. (–U) She would give the smallest sum to him.

The main difference is of course the ungrammaticality of negative-polarity

items (cf. 243) in contexts where the corresponding superlative simply lacks a
quantifying reading (245). This non-quantifying reading is not available for
negative-polarity items because they are idiomatic expressions in which the
function of expressing the low point on a scale has become [Link] To put it
differently: Minimal-unit expressions that are restricted to scale-reversing
contexts have been grammaticalized as inherently non-specific (cf. Ladusaw
1993), so they do not have the specific non-quantifying reading exemplified by
(225a) and (245).
 129

Perhaps not all negative-polarity items can be shown to express a low point
on a scale (cf. Fauconnier (1975b: 198), who points out cases like English yet),
but in the case of negative-polarity indefinites it is not difficult to make this
plausible considering the form of the indefinites. In particular, the expression
types in (246), to be discussed in greater detail in the following chapters, are
motivated by the basic value ‘low point on a scale’.

(246) a. ‘one, single’: Latin ullus, English any, etc. (§§ 7.5.2, [Link])
b. generic nouns: French personne, Hebrew iš , etc. (§§ 7.5.1, [Link])
c. scalar focus particles: Japanese nani-mo lit. ‘even what’,
Modern Greek kan-énas lit. ‘at least one’, Russian ni-kto
lit. ‘not even who’ (§§ 7.1, § 8.3.1)
d. minimal-unit expressions like ‘a trace’, ‘a tittle’ (§ [Link])
e. maximal-unit expressions like ‘my life’, ‘eternity’, ‘the world’
(§ [Link])

5.5.5. Free-choice indefinites express the low point on a scale

Fauconnier (1975a) has explored in great detail the parallels between the free-
choice indefinite any and quantifying superlatives. He observes that the (a) and
(b) sentences of (247-48) are virtually synonymous (1975a: 354-5).

(247) a. My uncle can hear the faintest noise.

b. My uncle can hear any noise.
(248) a. She would give the largest sum to him.
b. She would give any sum to him.

In addition to the near synonymy of these sentences, there is another striking

parallel between quantifying superlatives and free-choice indefinites, which
Fauconnier fails to mention but which is observed in Bakker (1988: 30-6) and
König (1991). Quantifying superlatives are possible in precisely those contexts
where free-choice indefinites are licensed, i.e. mainly possibility contexts,
generic contexts, and contexts expressing a sufficient condition (cf. § 3.2.5).
While (249-50) (a) have a quantifying reading, which can be paraphrased by
‘any’, (251-52) (a) do not have such a reading, and the corresponding sentences
with ‘any’ are unacceptable.

(249) a. (+U) The youngest beaver can swim.

b. Any beaver can swim.
(250) a. (+U) The smallest amount is enough.
b. Any amount is enough.
(251) a. (–U) The youngest beaver swam across the river.
b. *Any beaver swam across the river.
(252) a. (–U) The largest/smallest amount is necessary.
b. *Any amount is necessary.

Again, free-choice indefinites are licensed in contexts where superlatives give

rise to a quantifying reading. This completely parallel behavior of quantifying
superlatives and free-choice indefinites suggests that free-choice indefinites,
too, express a low point on a scale.
But we saw in the preceding section that negative-polarity items express the
low point on a scale, and clearly negative-polarity indefinites and free-choice
 130

indefinites are not synonymous. About half of the languages of my 40-language

sample employ different indefinite series for the free-choice function and
negative-polarity functions. On the other hand, also roughly half of the
languages allow the same indefinite series to be used to express free choice and
in negative-polarity contexts, which confirms the suspicion that there is a
strong similarity.
In the framework advocated here, the solution to this dilemma is quite
straightforward: Free-choice indefinites express the low point on a non-re-
versed scale, whereas negative-polarity indefinites express the low point on a
reversed scale. We saw in § 5.5.3 that scale-reversing contexts license negative-
polarity items, which follows from the fact that these express a low point on a
reversed scale.
Thus, we can say that in a sense both negative-polarity indefinites and free-
choice indefinites express the low point on a scale, albeit on opposite scales.
Already Fauconnier (1975a: 373) proposed to 'interpret the function of any as
being the indication of a low point on an arbitrary scale'.
We can now also solve the puzzle of ambiguous sentences like (253), which
have baffled proponents of the univocal universal any view, and which
advocates of the two-any view have cited as crucial evidence in their favor.

(253) If she can solve any problem, she’ll get a prize.

a. ('existential') ‘If there is any problem she can solve...’
b. ('universal') ‘If she can solve every problem...’

The scalar approach allows us to have our cake and eat it too. Consider again
the equivalent sentences with quantifying superlatives. There are two different
sentences, one corresponding to each of the readings of (254).

(254) a. ('existential') If she can solve the simplest problem, she’ll get a
prize.
b. ('universal') If she can solve the most difficult problem, she’ll get a
prize.

These paraphrases clearly show what is going on in (253). Any each time ex-
presses the low endpoint on a scale: the low endpoint of the non-reversed scale
in (253b) (= ‘the most difficult problem’), and the low endpoint of the reversed
scale, i.e. the opposite endpoint, in (253a) (= ‘the simplest problem’). Thus, we
can capture both the insight of the univocal universal any theory that the two
uses of any are closely related, and at the same time account for the ambiguity
of sentences like (253) which motivated the two-any theory.

5.5.6. Widening and strengthening

Another recent proposal for a unified account of negative-polarity and free-

choice indefinites is found in Kadmon and Landman (1993). These authors
propose that polarity-sensitive and free-choice any are the same element, which
induces semantic WIDENING along a contextual dimension and functions to
STRENGTHEN the statement that it occurs in. For example, they note that (255a)
differs from (255b) in that its domain is wider (for example, (255b) might still be
true if a wet match does not light, but in (255a) even wet matches are included
in the domain of quantification).
 131

(255) a. Any match I strike lights.

b. Every match I strike lights.

They propose that any is licensed only if the widening that it induces fulfills its
function of creating a stronger statement, i.e. if the statement on the wide
interpretation entails the statement on the narrow interpretation. This happens
in downward-entailing contexts. Thus, Kadmon and Landman can account for
why any is licensed precisely in such contexts. In addition, they propose that
free-choice any arises when widening and strengthening are combined with a
generic interpretation.
Although Kadmon and Landman make use of the completely new concepts
of widening and strengthening, their explanation is not all that different from
Fauconnier's. Their notion of widening corresponds to the extreme point on a
scale, the contextual dimension corresponds to the pragmatic scale, and their
notion of strengtehning corresponds to scalar implicature. On their account, a
sentence like *I have any matches is ungrammatical because strengthening is not
satisfied, i.e. the widened statement I have matches of some kind (dry or wet) does
not entail the statement on the narrower interpretation I have dry matches. In
Fauconnier's theory, one would say, quite analogously, that *I have any matches
is ungrammatical because any is inherently non-specific and expresses and
endpoint on a scale, but when the scale is not reversed, no universal scalar
implicature results and hence the inherently non-specific any cannot be
interpreted. Thus, Fauconnier's proposal and Kadmon's and Landman's
proposal are more similar to each other than the very different terminology
suggests.
In my view, Fauconnier's theory has three advantages: (i) It makes use of
theoretical constructs that are independently motivated (pragmatic scales and
scalar endpoints), whereas widening and strengthening are theoretical
innovations; (ii) in many cases, there is a direct connection between the
meaning of a scalar endpoint and the form of negative-polarity items (see 246),
whereas there is no such connection with widening; and (iii) the frequent
connection with focusing (see § 5.7) is also explained by Fauconnier's theory
(because scales are a typical instantiation of the range of alternations that
focusing evokes), whereas there is no such natural connection between
widening and focusing.
In any event, Kadmon's and Landman's analysis of any, which is more
detailed than Fauconnier's and addresses a number of notorious difficulties in
the analysis of negative polarity, is successful in demonstrating that a semantic-
pragmatic approach to negative polarity and free choice is on the right track.

5.6. Explaining the Implicational Map

After discussing the various theoretical perspectives on the functions of
indefinite pronouns, I will now put the pieces together and propose an
explanation for the implicational map of chapter 4, reproduced in (256).

(256) The implicational map for indefinite pronoun series

direct
question indirect negation
specific specific irrealis negation
known unknown non-specific
 132

conditional comparative
free-choice

Why are the various functions arranged in this particular configuration? The
present section will answer this question by demonstrating that the functions
that are most similar to each other semantically and pragmatically are adjacent
on the map.
There are four binary features that can be used to characterize all the nine
functions on the map. Two of the four features have just been discussed in de-
tail, and the other two are, I trust, self-explanatory.

(257) (i) known to the speaker vs. unknown to the speaker

(ii) specific vs. non-specific (cf. § 5.4)
(iii) scalar endpoint vs. no scalar endpoint (cf. § 5.5)
(iv) in scope of negation vs. not in scope of negation

In addition, there is one binary feature that applies only to the functions with
the feature value ‘scalar endpoint’:

(258) (v) endpoint on non-reversed scale vs.

endpoint on reversed scale (cf. § 5.5.5)

I assume that the implicational map should be arranged in such a way that all
functions that share some relevant characteristics, i.e. that have the same value
on one of the features in (257-58) form a contiguous area on the map. This
general requirement together with the distribution of the features over the map
explains a considerable part of the explanandum.
First, the feature ‘known vs. unknown’ is distributed as shown in (259).

(259) known vs. unknown

direct
question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

In only one function is the referent known to the speaker, while it is unknown
in all others, so this function must be at the periphery of the map (otherwise all
other functions could not form a contiguous area). As was observed in § 3.2.4,
all ‘non-specific’ functions are automatically also ‘unknown’, because what
exists only in an irrealis or a distributive space cannot be known in principle.
Second, the feature ‘specific vs. non-specific’ is distributed as shown in (260).

(260) specific vs. non-specific

direct
question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice
 133

As was also observed earlier, an expression denoting a scalar endpoint must be

non-specific in order to give rise to quantificational implicatures (§ 5.5.2), so not
only the ‘irrealis-non-specific’ function, but also all scalar-endpoint (i.e.
negative-polarity and free-choice) functions are non-specific. The ‘known’
function, by contrast, cannot be non-specific because identifiability by the
speaker and non-specificity are in principle incompatible.
Third, the feature ‘scalar endpoint vs. no scalar endpoint’ is distributed as
shown in (261). The subclassifying feature ‘scale reversal vs. no scale-reversal’
is also shown in (261).

(261) scalar endpoint vs. no scalar endpoint;

scale reversal vs. no scale reversal
direct
question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

We saw in § 5.5 that a negative-polarity expression with a quantificational read-

ing is possible only in a scale-reversing context or in a free-choice context. This
yields the basic binary distinction in (261): Only one of the six ‘scalar-endpoint’
functions has the value ‘no scale reversal’, so that value must again be in a
peripheral position.
Fourth, the feature ‘in scope of negation vs. not in scope of negation’ is
distributed as shown in (262).

(262) in scope of negation vs. not in scope of negation

direct
question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

Only the two negation functions have the value ‘in scope of negation’, so these
must also be at the periphery of the map. The feature 'scope of negation' can
also be subdivided into two subclassifying features, corresponding to the
Boolean properties anti-additive and anti-morphic, as discussed insightfully in
van der Wouden (1994: ch. 1) (a discussion of these notions would take us too
far afield here). Both direct and indirect negation have the property of anti-
additivity, but only direct negation is anti-morphic as well. Clearly, direct
negation is in some sense 'more strongly negative', which explains its
peripheral position on the map.
Given the distribution of the four main features and the two subclassifying
features over the nine functions, there are not many possibilities left. Only the
relative position of the functions 'question', 'conditional', and 'comparative' to
each other must be explained.
The explanation for the closer relationship between questions and negation
is straightforward: In questions, negation is neutralized, whereas it is not in
conditionals. Can you hear nothing? and Can you hear anything? have identical
truth conditions, but if you hear nothing is radically different from if you hear
 134

anything. And it is also clear why conditionals should be closer to free choice.
Conditionals with scalar-endpoint indefinites are pragmatically quite normal,
just more emphatic than conditionals with non-emphatic indefinites. However,
questions with scalar-endpoint indefinites are somewhat odd, and make best
sense as rhetorical questions. A question like Did you hear the slightest noise?
could hardly be meant as an information question, because it is very unlikely
that the speaker should be interested in information about an extreme [Link]
The crucial difference between the question and conditional functions and
the comparative function is that only the latter context allows both non-specific
indefinites (which are interpreted as scalar endpoints and yield a quan-
tificational reading) and specific indefinites (which are not interpreted as scalar
endpoints), whereas indefinite noun phrases are necessarily non-specific in
questions and conditionals. This results in the fairly dramatic semantic
difference between (263a) and (263b), whereas the difference between the (a)
and (b) readings in (264-65) is rather small (the truth conditions are not
affected, whereas (263a) and (263b) have very different truth conditions).

(263) a. Leoluca is smarter than anyone.

b. Leoluca is smarter than someone.
(264) a. If you hear something, wake me up.
b. If you hear anything, wake me up.
(265) a. Did you hear something?
b. Did you hear anything?

In this respect, the comparative is like the free-choice function: In free-choice

contexts, too, both ‘scalar-endpoint’ and ‘non scalar-endpoint’ indefinites are
admitted, but with a big meaning difference.

(266) a. Leoluca can do anything.

b. Leoluca can do something.

Another similarity between the free-choice function and the comparative

function is that in both contexts universal pronouns have a very similar
meaning to scalar-endpoint indefinites, whereas this is not the case in questions
and [Link]

(267) a. Leoluca can do anything (≈everything).

b. Leoluca is smarter than anybody (≈everybody).
c. If you hear anything (≠ everything), wake me up.
d. Did you hear anything (≠ everything)?

This completes my account of the implicational map. It has been shown that
there are independent semantic reasons why the map should be arranged
precisely in this fashion: Only in this way are semantically related functions
adjacent to each other. That indefinite pronoun series show only polysemy
patterns that cover adjacent areas on the map (as we saw in § 4.4) is completely
as expected, if we assume that polysemy patterns are not random but reflect
semantic relationships.
I should perhaps point out that the map has originally been established
inductively. I hope that this chapter has shown that the deductive perspective
leads to the same results, so that explanatory success has been achieved.
 135

5.7. Focusing and Sentence Accent

Before we look at the diachronic sources of indefinite pronouns in the next two
chapters, I would like to point out the intimate connection between focusing
and certain types of indefinite pronouns.

5.7.1. Sentence accent

In § 5.5.1 I illustrated the notion of a pragmatic scale with the example of scalar
and non-scalar focus particles. This was not accidental: Focusing plays an
important role in all scalarity phenomena. It seems that every pragmatic scale
presupposes focusing (whereas there is of course focusing without scalarity, as
in (223a)). Quite generally, a focus establishes a relation between the value of a
focused expression and a set of alternatives (see König (1991: 32), who cites
earlier work by Joachim Jacobs). In the case of a pragmatic scale, the non-
extreme values constitute this set of alternatives.
In this section, I would like to point out a common type of surface mani-
festation of focusing that is often observed in scalar-endpoint indefinites in
English and typologically similar languages: sentence accent. This is also
characteristic of other types of focusing, as illustrated in (268) (cf. König 1991:
12).

(268) sentence accent on focused element

a. (contrastive focus) COLLOR won the election, not Lula.
b. (focus particle) Only FRANCESCA bought a new bike.
c. (attitudinal verb) I regret that PALME was shot./
I regret that Palme was SHOT.
d. (question) Is MOTHER home?/Is mother HOME?

Constituents have to bear sentence accent in order to be interpreted as scalar

endpoints. This is true not only for true superlatives (269) and ‘pragmatic
superlatives’ (270), but also for free-choice indefinites (271). Sentences with
superlatives and pragmatic superlatives simply lose the scalar universal
interpretation when the stress is not on these elements (cf. 269b, 270b), but
sentences with free-choice indefinites become odd (271b).

(269) superlatives
a. (+U) She would review the CHEAPEST book for that prestigious
journal.
b. (–U) She would review the cheapest book for that prestigious
JOURNAL.
(270) ‘pragmatic superlatives’
a. (+U) The Dalai LAMA would apply force in such a situation.
b. (–U) The Dalai Lama would apply FORCE in such a situation.
(271) free-choice indefinites
a. You may invite ANYONE to our party.
b. ?*You may INVITE anyone to our party.

Free-choice indefinites contrast with specific indefinites in a particularly

striking way, because specific and other non-scalar-endpoint indefinites never
bear accent, even when they are in a position that normally bears sentence
 136

accent. In English, for example, the sentence accent is normally on the direct
object. (272) shows that free-choice indefinites are marked in that they attract
stress, and (273) shows that non-scalar-endpoint indefinites are marked in that
they reject stress.

(272) a. Ram may buy a BOOK.

b. ANYONE may buy a book. (?*Anyone may buy a BOOK.)
c. Someone may buy A BOOK.
(273) a. You may invite SANGITA.
b. You may invite ANYONE.
c. You may INVITE someone. (?*You may invite SOMEONE.)

I have not investigated stress systematically, but it appears that a similar sit-
uation obtains in many other languages. Below examples from Russian and
German are [Link]

Russian
(274) a. KTO UGODNO mož et kupit’ knigu.
(?*Kto ugodno mož et kupit’ KNIGU)
‘Anyone may buy a book.’
b. Kto-nibud’ mož et kupit’ KNIGU.
(?*KTO-NIBUD’ mož et kupit’ knigu.)
‘Someone may buy a book.’
(275) a. Ty mož eš ’ priglasit’ KOGO UGODNO.
(?*Ty mož eš ’ PRIGLASIT’ kogo ugodno.)
‘You may invite anyone.’
b. Ty mož eš ’ PRIGLASIT’ kogo-nibud’.
(?*Ty mož eš ’ priglasit’ KOGO-NIBUD’.)
‘You may invite someone.’
German
(276) a. IRGEND JEMAND kann ein Buch kaufen.
‘Anyone can buy a book.’
(≠ Irgend jemand kann ein BUCH kaufen.
‘Someone can buy a book.’)
b. Jemand kann ein BUCH kaufen.
(?*JEMAND kann ein Buch kaufen.)
‘Someone may buy a book.’
(277) a. Du darfst IRGEND JEMANDEN einladen.
‘You may invite anyone.’
(≠ Du darfst irgend jemanden EINLADEN.
‘You may invite someone.’)
b. Du darfst jemanden EINLADEN.
(?*Du darfst JEMANDEN einladen.)
‘You may invite someone.’

5.7.2. Emphatic vs. non-emphatic indefinites

Since the contrast with respect to sentence accent that was just exemplified is so
striking, I will occasionally use the terms EMPHATIC INDEFINITE and NON-
xviii
EMPHATIC INDEFINITE. These terms are used in a deliberately loose way here
because it is not always possible to pay attention to all details, but it is often
useful to have this distinction. The functions ‘specific (known or unknown)’
 137

and ‘irrealis non-specific’ require non-emphatic indefinites, and the functions

‘free choice’ and ‘comparative’ require emphatic indefinites. The functions
‘indirect negation’, ‘question’ and ‘conditional’ are interesting in that they
allow both emphatic and non-emphatic indefinites without a difference in the
truth conditions. In (278-79) the distinction between emphatic and non-
emphatic indefinites is only marked by the [Link]

(278) conditional
a. If you HEAR anything, wake me up.
b. If you hear ANYTHING, wake me up.
(279) questionxx
a. Can you SEE anything?
b. Can you see ANYTHING?

There is a clear meaning difference between the (a) and (b) sentences: In the (b)
case, a scale of alternative values is present of which the chosen value is the
endpoint. No such scale is present in the (a) sentences, where non-emphatic
any-indefinites are [Link]
Many languages use two different indefinites to render these two meanings.
If these indefinites are used in other functions as well, the emphatic indefinite
(corresponding to the (b) case) is used in other emphatic functions (comparative
and free choice), and the non-emphatic indefinite is used in other non-emphatic
functions (irrealis non-specific, specific). The following examples are
illustrative.

(280) French
a. Si quelqu’un vient, réveille-moi.
if someone comes wake-me
‘If anyone COMES, wake me up.’
b. Si qui que ce soit vient, réveille-moi.
if anyone comes wake-me
‘If ANYONE comes, wake me up.’
(281) Polish
a. Jez%eli co-ś zobaczysz, odrazu mnie obudź .
if what-INDEF see:2SG immediately me wake:IMPV
‘If you SEE anything, wakeme up immediately.’
b. Jez%eli co-kolwiek zobaczysz, odrazu mnie obudź .
if what-INDEF see:2SG immediately me wake:IMPV
‘If you see ANYTHING, wake me up immediately.’
(282) Hindi/Urdu
a. Agar koii fon kare, mujhe bataanaa.
if someone phone calls I:DAT tell:IMPV
‘If anybody CALLS, tell me.’
b. Agar koii bhii fon kare, mujhe bataanaa.
if someone INDEF phone calls I:DAT tell:IMPV
‘If ANYBODY AT ALL calls, tell me.’
(283) Chinese (Fengxiang Li, p.c.)
a. Wǒ bù xiangxin shénme rén lái le.
I not think what man come PFV
‘I do not think that anyone CAME.’
b. Wǒ bù xiangxin rènhé rén lái le.
I not think any man come PFV
‘I do not think that ANYONE came.’
 138

More examples of this can be found in Appendix A, e.g. in Russian, Hungarian,

Bulgarian, Italian.
One might say that these examples show that it is not sufficient to have one
question/conditional function, but that two such functions have to be
distinguished: first, ‘question/conditional + scalar endpoint’, and second
‘question/conditional + no scalar endpoint’. The first would be located next to
the irrealis-non-specific function, and the second would be located next to the
indirect-negation and comparative functions on the implicational map. I have
not adopted this proposal because the semantic distinction between the (a) and
(b) sentences in (278-83) is very subtle and is not easy to identify across
languages. However, future research along these lines should take this
possibility very seriously.

5.7.3. Further meaning differences associated with sentence accent

In addition to the semantic differences illustrated in (278-79), even more

dramatic distinctions can be coded by sentence accent. In particular, quite a few
languages have indefinites that may be used both in clearly emphatic and in
clearly non-emphatic functions. It appears that in such cases sentence accent in
general disambiguates the two readings. Thus, the German irgend-series is used
both in the free-choice function and in the irrealis-non-specific function. Which
function is intended is determined by the location of sentence accent.

(284) German
a. Du kannst irgend etwas KAUFEN.
you can INDEF something buy
‘You may buy something (or other).’
b. Du kannst IRGEND ETWAS kaufen.
you can INDEF something buy
‘You may buy anything.’

The situation is quite parallel in Portuguese:

(285) Portuguese
a. Você pode COMPRAR qualquer coisa.
you can buy INDEF thing
‘You may buy something (or other).’
b. Você pode comprar QUALQUER COISA.
you can buy INDEF thing
‘You may buy anything.’

In Spanish, a particularly dramatic distinction can be observed with the

expression en mi vida ‘in my life’ (and several related expressions). This phrase
has been idiomaticized as a negative-polarity item meaning ‘ever’, as illustrated
in (286a) (see also § [Link]). By analogy with indefinites like nunca ‘never’, en mi
vida may now also be used preverbally, and since preverbal negative indefinites
do not cooccur with verbal negation in Spanish (cf. § 8.2.2), there is no verbal
negation in (286b) (see Vermeylen 1984). However, (286b) has this negative
interpretation only when it is stressed, so that a scalar interpretation becomes
possible. Otherwise, it simply means ‘in my life’, as in (286c). (I am grateful to
Brenda Laca for native speaker judgments.)
 139

(286) Spanish
a. No lo he visto EN MI VIDA.
NEG it have:1SG seen in my life
‘I haven't ever (< in my life) seen it.’
b. EN MI VIDA lo he visto.
in my life it have:1SG seen
‘Never have I seen it.’
c. En mi vida LO HE VISTO.
in my life it have:1SG seen
‘In my life I have seen it.’

Hungarian has the lexical item valamennyi which has the two very different
meanings ‘some, several’ and ‘all’. There is probably no synchronic way in
which these two meanings can be unified, but I would propose the following
diachronic scenario to account for the synchronic polysemy: Originally the vala-
series had the free-choice function (as expected given its etymology: val- ‘be’, cf.
§ 6.2.3), which later developed into a non-emphatic meaning (‘some’) in the
manner outlined in § 6.4.2. But independently, it also developed into a
universal meaning, as outlined in § 6.5. Although the contemporary universal
meaning is no longer scalar, valamennyi preserved its original stress properties:
When it means ‘all’, it must bear sentence accent, but when it means ‘some’, it
cannot (Hunyadi 1981).

(287) Hungarian
a. Valamennyi fiú TANULT.
some boy learned
‘Some of the boys were learning.’
b. VALAMENNYI fiú tanult.
all boy learned
‘All of the boys were learning.’

Finally, in Modern Greek the indefinites of the típota-series may be stressed

or unstressed. If they are stressed (and thus focused, cf. Tsimpli and Roussou
1993), they are strictly limited to the direct-negation function, but they may also
have that function if they are unstressed, much like English any-indefinites. (See
Tsimpli and Roussou (1993: 147-50) for the subtle difference in interpretation
between the two variants of (288b).)

(288) Modern Greek

a. An dhís kanéna/ *KANÉNA,...
if see:2SG anyone anyone
'If you see anybody,...'
b. Dhen ídha kanéna/ KANÉNA.
not saw:1SG anyone anyone
'I didn't see anybody.'

All these facts once more underline the point that the notions of focusing and
scalarity are central for our understanding of both meaning and form of
indefinite pronouns.
 140

i
Veyrenc merely speaks vaguely of '-nibud’, indicateur de l’indéterminé virtuel du
particulier', as opposed to '-libo, indicateur de l’indéterminé du général', not even
attempting a detailed justification of his classification. To be fair, however, I must add
that Veyrenc’s prose text contains a lot of original useful observations and
interpretations, and that the classification in (165) is perhaps not the main point of his
paper.
ii
For the earlier discussion within the generative semantics debate, see Jackendoff
(1972a), Carden (1976), Aldridge (1982). For a later standard treatment within
Chomsky’s Government-Binding framework, see May (1985).
iii
Universal pronouns and determiners are therefore often called universal quantifiers in
the linguistic literature. The analogous use of existential quantifier for 'indefinite
pronoun' is much rarer, but is also occasionally found (e.g. König 1991: 32).
iv
Here I use restricted quantifiers (following McCawley 1981: §4.5) for convenience.
Strictly speaking, also the quantifiers in (166-168) should be restricted: (166a') (∃: thing
x) (Pavlina saw x); (167a') (∀: thing x) (Waldemar ate x); (168a') (∀: person x) (x
knows this secret). Natural languages have no expressions corresponding to variables
bound by unrestricted quantifiers – even the most general expressions, pronouns,
incorporate the meaning of an ontological category.
v
The situation is similar to the analysis of the focus particle even (cf. König 1991:
§4.2.1), which may characterize its focus value as maximal or minimal. In a question
like Can she even speak French?, either reading is possible, and here, too, a univocal
analysis would be highly desirable. See § 5.5, § 8.3.1 for more parallels between even
and any.
vi
Fodor (1970: ch. 2) claims that (186) allows a specific and a non-specific reading of a
friend of mine, whereas a sentence like A friend of mine is a busdriver allows only a
specific reading. I do not follow this. It is true that (186) may be construed in two ways,
(a) meaning that there is some particular friend of mine such that Mary believes he or
she is a busdriver, and (b) meaning simply that Mary believes that there is a friend of
mine who is a busdriver, though she may not know, or may not be concerned with,
which particular friend it is. But these two readings do not depend on the embedding in
a propositional-attitude verb and are also available in the simple sentence A friend of
mine is a busdriver. A similar example is discussed in Ioup (1977: 236): I talked to a
logician. This may mean (a) that I talked to a person who is independently identifiable
but happens to be described as a logician in (ii), or (b) that all that matters is that the
person I talked to was a logician. Ioup also characterizes these two readings as
specific/non-specific, but this terminological usage is infelicitous. The two readings
 141

should rather be distinguished as the VALUE reading and the ROLE reading, see the
discussion in Fauconnier (1985: § 2.2). Non-specific expressions can only have a role
reading, so there is a relation here, but the non-specific/specific distinction must not be
collapsed with the role/value distinction.
vii
A further insurmountable problem for the classical logical-semantic analysis is
discourse anaphora (cf. Jackendoff 1972a: 281-4, Fodor 1977: 186-93, Lyons 1977:
191-2). In response to these problems, Jackendoff's (1972a) theory of modal structure,
Fauconnier's (1985) mental-space theory, and Kamp's Discourse Representation Theory
(Kamp and Reyle 1993) were developed.
viii
Note that the more abstract notion of 'binding' that can account for both pronouns
and negative-polarity items has nothing to do with coreference – obviously, negative-
polarity items cannot be said to be coreferential with negation or an empty polarity
operator. (In addition, Progovac works within the 'generalized binding' framework of
Aoun (1986). The technical details need not concern us here.)
ix
Turkish is not a good example because hiç-indefinites are allowed in one non-neg-
ative environment, questions (cf. Appendix A, Section 23 for data). Progovac’s attempt
to explain this away by saying that the question particle mu/mü/mi/mı is 'negative' in
some way (1994: 151) is unconvincing. However, there are other languages where an
indefinite series is restricted to clausemate and indirect negation, e.g. the Maltese ebda-
series (Appendix A, Section 30), or the Hebrew af/š um-series (Section 31), etc.
Progovac could have chosen one of these languages to make her point.
x
Progovac does discuss free-choice indefinites (1994: ch. 7), but she makes the crucial
mistake of considering Serbian/Croatian bilo-indefinites as representing unambiguous
free-choice indefinites in all their uses, although these can be used in non-free-choice
functions like questions and conditionals (see Appendix A, Section 14 for the data).
xi
The same criticism applies to Chomskyan treatments of reflexive and ordinary pro-
nouns. These studies have rarely addressed the question why reflexive pronouns are so
often identical to emphatic reflexives (cf. König (1995) for an explanation in semantic-
pragmatic terms).
xii
There is also a suprasegmental difference between the two readings; see § 5.7.1
below for discussion. (Note, incidentally, that (225) shows the need to recognize
definite non-specific phrases, as argued in § 5.4.2 above.)
xiii
For a detailed discussion of pragmatic implicature in questions, see Fauconnier
(1980). (Note that questions are a problem for Ladusaw’s downward-entailment
approach, but they are not for Fauconnier’s pragmatic-implicature approach.)
xiv
Quantifying superlatives may also become idiomaticized or lexicalized and thus turn
into negative polarity items. For example, the Russian synthetic superlative malejš ij
‘the smallest’ can only be used in negative-polarity environments (the analytic
 142

superlative samyj malen’kij has no restrictions). An example is polnoe otsutstvie

malejš ix priznakov nacionalizma ‘the complete absence of the slightest signs of
nationalism’.
Hoeksema (1994) discusses the process of fixing or grammaticalization of
expressions which denote a scalar endpoint or are frequently associated with negative
polarity for other reasons.
xv
I am grateful to Ekkehard König for pointing this out to me.
xvi
The ambiguous status of any in comparatives between 'free-choice/universal any' and
'existential any' has sometimes been noted in the literature (Carlson 1981: 13,
Hoeksema 1983: 408-10). The intermediate location on the map is a reflection of this
status in my approach.
xvii
The 'focus gemination' that we find in Chechen indefinites (see (26) above) seems to
have a function similar to sentence accent, iconically reflecting the pragmatic emphasis.
xviii
Note that emphasis was the usual term for ‘focus’ until Chomsky (1970) used this
latter term. The term emphasis is of course associated with all sorts of other meanings,
but this is not a big disadvantage here because my use of this term is deliberately
somewhat loose.
xix
In the direct-negation function, too, both emphatic and non-emphatic indefinites
seem to be possible: I didn’t SEE anything vs. I didn’t see ANYTHING. However, it is not
clear to me whether one can say that only the second involves a scale. But intuitively,
the second sentence is ‘stronger’ than the first in much the same way as (278-79) (b) are
‘stronger’ than (278-79) (a).
xx
The same contrast is found with the Dutch cognate enig ‘any’, e.g. Heeft hij enig
success gehad met al zijn pogingen? This may mean (with enig stressed) ‘Has he had
any success at all with all his attempts?’, or it may mean (with enig unstressed) ‘Has he
had some success with all his attempts?’
xxi
This was pointed out to me by Ekkehard König. The relevance of stress in examples
of this type is also discussed in Kadmon and Landman (1993: 362-8).
 6
The Grammaticalization
of Indefinite Pronouns

6.1. Diachronic Typology

In this and the following chapter, I study the ways in which indefinite
pronouns arise and change over time in different languages and the regularities
in these changes. There are two main reasons for engaging in such a study of
diachronic typology. First, for the sake of diachrony itself. Language change is a
universal and essential feature of human language, and by studying the general
laws of language change, we learn a lot about human language.
But secondly, diachronic typology also helps us understand synchronic
language states better. All languages are constantly in a process of change, in a
kind of flux, and many features that do not fit neatly into a synchronic system
begin to make sense once a diachronic point of view is taken. This applies both
to recent innovations and to remnants of earlier regularities that are no longer
synchronically motivated. Languages can carry around such synchronic
irregularities for many generations of speakers, and if our goal is the
explanation of linguistic structures, we have to take diachronic explanations
into account.
Most importantly for my purposes, there is often a close correspondence
between the generalizations obtained from synchronic and from diachronic
typological studies, so that the results from such studies reinforce each other.
For instance, in Hawkins’s (1983) study of word order universals, the
implicational hierarchies that account for the cross-linguistic distribution of
word order patterns also make correct predictions about possible diachronic
changes, as Hawkins shows. If a language acquires new word order patterns, it
acquires them in accordance with the order of the implicational hierarchy.
Quite analogously, implicational maps that account for the cross-linguistic
distribution of different functions of grammatical categories also constrain the
possible diachronic changes: A category can acquire a new function only if that
function is adjacent on the semantic map to some function that the category
already covers. Semantic change of grammatical categories is 'incremental' (cf.
Croft et al. 1987), and grammatical categories gradually extend their uses along
the paths allowed by the map.
This correspondence between synchronic and diachronic typology can
also be observed in the case of indefinite pronouns. As I will show in this
chapter (especially § 6.4), the extension of indefinite series to new functions
proceeds along the paths permitted by the implicational map of chapter 4.
One problem for the diachronic-typological study of indefinite pronouns
at this stage is that there are very few specialized studies on diachronic change
in indefinite pronouns in individual languages. While I was able to make use of
quite a few studies on individual languages for the synchronic distribution, I
had to rely on other kinds of evidence for the diachronic study. First, some of
the large historical grammars of the major European languages do contain a
limited amount of relevant information. Second, I use comparative evidence
from closely related languages. And third, etymological information, which is
available for many languages, also gives us valuable insights, especially about
the source construction from which an indefinite pronoun was derived.
A key concept for understanding the genesis and later development of
indefinite pronouns is the concept of grammaticalization (cf. Lehmann 1982a,
Heine et al. 1991, Hagège 1993, Hopper and Traugott 1993, among many others).
Research in diachronic typology is to a large extent concerned with various
types of grammaticalization. Grammaticalization is the unidirectional gradual
diachronic change by which a lexical-syntactic source construction loses its
autonomy and is integrated into the grammar. In the following section (§ 6.2), I
identify four main source constructions from which the grammaticalization of
indefinite pronouns started. In a next step (§ 6.3-4), I will relate these changes to
general properties of grammaticalization, and discuss the consequences for
different theoretical accounts of grammaticalization changes.
In this and the following chapter mainly non-negative pronouns will be
treated. The discussion of the diachrony of negative indefinites will be reserved
for chapter 8 (especially § 8.3).

6.2. Source Constructions for Indefiniteness Markers

The grammaticalization processes are particularly interesting in the case of a
large subclass of interrogative-based indefinite pronouns. Accordingly, I will
concentrate the discussion on such indefinites in the following subsections. In
chapter 7, I will then discuss further diachronic sources of indefinite pronouns
that cannot be subsumed under grammaticalization.

6.2.1. The ‘dunno’ type

[Link]. The source construction. Some indefiniteness markers that combine

with interrogative pronouns have arisen from a clause with the meaning ‘I
don’t know’, or similar. This type is especially well-attested in European
languages. Some cases are shown in (289). (Here and in many of the following
examples, one indefinite pronoun, often the one denoting a person, stands for
the whole indefinite series.)

(289) ‘dunno’-indefinites from ‘I don’t know wh-’

a. Middle High German neizwer ‘somebody’
< ne weiz wer ‘(I) don’t know who’
b. Old English nthw‘somebody’
< ne wt hw‘(I) don’t know who’
c. Old Norse nekkver ‘somebody’
(> Swedish någon, Icelandic nokkur)
< *ne wait ik hwarir ‘I don’t know who’
d. Romanian (dialectal) nes3tine ‘some’
< Latin nescio quis ‘I don’t know who’
e. Bulgarian (dialectal) na(m)koj ‘somebody’ (cf. Paov 1965)
< ne znam koj ‘I don’t know who’
f. Old Church Slavonic nkto ‘somebody’
< *ne vkto ‘I don’t know who’
g. French je ne sais quel ‘some kind of’
cf. je ne sais (pas) quel ‘I don’t know which’
 A variant of the explicit negation ‘I don’t know’ is the rhetorical question
‘who knows?’, which by way of a conversational implicature renders the same
meaning. This type of source construction has been strongly grammaticalized
in Lithuanian, while in the other languages where I have found it it is quite rare
and highly expressive, i.e. it is still weakly grammaticalized.

(290) ‘dunno’-indefinites from ‘who knows wh-?’

a. Lithuanian kakas (kainkas, kanokas) ‘somebody’
< kas ino kas ‘who knows who’
b. Czech kdovíkdo
< kdo ví kdo ‘who knows who’
c. German wer weiß wer ‘someone or other’
cf. wer weiß wer ‘who knows who’

In Albanian, the strongly grammaticalized indefiniteness marker di- (e.g. di-

kush ‘somebody’, di-ç ‘something’, di-ku ‘somewhere’, etc.) is identical to the
root di- ‘know’. It is unclear whether this goes back to type (289) (with the
negation omitted) or to type (290) (with ‘who’ omitted) or perhaps to some
third type (e.g. a rhetorical polar question like ‘do I know who?’).
Another highly expressive variant of this type is an expression like ‘God
knows wh-’ or even ‘the devil knows wh-’. I know of no case where such a
source construction has been grammaticalized strongly.

(291) ‘dunno’-indefinites from ‘God (etc.) knows wh-’

a. Russian Bog vest’ kto ‘God knows who’
ert znaet kto ‘the devil knows who’
b. Serbian/Croatian bogzna ko ‘God knows who’
c. Slovak bohvie kto ‘God knows who’
d. French Dieux sait qui ‘God knows who’
e. Swedish Gud vet vad ‘God knows what’

The scarcity of attested cases of ‘dunno’-indefinites outside of Europe

might lead one to suggest that such indefinites are an areal feature typical of
Europe. But in view of the unsatisfactory documentation of most non-European
languages, this conclusion may turn out to be premature.
It is not difficult to see why an expression like ‘I don’t know’ should
come to mark an indefinite pronoun: Indefinite pronouns are typically used
when the referent is unknown to the speaker (though it need not be unknown,
cf. § 3.2.4). The diachronic scenario of its development is also easy to
reconstruct: The original source structure is an indirect parametric (or 'wh-')
question embedded in the matrix clause ‘I don’t know’ where the greater part
of the embedded question is omitted because it is obvious from the context (this
type of contextual omission was called 'sluicing' in Ross (1969); cf. also von
Bremen (1983: 118-19)).

(292) (She told him somethingi.) I don’t know what [iti was].

In a next step, this sentence is inserted into another sentence where the
interrogative pronoun occupies some syntactic position.

(293) She told him I don’t know what.

This is a kind of 'syntactic amalgam' of the type studied in Lakoff (1974) (cf.
Lakoff’s example I saw you’ll never guess how many people at the party). One could
propose a constituent structure as in (294) for it (cf. Espinal 1991: 748-9), where
there are two separate structures with their own root node that share one
constituent (the NP what).

(294) S S

NP VP NP VP

V NP V NP

She told him I don’t know what.

Whatever the best syntactic analysis of such constructions is, they are obviously
highly marked and prone to reanalysis. The erstwhile matrix clause ‘I don’t
know’ is reanalyzed as an indefiniteness marker, and when this way of
expressing indefinite pronouns becomes more frequent, it may undergo quite
radical phonological reduction, as shown by the Bulgarian, Romanian and
especially the Old Norse example.

[Link]. The original meaning. Now let us ask what kind of indefinite meaning
we get from the source ‘I don’t know’. Obviously we do not get the meaning
‘known to the speaker’, but can we be more specific? We can: In all the other
functions distinguished in § 3.3, the referent is unknown to the speaker, but this
source only gives rise to the specific-unknown function. The reason is that in all
the non-specific functions, it would be nonsensical for the speaker to state that
he or she does not know the referent because if the referent is non-specific,
nobody could possibly know it. This can also be illustrated by the impossibility
of adding I don’t know wh- when the referent of the indefinite pronoun is non-
specific. Thus, while (295a-b) are possible (like 292), (296a-c) are anomalous.

(295) specific
a. Susanne is thinking about something. I don’t know what.
b. She wants to marry an Ainu speaker. I don’t know whom.
(All I
know is that) she fell in love with him during fieldwork
sessions.

(296) non-specific
a. (irrealis: imperative) #Take some apple. I don’t know
which one.
b. (interrogative) #Did you see anybody? I don’t know whom.
c. (free choice) #You can take any apple. I don’t know which
one.

When we look at the way ‘dunno’-indefinites are used, we see that the
prediction that they have the specific-unknown function is confirmed.

(297) a. French
 Il nous reste encore je ne sais quel désir
vague,
it us remains still indef which desire vague
je ne sais quelle inquiétude. (Voltaire)
indef which restlessness
‘We are left with some kind of vague desire, some kind of
restlessness.’
b. Old English (Rissanen 1987: 417)
pær on innan giong / niDa nat-hwilc,
there on inside went [Link] indef-which
se pe neh gefeng / hæDnum horde.
(Beowulf 2214-16)
he that near caught heathen treasure
‘Some man or other crept inside it, who reached out
toward the
heathen treasure.’
c. Lithuanian (Pilka 1984: 29)
Ka-kas atejo.
indef-who came
‘Someone came.’

6.2.2. The ‘want/pleases’ type

[Link]. The source construction. Some indefiniteness markers that combine

with interrogative pronouns go back to an expression meaning ‘want’ or
‘pleases’ or similar. Examples are given in (298). (Again, the indefinite pronoun
denoting a person generally represents its series.)

(298) Latin qui-vis ‘anybody’ vis ‘you want’

qui-libet ‘anybody’ libet ‘it pleases’
Spanish cualquiera ‘any’ quiera ‘wants (subjunctive)’
Italian qualsivoglia ‘any’ voglia ‘wants (subjunctive)’
Russian kto-libo ‘anybody’ libo < ljubo ‘dear, pleasing’
kto ugodno ‘anyone’ ugodno ‘suitable’
Romanian cine-va ‘somebody’ va < vrea ‘wants’
Albanian kushdo do ‘wants’
Bulgarian koj-gode ‘anyone’ gode <
Old Church Slavonic god‘suitable’
Ossetic i-fændy ‘anyone’ fændy ‘you want’
Slovakvol’a-kto ‘anyone’ vol’a ‘wanting’
Serbian/Croatian ko mu drago ‘anybody’drago mu ‘dear to him’

For these indefinite pronouns, I hypothesize the source constructions in (299) as

the starting point of the grammaticalization process.

(299) a. You may take what you want [to take].

b. You may take what it pleases (you) [to take].

(299a) and (299b) are semantically equivalent and differ only in that the
addressee is the subject of the predicate ‘want’ in (299a), but the object of the
predicate ‘please’ in (299b). The predicate (‘want/pleases’) is in both cases the
predicate of a non-specific free relative clause which serves as an argument of
the main clause.
 In contrast to the previous case, this source construction does not require
any syntactic restructuring before it can be turned into a grammatical marker,
the reason being, of course, that the original construction consists of a
subordinate clause, rather than a superordinate clause as in § 6.2.1.
The hypothesized source constructions also explain why the
indefiniteness markers in this section are suffixes, whereas the indefiniteness
markers from ‘I don’t know’ are prefixes. And that these indefinites, too, are
based on interrogative pronouns finds a natural explanation in the fact that in
many languages, non-specific free relative clauses are formed with an
interrogative-based relative pronoun (cf. Lehmann 1984: 326)

[Link]. The original meaning. As in ‘dunno’-indefinites, it is pretty obvious

why source constructions like (299a-b) should be chosen for an indefinite
pronoun. A sentence like (299a) is a close paraphrase of a sentence with a free-
choice indefinite pronoun like You may take anything, and in § 3.3.3 we saw that
languages lacking free-choice indefinites use precisely such constructions to
render the same meaning.
Thus, we expect ‘want/please’-indefinites to express the free-choice
meaning initially, and this is indeed what we find in many cases:

(300) a. Spanish
Puedes traerme cualquier libro.
can:2sg bring:me any book
‘You can bring me any book.’
b. Latin
Utrum-libet elige; alterum incredibile est,
alterum nefarium
which-indef choose:impv other incredible is other
sinful
‘Choose either; one is incredible, the other sinful.’ (Quint.
81)
c. Russian
Ty moe’ sprosit’ to ugodno.
you can ask what indef
‘You may ask anything.’

6.2.3. The ‘it may be’ type

[Link]. The source construction. Many indefiniteness markers contain an

element that goes back to form of the verb ‘be’. Why this should be so is
perhaps not immediately obvious, but there is a straightforward account for it,
which will be presented immediately below. Consider the examples in (301).

(301) Russian kto by to ni bylo ‘anyone’ 'who it not be',

i.e. ‘whoever it may be’
kto-nibud’ ‘any-/some-one’ < kto ni budi
‘whoever it may be’
Bulgarian kojto i da e ‘anyone’ 'who also it be', i.e.
‘whoever it may be’
Serbian/Croatian ko bilo ‘anyone’ 'who it be'
French qui que ce soit ‘anyone’ ‘whoever it may be’
Icelandic hver sem er ‘anyone’ 'whoever it is'
Hebrew mi-e-hu ‘someone’ 'who that it (is)", i.e.
 ‘whoever it may be’
Korean mues-i-n-ka ‘someone’ mues ‘what?’, i-n- ‘be’, -ka
‘question particle’
Lezgian wuajit’ani ‘anyone’ ‘whoever it may be’
(wu‘who’, a-ji- ‘be’ -t’a conditional, -ni ‘also’)
Kannada yaar-aadaruu ‘anyone’ ‘whoever it may be’
(yaaru ‘who’, aa-d- ‘be’ -ar conditional, -uu ‘also’)
Czech kdo-si ‘someone’ < *kto sit ‘whoever it be’

The hypothesized source construction for these indefinites is a parametric

concessive conditional clause, as shown in (302) (cf. von Bremen 1983).

(302) You can take somethingi, whatever iti may be.

One might ask why the source construction cannot contain a non-specific free
relative clause, as in the ‘want’-type that we saw in the preceding section. This
would look as in (303).

(303) You can take whatever it may be.

The problem is that the subject ‘it’ (which appears overtly in Russian (to),
French (ce), and Hebrew (hu)) in this hypothetical source construction is
unmotivated, so I assume that (302) is correct. But eventually some kind of ill-
understood restructuring takes place and turns (302) into (303), which is then
grammaticalized, yielding the indefiniteness markers in (301). The replacement
of (302) by (303) may also be facilitated by the fact that in many languages,
parametric concessive conditional clauses and non-specific free relative clauses
are structurally quite similar (cf. Haspelmath and König (to appear)).
The typical structural features of parametric concessive conditional
clauses are also reflected in the resulting indefinite pronoun. Thus, the verb is
typically in some kind of subjunctive mood in concessive conditional clauses,
and this is what we find in the indefiniteness markers: French soit, Russian
bud’/by...bylo, Bulgarian da e, Proto-Slavic *sit (in Czech kdo-si) are all
subjunctive forms of the verb ‘be’. Russian parametric concessive conditional
clauses are marked by the pleonastic negator ni, and this is carried over to the
indefiniteness markers by to ni bylo and -nibud’.
A common formal feature of concessive conditional clauses is the focus
particle ‘also, even’ (cf. König 1988). In Lezgian and Kannada, this focus particle
follows the verb, which is in the conditional mood, cf. (304).

(304) a. Lezgian (Haspelmath 1993a: 399)

Hiniz wun fe-ji-t’a-ni, zun wa-qh galaz fi-da.
where you:abs go:[Link]-cond-even I:abs you-postess with
go-fut
‘Wherever you go, I will go with you.’
b. Kannada (Sridhar 1990: 912)
Aval5u es5t5u heel5i-d-ar-uu keel5-al-ee illa.
she [Link] tell-past-cond-even listen-inf-emph neg
‘However much she was told, she didn’t listen.’

Accordingly, the particle ‘also, even’ (-ni/-uu) also appearsin this place in the
Lezgian and Kannada indefiniteness markers. In European languages, by
contrast, the focus particle tends to occur after the interrogative pronoun,
before the verb, as for instance in Bulgarian:

(305) Bulgarian
Kakvo-to i da kae tja, toj te mâli.
what-indef also that says she he fut [Link]
‘Whatever she says, he will remain silent.’

As a result, we find the focus particle i in the Bulgarian indefiniteness marker

(WH-to) i da e ‘any-’ between the interrogative pronoun and the verb.
There are four further structural features that parametric concessive
conditionals commonly display, in addition to those already mentioned ((i)
subjunctive mood, (ii) pleonastic negator, (iii) focus particle ‘also, even’, (iv)
conditional marker): (v) a temporal adverb like ‘ever’, (vi) an additional general
subordinator (‘that’), (vii) an expression meaning ‘want’, (viii) other emphatic
particles like ‘now’, ‘only’ (cf. von Bremen 1983, Haspelmath and König (to
appear)). These are exemplified in the following sentences:

(306) temporal adverb

a. German (focus particle auch, temporal adverb immer),
Wohin auch immer du gehst, ich werde dich
nicht verlassen.
whither also ever you go I will you not
leave
b. English (temporal adverb ever)
Wher-ever you go, I will not leave you.

(307) general subordinator

a. Italian (che)
Quali che siano le sue protezioni, dovrà
scontare la pena.
which that be the his protections must:fut
serve the sentence
‘Whatever his protection is, he will have to serve his
sentence.’
b. Standard Arabic (maa)
Wa-ÌayQu-maa daxal-tum bayt-an fa-?aqiim-
uu fii-hi
and-where-sbor enter:perf-2pl house-acc and-remain:impv-2pl
in-3sg
?ilaa ?anna ta-rÌal-uu.
until sbor 2-depart:impf-pl
‘In what place soever ye enter into an house, there abide till
ye
depart from that place.’
(308) ‘want’
a. Spanish (-quiera from querer ‘want’, que general
subordinator)
No abras la puerta, quien-quiera que sea.
not open:impv the door who-want that be
‘Don’t open the door, whoever it may be.’
b. Hungarian (cf. akar ‘want’)
Akár-ki tanul-t, túl hangos-an tanul-t.
 want-who learn-past(3sg) too loud-adv learn-
past(3sg)
‘Whoever was learning, he was learning too loudly.’
(309) other emphatic particles
Swedish
Han måste lyckas, hur det nu skal gå till.
he must succeed how that now will come about
‘He has to succeed, however that will come about.’

It is not my task here to explain this astonishing variety of structural

means associated with parametric concessive conditionals (cf. Haspelmath and
König (to appear) for an attempt at an explanation). But we need to be aware of
the components of the source constructions in order to understand the
components of the resulting indefinite pronouns. Thus, the French
indefiniteness marker que ce soit contains a general subordinator (French que),
which is explained by the fact that it has arisen from a source construction with
a subordinator, cf. (307).
In quite a few cases, indefiniteness markers consist of the same structural
components as parametric concessive conditional clauses, but show no trace of
the verb ‘be’. Nevertheless, I claim that these, too, can be understood as arising
from such clauses. Examples are given in (310-14).

(310) temporal adverb

German wer auch immer ‘anyone’ immer ‘always’
Polish kto-kolwiek ‘anyone’ < *koli ‘ever’, wiek ‘age’
French quiconque ‘anyone’ < qui qu’ onques
(< Latin umquam ‘ever’)
quelque ‘some’ < quel que
M. Greek opjos-dhí-pote ‘anyone’ poté ‘(n)ever’
(311) focus particle ‘also, even’
German wer auch immer ‘anyone’ auch ‘also, even’
Dutch wie dan ook ‘anyone’ ook ‘also’
(312) general subordinator
Hebrew miehu ‘someone’ e ‘that’
(313) conditional marker
Turkish kimse ‘someone, anyone’ -se ‘conditional marker’
Udmurt kin ke ‘someone’ ke ‘conditional marker’
(314) ‘want’
Hungarian akár-ki ‘anyone’ akar ‘want’
Romanian ori-cine ‘anyone’ ori- < *volet ‘wants’
Spanish cual-quiera ‘any’ quiera ‘wants (3sg subjv)

A well-documented case of this development is French quelque, whose

evolution is traced in detail in Foulet (1919).
In descriptive grammars of European languages, the identity between
indefinite pronouns in (310-14) and the pronouns introducing parametric
concessive conditional clauses is often described as if the indefinite pronouns
were primary and their function in concessive conditionals secondary.
However, the opposite is in fact the case: The indefinite pronouns in (310-14)
are derived from the same kind of source construction that was discussed
above (see 302), with the difference that the verb ‘be’ has also been ellipted.
Although the verb ‘be’ cannot be reconstructed from the context (unlike the
ellipted parts in the earlier source constructions (see 292, 299)), it does not carry
much informational weight and is therefore dispensable.

(315) You can take something, whatever [it may be].

Such source constructions are not entirely hypothetical, as shown by (316) from
Dutch.

(316) Dutch (Paardekooper 1978: 572)

Je mag iemand anders vragen, wie dan ook.
you may somebody else ask who indef
‘(lit.) You may ask somebody else, whoever.’

[Link]. The original meaning. What kind of indefiniteness meaning do we get

from a parametric concessive conditional source construction? As in the ‘want’
type of the previous section, the original meaning is that of free choice. While
the source construction in the ‘want’ type explicitly leaves the choice to the
hearer, the same effect is achieved more indirectly by the concessive conditional
source construction. Concessive conditionals express a conditional relationship
between a consequent and a series of antecedent conditions, but in contrast to
ordinary conditionals, they entail their consequent. The antecedent conditions
are therefore irrelevant for the consequent, and concessive conditionals have
also been called 'irrelevance conditionals' (cf. König 1985). Thus, a parametric
concessive conditional clause like whoever it may be states that the identity of the
person in question is irrelevant, which amounts to the same as free choice. This
close semantic relationship between the two meanings is also reflected in
similar cooccurrence restrictions. For example, neither free-choice indefinites
nor the two source constructions can be used in contexts that allow only
specific reference (in (318-19), a non-habitual reading of the past tense is
intended):

(317) a. You may take any apple.

b. *You took any apple.
(318) a. You may take whichever apple you want.
b. *You took whichever apple you want(ed).’
(319) a. You may take an apple, whichever it may be.
b. *You took an apple, whichever one it may be.

Here are some examples of ‘it may be’-indefinites in their original (i.e. free-
choice) meaning:

(320) a. Bulgarian
Moe da vzeme koj-to i da e moliv.
you:may sbjv you:take which-indef pencil
‘You may take any pencil.’
b. Lezgian (Haspelmath 1993a: 195)
Bilbildi-kaj mani ni-waj ajit’ani tük’ür-iz
e-da.
nightingale-subel song who-adel indef compose-inf
can-fut
‘Anyone can compose a song about a nightingale.’
c. Dutch
Je mag wie dan ook uitnodigen.
 you may who indef invite
‘You may invite anyone.’

6.2.4. The ‘no matter’ type

[Link]. The source construction. In indefinite pronouns of this type, the

indefiniteness marker is derived from an expression meaning ‘it does not
matter wh-’, ‘it’s all the same wh-’. Some examples are given in (321).

(321) French n’importe qui ‘anyone’

cf. il n’importe (pas) ‘it does not matter’
Dutch onverschillig wie ‘anyone’
cf. onverschillig ‘indifferent’
German gleich welcher ‘any’
cf. gleich ‘equal, same’
English no matter who

I know of no indefinite pronoun of this type that has been strongly

grammaticalized, so the indefiniteness markers of this type are fairly
transparent. The source construction is shown in (322).

(322) You may take something. It does not matter what [you take].

Since the future indefiniteness marker is a superordinate clause here, as in the

‘dunno’-type of § 6.2.1, the source construction has to be 'amalgamated',
yielding a structure as in (323).

(323) You may take it does not matter what.

As the it does not matter clause is turned into an indefiniteness marker, it may
undergo simplification, especially omission of the copula (thus, German es ist
gleich w- ‘it is the same wh-’ becomes gleich wer).
In this type we see again an overlap with an earlier type: In some
languages, parametric concessive conditionals are expressed by means of a ‘no
matter’ expression, as for example in (324).

(324) Dutch, English

a. Onverschillig wat je wilt kopen je zult
het te duur vinden.
[Link] what you want buy you will it
too expensive find
b. No matter what you want to buy, you will find it too expensive.

However, it seems an unnecessary complication to assume that the

development of the indefiniteness markers of the ‘no matter’ type proceeded
via concessive conditionals. The source construction in (322) must be posited
anyway because not all languages with ‘no matter’ indefinites have ‘no matter’
concessive conditionals. For example, in French there is no evidence that a
n’importe-concessive conditional ever existed that could have given rise to the
indefiniteness marker n’importe:

(325) *N’importe qui vient, n’ ouvre pas la porte!

[Link] who comes neg open not the door
 ‘No matter who comes, don’t open the door!’

[Link]. The original meaning. Since ‘no matter’ indefinites are all weakly
grammaticalized, they all have the expected free-choice meaning
synchronically, as exemplified by (326).

(326) a. Dutch (Geerts et al. (eds.) 1984)

Kinderen van onverschillig welke leeftijd,kunnen
hier hun
children of irrelevant which age can here
their
hart ophalen.
heart [Link]
‘Children of any age can enjoy themselves here.’
b. French
N’importe qui peut venir à la réunion
demain.
indef who can come to the meetingtomorrow
‘Anyone can come to the meeting tomorrow.’

6.3. Grammaticalization
6.3.1. Grammaticalization theory

Let us now see in what way the changes described in the previous section can
be subsumed under the general phenomenon of grammaticalization. A
comprehensive and systematic description and discussion of the various
individual aspects of grammaticalization changes is Lehmann (1982a),
summarized in Lehmann (1985). Lehmann identifies three main parameters:
weight, cohesion, and variability, each of which has a paradigmatic and a
syntagmatic aspect. The six resulting parameters are shown in Table 7.

Table 7. Lehmann’s (1982a, 1985) six parameters of grammaticalization

(cover terms)
paradigmatic
syntagmatic
(– weight)
– integrity
– scope
(+ cohesion)
+ paradigmaticity
+ bondedness
(– variability)
– paradigmatic
variability
– syntagmatic
variability

A plus sign in front of a parameter means that with increasing

grammaticalization, the degree to which this parameter is present increases,
and a minus sign means that the degree to which a parameter is present
decreases. In principle, all of these parameters are affected simultaneously by
grammaticalization changes, and there is a high degree of correlation among
them. However, in each particular case there may be circumstances that make a
parameter inapplicable, so not all parameters can be observed in every change.
But what is strictly disallowed by Lehmann's theory is for different parameters
to change in opposite directions.
Three well-known paradigm cases of grammaticalization are the change
from a modal construction like cantare habeo ‘I have to sing’ in late Latin to the
Romance synthetic future, e.g. Portuguese cantarei ‘I will sing’; the development
of a suffixed definite article in Bulgarian (kniga-ta ‘the book’) from a
demonstrative determiner (cf. Old Church Slavonic kniga ta ‘that book’); or the
development of a comitative/instrumental case suffix in Turkish (aaçla ‘with
the tree’) from an earlier postposition (aaç ile). These examples serve to illustrate
what is meant by each of the six parameters.
Integrity is the most conspicuous parameter of grammaticalization. It
has two aspects, a phonological one and a semantic one. Loss of phonological
integrity or erosion means that an expression loses phonological substance (loss
of segments or whole syllables) or distinctiveness (loss of stress, assimilation),
as exemplified by all three examples above. Loss of semantic integrity or
desemanticization means that an expression loses semantic features, is
'bleached', generalized, or weakened (for alternative views, see § 6.4). For
example, the meaning of the definite article is weaker than and included in the
meaning of a demonstrative pronoun.
Reduction of (syntactic) scope means that an item that earlier combined
with constituents of arbitrary complexity is increasingly restricted to a word or
stem. For instance, the Latin verb habeo ‘have’ in its modal sense combined with
a verb phrase, whereas the Portuguese suffix -ei combines with a verb stem.
Increasing paradigmaticity means the integration into an increasingly
small and tightly organized paradigm. For example, the Turkish postposition
ile was part of a large paradigm of postpositions with little coherence, but by
becoming a case affix it joins the small Turkish case paradigm which consists of
only six cases (including the new comitative/instrumental).
Increase of bondedness means that an item is more tightly attached to a
host. In all three examples cited above, an erstwhile independent word
becomes an inseparable affix.
Loss of paradigmatic variability means that an item is increasingly
obligatory, more dependent on grammatical rules than on communicative
intentions. Thus, Bulgarian nouns have to have a definite article when its
conditions (uniqueness and inclusiveness) are met, independently of the
speaker’s communicative intentions.
And finally, loss of syntagmatic variability means an increasingly fixed
word order. In Latin, the verb habeo could precede or follow its complement,
but in the Romance future, the future suffixes may only follow the stem.
Let us now consider the way in which the various parameters of
grammaticalization are manifested in the development of indefinite pronouns
as presented in § 6.2. First of all, grammaticalization changes are always
unidirectional, and indefiniteness markers are no exception. No changes
whereby an indefiniteness marker turns into a superordinate clause of the
‘dunno’ or ‘no matter’ types, or into a free relative or concessive conditional
clause, have been attested. In the following sections we will consider each of
Lehmann’s six parameters of grammaticalization in turn.
6.3.2. Integrity

The semantic aspect of loss of integrity, desemanticization, will be treated in

detail in the next section (§ 6.4). So far we have only seen the original meanings
of the indefiniteness markers: ‘specific-unknown’ for the ‘dunno’-type (§ 6.2.1),
and ‘free choice’ for the other three types (§ 6.2.2-4). These are the meanings
that I assume for the earliest stage of the new indefiniteness markers, and they
are not very far away from the meanings of the source constructions yet. But
later desemanticization is amply attested, as shown in § 6.4.
Phonological attrition (the phonological aspect of the loss of integrity) is
especially radical in ‘dunno’-indefinites. We find phonological changes like ne
weiz > neiz (Middle High German), ne znam > nam (Bulgarian), *ne v> n, kas ino
kas > kakas (Lithuanian), or even *ne wait ik hwarir > nekkver (Old Norse). These
reductions go well beyond regular sound changes, but this is a frequent feature
of phonological attrition as part of grammaticalization changes and is therefore
not surprising. Phonological attrition is less conspicuous in the other three
types of indefinites, although it can be observed in some cases (e.g. Romanian -
va < vrea). The reason for this probably has to be sought in their basic free-
choice meaning. As was mentioned above (§ 3.2.6, § 5.7), free-choice indefinites
are typically stressed, and stressed expressions are naturally more resistant to
phonological reduction. This seems to apply also to the expressions of the type
‘God knows wh-’ (cf. 291) which have an emphatic value, are therefore stressed
and are not reduced phonologically.
However, free-choice indefinites are of course not immune to semantic
change, as will be documented below (§ 6.4.2). Once they are no longer
restricted to the free-choice function, they are unstressed and therefore subject
to more substantial changes. Thus Romanian -va (< vrea) is a general indefinite
with non-specific and specific uses and no longer has the free choice meaning.
Similarly, Russian -libo (< ljubo) and -nibud’ (< ni budi) no longer have the free
choice meaning and show some phonological reduction. Another suggestive
case from the ‘dunno’-type is Lithuanian kakas (cf. 290), which contrasts with
the other indefinites of its type (Czech kdovíkdo, German wer weiss wer, and
others) both formally (it is reduced) and functionally (it is not emphatic, cf.
Appendix A, Section 17). In order to prove the correlation between
desemanticization and phonological attrition one would have to study a large
number of examples, devise a measure for the degree of deseman-ticization and
phonological attrition, and perform a statistical analysis (as is done for futures
in Bybee et al. 1991). Such an investigation is beyond the scope of the present
work, but the few examples just cited show that this approach seems to be
promising.

6.3.3. Scope

The next parameter of grammaticalization is the reduction of syntactic scope.

This can also be observed in indefiniteness markers. For example, in older
French the indefiniteness marker n’importe combined with prepositional
phrases, as is evident from its position in front of the preposition in (327a)
(Grevisse 1986: § 373).

(327) a. (older French) n’importe à quelle heure ‘at any hour’

b. (modern French) à n’importe quelle heure
In the contemporary language, only (327b) is possible, where the indefiniteness
marker is combined directly with the noun phrase (or the pronoun). Thus, its
scope, which used to extend over a prepositional phrase, has been reduced.
Similarly, in Old Church Slavonic the indefiniteness marker n- could
precede a preposition, as shown in (328a). In modern Russian, this is no longer
possible (328b). Again, the scope of n- has been narrowed.

(328) a. Old Church Slavonic

nna koje msto ‘at some place’
b. Russian
na nekoe mesto (*ne na koe mesto)

Indefiniteness markers show a tendency to stand as close as possible to

the pronominal stem, i.e. to have the narrowest possible scope. Suffixal
indefiniteness markers often switch places with suffixal case markers after the
indefiniteness marker has become an affix. Thus, the Georgian suffixed
indefiniteness marker -me, which used to be a case-external extrafix (cf. § 3.1.1),
may now also occur in internal position. This type of change, the
externalization of inflection, is not motivated by grammaticalization, but
grammaticalization is responsible for creating the structures that are affected by
it (cf. Haspelmath 1993b for detailed discussion).

(329) Georgian (Vogt 1971: 44-6)

old new
Nom. ra-me ra-me
Dat. ra-s-me ra-me-s
Instr. r-it(i)-me ra-me-ti
Adv. ra-d-me ra-me-d

Again, it should in principle be possible to rigorously test the prediction

that the parameter of scope correlates with the others. Some anecdotal but
suggestive evidence for a correlation comes from Lezgian, which has two non-
negative indefiniteness markers, jat’ani and ajit’ani. Both belong to the ‘it may
be’-type (cf. 301), and the only difference is that jat’ani is based on the copula ja
‘be’, whereas ajit’ani is based on the full verb un ‘become, be’. However, there is
a striking meaning difference between them: ajit’ani, evidently the younger
form, has free-choice meaning, but jat’ani has only functions further to the left
on my implicational map. This correlates with different scopes: The
functionally younger form ajit’ani occurs in phrase-external position, i.e. has
scope over the phrase, whereas the older form jat’ani always stands next to the
interrogative word, i.e. has scope only over this word.

(330) Lezgian (Haspelmath 1993a: 194)

a. hitin mani ajit’ani
[Link] song indef
‘any kind of song’
b. sa hitin jat’ani mani
one [Link] indef song
‘some kind of song’

Analogously, in Japanese the indefiniteness marker -ka occurs inside case

particles, whereas -mo and -demo occur outside. The marker -ka, which has uses
to the left of the implicational map and is presumably older, has narrower
scope:

(331) Japanese
a. dare-ka-ni ‘to somebody’
b. dare-ni-mo ‘to nobody’
c. dare-ni-demo ‘to anybody’

6.3.4. Paradigmaticity

There is usually a very small set of indefiniteness markers (at most four or five)
that are strongly grammaticalized, but there is a much larger set of
indefiniteness markers with a low degree of grammaticalization in many
languages. For example, English has a core system consisting only of some-, any-
and no-, but at the periphery there are various forms such as wh-ever, God knows
wh-, no matter wh-, and others. Sometimes these weakly grammaticalized,
peripheral forms are not fixed in their internal structure. For example, in older
Latin there is not only the form qui-vis ‘anyone’, with the indefiniteness marker
vis (2nd singular indicative of velle ‘want’), but similar expressions with other
tense-mood forms of velle are attested as well:

(332) Latin
a. Dominus vino quid volet faciet. (Cato R.R. 47)
lord wine:dat what wants do:fut:3sg
‘The lord will do what he wants (=anything) with the wine.’
b. Facile cuj velles tuam causam probares. (C. Verr.
4.28)
easy whom:dat want:2sg your cause prove:2sg
‘You can easily prove your cause to anyone (=to whom you
want).’

Similarly, in modern French there is not only the weakly grammaticalized je ne

sais qu- ‘I don’t know wh-’ (e.g. 297a), but other pronominalized subjects (on
‘one’, elle ‘he’) are also possible, or different tense forms of the verb savoir
‘know’ (Grevisse 1986: § 373).

(333) French
a. J’ ai peur du sommeil comme on a
peur d’ un grand
I have fear [Link] sleep as one has fear
of a big
trou... menant on ne sait où. (Baudelaire)
hole leading one neg knows where
‘I’m afraid of sleep as one is afraid of a big hole that leads
one
doesn’t know where.’
b. Ce mariage, c’ était un point de
départ vers
this wedding it was a point of
departure toward
elle ne savait quelle vie.
she neg knew what life
‘This wedding was the beginning of she didn’t know what
kind of
 life.’

This freedom of choice is no longer possible in indefiniteness markers with a

high degree of grammaticalization.

6.3.5. Bondedness

Increase of bondedness between two expressions is the gradual transition from

juxtaposition (where both are independent words) to cliticization and affixation,
and possibly on to internal modification. It is often hard to tell which degree of
bondedness an indefiniteness marker has, but there is no doubt that
indefiniteness markers start out as (sequences of) independent words and end
up as affixes. To give just one example, the Russian marker by to ní bylo (e.g. in
gdé by to ní bylo ‘anywhere’) still has its own stress, whereas the marker -nibud’
(e.g. in gdé-nibud’ ‘somewhere, anywhere’) is stressless and thus at least a clitic,
if not a suffix. The higher degree of bondedness of -nibud’ (which is also
reflected in its spelling) again correlates with its higher degree of
desemanticization.

6.3.6. Paradigmatic and syntagmatic variability

With increasing grammaticalization, the items undergoing the change become

increasingly obligatory, and there is less and less choice between different
members of a paradigm. This parameter is not so easy to illustrate in the
domain of indefiniteness markers because there are no syntactic environments
that require an indefinite pronoun (in contrast to cases or agreement categories,
for instance, which are obligatory under certain syntactic conditions,
independently of their meaning). But one might perhaps say that, for example,
weakly grammaticalized free-choice pronouns leave the speaker more freedom
to select between them (e.g. between Dutch wie dan ook, wie ook, wie ook maar,
onverschillig wie, om het even wie, gelijk wie, all meaning ‘anyone’), whereas
strongly grammaticalized ones allow fewer alternatives (e.g., there is no
alternative to Dutch iemand ‘someone’ in iemand is gekomen ‘someone came’).
Syntagmatic variability, i.e. increasing fixation of order, cannot be
observed in indefiniteness markers because the source constructions already
show fixed word order, so this parameter is vacuous in our case.

6.3.7. The explanatory power of grammaticalization

The paths of grammaticalization that we have seen in this section allow us to

formulate two types of diachronic explanations: parochial and universal
explanations. The parochial explanations concern facts of individual languages
that become clear once they are viewed against the background of
grammaticalization theory, e.g. the fact that the Russian indefiniteness marker -
nibud’ contains the negative marker ni and the root bud’ ‘be’, or the fact that the
Albanian indefiniteness marker -do is homonymous with do ‘want’. These facts
are of course not part of the linguistic knowledge of the speakers of these
languages, but in the present context that does not mean that they are
automatically uninteresting.
At another level, grammaticalization also explains universal or cross-
linguistically widespread properties of indefinite pronouns. For example, the
fact that indefinite pronouns are so often based on interrogative pronouns is
explained (at least in part) by the source structures identified in § 6.2, which
contain embedded interrogative, free relative and parametric concessive
conditional clauses, and these three clause types usually contain an
interrogative pronoun. But the strongest predictions are made by the general
tenet of grammaticalization theory that there is a correlation between the
degrees of grammaticalization of the six parameters of § 6.3.1. The overall
prediction is that if an element is more grammaticalized than another element
on some parameter, then it is grammaticalized also on all other parameters (or
at least not less grammaticalized). We have already seen some examples of such
correlations in § 6.3.2-6, and some more evidence will be presented in § 6.4.

6.4. Desemanticization:
The Semantic Side of Grammaticalization
6.4.1. Three competing theories of semantic grammaticalization

In Lehmann’s systematization of grammaticalization processes, the semantic

aspects are subsumed under 'loss of integrity', i.e. Lehmann stresses those
aspects of semantic development that have been characterized as 'bleaching'
(Givón 1975), 'generalization', or 'weakening' (Bybee and Pagliuca 1987) of
meaning. Other authors have put the emphasis on other aspects of semantic
grammaticalization. Thus, Heine et al. (1991) see metaphorization as the main
driving force behind semantic change associated with grammaticalization. For
example, often concrete spatial expressions like ‘head’ come to be used as local
relators with abstract meanings like ‘on’, ‘in front of’, or ‘before’, or a concrete
spatial verb like ‘go (to)’ comes to express future meaning. Traugott (1988), by
contrast, emphasizes the role of 'pragmatic strengthening' in the development
of grammatical meaning. For instance, she assumes that the semantic extension
of English while from a strictly temporal subordinator to an adversative one
(‘whereas’) has to do with grammaticalization. It would be nice if the evidence
of grammaticalization of indefinite pronouns helped to resolve the theoretical
issue, and indeed it does.
Most of the semantic changes that are found in indefinite pronouns and
that are part of the grammaticalization process are clear instances of weakening
or generalization of meaning, so the evidence from indefinite pronouns favors
the views of Lehmann, Bybee and Pagliuca, and Givón. In contrast, there is no
evidence whatsoever for metaphorization in indefinite pronouns – the source
structures of § 6.2 are already rather abstract, so there is no change from
concrete to abstract here. And those instances of pragmatic strengthening that
we do find in indefinite pronouns are not connected with grammaticalization –
appreciative meanings like ‘some important person’ for someone, and
depreciative meanings like ‘an unimportant person’ for anyone are not linked to
grammaticalization. Indeed, they are rarely conventionalized, so that there is
rarely any semantic change. (These senses are discussed in § 7.5.4 below.)
None of the three theories of semantic grammaticalization has made
crucial use of indefinite pronouns (although Lehmann (1982a: 51-6) discusses
them), so the fact that the weakening/generalization hypothesis can
accommodate this new case is a point in its favor.
The semantic development of indefinite pronouns can be characterized
as 'generalization' in that indefinite series are often diachronically extended to
more functions on the implicational map of chapter 4, thus becoming more
general. As mentioned in § 4.4, the implicational map makes predictions about
the route of the change: Indefinite series can extend only to functions that are
adjacent to those that it already covers, and the extension is incremental, i.e. the
marker is extended only to one new function at a time (cf. Croft et al. (1987) for
a programmatic proposal along these lines).

6.4.2. Extension from ‘free choice’

The clearest cases of generalization of the meaning of indefinite pronouns

involve indefinites whose original function was that of free choice, and which
were then extended to other functions to the left of ‘free choice’ on the map.
The extension thus proceeds from right to left on the scale in Figure 19. This
scale is identical to the map of chapter 4 except that the two negation functions
are missing. Below the scale, I give some examples of indefiniteness markers
whose original function must have been ‘free choice’ because they come from
one of the sources in § 6.2.2-4.

Figure 19. Diachronic extension of indefinite series from ‘free choice’

specific irrealis- question/ comparative free choice
non-specific conditional
|———————|
Russian ugodno
|—————————————|
Czech -koli, Lezgian ajit’ani
|——————————————————|
French que ce soit
|—————————————————|
Russian -libo
|———————————|
Russian -nibud’
|————————————————|
Lezgian jat’ani, French quelque
|—————|
Czech -si

The first three examples show different stages of extension to the left. The
fourth example, Russian -libo, covers one more function on the left, but on the
other hand, it no longer has the original free-choice function. Russian -nibud’
has even lost the comparative function. The next two examples, Lezgian ajit’ani
and French quelque, in addition have the ‘specific’ function (see Foulet (1919) for
the semantic development of quelque), and Czech -si only has the ‘specific’
function (but recall that its etymology is not certain, cf. § [Link]).
Thus, an original free-choice indefinite may be extended all the way to
the opposite end of the scale. However, generalization of the meaning is clearly
not the whole story, because indefinites that have been extended to functions in
the left part of the scale lose some of the original functions in the right part.
Rather than growing larger and larger, the area covered by an indefinite shifts
to the left, like a window that opens up the view on a limited area of the
semantic space. But the shift of the window from right to left is apparently
unidirectional – indefinites do not acquire the comparative or free-choice
functions by semantic extension. How can we understand this leftist orientation
of indefinites if it is not motivated by generalization?
 Here we have to make use of the notion of 'weakening'. The functions on
the right side of the scale are in some sense 'stronger' than the functions in the
middle and on the left, and a change from right to left means a loss of 'strength',
just as predicted by the weakening view of semantic grammaticalization. To
express this visually, the scale in Figure 19 can be represented as a 'trough' (cf.
Figure 20), where the stronger functions are above the weaker functions. The
diachronic extension is restricted to a downward movement in this
visualization.

Figure 20. Diachronic extension of indefinite functions from ‘free choice’:

the 'trough' model

free choice

comparative

question/
conditional
weakening
irrealis
specific non-specific

6.4.3. Semantic change as weakening

After these rather abstract considerations, let us now ask what 'weakening'
means in semantic terms.

[Link]. Loss of focusing and scalarity. If we just look at the free-choice function
and at the irrealis-non-specific function, it is immediately clear that semantic
substance has been lost. Consider the minimal contrast in (334).

(334) a. (free choice) You may invite anyone.

b. (irrealis non-specific) You may invite someone.

The free-choice indefinite in (334a) expresses the endpoint of a pragmatic scale,

as we saw in § 5.5.5, but no pragmatic scale is associated with the simple non-
specific indefinite in (334b). Both share the feature of non-specificity – recall
that free-choice indefinites are also non-specific (cf. § 3.2.6). In those cases
where an original free-choice indefinite has acquired the irrealis non-specific
function, we can say that the semantic development consists in the loss of
focusing and thereby of the semantic component of the pragmatic scale and its
endpoint. Only the semantic component of non-specificity is preserved. This
development must have taken place, for example, in Russian -nibud’- and -libo-
indefinites.

[Link]. Loss of non-specificity. Desemanticization may also be understood as

loss of semantic substance in the further development from the simple non-
specific function to the ‘specific-unknown’ function. Consider the minimal
contrast in (335).
(335) a. (irrealis non-specific) She may go somewhere.
b. (specific unknown) She went somewhere (I don’t know
where).

If an indefinite that has the simple non-specific function is extended to the

‘specific-unknown’ function, it loses the feature of non-specificity. Since the
referents of non-specific indefinites are also necessarily unknown to the speaker,
the ‘unknown’ feature is common to both, and it is the only feature that is left
after non-specificity has been lost. This development must have taken place in
Portuguese qualquer-indefinites, for example.

[Link]. Loss of unknownness. Finally, even the feature of unknownness may

be lost, and then the indefinite may even be used in the specific-known
function. This development must have taken place, for example, in Romanian -
va-indefinites. (Of course, one could also think of this change as the acquisition
of the new feature 'unknown', so this extension is not a strong argument for the
view that semantic change in indefinites generally means loss of semantic
features.)

[Link]. The comparative. In the preceding subsections, I hope it has become

plausible that the extension of the functions of original free-choice indefinites
can be understood as the loss of semantic substance. The developments are
summarized in (336).

(336) loss of semantic substance of original free-choice indefinites

free choice > simple non-specific : loss of focusing
and scalarity
simple non-specific > specific unknown : loss of specificity
specific unknown > specific known : loss of
unknownness

But so far I have not accounted for the two intermediate functions ‘comparative’
and ‘question/conditional’. Unfortunately, the semantics of the comparative
function is not clear to me, so I cannot say much on its semantic development
here. However, it has been remarked several times in the literature that the
comparative function is intermediate between free choice and negative polarity
(cf. § 5.6), and if that is correct, then whatever explains the shift from free choice
to negative polarity will also explain the shift from the free-choice function to
the comparative function, and from the comparative to the
question/conditional function.

[Link]. Negative polarity: questions and conditionals. The semantic difference

between free-choice and negative polarity has been described above as the
difference between non-reversed and reversed pragmatic scales. It is not clear
that this difference can be described as 'weakening' in any way, and I can see no
other reason why there should be a unidirectional development from non-
reversed to reversed scales. Thus, I have to admit that I have no good
explanation for the fact that free-choice indefinites commonly acquire the
negative-polarity function, whereas negative-polarity indefinites are not
generally extended to the free-choice function.
However, semantic weakening can be observed in questions and
conditionals in a different way. Recall from § 5.7.2 that both emphatic and non-
emphatic indefinites may occur in these contexts, where they have the same
truth conditions but subtly different meanings, as in (337).

(337) Russian
a. Esli ty uslyi to-nibud’, razbudi menja.
if you hear what-indef wake:impv me
‘If you hear anything, wake me up.’
b. Esli ty uslyi to by to ni bylo, razbudi
menja.
if you hear what indef wake:impv me
‘If you hear anything at all, wake me up.’

In § 5.7.2 we saw that the meaning difference between these sentences can be
characterized as 'presence vs. absence of scalarity'. In (337b), to by to ni bylo is
the endpoint of a pragmatic scale, hence a more emphatic reading results ('if
you hear the slightest noise, e.g. if a cat meows'). At some stage the indefinite in
(337b) may lose its emphatic value and become equivalent to (337a) – this must
indeed have happened in Russian, because etymologically -nibud’ is completely
analogous to by to ni bylo (§ [Link]).
Thus, the question and conditional functions serve as a bridge between
the obligatorily scalar free-choice and comparative functions, and the
obligatorily non-scalar irrealis-non-specific function.

6.4.4. Extension from ‘dunno’

In addition to the source constructions that yield scalar-endpoint indefinites (§

6.2.2-4), there is also one source construction that yields specific indefinites: the
‘dunno’ type of § 6.2.1. As I showed in § [Link], the original function of recently
grammaticalized ‘dunno’-indefinites is ‘specific unknown’. Like the scalar-
endpoint indefinites, ‘dunno’ indefinites undergo semantic extension when
they are strongly grammaticalized, and as a result they may come to cover a
substantial portion of the implicational map, and perhaps lose their original
function. However, here it is much more difficult to make generalizations
because such cases are not common. In the modern European languages, only
Lithuanian ka-, and Albanian di-, and perhaps Slavic n-, Scandinavian någon
(Swedish)/nokkur (Icelandic) represent strongly grammaticalized ‘dunno’
indefinites. It appears that ‘dunno’ indefinites resist stronger
grammaticalization for some reason.
But Lithuanian ka- and Albanian di- can also be used in non-specific
functions and in questions and conditionals, and so can neiz-indefinites in older
German. This is exemplified in (338-39).

(338) Lithuanian
a. Ar tu ka-ka matai?
q you indef-what you:saw
‘Did you see anything?’
b. Jei tu ka-ka matai, pasaky-k man.
if you indef-what you:saw tell-impv me
‘If you see anything, tell me.’
(339) older German neiz- (neusz) (Grimm’s dictionary, s.v. neizwaz)
a. Nach solchem rumor... fraget er seinen
nachbarn, ob ihm
 according such rumor asks he his neighbor
whether he
neusz was gebrest.
indef what lacks
‘According to this rumor he asks his neighbor whether he
lacks
anything.’
b. Ich müeste mich wol imer schamen, solte ich

I should refl prt always [Link] should I

fürhten neizwaz.
fear something
‘I would have to be ashamed forever if I were afraid of
anything.’

Thus, in addition to the right-to-left extension of Figure 20, we also have to

assume left-to-right extension as shown in Figure 21.

Figure 21. Diachronic extension of indefinite functions from ‘dunno’

specific irrealis question/

unknown non-specific conditional
—––––––––––––––––––––––––––––––––––––>

If there is both right-to-left extension and left-to-right extension on the map,

does that mean that the semantic grammaticalization is not unidirectional? I do
not think so. I have argued above that the most important semantic change in
the grammaticalization of indefinites is semantic weakening, and this is a
dimension that is not directly represented in Figure 21. In the weakening of the
original ‘dunno’ meaning, the most important change is that the speaker’s lack
of knowledge is no longer emphasized, and similarly, in the weakening of free-
choice indefinites, the most important change is the loss of scalarity. Thus,
instead of two opposite movements in one dimension, what seems to be going
on is unidirectional movement in two dimensions, from strong (or emphatic) to
weaker (or less emphatic). This can be represented by the ‘trough model’ in
Figure 22.

Figure 22. Diachronic extension of indefinite functions from ‘free-choice’ and ‘dunno’:
a trough model

emphatic/ emphatic free choice

stronger unknown
comparative
specific irrealis question/
weaker/ non-specific conditional
non-emphatic

It must be admitted that this visual representation cannot capture all the
relevant details. But it does seem useful as an approximation, and it drives
home the point that semantic grammaticalization of indefinite pronouns is
primarily weakening of emphasis, not metaphorization or pragmatic
strengthening.
 In addition, the model in Figure 22 helps us understand an observation
that was made above in § 4.5 (Principle 1): In the middle of the implicational
map, indefinite pronoun series always express more than two adjacent
functions. This seems to be due to the fact that the functions in the middle,
which are the functions at the bottom of the trough in Figure 6.4, are ‘weakest’,
i.e. least distinctive, and hence least likely to be expressed by a unique
indefinite series that has no other functions.

6.5. From Free-Choice Indefinite to Universal Quantifier

In some languages, the indefinites that express the free-choice functions can
also be used as true universal quantifiers, corresponding to English every,
everyone, everything, etc. Since the meaning distinction between ‘any’ and ‘every’
is often quite subtle, we can be sure that we are dealing with a truly universal
use only if the expression can be used in contexts that do not allow free-choice
indefinites, as in (340).

(340) German
a. Der Vater gab jedem Kind einen Apfel.
‘The father gave every child an apple.’
b. Jeder Mensch muss sterben.
‘Every human being must die.’

But how can we be sure that the German determiner jeder is also a free-choice
indefinite? True, it can be used to translate English free-choice any, as in Jeder
Idiot würde das sehen ‘Any idiot would see that’. But we saw in § 3.3.4 that some
languages do not have special free-choice indefinites and use universal
quantifiers in their place. In the case of German, a convincing argument is that
jeder may also be used in the indirect-negation function, where it cannot be
paraphrased by ‘every’ (cf. 341). In such cases, jeder also combines with non-
countable nouns.

(341) German
a. ohne jede Hilfe
‘without any help’
b. Sie leugnete jede Schuld.
‘She denied any guilt.’

Thus, German jeder spans the four functions ‘indirect negation’, ‘comparative’,
‘free choice’, and ‘universal’. We thus have reasons to add an additional
function to our implicational map, which would now look as in Figure 23,
where the distribution of jeder is shown.

Figure 23. German jeder on the extended implicational map

direct
question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice universal
I have not found many languages where one expression covers both the
indirect-negation function and the universal function, but it is intuitively highly
plausible that ‘universal’ should be located next to ‘free choice’ on such an
extended implicational map.
This would also account for the fact that free-choice indefinites may
diachronically evolve into universals, as I argued in Haspelmath (1995). Again,
the evidence for this development is rather indirect, but a number of languages
have universal quantifiers which consist of the same formal elements as (free-
choice) indefinite pronouns and hence must go back to one of the source
constructions identified in § 6.2.2-3 and § 7.1. Compare the following cases:

(342) a. Romanian fiecare ‘every(one)’

care ‘who, which’ fie ‘be (3sg subjunctive)’
b. Latin quisque ‘every’
quis ‘who, which’ -que ‘and, also’
c. Gothic hwarjizuh ‘every’
hwarjis ‘which’ -uh ‘and, also’
d. Old Church Slavonic
ku-ido ‘every’
ku ‘which’ -ido [meaning unknown]
e. English each
< Old English ælc < *a-whilc ‘ever-which’
(the strengthened form ever-each later became every)

The semantic change from ‘any’ to ‘every’, which must be posited to account
for these cases, also involves the loss of the semantic feature of scalarity, like
the change from ‘any’ to ‘some’ that was posited in § 6.4.2. However, in this
case the semantic feature of non-specificity is also lost, and the implicature of
universal quantification is strengthened to a semantic component (in contrast to
the change from ‘any’ to ‘some’, where the universal implicature is also lost, but
non-specificity is preserved).
Thus, the meaning of free choice can develop in two directions: to ‘some’,
and to ‘every’. Both these developments seem to be unidirectional. ‘Some’
cannot develop into free-choice ‘any’, and ‘every’ cannot develop into ‘any’,
either.
 194

Further Sources of Indefinite Pronouns

This chapter deals with diachronic sources of indefinite pronouns that cannot
be accounted for by grammaticalization. First I discuss indefinite pronouns
marked by scalar focus particles like ‘even’ and ‘at least’ (§ 7.1), and then I
discuss and reject the possibility that the disjunctive conjunction ‘or’ may be
used as an indefiniteness marker (§ 7.2). In § 7.3 I look at bare interrogatives
that are used as indefinites, and in § 7.4 I treat reduplicated indefinite
pronouns. These last two types are not so common in European languages, but
they are widespread elsewhere. Although I do not have a good explanation for
the form-meaning relation in either of these form types, I formulate a number
of interesting generalizations. Finally, § 7.5 deals with some further diachronic
issues, such as the change from generic nouns and ‘one’ into indefinite
pronouns, and the borrowing of indefinite pronouns.

7.1. Indefinite Pronouns Marked by Scalar Focus Particles

7.1.1. The facts

In many languages there are indefiniteness markers that are identical to scalar
focus particles meaning ‘even’. Since normal additive focus particles like ‘also’
often have a scalar value, too (König 1991: 68), it is not necessary or even
possible to strictly separate the meanings ‘also’ and ‘even’. As a rule, when an
indefiniteness marker is identical to a form that is glossed ‘also’ in a grammar
or dictionary, I assume that its scalar use (‘even’) is responsible for its
functioning as an indefiniteness marker. (Such particles often also have the
meaning ‘and’, König 1991: 6-66.)
Below is a selection of such cases from a wide variety of languages
(supplementing the earlier lists in Coyaud and Aït Hamou (1972, 1976), König
(1991: 67), Gil (1993)). In most cases, focus particles are added to interrogative
pronouns.

(343) additive focus particles on interrogative pronouns

Serbian/Croatian i-ko ‘anyone’ i ‘and, also, even’
Indonesian siapa-pun ‘anyone’ -pun ‘also, even’
Tagalog kahit na sino ‘anyone’ kahit (na) ‘even’
Hittite kuiš -ki ‘someone’ -ki ‘and, also’
Even Ni-de ‘someone’ -da/-de ‘and, also’
Kannada yaar-uu ‘anyone’ -uu ‘and, also’
Ancash Quechua ima-pis ‘anything’ -pis ‘also, even’
Selkup ämtä kuty ‘nobody’ ämtä ‘even’
Nivkh aN-hagin ‘nobody’ hagin ‘even’
Gooniyandi ngoorndoo-ngaddaya ‘someone’ -ngaddaya ‘also’
Sanskrit kaś cana ‘anyone’ cana ‘even’
Japanese nani-mo ‘nothing’ -mo ‘also’
nan-demo ‘anything’ -demo ‘even’
 195

Less frequently we find scalar additive focus particles added to generic nouns
or the numeral ‘one’ (cf. 344), or to (non-derived) indefinite pronouns (cf. 345).

(344) additive focus particles on generic nouns and ‘one’

Chechen addam a ‘nobody’ cf. adam ‘person’
a ‘also’
Hebrew af eÌad ‘nobody’ cf. eÌad ‘one’
af ‘even’
Lezgian (sa) kas-ni ‘nobody’ cf. (sa) kas ‘(one) person’
-ni ‘and, also’

(345) additive focus particles on (non-derived) indefinite pronouns

Hindi/Urdu koii bhii ‘anybody’ cf. koii ‘someone’
bhii ‘also, even’
Dutch ook maar iemand ‘anybody’ cf. iemand ‘someone’
(cf. Paardekooper 1979) ook maar ‘even, at least'

While the use of a focus particle meaning ‘even’ as an indefiniteness marker

is probably most common, some languages use a scalar focus particle meaning
‘at least’. A typical example is Russian xot’, which is used as in (346) (cf. Parš in
(1987, 1988) for detailed discussion of Russian xot’).

(346) Russian (cf. Parš in 1987: 72)

a. Skaž ite xot’ slovo.
say [Link] word
‘Say at least a word.’
b. Ja rabotaju s utra do več era, soderž u tebja, – mogu ž e ja

I work from morning till evening support you can PRT I

imet' xot’ to udovol’stvie, č tob obo mne zabotilis’.
have [Link] that pleasure that about me care:3PL
‘I work from morning till evening, I support you, – so I should at
least have the pleasure that somebody cares for me.’
c. Ona gotova xot’ v pustynju bež at’ so mnoj.
she ready even in desert run with me
‘She is ready to run away even into a desert with me.’

As example (346c) shows, xot’ can also mean ‘even’ in certain environments. In
contrast to English even and corresponding expressions in other languages like
German sogar, French même (also Russian daž e), which characterize the value of
its focus as high on a pragmatic scale, Russian xot’ can also be used when the
value of its focus is characterized as low on a pragmatic scale, as in (346a-b), i.e.
only medial scalar values are not possible.
In several languages, words that also serve as focus particles of the xot’ type
are used as indefiniteness markers, as the examples in (347) show.

(347) xot’-type focus particles on interrogative pronouns

Russian xot’ kto ‘anyone’ xot’ ‘at least’
Finnish vaikka kuka ‘anyone’ vaikka ‘at least’
Modern Greek kan-énas ‘anyone’ kan ‘at least’
Latvian kaut kas ‘something’ kaut ‘at least’
Hungarian akár-ki ‘anybody’ akár ‘at least’
Nenets xibja-xart ‘nobody’ -xart ‘at least’
 196

Lezgian wuž x̂ ajit’ani ‘anybody’ x̂ ajit’ani ‘at least’

Kannada yaar-aadaruu ‘anybody’ -aadaruu ‘at least’
W. Greenlandic suna-luunniit ‘anything’ -luunniit ‘at least’
Yakut kim eme ‘somebody’ eme ‘at least’

7.1.2. Two possible pathways

It is not immediately obvious why a focus particle meaning ‘even’ or ‘at least’,
when combined with an interrogative pronoun, should yield an indefinite
pronoun. One avenue of explanation that one might try is to relate the use of
focus particles as indefiniteness markers seen in § 7.1.1 to the well-understood
cases of § 6.2.2-3. Indeed, additive focus particles are also widely used in
parametric concessive conditional clauses (e.g. 304-06), so perhaps the cases of
additive focus particles in (343) can be accounted for in the same way as the
indefiniteness markers in (310-14). That is, the focus particles are used to mark
concessive conditional clauses, and then become associated with the
interrogative word. Indefinite pronouns containing the focus particle result
from the reduction of concessive conditionals. König (1991: 67-8) seems to opt
for such an explanation.
Similarly, based on the observation that many of the xot’-type focus particles
in (347) derive from ‘want’ or ‘it may be’, one could suggest that these cases
should also be related to the well-understood cases of § 6.2.2.-3. The etymology
of some of the xot’-type focus particles of (347) is given in (348-50).

(348) ‘want’
Russian xot’(< xotja) < converb of xotet’ ‘want’
Hungarian akár < some form of akar ‘want’
Lithuanian nors < noris, converb of noreti ‘want’

(349) ‘it may be’

Lezgian x̂ ajit’ani < x̂ a-ji-t’a-ni (lit.) ‘even if it is’
Kannada -aadaruu < aa-d-ar-uu (lit.) ‘even if it becomes/is’

(350) ‘even if’

M. Greek kan < Classical Greek kán < kaì eán ‘even if’

Thus, the similarity between indefiniteness markers and xot’-type focus par-
ticles noted in (347) could be explained through their common origin.
Schematically, we have the following possible pathways that would explain the
formal parallels between indefiniteness markers and focus particles:

(351) ‘even’-type focus particles: two possible pathways

a. focus particle → indefiniteness marker
b. focus particle → marker of non-specific relative pronoun/
concessive conditional clause → indefiniteness marker

(352) xot’-type focus particle:i two possible pathways

a. focus particle → indefiniteness marker
focus particle
b. ‘you want/it may be’
indefiniteness marker
 197

Option (b) of (351) and (352) is more attractive than option (a) for two reasons:
First, option (b) would allow us to reduce the occurrence of focus particles as
indefiniteness markers to the pathways established in § 6.2, so that no
additional source for indefinite pronouns has to be posited. Secondly, option (a)
is not as well understood semantically. In Haspelmath (1991, 1995) I assumed
that the pathway in (b) was universally correct, but I now realize that option (a)
must be possible, too. Several reasons speak in favor of this.
First, it will be shown in § 8.3.1 that negative scalar focus particles play an
important role in deriving negative indefinites, and in that case the direct
combination of a focus particle with its base (corresponding to the (a) option
above) is the only plausible possibility. So there is independent evidence that
indefinite pronouns can be formed by simply combining a focus particle and its
base, without an intermediate source construction as in § 6.2.
Second, the source constructions of § 6.2 can only explain indefinite pro-
nouns based on interrogative pronouns, not indefinite pronouns based on
generic nouns or (non-derived) indefinite pronouns that we saw in (344-45).
(But of course, it could also be that interrogative-based indefinites are derived
according to (b), whereas non-interrogative-based indefinites are derived
according to (a).)
Third, and perhaps most importantly, the clausal source constructions of §
6.2 cannot account for the forms of the resulting indefinite pronouns.
Consider first the case of an additive focus particle, which according to op-
tion (b) is first used in a parametric concessive conditional clause, which is later
reduced, as shown schematically in (353) (in pseudo-English).

(353) You can go [somewhere], where even it may be. >

You can go where-even.

This scenario can explain indefinites like Dutch wie dan ook, German wer auch
immer, because parametric concessive conditional clauses do show the additive
focus particle ook/auch in the same place, i.e. after the WH-word. But a problem
arises with indefinites like Tagalog kahit na sino and Serbo-Croatian i-ko, where
the focus particle precedes the WH-pronoun. In these languages, parametric
concessive conditional clauses are not marked by additive focus particles
preceding the WH-pronoun, and there is no evidence that they ever were.
Fourth, in many verb-final languages, parametric concessive conditional
clauses are formed by suffixing an additive focus particle to the whole clause,
i.e. to the verb, as exemplified in (354) (see Haspelmath and König (to appear)).

(354) Huallaga Quechua (Weber 1989: 370)

May-ta aywa-r-pis yus-ninchi ayura-yka:-ma-nchi.
where-ACC go-CONV-even [Link] [Link]
‘Wherever we go, God is helping us.’

It seems somewhat less likely that a clause like mayta aywar-pis in (354) should
be reduced to mayta-pis, because the WH-word mayta and the focus particle -pis
are not adjacent in the presumed source structure. And consider the case of
Kannada, where concessive conditionals have the same form as in Quechua:

(355) Kannada (Sridhar 1990: 312)

Avalu estu heeli-d-ar-uu keel-al-ee illa.
she [Link] tell-PAST-COND-even listen-INF-EMPH NEG
‘No matter how much she was told, she didn’t listen.’
 198

By turning such a clause with ‘be/become’ as its predicate into an indefinite

pronoun, we get yaar-aadaruu ‘anyone’, completely as expected. But notice that
Kannada also has yaar-uu, with the additive focus particle suffixed directly to
the WH-pronoun. It seems unlikely that the same source construction should
have led to two quite different resulting indefinites.
Fifth, consider indefiniteness markers that are identical to xot’-type focus
particles. One problem for option (6352b) is again the position of the indefi-
niteness marker. If indefiniteness markers of the types ‘want’/‘it may be’ arise
through the scenario in (356), as proposed in § 6.2.2-3, then one would expect
them to become frozen in the position after the interrogative pronoun.

(356) a. You can go [somewhere], where(ever) it may be. >

You can go where-it-may-be.
b. You can go where(ever) you want.
You can go where-you-want.

This is indeed the case in five languages in (347) (Nenets, Lezgian, Kannada,
West Greenlandic, Yakut), but not in Russian, Finnish, Latvian, and Hungarian,
which argues against option (352b) at least for these languages.
Sixth, according to option (352b), the same source construction would have
led independently to the indefiniteness marker and the focus particle in all the
cases in (347). It is certainly possible to imagine this, but it would constitute a
remarkable coincidence. For even if a language chooses essentially the same
source construction for both structures, it is by no means guaranteed that they
will end up having the same form. For example, in French and in Spanish, the
resulting forms are somewhat different despite the similar source constructions:

(357) French (< ‘it may be’)

a. (indefiniteness marker) qui que ce soit
b. (focus particle)
L’ homme doit oublier, ne serait-ce que pour un bref instant,

the man must forget [Link] for a brief moment

son souci du pain quotidien.

his concern of:the bread daily
‘A human being has to forget, at least for a short while, his or her
concern for the daily bread.’

(358) Spanish (< ‘want’)

a. (indefiniteness marker) cual-quiera
b. (focus particle)
Diga siquiera una palabra.
say [Link] a word
‘Say at least a word.’

By contrast, in all nine cases of (347) the form of the indefiniteness marker is
exactly identical to that of the focus particle.
Seventh, some of the forms of (347) clearly show that they originated as focus
particles. In (357-58) (b), the focus particle is still overtly marked as conditional:
In Spanish by means of the prefix si- (cf. si ‘if’), in French by means of the
conditional verb form serait and subject-verb inversion in serait-ce (cf. English
 199

were it). The scenarios in (369) seem to represent plausible paths by which such
focus particles arise.

(359) a. Say [something], (even) if it is (only) one word. >

Say if-it-is(-only) one word. (= ‘Say at least (xot’) one word.’)
b. Say something, (even) if you want (only) one word. >
Say if-you-want(-only) one word. (= ‘Say at least (xot’) one word.’)

Three of the forms in (347) show traces of the source constructions in (359): The
clearest case is Greek kan, which comes from kaì eán ‘even if’ and is derived via
(359a) by omission of ‘it is’. Russian xot’ and Hungarian akár are also naturally
explained in terms of (359b), because both forms are originally imperative
forms of ‘want’.ii Now an imperative form of the prototypically non-agentive
verb ‘want’ seems to be a strange thing, until we recall that imperative
sentences are commonly interpreted as conditionals if the right pragmatic
conditions are present (cf. also König (1985) on the concessive conditional
interpretation of imperative sentences). Thus, for several of the forms of (347)
we have conclusive evidence that they were originally created by a scenario like
(359) as focus particles, so option (352a) must be correct for these forms.
All these arguments only show that at least SOME of the indefiniteness
markers of (343) and (347) must have been created by direct combination of the
base and the focus particle. For several indefiniteness markers, both scenarios
seem equally plausible, e.g. for Lezgian x̂ ajit’ani and Kannada -aadaruu.
These both have a conditional component (Lezgian -t’a ‘if’, Kannada -ar- ‘if’),
but this is not decisive evidence for the focus-particle-first account because in
these languages, parametric concessive conditional clauses also make use of the
conditional verb form, unlike in typical European languages.

7.1.3. Towards an explanation

As the discussion of the preceding section has shown, some indefinite pro-
nouns are created by combining an interrogative pronoun with a scalar focus
particle of the type ‘also, even’ or of the type 'at least’. It is somewhat puzzling
that interrogative pronouns should be used in such combinations, and I cannot
offer a completely satisfactory explanation for this fact here. The use of
interrogative pronouns has been satisfactorily explained for the indefiniteness
markers arising from one of the source constructions in § 6.3, hence the
temptation to reduce the cases of § 7.1.1 to these source constructions (to which
I succumbed in Haspelmath 1991, 1995). But as we have just seen, this does not
account for several formal properties of the resulting indefinite pronouns.
In approaching the question of why interrogative pronouns should be used
with focus particles, one has to take into account the widespread use of bare
interrogative pronouns as indefinite pronouns in the world’s languages, as
documented and discussed in § 7.3. No conclusive explanation for this will be
offered there, but at least one puzzle can be subsumed under another,
independently existing puzzle. We can thus assume that in the cases of (343)
and (347), we are not dealing with combinations of an interrogative pronoun
plus a focus particle, but with combinations of an indefinite pronoun (which
happens to be identical to an interrogative pronoun) plus focus particle. In this
way, the indefinites of (343) and (347) are assimilated to the indefinites of (344-
45), which clearly consist of indefinites plus focus particle.
 200

Why scalar focus particles should combine with indefinite pronouns to yield
indefinite pronouns with a special meaning is much easier to understand. We
saw in § 5.5.5 that free-choice indefinites must be understood as denoting the
low point on a pragmatic scale. But this is precisley the function of scalar focus
particles: expressing an extreme point on some scale. If we look at the meanings
of the indefinites in (343) and (347), we see that most of them express free-
choice functions, negative functions, or negative-polarity functions, i.e.
precisely those functions that are associated with a scalar endpoint. This is not
true for all such indefinites, however. For example, Evenki Ni-de means
‘someone, kto-to’ (‘anyone/kto-nibud’ is Ni-wul), and Latvian kaut kas may mean
‘something’ or ‘anything’ (cf. Appendix A, section 18 for details). In such cases,
it must be assumed that a semantic change analogous to that described in § 6.4
has intervened.
Let us now consider how focus particles that are used as indefiniteness
markers contribute to the meaning of indefinite pronouns. It does not seem to
be an accident that many of the indefinites in (343-45) are primarily used in
negative functions. In a negative context, pragmatic scales are reversed, and
expressions denoting the low point on a scale acquire the meaning of non-
existence by implicature:

(360) Jebba did not see even something.

= ‘Jebba did not see anything.’

In this respect, indefinite pronouns (especially, apparently, bare interrogatives

used as indefinites) may function like minimal-unit expressions (‘a jot’, ‘a soul’,
etc.). While such minimal-unit words are generally more colorful expressions
for insignificant entities, indefinite pronouns are also suitable for this function
because their semantic content is also minimal. For almost any X, the statement
Jebba saw X tells us more about Jebba’s visual impressions than the statement
Jebba saw something. Thus, the negation of this statement, Jebba did not see
something, can carry the implicature that Jebba saw nothing, especially if the
scalar-endpoint value is reinforced by the scalar focus particle ‘even’. The same
can of course be said for other negative-polarity (i.e. scale-reversing)
environments like conditionals and questions. Xot’-type focus particles are
semantically quite similar to ‘even’-type focus particles, and in some languages
they are not even distinguished (cf. (64c), where Russian xot’ is translated by
English ‘even’).
It is of course not an accident that similar formal elements appear in the two
different pathways that I have contrasted in this section. The fundamental
semantic component contributed by the indefiniteness markers considered here
is the endpoint of a pragmatic scale, and although this may be expressed by
syntactically different means (concessive conditional clauses, free relative
clauses, focus particles), the basic building blocks for expressing this idea are of
the same type. Thus, ultimately the two distinct pathways of (351-52) (a-b) are
closely related.

7.2. Indefinite Pronouns from ‘Or’?

It has long been known to logicians that there is a close relationship between
universal quantification and logical conjunction on the one hand, and between
existential quantification and logical disjunction on the other (cf. Reichenbach
 201

(1947: 92), Rohrer (1973), among many others). For a finite number of entities x,
a universal statement is equivalent to a conjunction, and an existential
statement is equivalent to a disjunction:

(361) a. (∀x) f(x) ≡ f(x1) & f(x2) ... & f(xn)

b. (∃x) f(x) ≡ f(x1) ∨ f(x2) ... ∨ f(xn)

Now it looks as if in natural languages there may also be a connection be-

tween these logical types. A particularly intriguing example comes from
Japanese, where both universal pronouns and indefinite pronouns are derived
from interrogative pronouns with markers that are identical to coordinating
conjunctions:

(362) Japanese
a. dare-mo ‘everyone’ mo ‘and, also’
nani-mo ‘everything’
b. dare-ka ‘someone’ ka ‘or’
nani-ka ‘something’

The distribution of mo ‘and’ and ka ‘or’ is exactly as suggested by (361). At first

blush, it seems unlikely that this should be an accident. But can we say that
Japanese mo expresses an extremely general meaning ‘universality/ con-
junction’, and ka expresses an extremely general meaning ‘existentiality/
disjunction’?iii In this section I argue against such a hypothesis, and I will in-
deed claim that the pattern in (362) is a coincidence (at least to some extent).
In § 6.2.3 and § 7.1 we saw possible ways of explaining the use of ‘and, also’
in indefinite pronouns (especially of the free-choice type), and in § 6.5 we saw
that free-choice indefinites may become universal pronouns. The explanations
offered there took as their point of departure the use of ‘and, also’ as scalar
focus particles (‘even’), not the use of ‘and, also’ as a coordinating conjunction.
This analysis is thus incompatible with the hypothesis that there is a close
connection between the pattern in (362) and (361). When we go beyond the
Japanese data, the empirical evidence confirms that there is no direct formal
connection between 'or' and existential indefinites, and between 'and' and
universal pronouns.
First of all, although many languages have indefiniteness markers that are
formally identical to disjunctive conjunctions, the situation in Japanese is quite
exceptional. Consider the examples of indefiniteness markers that are identical
to ‘or’ in (363).

(363) Indefinites whose marker is identical to ‘or’ (all interrogative-based)

Japanese dare-ka ‘someone’ ka ‘or’
Kannada yaar-oo ‘someone’ -oo ‘or’
Korean nwukwu-na ‘anyone’ -(i)na ‘or’
nwukwu-tunci ‘anyone’ -(i)tunci ‘or’
Russian kto-libo ‘anyone’ libo ‘or’
Hungarian akár-ki ‘anyone’ akár...akár ‘whether..or’
Portuguese qual-quer ‘any’ quer...quer ‘either...or’
Basque edo-nor ‘anyone’ edo ‘or’
Latvian jeb-kas ‘anything’ jeb ‘or’
Rumanian ori-cine ‘anybody’ ori ‘or’
Ossetic č i-fændy ‘anyone’ fændy ‘or’
 202

Nanay uj-nuu ‘someone’ -nuu ‘or’

Hausa kóo-mée ‘anything’ kóo ‘or’
W. Greenl. sumi-luunniit ‘anywhere, somewhere or other’
-luunniit ‘or’

Most of the ‘or’-indefinites in (363) are primarily free-choice indefinites, not

non-emphatic indefinites like Japanese WH-ka. The only exceptions are
Kannada and Nanay, which are specific (and partially Russian and West
Greenlandic, insofar as these forms can also be used in irrealis-non-specific
functions). Now this is not necessarily a conclusive argument against a
connection of ‘or’-indefinites with (361b), as long as one assumes that free-
choice indefinites express existential quantification rather than universal
quantification (cf. § 5.2.1).iv
But not all languages with ‘or’-indefinites also have ‘also’-indefinites, and in
those that do, the meaning relation between the two only occasionally
corresponds to the equivalences in (361). In Figures 24.1-6, I repeat the
distribution of ‘or’- and ‘also’-indefinites in the six languages of the 40-
language sample that have them.

Figure 24.1. Japanese

direct
question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free choice

Figure 24.2. Korean

direct
question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free choice

Figure 24.3. Kannada

direct
question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free choice

Figure 24.4. Nanay

direct
question indirect negation
specific irrealis negation
non-specific
conditional comparative
free choice
 203

Figure 24.5. Ossetic

direct
question indirect negation
specific irrealis negation
non-specific
conditional comparative
free choice
Figure 24.6. Yakut
direct
question indirect negation
specific irrealis negation
non-specific
conditional comparative
free choice

It is quite striking how different the ‘or’-indefinites and the ‘also’-indefinites are
in different languages. While Kannada, Nanay and Yakut are like Japanese at
least in that the ‘or’-indefinite is further to the left on the map and the ‘also’-
indefinite is further to the right, not even that is true for Korean and Ossetic.
Thus, the Japanese situation does not even seem to represent a tendency.
But if (361) is not the explanation for ‘or’-indefinites, why is ‘or’ used in
indefinite pronouns at all? Again, as in the discussion of xot’-type focus par-
ticles in § 7.1, there is an obvious alternative to the most straightforward option
(364a). It could be that both ‘or’ and the indefiniteness marker arise from the
same source independently, as shown in (364b).

(364) a. ‘or’ → indefiniteness marker

b. ‘or’
X
indefiniteness marker

In this case, I will argue that (364b) is correct. The 'X' of (364b) is to be filled by
one of the two expressions ‘want’ or ‘it may be’ that we have already seen twice
(§ 6.2, § 7.1).
Explaining how ‘or’ arises from ‘want’ or ‘it may be’ is a fairly straightfor-
ward matter. An abstract scenario like (365) seems immediately plausible.v

(365) a. Jebba (may come), if you want Jawro may come. >
Jebba or Jawro may come.
b. (Someone) may come, it may be Jebba, it may be Jawro. >
Jebba or Jawro may come.

It is not difficult to find cases of ‘or’ that evidently go back to similar source
constructions:

(366) a. ‘or’ from ‘if (you) want’

Ossetic fændy ‘or’ < fændy ‘you want’
Russian libo ‘or’ < ljubo ‘pleasant ("one wants")’vi
Rumanian ori ‘or’ < late Latin *volet (< vult) ‘wants’
Hungarian akár ‘or’ < imperative of akar ‘want’
Latin vel ‘or’ < imperative of velle ‘want’
Portuguese quer ‘or’ < quer ‘wants’
 204

b. ‘or’ from ‘it may be’

Korean -(i)na ‘or’ < -na-mood of i- ‘be’
French soit ‘or’ < soit 3sg subjunctive of être ‘be’
Lango òˆò ‘or’ < 3sg perfective of ˆòo ‘be possible’

Now I have to show that for those cases of ‘or’/indefiniteness marker whose
etymology is known the change from ‘want’/‘it may be’ to an indefiniteness
marker is plausible. The most straightforward case are indefiniteness markers
that follow the interrogative root, because these have already been explained in
§ 6.2.2-3.

(367) a. indefiniteness marker (identical to ‘or’) from ‘want’

Russian kto-libo ‘anyone, someone’
Ossetic č i-fændy ‘anyone’
Portuguese qual-quer ‘any’

b. indefiniteness marker (identical to ‘or’) from ‘it may be’

Korean nwukwu-na ‘anyone’

We can safely assume a similar explanation for those ‘or’-indefiniteness

markers which are suffixed and whose origin is unknown, e.g. Japanese -ka,
Kannada -oo, Nanay -nuu, West Greenlandic -luunniit. But what about the
prefixed markers, whose function as indefiniteness markers cannot be ex-
plained by the source construction in § 6.3.2-3 (e.g. Hungarian akár-, Basque edo-
, Latvian jeb, Rumanian ori-, Hausa kóo)?
For these cases, my hypothesis is that they developed their function as
indefiniteness markers via a function as focus particles of the type ‘at least’, so
that the complete set of possible changes looks as in (368).

(368) ‘or’
‘want’/‘it may be’ indefiniteness marker
xot’-type focus particle → indefiniteness
marker

This hypothesis not only accounts well for the data without any additional
assumptions, it is also independently motivated by the fact that some of the
particles of (363) are attested as having the additional meaning ‘at least’, in
particular Hungarian akár, West Greenlandic -lunniit (cf. § 7.1), and probably
also Hausa kóo (although Meyers (1974) describes only the meaning of positive
‘even’, which is more salient for speakers of English, since English lacks a
precise equivalent of Russian xot’). For the remaining cases, in particular
Latvian jeb and Basque edo, it can be assumed that these had the meaning ‘at
least’ at an earlier stage.
This completes my proposed account of the observed homonymy of ‘or’ and
indefiniteness markers. It is essentially a diachronic-typological account: This
homonymy is so common because the sources and pathways of change are so
common, not because there is a synchronic common meaning that must be
captured by a synchronic semantic analysis. Seen in this way, the massive
parallel homonymy of (363) is not an accident (which would be excluded by the
laws of statistics, as observed by Haiman (1974)), because there is a diachronic
semantic relation, but in each individual language it may well be that speakers
 205

no longer see the similarities. Therefore indefiniteness markers may change

their meaning (along the lines of § 6.4) independently of other meanings that
are historically closely related, which results in the quite diverse patterns that
were illustrated in Figure 24.1-6.

7.3. Bare Interrogatives as Indefinites

7.3.1. Bare interrogatives and strategies for disambiguation

In many languages, some indefinite pronouns are formally identical to in-

terrogative pronouns. In other words, these languages use bare interrogatives
(as opposed to interrogatives plus an indefiniteness marker, the most
widespread type of indefinite in my 40-language-sample) as indefinite pro-
nouns. Some examples of such languages are given in (369).

(369) a. Classical Greek

tís ‘who?’ tis ‘someone’
pou$ ‘where?’ pou ‘somewhere’
b. Chinese
sheí ‘who?’ sheí ‘someone’
shénme ‘what?’ shénme ‘something’
c. Hopi (Uto-Aztecan; Malotki 1979: 110)
hak ‘who?’ hak ‘someone’
haqam ‘where?’ haqam ‘somewhere’
d. Newari (Sino-Tibetan; Korolev 1989: 62)
su ‘who?’ su ‘nobody (with verbal neg.)’
chu ‘what?’ chu ‘nothing (with verbal neg.)’
e. Dyirbal (Pama-Nyungan; Dixon 1972: 265)
wanya ‘who?’ wanya ‘someone’
minya ‘what?’ minya ‘something’
f. Khmer (Austro-Asiatic; Huffman 1967: 153-6)
qw´y‘what?’ qw´y ‘something’
naa ‘where?’ naa ‘somewhere’

Disambiguation is usually achieved by syntactic differences and/or

suprasegmental means. For instance, in Classical Greek and in Modern
German, the interrogative pronoun is usually clause-initial, whereas the
indefinite pronoun must cliticize to the preceding word and hence cannot be
clause-initial.

(370) Classical Greek

a. Tís e$lthen?
who came
‘Who came?’
b. Elthén tis. (*Tis elthen.)
came who
‘Someone came.’
(371) Modern German
a. Wer kommt da?
‘Who is coming?’
b. Da kommt wer. (=Jemand <*wer> kommt da.)
 206

‘Someone is coming.’

In Lakhota, questions are marked by a sentence-final particle he, which

marks the difference in (372).

(372) Lakhota (Van Valin 1990: 210)

a. Wič háš a ki táku nax?ú.
man the what heard
‘The man heard something.’
b. Wič háš a ki táku nax?ú he?
man the what heard Q
‘What did the man hear?’

In Chinese, only intonation disambiguates the two sentences in (373) (Frei

1940: 14). Rising intonation indicates a question, and normal falling intonation
indicates a statement.

(373) Chinese (Tsai 1990: 41)

a. Tā bǎ shénme shū diū le?
she ACC what book throw PFV
‘What books did she throw away?’
b. Tā bǎ shénme shū diū le.
she ACC what book throw PFV
‘She threw away a certain book.’

In Classical Greek, the interrogative tís is always stressed, and the indefinite tis
is always unstressed, as shown in the spelling. According to Kang (1988: 188),
stress is also the main factor in disambiguating sentences with interrogative-
indefinites in Korean.

7.3.2. Bare interrogatives in Indo-European

Bare interrogatives as indefinites are an option in several western Indo-

European languages which normally use interrogative-derived indefinites:
Vedic Sanskrit, Latin, German, Gothic (Behaghel 1917), Lithuanian, and most of
the Slavic languages (cf. Kř íž ková (1974) for a special study of bare in-
terrogatives in Slavic languages). Since Classical Greek also uses bare inter-
rogatives, this could be a very old feature, although it is doubtful that it could
be reconstructed for Proto-Indo-European, because the oldest Indo-Iranian texts
show it only sporadically (Delbrück 1893: 511). Independently of the genetic
relationships, the conditions for the use of bare interrogatives in these
languages are remarkably similar. It does not seem possible to attribute these
similarities only to common inheritance, especially since indefinite pronouns
are otherwise known to change at a fast rate.
In most of the western Indo-European languages, bare interrogatives may be
used as indefinites in conditional clauses, e.g.

(374) a. Latin
Si quis mortuus fuerit non habens filium,... (NT, Mt 22.24)
if who dead becomes not having son
‘If someone dies, having no children...’
b. Old Church Slavonic (Kř íž ková 1974: 91)
Aš te kŭ to xoš tetŭ po mĭ ne iti... (NT, Mt 16.24)
 207

if who wants after me go:INF

‘If anyone wants to come after me...’
c. Ukrainian (Kř íž ková 1974: 91)
Koly maeš suproty mene jakyj sumniv — skaž y...
if you:have against me which doubt say:IMPV
‘If you have any doubt with respect to me, tell me...’

They may occur in questions:

(375) a. Old High German (Behaghel 1917)

Habet ir hier waz, thaz man ezzan mugi?
have you here what that one eat might
‘Do you have anything to eat here?’
b. Slovene (Kř íž ková 1974: 90)
Se je zatreskala v kakš nega fanta?
REFL is fallen in which [Link]
‘She fell in love with some young man.’

Bare interrogatives as indefinites also occur in other non-specific functions,

such as non-specific free relative clauses (cf. 376), other irrealis clauses (cf. 377),
and indirect negation (cf. 378).

(376) non-specific free relative clauses

Old High German (Behaghel 1917; O. III, 16, 19)
Ther fon imo saget waz, ther suachit io thaz sinaz.
who from him says what he seeks always the own
‘Whoever says something of him, he always seeks his own (thing).’
(377) irrealis subordinate clauses
a. Old High German (Behaghel 1917; Tat. 145,3)
Gisehet ir, thaz wer iuuuih ni forleite.
see you that who you not [Link]
‘Take care that someone does not lead you astray.’
b. Ukrainian (Kř íž ková 1974: 91)
Ozyralasja, šč oby xto ïï ne pidsluxyvav.
[Link] that who her not eavesdrop
‘She looked around so that someone would not eavesdrop on her.’
(378) indirect negation
a. Latin
neque Patrem quis novit, nisi filius (NT, Mt 11.27)
and:not Father:ACC who knows if:not son
‘neither knoweth any man the Father, save the Son’
b. Gothic
Ni manna in analaugnein hwa taujip (NT, John 7.4)
not man in secret what does
‘Nobody does anything in secret.’

Further non-specific functions where bare interrogatives occur are the

imperative (cf. 379) and future and uncertain statements (cf. 380).

(379) imperative
a. Slovene (Kř íž ková 1974: 91)
Piš i kaj iz Pariza.
write:IMPV what from Paris
‘Write something from Paris.’
 208

b. Belorussian (Kř íž ková 1974: 91)

Njaxaj jana paprosic’ kago pamagč y.
let her ask whom to:help
‘Let her ask somebody to help her.’
(380) future/uncertain statements
a. Gothic
skal pus hwa qipan (NT, Lk 7.40)
I:must to:you what say
‘I must tell you something.’
b. Polish
Moz%e on jeszcze kiedy przyjedzie.
maybe he still when will:arrive
‘Perhaps he will still arrive sometime.’

Thus, western Indo-European bare interrogatives may occur practically in all

non-emphatic non-specific functions. They are generally excluded from past or
current present affirmative declarative clauses, where indefinites must be
specific.
This situation is rather puzzling. Why should bare interrogatives used as
indefinites be restricted in this way? I have no answer to this question, but the
facts from the western Indo-European families which I have cited are rather
robust, and the generalizations cannot be due to coincidence.
Moreover, a restriction of bare interrogatives to non-specific functions can be
found outside of Indo-European as well. Li (1992) reports that Chinese bare
interrogatives (especially shénme ‘what, which’) may be used as indefinites in
questions, conditionals, negative sentences and other contexts where it can be
interpreted non-specifically (in embedded clauses of non-factive verbs), but not
in ordinary declarative sentences:

(640) Chinese (Li 1992: 127-9)

a. (question) Tā xǐ huan shénme ma?
he like what Q
‘Does he like anything?’
b. (negation) Tā bù wèi shénme rén zuò shi.
he not for what person work thing
‘He does not work for anyone.’
c. (non-factive) Wǒ xī wàng ni xǐ huan shénme (dō ngxi).
I hope you like what thing
‘I hope that you like something.’
d. (factive) *Wǒ hòuhuǐ zuo shénme (shìqing).
I regret do what thing
‘I regret having done something.’

The similarity between the Chinese situation and the one in Indo-European is
too striking to be due to accident. The only explanation for these restrictions
that I can think of is that bare interrogatives are perhaps less easily mistaken for
true question pronouns in such environments, but it is hardly sufficient.

7.3.3. Why are indefinites often identical to interrogatives?

We have seen that in many cases, interrogatives are derived from indefinites
because they arose by grammaticalization from an indirect question or a non-
specific free relative clause (cf. chapter 6). So one possible way of explaining the
use of bare interrogatives as indefinites is the hypothesis that bare
 209

interrogatives also come from these sources but their indefiniteness marker has
gotten reduced to the point of disappearing. However, this hypothesis must be
rejected.
One problem with it is that there is absolutely no positive evidence for it. I
do not know of a single case of a language whose earlier stages are attested
where we see an indefiniteness marker being reduced to zero, resulting in a
bare interrogative pronoun.
Secondly, one does not get the impression that bare interrogatives arise as an
accident of language change, and that they are subject to quick change.
Speakers seem quite happy with them and not too concerned about removing
them (although they did disappear in modern Romance, modern Greek, and
modern Scandinavian).
Thirdly, and most importantly, bare interrogatives are extremely widespread
in the world’s languages. In my 100-language sample, there are 64 languages
whose indefinites are based on interrogatives. Of these 64 languages, 31
languages or almost one third of the whole sample have bare interrogatives. It
is extremely unlikely that zero-grammaticalization should happen so often, and
so systematically. The typological facts show conclusively that we are dealing
with systematic polysemy rather than with accidental homonymy.
Since it cannot be shown that bare interrogatives as indefinites are di-
achronically derived from interrogatives, we are well advised to look for
synchronic functional similarities that justify the systematic polysemy. Such
functional similarities are not hard to come by: It is clear that the sentences
(382a) and (382b) have a lot in common.

(382) a. Someone stole my bike.

b. Who stole my bike?

As pointed out by Karcevski (1969) and many others (e.g. Wierzbicka 1980b),
the common element in the meaning of indefinite and interrogative pronouns is
ignorance (hence Karcevski’s term 'ignorative pronouns'). Similarly, Bhat (1993)
says that both are characterized by an 'information gap'. In indefinites such as
(382a), the speaker either does not know who stole the bike or does not want
the hearer to know who stole it. In interrogatives such as (382b), the speaker
does not know who stole the bike either, but here she or he wants the hearer to
supply the missing information.
Given this common meaning element, one might suspect that an inter-
rogative-indefinite pronoun could actually be vague rather than polysemous.
Indeed, Dixon (1980: 372) reports that speakers of Australian languages will
often translate a sentence with an interrogative-indefinite pronoun as in (383),
suggesting that it is simultaneously an indefinite specification, and a request for
further [Link]

(383) Someone stole my bike – who did it?

But vagueness at the utterance level is not the general case – in most languages,
it seems, sentences containing interrogative-indefinite pronouns are clearly
either interrogative or indefinite (cf. the disambiguation strategies mentioned in
§ 7.3.1).
Another way of formulating the similarity between (382a) and (382b) is to
say that (382a) is presupposed by (382b) (Lyons 1977: 758), so in a way the
indefinite meaning is contained in the interrogative meaning. Attempting to ac-
 210

count for the formal relationship between interrogatives and indefinites found
in many languages, Lyons suggests that the interrogative variant of (382a), the
polar question (384), might also be construed as a parametric question,
presupposing (382a) and expecting the hearer to respond by supplying a value
for ‘someone’ (cf. already Paul (1920: 136) for similar considerations).

(384) Someone stole my bike?

Somewhat similarly, Klima (1964: 252-3) derived interrogative pronouns like

what by 'incorporation' of indefinites (in this case, something) into the abstract
question operator wh (wh...something is transformed to wh+something, which is
spelled out as what).7a
In this line of reasoning, interrogatives are derived from indefinites. It is
therefore natural to ask whether the formal similarities between interrogatives
and indefinites can perhaps be explained (even if only partially) by a diachronic
change from an indefinite to an interrogative pronoun. Such a change was
indeed posited speculatively by several linguists several decades ago (e.g.
Meunier 1875, Gardiner 1932: 307-8; cf. the discussion in Frei 1940: 7-9). This
hypothesis is interesting, but it must be rejected. There is no positive evidence
for it, although it should be quite easy to find such evidence: There should be
plenty of languages whose interrogative-indefinite pronouns go back
etymologically to generic ontological-category nouns or to the numeral ‘one’.
These are the sources of indefinites in many languages, and we would expect
cases where such indefinites become interrogatives. Furthermore, we would
expect at least some of the numerous indefinites consisting of root plus
indefiniteness marker to become interrogative pronouns, but we never observe
any traces of such earlier indefiniteness markers in interrogatives. What we
find, instead, is that in language family after language family, interrogative
pronouns prove resistant to etymological analysis. In fact, interrogative
pronouns are among the slowest-changing elements in any language. Consider,
for instance, interrogatives in Indo-European languages: The reconstructed
original root is *kwi-/*kwo-, and reflexes of this root are still found in all modern
languages, e.g. English what, Russian č to, Modern Greek ti, Yazgulami č iǵ ,
Irish cad, Punjabi ki, Italian che, etc. These words have undergone phonological
and morphological changes, but the root was never replaced. Interrogative
pronouns are generally so old that their root cannot be [Link]
So neither can interrogative-indefinite pronouns go back to interrogative
pronouns in a well-understood way, nor can they go back to indefinite pro-
nouns in the way just sketched. We can only conclude that somehow the
functional similarity between them which was noted above must be responsible
for the systematic polysemy.
But one thing we know for sure: The interrogative function is always
primary, and the indefinite function is secondary. An interrogative pronoun
may lose its indefinite function (e.g. Latin quis/quem ‘who(m)’, which survived
into modern Romance, but can no longer be used as an indefinite; similarly,
Classical Greek tís ‘who?’ can no longer be used as an indefinite in Modern
Greek). Unfortunately, I know of no good case where an interrogative pronoun
that lacks the indefinite sense acquires it at some later stage. Such cases could
give us hints for explaining the mechanism the rise of indefinites from
interrogatives.

7.3.4. The multiple partitive use of bare interrogatives

 211

There is one special type of indefinite use of bare interrogatives which is more
widespread than the use of bare interrogatives as ordinary indefinites, as
discussed in the preceding sections (§ 7.3.1-3). For lack of a better term, I call
this the MULTIPLE PARTITIVE construction. It occurs in three variants.
A typical example of the first variant is (385).

(385) Georgian
Vin pul-s eloda, vin c’eril-s, vin gazet-eb-s.
who money-DAT waits who letter-DAT who newspaper-PL-DAT
‘Some (people) are waiting for money, some for a letter, some for
newspapers.’ (Lit. ‘Who is waiting for money, who for a letter,...’)

The bare interrogative in this construction always occurs together with at least
another instance of the same interrogative pronoun in a parallel coordinate
clause (hence 'multiple'), and each indefinite denotes a subset (hence 'partitive')
of a larger contextually given set, so the translation ‘some (people)’ is more
appropriate than ‘someone’. The verb is often gapped in all clauses except one.
The multiple partitive construction can be found in a wide variety of languages,
a selection of which is illustrated in (386).

(386) a. French (Grevisse 1986: §729)

Qui apportait un fromage, qui un sac de noix, qui un quartier de chèvre.
‘Some brought a piece of cheese, some a bag of nuts, some a piece of
goat meat.’
b. Russian
Gde ubirajut sveklu, gde paš ut loš ad’mi.
where reap:3PL beet where plow3PL horses:INSTR
‘In some places people are reaping beet, in some places people are
plowing with horses.’
c. Finnish (Karttunen and Peters 1980: 194)
Tuuli puhaltaa milloin länne-stä, milloin idä-stä.
wind blows when west-ELAT when east-ELAT
‘The wind blows sometimes from the west, sometimes from the
east.’
d. Mansi (Finno-Ugrian; Beke 1913-14: 150)
Xot xum xańś i, xot xum at xańś i.
who person knows who person not knows
‘Some people know, some people don’t know.’
e. Turkish (Xovratovič 1989: 35)
Kimi adak, kimi matem için topla-n-dı.
who sacrifice who mourning for gather-REFL-PAST(3SG)
‘Some gathered for a sacrifice, some for mourning.’
f. Kilivila (Austronesian: Oceanic; Senft 1986: 63)
Te-vila e-sisu-si va simla te-vila
G(MALE)-[Link] they-stay-PL at island G(MALE)-[Link]
e-sila-si.
[Link]-PL
‘Some of them stay at the island, the others keep on sailing.’

In the second variant of this construction, the multiple partitive indefinite

pronoun is not an independent element of the clause, but a kind of
discontinuous modifier of another (semantically plural) phrase.
 212

(387) a. French (Grevisse 1986: § 729)

Les clients de l’ hôtel prenaient, qui du thé, qui du porto,
the guests of the hotel take who ART tea who ART port
qui un cocktail.
who a cocktail
‘Of the hotel guests, some drank tea, some port, some a cocktail.’
(Lit.: ‘The hotel guests drank, who tea, who port,...’)
b. Hebrew (Glinert 1989: § 7.8)
Et zot ro?im, mi be-mida raba u-mi be-mida peÌuta,
ACC this they:see who in-degree big and-who in-degree less
ke-¿erex bifney ¿acmo.
as-value before itself
‘They regard this as a value in itself, some to a greater and some to
a lesser extent.’
c. Hungarian (Beke 1913-14: 150)
Az ország-á-ban lak-ó ember-ek, ki kocsi-n, ki szekér-en
the country-3SG-IN live-PTCP person-PL who cart-by who coach-by

hoz-t-ák a temérdek tülk-öt.

bring-PAST-3PL the loads horn-ACC
‘The people living in his country brought the innumerable horns,
some by cart, some by coach.’

In the third variant of this construction, two different bare interrogatives are
used in this way, i.e. not as independent elements of the clause. In this
construction, there are not two parallel coordinate structures, but the resulting
meaning is similar.

(388) a. Russian
Oni rassypalis' kto kuda.
they dispersed who whither
‘(Lit.) They dispersed, some in some direction, some in another
direction (i.e. different people in different directions).’
b. Finnish (Karttunen and Peters 1980: 194)
Tuuli puhaltaa milloin mistäkin sunna-sta.
wind blows when which direction-ELAT
‘(Lit.) The wind blows, sometimes from some direction, sometimes
from another direction (i.e. from different directions at different
times).’
c. Romanian
Acestia se împră stiară ca puii de potîrniche,
these REFL dispersed like young of partridges
caresi pe unde.
which to where
‘These dispersed like young partridges, different people in different
directions.’

I note these constructions here without offering so much as the beginning of

an explanation. I find them quite puzzling, but their wide cross-linguistic
distribution (at least in Europe and northern Asia) makes it impossible to
regard them merely as marginal curiosities.
 213

7.4. Indefinite Pronouns Derived by Reduplication

In many languages, indefinite pronouns have a reduplicated form. The most
common case are indefinites consisting of two full instances of the
corresponding interrogative pronoun. I know of no case of partial reduplica-
tion, or of triple reduplication. Reduplicated indefinites of this form occur
commonly all over the world (cf. Pott 1862, Gonda 1949, 1954-55, Coyaud and
Aït Hamou 1971, 1972, 1976, Moravcsik 1978: 319).

(389) a. Moksha Mordva (Finno-Ugrian; Feoktistov 1966: 207)

kije ‘who?’ kije-kije ‘somebody, anybody’
meze ‘what?’ meze-meze ‘something, anything’
kodama ‘which?’ kodama-kodama ‘any’
b. Ila (Bantu)
oni ‘who?’ oni-oni ‘whoever, anyone’
ukwi ‘where?’ ukwi-ukwi ‘anywhere’
c. Vietnamese (Coyaud and Aït Hamou 1972)
ai ‘who?’ ai ai ‘anybody, everybody’
d. Latin
quis ‘who?’ quis-quis ‘anybody, whoever’
ubi ‘where?’ ubi-ubi ‘anywhere, wherever’
e. Bugis (Austronesian; Sirk 1979: 116)
niga ‘who?’ niga-niga ‘anyone’
aga ‘what?’ aga-ga ‘anything’
f. Ainu (Refsing 1986: 104)
nen ‘who?’ nen-nen ‘whoever, nobody’
g. Khasi (Mon-Khmer; Moravcsik 1978: 319)
ka?ey ‘who?’ ka?ey-ka?ey ‘someone’
kumnu ‘how?’ kumnu-kumnu ‘somehow’
h. Kristang (= Malacca Creole Portuguese; Baxter 1988: 54)
keng ‘who?’ (< quem) keng keng ‘anybody’
ki ‘what?’ ki ki ‘anything’

But general nouns are also occasionally reduplicated to form indefinite

pronouns:

(390) Yoruba
e5ni ‘person’ e5ni-k-e5ni ‘anybody’
ohun ‘thing’ ohun-k-ohun ‘anything’

And one also finds reduplicated combinations of an interrogative pronoun

and an indefiniteness marker (na also means ‘or’ and ‘although’, cf. § 7.2):

(650) Malagasy
iza ‘who?’ na iza na iza ‘anyone’
inona ‘what?’ na inona na inona ‘anything’

How can this formal strategy for creating indefinite pronouns be related to
their meaning? Unfortunately, the answer is not clear to me, although
reduplication is sometimes regarded as one of the most iconic ways of signaling
meaning. According to Moravcsik (1978: 317), 'the most outstanding single
concept that reduplicative constructions recurrently express in various lan-
 214

guages is the concept of increased quantity'. She distinguishes two subtypes of

this meaning, 'quantity of referents' and 'amount of emphasis', and subsumes
indefinite pronouns under the first of these two general categories, noting
further that the plurality of referents expressed by reduplicative constructions
is usually distributive rather than [Link] The iconic motivation of
reduplication for the expression of distributive plurality is easy to see. But in
what respect is the meaning of indefinite pronouns related to distributive
plurality?
When we look at the meanings of reduplicated indefinites more closely, we
see that most of them have a free-choice or at least negative-polarity meaning,
rather than a specific meaning. The similarity between the free-choice function
and the distributive universal function has been highlighted earlier (cf. § 3.2.6, §
6.5), and indeed reduplication is widely used to express the meaning ‘every’, as
illustrated in (392-93) (cf. Gonda 1949: 176).

(392) Tagalog taon-taon ‘every year’

Manggarai leso leso ‘every day’
Sanskrit ahany-ahani ‘every day’
(393) Hungarian ki ki ‘everyone’ (ki ‘who’)
Hindi/Urdu ek ek ‘every’ (ek ‘one’)

But there are some additional facts that suggest that reduplication may
somehow express the notion of irrelevance by itself. Thus, in some languages,
reduplicated interrogatives are used primarily or exclusively as non-specific
free relative pronouns, in free relative clauses or in parametric concessive
conditional clauses, e.g.

(394) a. Da’a (Austronesian, Central Sulawesi; Barr 1988)

Sema-sema na-rata ri ja’i-na kana ni-tarima-na.
who-who REAL-come [Link] must [Link]-receive-he
‘Whoever comes to him will be received by him.’
b. Latin
Quid-quid id est, timeo Danaos et dona ferentes.
what-what that is I:fear Danaans even gifts bringing
‘Whatever that is, I fear the Danaans even if they bring gifts.’
c. Bagandi (Australian; Hercus 1982: 171)
Gila yuri-wa-yiga-ayi, mina-mina yawara Nadu
not [Link] what-what wordI:ERG
gulba-ra-na-ama.
[Link]
‘They don’t understand me, whatever words I may be saying to
you.’

Not only interrogative pronouns may be reduplicated in such sentences, but

also relative pronouns, as in Sanskrit (whose interrogative and relative
pronouns are distinct), and even verbs, as in Sicilian:

(395) a. Sanskrit
Yo yas tad karoti sa (sa) mū rkhah5.
who who that does that (that) fool
‘Whoever does this is a fool.’
b. Sicilian (Bollée 1978: 329)
Unni vaju vaju, tutti mi salutunu.
 215

where I:go I:go all me they:greet

‘Wherever I go, everyone greets me.’

As we saw in § 6.2.3, indefinite pronouns may have their source in non-

specific relative pronouns. So a possibility is that the reduplicated indefinites of
(389) in general come from non-specific relative pronouns as in (394-95). At
least for Latin, where the use of reduplicated interrogatives as indefinite
pronouns is quite limited and evidently secondary, this looks like a plausible
scenario.
However, in view of examples like (390), where a non-interrogative-based
indefinite is formed by reduplication, one should perhaps resist the temptation
to derive interrogative-based indefinites in general from clausal sources. (Cf.
also the completely analogous dilemma in § 7.1 and the discussion there.) But
even if the non-specific clauses in (394-95) cannot be regarded as the single key
to reduplicated indefinites, they suggest that the meaning of distributive
plurality is not the only starting point for understanding reduplicated
indefinites.
While distributive plurality can probably also be detected in (394-95), an-
other important meaning element is that of free choice, negative polarity, and
non-specificity. There is independent evidence that these meanings are also
somehow inherent in reduplicated constructions. Steever (1988) describes
verbal 'echo compounds' (i.e. partial reduplications) in several Dravidian
languages (Tamil, Toda, Kannada, Kodagu, Telugu), which are used only in
'affective contexts such as questions, negatives and modals' (1987: 63). An
example from Tamil is given in (396).x

(396) Tamil
Pā ttiratai ut5 ai-tt-ā y kit5 ai-tt-ā y en2r2-ā l tō lai uri-pp-ē n2.
pots break-PAST-2SG RED:break-PAST-2SG say-COND you skin-FUT-1SG
‘If you broke the pots or did any such stupid thing, I’ll skin you alive.’

7.5. Some Further Diachronic Issues

7.5.1. Generic nouns turned pronouns

As I have already observed, many languages use generic nouns like ‘person’,
‘thing’, ‘place’, ‘time’, etc. to express notions like ‘someone’, ‘something’,
‘somewhere’, ‘sometime’, etc. (but not, of course, for the free-choice sense of
‘anyone’, ‘anything’, etc.). In my 100-language sample, 42 languages use this
strategy. As was also mentioned above, it is not always easy to tell whether we
are really dealing with pronouns, or whether true nouns are used as
alternatives to indefinite pronouns (cf. § 3.3.1), because the differences are
sometimes quite subtle (cf. also § 2.2.2).
The diachronic process by which a generic noun is turned into an indefinite
pronoun is quite straightforward, so not much needs to be said here. A generic
noun is first used in a noun phrase without modifiers to render meanings like
‘somebody’, ‘something’, and it gradually acquires phonological,
morphological and syntactic features that set it off from other nouns. The last
doubts about its new pronoun status are dispelled when the original generic
noun falls into disuse, as in the case of English somebody/anybody, where body
can no longer be used in the sense ‘person’; French rien ‘nothing’, which can no
 216

longer be used in the sense of ‘thing’ (cf. Latin rem ‘thing(ACC)’); or Maltese xi
mkien ‘somewhere’, where mkien (cf. Arabic makaan ‘place’) can no longer be
used in the sense of ‘place’.
In some cases, an indefinite determiner like ‘one’, ‘some’ or ‘any’ is com-
bined with a generic noun in a new pronominal form, as in Italian qualcosa
‘something’ (< qualche ‘some’ + cosa ‘thing’), Maltese xi mkien ‘somewhere’ (xi
indefiniteness marker < š ay? ‘thing, something’ + mkien ‘place’). Again, this is
so straightforward that it need not be commented on further.
It should be pointed out, however, that there is an interesting tendency for
generic-noun-based indefinites to be restricted to negative-polarity or negative
functions, e.g. French personne ‘person; nobody’, Hebrew davar ‘thing; anything,
nothing’. This tendency will be discussed further in § 8.3.2.

7.5.2. Indefinite pronouns from ‘one’

Not much needs to be said on the numeral ‘one’ as a source of indefinite

pronouns either. This source is also very widespread, but ‘one’ is usually re-
stricted to the use as an indefinite determiner or an indefinite pronoun of the
ontological category ‘person’ (presumably because individuality is generally
associated with people). It is often used in the same series as generic nouns, e.g.

(397) 'one': generic noun:

French quelqu’un quelque chose
English some-one some-thing
Catalan ningú ‘anybody’ res/gens ‘anything’
(< nec-unus)
Maltese xi Ìadd ‘someone’ xi Ìaga ‘something’
(Ìadd < ‘one’) (Ìaga ‘thing’)
Kabyle yiwen ‘one; someone’ kra ‘thing; something’
Dongolawi wEr ‘one; someone’ hāǧ a ‘thing; something’
Welsh rhyw-un ‘someone’ rhyw-beth ‘something’
(un ‘one’) (peth ‘thing’)

But ‘one’ is occasionally also found in series that are otherwise based on in-
terrogative pronouns, e.g.

(398) M. Greek kan-énas ‘anybody’ (énas ‘one’)

tí-pota ‘anything’ (ti ‘what’)
Latvian ne-viens ‘nobody’ (viens ‘one’)
ne-kas ‘nothing’ (kas ‘what’)

Somewhat surprisingly, ‘one’ is sometimes also found as an indefiniteness

marker together with interrogative pronouns, e.g.

(399) Pashto yaw cok ‘somebody’ yaw ‘one’

cok ‘who; somebody’
Lezgian sa wuž jat’ani ‘somebody’ sa ‘one’
wuž ‘who’
Udmurt og-kin ‘somebody’ og ‘one’
kin ‘who’
Icelandic ein-hver ‘somebody’ ein ‘one’
hver ‘who’
 217

Uighur bir nemä ‘something’ bir ‘one’

nemä ‘what’
Khmer qw´y-mu´y ‘something’ qw´y ‘what’
mu´y ‘one’

An interesting case of reanalysis in terms of 'one' is reported by Haegeman

(1991) for a West Flemish dialect, where the older opaque indefiniteness marker
e(n)t- was reanalyzed as containing the numeral een 'one', so that besides eentwa
'something' (i.e. een-twa, earlier e(n)t-wa) it is also possible to say een schuon twa
'something nice', lit. 'a nice something'.
Like generic nouns, ‘one’ has a tendency to become restricted to negative-po-
larity and negative functions – see § [Link].

7.5.3. Borrowing of indefiniteness markers

Although indefinite pronouns are part of the grammatical system and

grammatical markers are relatively resistant to borrowing, a surprisingly large
number of cases have been attested where an indefiniteness marker was taken
over from another language, either by direct borrowing or by calquing.

[Link]. Direct borrowing. I know of no cases where a complete indefinite

pronoun has been borrowed,xi but some languages with interrogative-based
indefinites have indefiniteness markers that were adopted from other
languages with interrogative-based indefinites. This usually happens only in
languages that are under very strong influence of another language (cf., e.g.,
Majtinskaja (1969: 250-1) and Beke (1913-14: 174-7) for Finno-Ugrian
languages).

(400) Hungarian né- (in né-mi ‘some’, né-hol ‘here and there’, etc.)
< Slavic ně - (e.g. Russian ne-č to ‘something’, Czech ně co,
Serbo-Croatian neš to)
Mordva koj- (in koj-kov ‘somewhere’, koj-koš ta ‘sometime’, etc.)

< Russian koe- (e.g. koe-gde ‘somewhere’)

-buti (in kije-buti ‘somebody’, meze-buti ‘something’, etc.)
< Russian -(ni)bud’ (e.g. kto-nibud’ ‘somebody’)
Hill Mari ta- (in ta-ma ‘something’, tä-gü ‘somebody’, etc.)
< Chuvash ta- (e.g. ta-kan ‘somebody’, te-mĕ n ‘sth’)
Udmurt kot’- (in kot’-kin ‘anybody’, kot’-ma ‘anything’)
< Russian xot’ (e.g. xot’ kto ‘anyone’)
Slovak bár- (in bár-kto ‘anyone’)
< Hungarian bár- (e.g. bár-ki ‘anyone’)

Widespread borrowing of indefiniteness markers has been documented for

Romani dialects of eastern Europe by Boretzky and Igla (1991). Romani dialects
have adopted indefiniteness markers from practically all the surrounding
languages of the region, e.g.
 218

(401) Romani indefinites based on interrogatives

(ko ‘who’, so ‘what’, kaj ‘where’, etc.) (Boretzky and Igla 1991: 21-3)
bilo kaj ‘anywhere’ < Serbo-Croatian bilo
i-so ‘anything’ < Serbo-Croatian i-
kaj-godi ‘anywhere’ < Bulgarian -gode
vare-so ‘something’ < Rumanian oare-
choč -so ‘anything’ < Polish choć (= Russian xot’)
so-nebud’ ‘anything’ < Ukrainian -nebud’ (= Russian -nibud’)
akar-so ‘anything’ < Hungarian akár

[Link]. Calquing. Indefiniteness markers are often relatively weakly gram-

maticalized and can still be related to other elements in the grammar, so that
calquing is possible.
An example comes from Yiddish, which has been heavily influenced by
Slavic languages (espacially Polish and Ukrainian). For non-emphatic non-
specific uses, Yiddish employs the indefiniteness markers -(s’)-nit-iz (lit. ‘not is’)
and -es-iz (lit. ‘it is’), which are probably modeled on Slavic forms like
Ukrainian -nebud’ and Polish (nie) ba4dź (lit. ‘(not) be’) . Of course, it cannot be
excluded in principle that these forms arose independently in the manner
described in § 6.3.3, but the fact that German, the closest relative of Yiddish, has
nothing of this sort, and that Slavic influence is strong elsewhere in Yiddish
grammar, makes it highly plausible that this is a calque. Examples are given in
(402).

(402) Yiddish (Š apiro et al. 1984: 103, 206)

a. Mir veln vu-nit-iz ibernekhtikn.
we want where-INDEF [Link]
‘We’ll spend the night somewhere.’
b. mer vi ven-s’-nit-iz
more than when-INDEF
‘more than ever’

A particularly intriguing example of calquing is described by Subbarao and

Arora (1988) for the Dakkhini dialect of Hindi/Urdu. This dialect has been in
close contact with Telugu (Dravidian) for centuries, and its grammatical system
is now in many ways identical to Telugu. At some point, Dakkhini speakers
equated the Hindi/Urdu particle ki (‘that, whether’, a subordinator borrowed
originally from Persian ke) with the Telugu particle -oo, which is used as a
subordinator in embedded questions, as a disjunctive coordinator (‘or’), and as
an indefiniteness marker on interrogatives (cf. the Kannada cognate -oo in
Appendix A, Section 35). As a result, Dakkhini now uses interrogative-based
indefinites marked with -ki, in contrast to standard Hindi, which has a special,
synchronically non-derived series of indefinites (cf. Appendix A, Section 22,
and (28c) in § 3.1.1). Compare the parallel examples in (403-04) (standard Hindi
would use koii and kuch here).

Dakkhini Hindi (D), Telugu (T) (Subbarao and Arora 1988: 105)
(403) D. Kon ki aayaa.
who INDEF came
T. Ewaru-oo wacceru.
who-INDEF came
‘Someone came.’
(404) D. Mere abbaa kyaa ki leko aaye.
 219

my father what INDEF taking came

T. Maa naannagaaru emiti-oo tecceru.
my father what-INDEF brought
‘My father brought something.’

Another example of calquing, provided to me by David Gil, comes from

Singlish, a variety of English spoken in Singapore which exhibits strong
influence from southern Chinese. In Singlish, the combination which also is used
in the free-choice function:

(405) Singlish
Which way also can? [from a taxi driver]
'Can (I take) any route?'

This construction is evidently based on a Chinese construction corresponding

to Mandarin shénme yě 'which also, i.e. any' (cf. Appendix A, Section 36).
Calquing probably occurs widely in indefinite pronouns, but in most cases it
is hard to prove that language contact played a role in the creation of a
particular type of indefiniteness marker because it might as well have been
arisen independently. Thus, rather than being an exotic phenomenon, the clear
cases of Yiddish, Dakkhini and Singlish may well represent the tip of an
iceberg.

7.5.4. Semantic enrichment by implicature:

Appreciative and depreciative meanings

In many languages, indefinite pronouns may be used with secondary quali-

tative meanings, both positive (‘someone important’, ‘something remarkable’)
and negative (‘an unimportant person’, ‘in a negligent manner’). I will call these
uses APPRECIATIVE and DEPRECIATIVE (following Stoffel’s (1899) terminology).
These uses have often been noted in the literature, but no cross-linguistic
comparison has ever been made to my knowledge.
Appreciative and depreciative uses of indefinite pronouns are interesting for
linguistic theory because they are an example of semantic enrichment based on
conversational implicatures. More specifically, Grice’s first maxim of Quantity,
'make your contribution as informative as is required for the current purposes
of the exchange', comes into play here. Indefinite pronouns are intrinsically
uninformative because their semantic content is quite minimal in comparison to
most other words, and when speakers nevertheless use them in situations
where they do not contribute any additional information, hearers are entitled to
make additional inferences. Such interpretative enrichment may then be
conventionalized and lead to semantic change. However, I know of no case
where a whole indefinite series is only used in appreciative or depreciative
senses. Clearly, semantic change by conversational implicature is not nearly as
widespread and important for indefinite pronouns as semantic change by
grammaticalization (§ 6.4), and languages do not differ as radically in this
respect as they differ with respect to the patterning of other indefinite pronoun
meanings. It appears that most the examples cited in this section generally
illustrate special contextual interpretations based on converstional implicature
rather than additional truly semantic readings.

[Link]. Appreciative meanings. The examples in (406) illustrate the apprecia-

tive interpretation of indefinite pronouns. In general, the meaning of ap-
 220

preciative indefinites can be described as ‘remarkable, important’. The fact that

all the examples in (406) are from European languages should not be taken to
imply that this phenomenon is restricted to European languages, or even
particularly characteristic for them. It just reflects the fact that I happen to have
much better access to data from European languages, and that appreciative and
depreciative meanings are not often noted in grammars of non-European
languages. I would not expect languages to differ significantly in the extent to
which they show such meanings.

(406) a. German
Wir sind wieder wer.
‘We are somebody (important) again.’
b. French
Il lui sembla qu’ il entrait dans une vie nouvelle et charmante,
it him seemed that he entered in a life new and charming
qu’ il devenait quelqu’un, qu’ il était sauvé.
that he became somebody that he was saved
‘It seemed to him that he began a new and charming life, that he
became somebody (important), that he was saved.’
c. Polish
On byl kiedy-ś czym-ś ,ale dziś jest niczym.
he was when-INDEF what-INDEF but today is nothing
‘He was something (important) once, but today he is nothing.’
d. English (informal, especially American) (Quirk et al. 1985: 6.52[c])
That was SOME meal! (‘a very good meal’)
e. Latin
Nunc iste se ali-quem putat. (Seneca Dial. 5.37.3)
now that self INDEF-whom considers
‘Now he considers himself (to be) somebody.’
f. Basque (Aulestia 1989: 432)
Nor-bait zen alkatea gu-re herri-an.
who-INDEF he:was mayor:the we-GEN country-LOC
‘The mayor in our country was someone (important).’
g. Hungarian
Ö vala-ki a gyár-ban.
he INDEF-who the factory-at
‘He is somebody at the factory.’

In all these cases the indefinite pronoun is (part of) the predicate nominal in
a copular clause with some kind of copular verb (‘be’ in (a), (c-e), (g-h),
‘become’ in (b), ‘consider’ in (f)). As predicate nominal, an indefinite pronoun is
not just relatively uninformative, but actually creates a tautology, a blatant
violation of the first maxim of Quantity (cf. Levinson 1983: 110-11 on
tautologies). The informative inference in all these examples is 'appreciative',
i.e. the speaker is understood as referring to something important or special.
Furthermore, all these indefinite pronouns also have the ‘specific’ function.
Non-specific indefinites are apparently never used appreciatively, but tend to
have depreciative meaning (cf. § [Link]). This is particularly nicely illustrated in
(407) from Latin, where the specific indefinite quidam ‘some(one)’ is contrasted
with the free-choice indefinite quilibet ‘any(one)’. Example (407) is not a copular
sentence, but it is also a tautology, because necessarily life is led in some way or
other.
 221

(407) Latin
Vita agenda est certo genere quo-dam, non quo-libet.
life [Link] is certain way which-INDEF not which-INDEF
‘Life must be led in a certain (special) way, not (just) anyhow.’
(Cicero Fin. 3.24)

Fijn van Draat (1898) observes that English any-indefinites can also have an
appreciative interpretation. However, all his examples involve negative-po-
larity contexts, where we would expect any-indefinites anyway, e.g.

(408) English (Fijn van Draat 1898: 153-4)

a. Nobody who is anybody doubts it.
b. an illustrated paper, which is the first to publish portraits of everybody who
becomes anybody
c. There was absolutely no natural foundation on which a building of any
size could be erected.
d. Stamps are liable to deterioration when exposed to view for any length of
time.

Here, too, examples (c-d) are not copular clauses, but the pragmatic enrichment
is also due to a tautology because ‘size’ and ‘length (of time)’ specify only a
dimension and do not constitute information in themselves (necessarily, any
building has a size, and any event take some time).

[Link]. Depreciative meanings. When an indefinite pronoun with primarily

non-specific functions is used in a specific tautological context, the inter-
pretative enrichment is generally in the opposite direction, i.e. a depreciative
meaning results (‘unimportant’ or ‘bad’). This is illustrated in (409).

(409) a. Russian
On vsë delaet kak-nibud’.
he all does how-INDEF
‘He does everything badly (‘anyhow’).’
b. Lithuanian
Jis viska daro bet kaip/ kaip nors.
he all does INDEF how how INDEF
‘He does everything badly (‘anyhow’).’
c. Spanish (Lombard 1938-39: 196)
No sean tan cualquieras.
not be(IMPV) so any
‘Don’t be such good-for-nothings.’
d. Basque (Aulestia 1989: 153)
Edo-nola ezarri du.
INDEF-how leave she:it:AUX
‘He has left it in a mess (‘anyhow’).’
e. Yakut (Ubrjatova (ed.) 1982: 201)
Baahynajdar sebileniilere daVany xannyk eme ete.
peasants arms PT [Link] INDEF were
‘The arms of the peasants were pretty bad (‘any kind’).’
f. English (Stoffel 1899: 147)
Who on earth can it be? It is somebody for us to a certainty, and my hair is
anyhow, and my eyes are red,...
 222

Note that most of these indefinites would not be grammatical in their normal
meaning, because they are not allowed in contexts that do not admit a non-
specific interpretation. Thus, the additional meaning also leads to additional
cooccurrence possibilities.
Stoffel (1899: 147) attempts a pragmatic explanation of this interpretative
enrichment: 'Taught by sad experience, speakers are convinced that, with
respect to qualitative statements, their hearers are apt to suspect the worst',
hence the depreciative interpretation. But this not only presupposes that all
hearers are pessimists, but also leads one to expect that all interpretative en-
richments are depreciative. As we saw in the preceding section, this is not the
case.
The cross-linguistically observable correlation between specific functions
and appreciative interpretations, and between non-specific functions and
depreciative interpretations suggests an alternative explanation: Non-specific
indefinites, especially free-choice indefinites, refer to an arbitrary element of
their class. Given that all people are choosy (a much likelier supposition than
that all people are pessimists), it is normal that hearers expect the worst if they
are told that the referent has been selected randomly.
However, occasionally specific indefinites must also be interpreted in a
depreciative sense.

(410) a. Russian
Rabota sdelana koe-kak.
work done INDEF-how
‘The work was done badly.’
b. Macedonian
Se druž i so ne-kakov neranimajko.
REFL befriends with INDEF-which [?]
‘She is friends with some bad person.’

Some individual indefinites seem to have conventionalized the depreciative

meaning, e.g. English anyhow ‘in a bad manner’ or Russian koe-kak ‘anyhow’.
Dictionaries generally list these forms and translate them, whereas this is the
case much more rarely with appreciative meanings as in (664).

[Link]. Anti-depreciative meanings. The third type of interpretatively enriched

reading, called ANTI-DEPRECIATIVE here, represents the negation of the
depreciative sense that we saw in the preceding subsection. It comes close to
the meaning of appreciative indefinites (‘She is not anybody’ ≈ ‘She is
somebody.’). Anti-depreciative readings are treated separately here because
they seem to be much more readily available than depreciative readings. For
example, in German (411a) the depreciative reading is hard to get, but in (411b)
is it the only possible reading.

(411) German
a. Sie hat die Arbeit irgendwie gemacht.
‘She did the work somehow.’/??‘She did the work anyhow.’
b. Sie hat die Arbeit nicht irgendwie gemacht.
‘She did the work not anyhow.’

This fact, too, points to a stronger tendency towards lexicalization of the de-
preciative interpretation.
 223

Like the depreciative interpretation, the anti-depreciative interpretation is

generally found with non-specific and free-choice indefinites, as illustrated by
the examples in (412) (where Chuvash takam ‘somebody (specific)’ is the only
exception).

(412) a. German
Ich verpacke doch nicht irgendwelche Brücken und Gebäude, einfach so.
I wrap PRT not any bridges and buildings simply so
‘I don’t wrap just any bridges and buildings, just like that.’ (Christo)
b. Latin
...virtutes-que non quas-libet faciebat Deus per manum Pauli.
miracles-and not which-INDEF did God through hand of:Paul
‘And God wrought special (‘not just any’) miracles by the hands of
Paul.’ (Act. 19.11)
c. Bulgarian (Guentchéva 1981: 411)
Č ovek-ât ne iska kakvato i da e ovca, a si târsi ovca-ta.
man-the not wants which INDEF shep but REFL seeks sheep-the
‘The man doesn’t want just any sheep, but is looking for his sheep.’
d. Serbian/Croatian (Progovac 1990: 130)
Milan nije video bilo koga, već predsednika.
Milan NEG:has seen INDEF whom but president
‘Milan did not see just anyone, but the president.’
e. Chuvash (Skvorcov (ed.) 1985: 438)
Vă l takam marś ke, xamă r ś yn.
he someone [Link] ourselves person
‘He is not just anyone, but one of ours.’

These anti-depreciative cases also differ from the depreciative cases of § [Link]
in that they are not restricted to tautological environments. Apparently the
negation somehow licenses the occurrence of free-choice indefinites in these
environments.
In languages like English where free-choice indefinites are also used in the
direct-negation function, sentences with anti-depreciative indefinites threaten
to be ambiguous between the negative and the anti-depreciative interpretation.
However, there is an intonational difference between the two interpretations
(see Ladd 1980: 145-62, Horn 1989: 230): The anti-depreciative reading is
associated with a fall-rise (cf. 413a), while the negative meaning is associated
with a simple falling contour (cf. 413b).

(413) English
a. I don’t want to talk to ˇ anyone.
b. I don’t want to talk to ´ anyone.

This fall-rise contour of anti-depreciative indefinites is also found in languages

in which there is no potential ambiguity, e.g. in German. The simple falling
contour is simply odd in (414).

(414) German
Ich möchte nicht mit ˇ irgend jemandem sprechen.
‘I don’t want to talk to ˇ anyone.’

The similar facts in Hungarian are discussed by Hunyadi (1987: 133).

 224

i
In a few languages, another possibility is that a xot’-type focus particle is used to mark
concessive conditionals, e.g. in Hungarian (cf. 308b) and Romanian. A route of
development analogous to (351b) is also a possible option for these languages.
ii
Compare also Latin vel ‘even; at least; or’, which derives from the imperative of velle
‘want’. However, vel is not used to form indefinite pronouns.
iii
The most explicit defense of a unified semantic analysis in the literature is Gil (1993),
who analyzes elements such as Japanese mo as 'conjunctive operators'. However, he
does not say whether there is a unified analysis of elements like ka as 'disjunctive
operators'.
iv
Indeed, Manoliu-Manea (1966: 445) and Jackendoff (1972b) suggest that the
distinction between ‘every’ and ‘any’ can be captured by paraphrases along the lines of
(361): (i) You can invite any friend. = You can invite Joan or Bill or Henry or Julie
or... (ii) You can invite every friend. = You can invite Joan and Bill and Henry and
Julie and... See also Hunyadi (1987) on Hungarian mind- 'every; and' and akár 'any; or'.
v
It is interesting that Sannikov (1989), in the only semantic analysis of ‘or’ (Russian
ili) that I am aware of, proposes to analyze ‘A or B’ as ‘A is possible, B is possible’.
The source construction in (365b) comes quite close to Sannikov’s analysis (which was
arrived at completely independently, without regard to the form of ‘or’), while (365a) is
basically a hearer-oriented variant of (365b).
vi
Vasmer (1953-58) rejects this etymology, found in Miklosich (1886), for
phonological reasons and derives it from li ‘whether’ and the particle bo. If that were
the case, the homonymy between -libo (indefiniteness marker) and libo ‘or’ would be
accidental, which is of course possible. I have no stake in the etymology of libo ‘or’
offered in (366), but I am convinced that Vasmer is too strict in demanding
phonological regularity even for such grammatical items as ‘or’ (cf. also Haiman
(1974), who makes the same point).
vii
Compare also Givón’s (1990: 818) continuum between declarative and parametric
question, moving gradually from certainty through uncertainty to question: Joe
called/What’s-his-name called/Whoever it was that called.../I don’t know who
called/Who knows who called/Who called?
7a
Li (1992) and Postma (1994) are more recent attempts of configurational analyses
within formal syntactic frameworks which assign the same meaning to WH-expressions
and derive the different resulting meaning from the syntactic structure.
viii
I know only of four cases where one could think of a diachronic origin of an
interrogative root in an indefinite pronoun:
 225

(i) Italian cosa ‘what?’ is of course related to cosa ‘thing’, but it derived from che
cosa? ‘which thing?’ (cf. the analogous reduction in qualcosa ‘something’ from
qualche cosa);
(ii) Maltese xi ‘what?’, which seems to be related to Arabic š ay? ‘thing’ – probably
a similar explanation as in the Italian case can be given;
(iii) Coptic u ‘what?’ has been claimed to go back to Ancient Egyptian w¿j ‘one’
(Gardiner 1932: 307-8). This etymology is anything but certain.
(iv) Acehnese (Austronesian) pat ‘where?’ apparently derives historically from
teumpat ‘place’ (Mark Durie, personal communication). This case is the most intriguing
one, since it is part of a complex system of ‘epistemological classifiers’ (Durie 1985:
ch. 6), where pat can also mean something like ‘place’, e.g. with dum ‘every’ (dum=pat
‘everywhere’) and sa= ‘one, same’ (e.g. sa=pat ‘same place’).
ix
In some languages, reduplicated interrogatives have simple plural meaning, e.g. in
Ainu, Khmer, Mongolian (juu juu 'what what', i.e. 'what kinds of things'), and Tagalog
(sinu-sino 'who-who', i.e. 'who(PL)').
x
In addition to being restricted to negative-polarity contexts, these Dravidian echo
compounds also express the nuance of contempt. This is a widespread function of
reduplication (cf. Moravcsik 1978), which is also represented in English in the
construction book-shmook borrowed from Yiddish. Contempt is semantically close to
irrelevance and hence to free choice, so there is another possible connection here.
Unfortunately, I have no concrete proposal for a way of integrating all these
interconnections and must leave this topic for future research.
xi
Except for some languages where a generic noun was borrowed, e.g. Turkish bir sey
‘something’ < ‘one thing’ (sey from Arabic š ay? ‘thing’) or Dongolawi hāǧ a
‘something, thing’ (from Arabic Ìaaǧ at ‘thing’)
 239

Negative Indefinite Pronouns

While this work is the first in-depth typological study of indefinite pronouns in
general, negative pronouns in particular have been the object of two recent
large-scale cross-linguistic investigations, Bernini and Ramat (1992) and Kahrel
(1989). I will focus my attention on the relation between negative indefinites
and other indefinites, as well as on explanations of the observed universal
tendencies that were not yet given in the works cited.
In § 8.1, I discuss the received taxonomy of negated indefinites, demonstrat-
ing that it is inadequate and that the implicational map of chapter 4 provides
for a better classification. § 8.2 deals with one important aspect of the syntax of
negative indefinites, the cooccurrence with a negative element associated with
the verb. I will formulate a number of cross-linguistic generalizations and will
propose functional explanations for them. Finally, in § 8.3 I examine various
diachronic sources of negative indefinites, in particular negative scalar focus
particles and minimal-unit expressions.

8.1. The Main Syntactic Types of Negative Indefinites

8.1.1. The consensus: four types

There is general agreement that there are four main syntactic ways of ex-
pressing negated indefinites, or in my terms, the direct-negation function of
indefinite pronouns (Dahl 1979, Kahrel 1989: 3,i Bernini and Ramat 1992: 109-
10). These four strategies are illustrated in (415a-d).

(81) a. Verbal negation plus (ordinary) indefinite

Turkish (cf. bir s3ey 'something')
Bir s3 ey duy-ma-dı-m.
something hear-NEG-PAST-1SG
‘I didn’t hear anything.’
b. Verbal negation plus 'special indefinite'
Basque (cf. nor-bait 'somebody', i-nor 'anybody')
Ez dutinor ikusi.
NEG I:have:him anybody seen
‘I have not seen anybody.’ (Saltarelli 1988, ex. 525)
c. Verbal negation plus 'negative indefinite'
Polish (cf. kto-ś 'somebody')
Nikt nie przyszedΩ.
nobody NEG come:PAST:3SG
‘Nobody came.’
d. 'Negative indefinite' without verbal negation
Latin
Nihil obstat.
nothing hinder:PRES:3SG
‘Nothing stands in the way.’
 240

While this typology is useful at a rather superficial level, it is problematic in

several ways. In order to apply it consistently, one would have to know
precisely (a) what counts as verbal negation, (b) how a 'negative indefinite' is
defined, and (c) how a 'special indefinite' is defined. In the following sections,
we will see that none of these questions are easily answered. The discussion of
question (a) will be deferred till § 8.2.1, and questions (b) and (c) will be
discussed first.

8.1.2. Negative indefinites and elliptical contexts

How should we define 'negative indefinite pronouns'? Intuitively, what is

meant by saying that an indefinite pronoun is negative is that it somehow
expresses the negative sense by itself, inherently, without needing an additional
verbal negation. Brown (1985) tries to make this intuition explicit by proposing
that negative indefinites like nobody and nothing refer to an empty set. However,
this semantic description does not seem to conform to our intuitions. It is not
the same to say I heard nothing and I heard an empty set (of sounds). In the first
sentence it is conveyed that no hearing event took place, whereas in the second
sentence, it is implied that such an event took place. In natural languages, the
use of negated indefinite pronouns invariably leads to the first sense, whether
the indefinite is inherently negative or not. Thus, despite the widespread
intuition that there is something 'inherently negative' about (e.g.) Polish nikt
‘nobody’ as opposed to Basque inor ‘anybody’, it seems difficult to make it
explicit in semantic terms. Since indefinite pronouns are interpreted within
clauses, and at the clause level inor and nikt have the same meaning, the
semantics does not seem to help.
A better way of identifying inherently negative pronouns is the purely
syntactic criterion of negative meaning in ELLIPTICAL CONTEXTS. Bernini and
Ramat (1992: 115) explicitly adopt this criterion to distinguish negative from
non-negative indefinites. According to this criterion, indefinites are considered
inherently negative if they may occur in elliptical contexts (where the verb with
its negation is ellipted) and still convey the negative meaning. I will also speak
of free-standing indefinites in such cases. A typical elliptical context is provided
by negative answers to parametric questions. In Polish and Spanish, the
indefinite can be used in this way (cf. 416), but in Turkish and Hindi/Urdu, it
cannot (cf. 417).

(416) free-standing indefinites conveying negative meaning

a. Polish
A: Kto przyszedΩ? B: Nikt.
who came nobody
‘Who came? Nobody.’
b. Spanish
A: ¿Qué viste? B: Nada.
‘What did you see? Nothing.’
(417) free-standing indefinites not conveying negative meaning
a. Turkish
A: Ne duy-du-n? B: *Bir s3 ey.
what hear-PAST-2SG one thing
‘What did you hear? Nothing.’
 241

b. Hindi/Urdu
A: Aaj kaun aayaa? B: *Koii bhii.
today who came someone INDEF
‘Who came today? Nobody.’

Another context where the verb with its negation is typically ellipted is in
the standard of comparison. Thus, Rumanian nimeni and Russian nikogda are
inherently negative by this criterion, as shown in (418).

(418) free-standing negative indefinite in the standard of comparison

a. Romanian
M’ a nenorocit ca nimeni altul.
me has [Link] like nobody other
‘She has made me unhappy like nobody else.’
b. Russian
Ona č uvstvuet svoju silu kak nikogda ran’š e.
she feels her strength like never before
‘She feels her strength like never before.’

In addition, there are various other elliptical contexts that allow free-standing
inherent negatives. For example, in French the following structures are
possible:

(419) French (Gaatone 1971: 135, 160, 173)

a. à une vitesse jamais atteinte de mémoire d’ homme
at a speed never reached from memory of man
‘at a speed never reached in human memory’
b. Je vous assure, je ne me permets rien. Absolument rien.
I you assure I NEG me permit nothing absolutely nothing
Aucune désinvolture...
no casualness
‘I assure you, I allow myself nothing. Absolutely nothing. No
casualness...’

The ellipsis criterion yields clear results in most cases. However, it has a
decisive drawback that Bernini and Ramat do not address: Indefinites that must
be characterized as inherently negative by this criterion are not associated with
negative interpretations in all contexts. Consider the cases of Catalan and
Turkish. Both languages have a series of indefinites which are inherently
negative by the ellipsis criterion (Catalan ningú ‘nobody/anybody’, gens/res
‘nothing/anything’, etc.; Turkish hiç kimse ‘nobody/anybody’, hiç bir s3ey
‘nothing/anything’, etc.).

(420) a. Catalan (Lleó 1983: 298)

Ells ho tenen tot, nosaltres res.
they it have all we anything/nothing
‘They have got everything, we nothing.’
b. Turkish
A: Kim geldi? B: Hiç kimse.
who come-PAST(3SG) INDEF anyone/no one
‘Who came? Nobody.’

But in both languages, this 'negative indefinite' is also found in non-negative

sentences, e.g. in questions, and in Catalan also in conditionals.
 242

(421) Catalan
a. Hi ha res de nou?
there exists anything of new
‘Is there anything new?’
b. Si dius res, et castigaran.
if you:say anything you they:will:punish
‘If you say anything, they’ll punish you.’
(422) Turkish
Hiçkimse gel-di mi?
INDEF anyone come-PAST(3SG) Q
‘Did anyone come?’

An analogous situation is also found, e.g., in French (the personne-series), Italian

(the nessuno-series), Spanish (the ninguno-series), and Persian (the hič -series).
Thus, it would seem that in these languages, the indefinite series in question is
not inherently negative after all, and that the negative meaning in elliptical
situations like (420a-b) results from the fact that when the ellipted part is
mentally restored by the hearer during the interpretation process, she or he
restores the verb together with the negation, so that the correct negative
meaning is obtained. This is possible because the association with verbal
negation is strong enough in these cases, the non-negative uses of (421-22) be-
ing rather rare. By contrast, in the cases in (417) the non-negative uses of the
indefinite (Turkish bir-series, Hindi/Urdu bhii-series) are much more
prominent, so that the negation cannot be restored in ellipsis.
This hypothesis, which makes use of the seemingly vague notion of
'association', receives support from the validity of the two implicational
generalizations in (423).

(423) Negative interpretation of elliptical indefinites: absolute universals

a. (negative) If an indefinite series that is used in the direct-negation
function is also used in the free-choice function or a specific
function, it may not be used elliptically with a negative
interpretation.
b. (positive) If an indefinite series that is used in the direct-negation
function is not used in any other function, it may be used elliptically
with a negative interpretation.

In fact, the observed regularities are even stronger, but (423a-b) are the
strongest statemenst that are exceptionless in my data. Implication (423a)
cannot be formulated more strongly because of Modern Greek. In this
language, the típota-series is used in the irrealis-non-specific function (cf. 424a),
but it has a negative interpretation when it is free-standing (cf. 424b).

(424) Modern Greek

a. Pés mu típota.
say:IMPV me anything
‘Tell me something!’
b. Tí tis ípes? — Típota!
what her you:told anything
‘What did you tell her? — Nothing!’
 243

Modern Greek is my only counterexample to the stronger generalization that a

free-standing direct-negation indefinite may not be interpreted negatively if it
is also used in a function other than the negative-polarity [Link]
Implication (423b) cannot be formulated more strongly because of Icelandic,
Italian and Basque. In Icelandic, the indefinite nein- is only used for direct and
indirect negation (cf. 425a-b), but it cannot be used with a negative in-
terpretation in elliptical answers (cf. 425c). The same is true for Italian alcuno
‘any(body), (no)body’.

(425) Icelandic (Halldór SigurDsson, p.c.)

a. Ég sá ekki neinn.
I saw not anybody
‘I saw nobody.’
b. Ég held ekki aD neinn hafi komiD,
I think not that anybody have come
'I don't think that anybody has come.'
c. Hver er er viD dyrnar? — *Neinn.
who is there at door:the anybody
‘Who is at the door? — Nobody.’

Icelandic and Italian are my only counterexamples to the stronger general-

ization that a free-standing direct-negation indefinite may be interpreted
negatively if it is only used in a negative function. And Basque is my only
counterexample to the even stronger statement that a free-standing direct-
negation indefinite may be interpreted negatively if it is only used in a nega-
tive-polarity function (the i-series is not used in a free-choice or irrealis-non-
specific function, but it may still not be used in elliptical answers with a
negative interpretation).
Thus, the generalizations in (423) are the strongest descriptive statements
that seem to hold universally. But given the rarity of some other cases, as just
discussed, the stronger formulation in terms of tendencies in (426) captures the
linguistic reality better.

(426) Negative interpretation of elliptical indefinites: tendencies

a. (negative) The more additional functions a direct-negation
indefinite is used in, the less likely it is that it may be used
elliptically with a negative interpretation.
b. (positive) The fewer additional functions a direct-negation
indefinite is used in, the more likely it is that it may be used
elliptically with a negative interpretation.

The generalizations in (423) and (426) are not sufficient to make exact predic-
tions about the availability of a negative interpretation for free-standing in-
definites in every case. For example, the Basque i-indefinites have the same
range of functions as the Catalan cap-indefinites, but only the latter can be used
elliptically. Thus, there is some degree of language particular conven-
tionalization.
This is also confirmed by cases of language-internal variation, as exhibited
by the Hebrew af-series (Glinert 1982: 454). This series is restricted to the direct-
and indirect-negation functions, and according to (426b) this makes it quite
likely that it may occur elliptically with a negative interpretation. This is true in
the colloquial language, but purists favor the use of lo ‘not’ in elliptical answers:
 244

(427) Hebrew (Glinert 1982: 454)

A: Mi ba? — B: Af eÌ ad/ af eÌ ad lo.
who came INDEF one INDEF one NEG
‘A: Who came? — B: Nobody.’

That conservative speakers (purists) favor the construction with lo ‘not’ can be
understood as a reflection of the relatively recent restriction of af eÌad to
negative functions. Af originates in (and is still homonymous with) the focus
particle af ‘even’, so formally there is nothing inherently negative about this
series. However, the new constraints on the non-elliptical uses of this series
have made the elliptical use in (427) possible, which is a nice confirmation of
the reality of the generalization in (426b).
The generalizations in (423) and (426) have a straightforward functional ex-
planation: If an indefinite series also has another function in addition to those
listed above (i.e. free-choice, irrealis-non-specific, or specific), elliptical
expressions would be potentially ambiguous, so they cannot receive a negative
interpretation. Thus, the English any-series, while being the major strategy for
expressing negated indefinites in English, cannot be used in elliptical answers
because it also has the free-choice function:

(428) English
A: What would you do? B: Anything.
(Only: ‘I would do anything’; Not: ‘I wouldn’t do anything.’)

From the preceding discussion I draw the conclusion that it does not make
much sense to single out a special type (415c) 'verbal negation plus negative
indefinite'. One can of course define 'negative indefinite' in various ways, e.g.
using the ellipsis criterion, or as 'indefinite whose functions include the direct-
negation function', but these definitions are incapable of distinguishing
between the three types (415a-c).
In this chapter, I use the term negative indefinite pronoun in the deliberately
vague sense 'indefinite pronoun that has ‘direct negation’ as an important func-
tion'.

8.1.3. 'Special indefinites'

Even more problematic is the definition of 'special indefinites',iii because there is

not even an intuitive basis for it. Bernini and Ramat (1992: 124-6) define them as
indefinites that occur in affirmative or negative questions, but do not occur in
negative elliptical answers or in affirmative elliptical answers. This seems to be
a rather arbitrary definition, and one might suspect that it is due to the
influence of English (a suspicion which is confirmed by the abbreviations used
by Bernini and Ramat: A stands for 'generic quantifier' (any), N stands for
'negative quantifier' (no), and S stands for 'existential quantifier' (some), cf. § 4.6)
In the data of my 40-language-sample (§ 4.4), it is easy to see that there are
many ways in which an indefinite series that also expresses the direct-negation
function can be 'special': it may express also the indirect-negation function (e.g.
Portuguese n-), the question function (e.g. Italian n-), the comparative function
(e.g. Finnish -kaan), the conditional function (e.g. Catalan cap/res/ningú), the
irrealis-non-specific function (e.g. Quechua -pis, Greek típota), the specific-
unknown function (e.g. Portuguese qualquer, Turkish herhangi), and even the
specific-known function (Hindi/Urdu koii, Persian -i, Swedish någon). If one
 245

wishes, one can of course arbitrarily divide this list of functions into three sets
and assign the three types in (415a-c) to these sets, but I see no motivation for
such a move.
Thus, the three types of (415a-c) can be dispensed with and reduced to the
implicational map of chapter 4. There is no significant further distinction that is
not already captured by the map. In addition, the approach adopted here al-
lowed me to formulate implicational generalizations about the possibility of
negative elliptical constructions (see 423, 426) that cannot be detected if the pos-
sibility of elliptical constructions is used as a classifying criterion.
The only aspect of the taxonomy in (415) that is not captured by my implica-
tional map is the (non-)occurrence with verbal negation. The next section (§ 8.2)
will be devoted to a discussion of this parameter.

8.2. Negative Indefinites and Verbal Negation

8.2.1. What counts as verbal negation

In this section, we first have to take up the first question posed in § 8.1.1: What
counts as verbal negation? Many languages require two separate (often
discontinuous) elements for verbal negation (cf. Dahl 1979: 88-9), and
sometimes one of them cooccurs with a negative indefinite, while the other one
must be absent. This is the case, for example, in French and Maltese:

(429) French
a. Phuoc ne la voyait pas.
Phuoc NEG her saw not
‘Phuoc did not see her.’
b. Phuoc ne voyait (*pas) rien.
Phuoc NEG saw not anything
‘Phuoc did not see anything.’

(430) Maltese
a. Patricia ma rat-x lit-tifel.
Patricia NEG she:saw-NEG ACC:the-boy
‘Patricia did not see the boy.’
b. Patricia ma rat(*-x) xejn.
Patricia NEG saw nothing
‘Patricia did not see anything.’

Do these cases show verbal negation or not? Bernini and Ramat (1992: 203) as-
sign them to the type that lacks verbal negation, claiming that the postverbal
negative element, which is absent when a negative indefinite is present, is the
main part of the negation. This is plausible for French, where ne is never
stressed and is often omitted in the colloquial language,iv but less so for
Maltese, where the postverbal negative element -x clearly has less phonological
weight than preverbal ma. Anyway, since both the French and the Maltese
situation is due to Jespersen’s Cycle (§ [Link]), which seems to reflect a gradual
change, it is quite likely that at least in some languages we are dealing with real
ambiguity (i.e. ambiguity for speakers, not just for linguists).
Thus, the classificatory criterion of cooccurrence with verbal negation is not
unproblematic either. But the source of the fuzziness is well-understood, and
the distinction cannot be reduced to anything else, so it remains as an
 246

important distinction that will be discussed in greater detail in the next sec-
tions.

8.2.2. Main subtypes with respect to verbal negation

Negative indefinites can be divided into three subtypes depending on their

relation to verbal negation:

(431) Type NV-NI

negative indefinites that always cooccur with verbal negation
e.g. the Polish ni-series (nikt ‘nobody’, nic ‘nothing’, etc.)

a. Nikt nie przyszedΩ. ‘Nobody came.’

nobody NEG came
b. Nie widziaΩam nikogo. ‘I saw nobody.’
NEG saw nobody

(432) Type V-NI

negative indefinites that never cooccur with verbal negation
e.g. the standard English no-series

a. Nobody came.
b. I saw nobody.

(433) Type (N)V-NI

negative indefinites that sometimes cooccur with verbal negation
and sometimes do not
e.g. the Spanish n-series

a. Nadie vino. ‘Nobody came.’

nobody came
b. No vi a nadie. ‘I saw nobody.’
not I:saw ACC nobody

In cases like (431), (433b), traditional linguistics has often spoken of 'double
negation'. This is a rather misleading way of talking, because semantically there
is of course only one negation in such sentences. However, in languages like
Latin and standard English, which represent type V-NI, it is possible (though
rather unusual) to express semantically double negation in one sentence:

(434) English
Nobody didn’t come.
(435) Latin
Nemo non venit.
nobody NEG came
‘Nobody didn’t come.’

Thus, from the point of view of the Latin norm, which has long dominated
grammatical thinking in Western linguistics, a Polish sentence like Nikt nie
przyszedΩ (431a) sounds as if the opposite were meant. But rather than being
'illogical', Polish speakers just speak a language of a different type.
When we turn to the cross-linguistic distribution of the three types in (431-
33), we are immediately struck by the strong skewing in favor of type NV-NI,
 247

and the rarity of types V-NI and (N)V-NI. In my biased 40-language-sample, 32

languages have negative indefinites of type NV-NI, 7 languages have
indefinites of type V-NI, and 3 languages have indefinites of type (N)V-NI.v In
Bernini and Ramat’s (1992) study, which is restricted to European languages,
the skewing is less strong: 28 languages belong to type NV-NI, 10 languages
belong to type V-NI, and 7 languages belong to type (N)V-NI.
In my sample, the Latin type (V-NI) is only represented by European lan-
guages, suggesting that it is an areal phenomenon. This idea is confirmed by
the distribution within Europe, which is almost confined to a contiguous area
from Iceland to the Alps and southern France: Within Europe, the 10 languages
of this type are Icelandic, Norwegian, Swedish, Danish, English, Frisian,
German, French, Occitan, and Maltese (Bernini and Ramat 1992: 205). That the
type V-NI is rather rare is most convincingly shown by Kahrel’s (1989)
representative world-wide sample of 38 languages. In this sample, only four
languages show this type (Dutch, Mangarayi (Australian), Yidiny (Australian),
and Nama (Khoisan)), while 34 languages show type NV-NI and (N)V-NI
(Kahrel does not distinguish between these two). Thus, type V-NI is so rare that
we must ask why it is disfavored. An answer to this question will be proposed
in the next section (§ 8.2.3).
It is ironic that the cross-linguistic investigation reveals a picture that is
almost opposite to the older prejudice against the 'illogical' type NV-NI. The
typological perspective leads to a rather dramatic change of our world view:
What used to be regarded as the norm of correct thinking turns out to be a
rather rare phenomenon in some peripheral areas such as southern Africa,
northern Australia, and the extreme west of Eurasia. The fact that speakers of
Latin, English, French and German, all V-NI languages, were extremely
successful economically, militarily, and culturally, has meant that the general
preference for the type NV-NI has gone unnoticed for a long time.
Since the type NV-NI seems to be the default, its discussion will be deferred
to § 8.2.5. The next two sections are devoted to type V-NI (§ 8.2.3) and (N)V-NI
(§ 8.2.4).

8.2.3. Type V-NI: No cooccurrence with verbal negation

I would like to advance the hypothesis that the relative rarity of type V-NI can
be explained by the discrepancy between the semantics, which is that of
ordinary sentence negation (or nexus negation, in Jespersen’s terms), and the
surface expression of negation, which is on a participant rather than on the verb
in this [Link] This form-meaning mismatch is tolerated by some languages, but
since such mismatches are generally dispreferred (as expressed in Haiman’s
(1980) principle of isomorphism), the proportion of languages that violate the
preference for a close association between form and meaning is rather small.
But if the type V-NI is dispreferred, why does it exist at all? I would like to
suggest a diachronic explanation: An unrelated diachronic change which led to
a preferred structure elsewhere in the grammar has had the undesirable side
effect of creating the V-NI pattern. Since language change generally results only
in local optimization, such apparently counterproductive changes are possible
and do not contradict the general rule that language change is language
improvement (cf. Vennemann 1988).
There seem to be two types of diachronic change that lead to the V-NI type:
Jespersen’s Cycle (§ [Link]) and diachronic negative absorption (§ [Link]).
 248

[Link]. Jespersen’s Cycle and type V-NI. As Bernini and Ramat (1992: 205) ob-
serve, the ten V-NI languages in Europe form not only an areally compact
group, but there is also an interesting typological correlation within Europe: All
and only the V-NI languages have postverbal negation elements that arose by
Jespersen’s Cycle. From this correlation it is only a small step to the hypothesis
that Jespersen’s Cycle is responsible for the V-NI pattern in these languages.
Jespersen’s Cycle is the name for the well-known cyclic weakening and
simultaneous reinforcement (i.e., the grammaticalization) of the sentence
negator, as exemplified by the history of French and of English (Jespersen
(1917, 1924: 335-6); see also the discussion in Horn (1989: 452-9), Bernini and
Ramat (1992: ch. 1-3), Ladusaw (1993) and the literature cited there). This cyclic
development is shown schematically in (436).

(436) Jespersen’s Cycle: French English

I. original situation: jeo ne di ic ne secge
II. negation reinforced: je ne dis pas I ne seye not
III. loss of original negator: je dis pas I say not

Like all grammaticalization changes, this change is motivated by the tension

between a least-effort tendency toward weakening and an information-pre-
serving tendency toward strengthening (cf. Horn 1989: 457, Lüdtke 1980), i.e. it
has identifiable motivations that are unrelated to the syntax of negation or
negative pronouns. But it has an effect on these: Since the reinforcement of the
negation is achieved by grammaticalizing minimal-unit expressions (e.g.
French pas) or indefinite pronouns (e.g. English naught > not), which occur
postverbally in VO languages, the language ends up with a postverbal negator
once the old negator (ne) is lost. It can be shown that the postverbal negator
derives from a strengthening element that underwent Jespersen’s Cycle in all
the ten European languages that have postverbal negation (West Germanic and
Gallo-Romance).
The effect of Jespersen’s Cycle on indefinite pronouns is the creation of the
V-NI pattern: Like minimal-unit expressions, indefinite pronouns denote an
extreme value on a pragmatic scale, so at stage II of Jespersen’s Cycle, sentences
with negative indefinites do not need the reinforcing element. Thus we have,
side by side, (437) (a) and (b), (438) (a) and (b).

(437) French
a. Je ne dis pas.
I NEG say not
‘I don’t say.’
b. Je ne dis rien.
I NEG say nothing
‘I don’t say anything.’
(438) Middle High German (Wolfram von Eschenbach, Parzival)
a. Ich en tuons niht. (266.10)
I NEG do:it not
‘I won’t do it.’
b. Er en-dorfte niemen vrâgen wâ. (69.18)
there NEG-needed nobody ask where
‘No one needed to ask where.’

At stage III, after the loss of the original negator, the V-NI pattern results:
 249

(439) Colloquial French

c. Je dis rien.
I say nothing
‘I don’t say anything.’
(440) Modern German
c. Es brauchte niemand fragen wo.
it needed nobody ask where
‘No one needed to ask where.’

Thus, the introduction of type V-NI in European languages can be understood

as an undesirable but unavoidable side effect of the desirable reinforcement of
the negation. From a functional point of view, we now expect speakers to do
something about the undesirable side effect, and a straightforward remedy
seems available: The addition of the new negator (French pas, German nicht,
etc.) to negative sentences with negative indefinites. This is indeed what we see
in many languages in the later development (441-44), although such
constructions were not in general admitted into the standard languages – the
Latin model, defended as 'logical' by purists, seems to have been a very
powerful influence (see Vachek 1947, 1962).

(441) African-American Vernacular English (Labov 1972: 785)

Ain’t nobody ever thought about pickin’ up nothin’.
(442) Broken (= English-based Torres Strait creole; Shnukal 1988: 72)
a. Nobody no go kam. ‘Nobody will come.’
b. Ai no speak nating. ‘I said nothing.’
(443) Quebec French (Muller 1991: 262)
Le samedi soir au mois de juillet, y a pas personne
the Saturday night [Link] month of July there has not nobody
en ville à Québec.
in town at Quebec
‘On Saturday nights in July there’s nobody in Quebec city.’
(444) dialectal German
Nichts Genaues weiss man nicht.
nothing precise knows one not
‘Nothing precise is known.’

[Link]. Diachronic negative absorption and the Negative First Principle. It

seems that not all cases of type V-NI can be explained by Jespersen’s Cycle, so
some additional explanation must be sought. In this sub-section I will try to
motivate another diachronic mechanism by which the V-NI pattern may arise:
NEGATIVE ABSORPTION.
It appears that the following generalization holds: All V-NI indefinites,
except possibly in those languages that recently underwent Jespersen’s Cycle,
contain a negation marker that is identical to the verbal negator. This can be
explained if a diachronic scenario analogous to Klima’s transformations of
negative attraction and incorporation is assumed, where the verbal negation is,
so to speak, 'absorbed' into the indefinite. Schematically:

(445) Diachronic negative absorption

a. ‘Nobody came’:
I. Person not-came. >
II. Not-person came.
b. ‘She saw nothing.’
 250

I. She thing not-saw. >

II. She not-thing saw.

The motivation for this shift of the negator seems to be Jespersen’s Negative
First Principle (so dubbed by Horn 1989: 293):

(446) The Negative First Principle (Jespersen 1917)

There is a tendency to put the negative element as early as possible in
an utterance because the contribution of the negation to the meaning
of a larger constituent is particularly dramatic and the hearer needs to
get this information as soon as possible.

If this principle is correct, we expect the shift in (445) to occur only where the
negative indefinite precedes the verb (and thus also precedes the negator
associated with the verb), i.e. we expect it to occur with all indefinites in verb-
final languages, only with pre-verbal indefinites in verb-medial languages, and
not at all in verb-initial languages. This prediction seems to be borne out by
what diachronic evidence I have.
Consider first Arabic. Classical Arabic belongs to type NV-NI, i.e. the verb is
always accompanied by a negator (laa, lam, or maa, depending on the tense).
The basic word order is VSO.

(447) Classical Arabic

Lam yasta? ǰ ir-naa ? aÌ adun. (NT, Mt 20.7)
NEG hired-us one
‘Nobody hired us.’

In modern Arabic dialects, the verbal negator is usually maa, regardless of the
tense, and the basic word order is SVO (cf. 448a). In Baghdad Arabic, when the
indefinite is in subject position, the negator ma is prefixed to the indefinite, not
to the verb (cf. 448b), but when the indefinite is in post-verbal position, the
negation is on the verb (cf. 448c).

(448) Baghdad Arabic (Ali 1972: 54, 53, 48)

a. Saalim ma raÌ i-š uf-ni hnak.
Salim NEG FUT he-see-me there
‘Salim will not see me there.’
b. Ma-Ì ad kis´r il š ibbač .
NEG-one broke the window
‘No one broke the window.’
c. Saalim ma š af ? æy-waÌ id hnak.
Salim NEG saw INDEF-one there
‘Salim did not see anyone there.’

Similarly, in Cairene Arabic the discontinuous verbal negation ma ...-

š (cf. 449a) may be associated with the indefinite only if it is in pre-verbal
position, as in (449b).

(449) Cairene Arabic (Mitchell 1962: 107, 111; Bernini and Ramat 1992: 181)
a. Ma t-xaf-š !
NEG 2SG-worry(IMPV)-NEG
‘Don’t worry!’
b. Ma Ì addi-š yi-¿raf yi-?ra xåt5t5-i.
 251

NEG one-NEG 3SG.M-can(IMPF) 3SG.M-read(IMPF) writing-1SG

‘Nobody can read my writing.’
c. Ma š uf-ti-š Ì add. (*Š ufti ma Ì addiš .)
NEG saw-1SG-NEG one
‘I didn’t see anybody.’

The Arabic case is particularly interesting because it is clear that the negator
on the indefinite pronoun is the verbal negator: Constituent-focusing negation
has a different form (mu in Baghdad Arabic, muš /miš in Cairene Arabic). This
is clear evidence that we are really dealing with a shift of the verbal negation to
the indefinite, not with some kind of constituent-focusing negation or a
negative focus particle (as in the cases of § 8.3.1).
Another case that can be explained by diachronic negative absorption is the
situation in several Uto-Aztecan languages (Tümpisa Shoshone, Hopi, Yaqui).
In Tümpisa Shoshone (Dayley 1989: 320-3), whose basic word order is SOV, the
sentence negator kee usually stands immediately after the subject, as in (450a).
When it occurs together with an indefinite pronoun in object position, no word
order change is necessary to yield a sentence like (450b). But according to
Dayley, the combination kee-hinna in (450b) is a lexicalized compound. Sentence
(450c) shows a negative indefinite containing kee- in clause-initial position,
demonstrating that a shift like (445a) has taken place (though exceptionally kee
may also occur clause-initially elsewhere, Dayley 1989: 321-2).

(450) Tümpisa Shoshone (Uto-Aztecan; Dayley 1989: 321, 324, 44)

a. Nü kee sukkwa punitü.
I NEG that:ACC see
‘I didn’t see that.’
b. Nü kee-hinna punitü.
I NEG-what:ACC see
‘I saw nothing.’
c. Kee-hii nanangkanna.
NEG-what [Link]
‘Nothing is making noise.’

Although no diachronic evidence is available in this case, the most plausible

explanation for the V-NI pattern of Tümpisa Shoshone is clearly negative
absorption.
The situation is similar in the related language Hopi. In Hopi, sentences are
negated by putting qa ‘not’ immediately before the predicate (cf. 451a). When a
negative indefinite is present in the sentence, qa- is attached to the indefinite (cf.
451b-c). In contrast to Tümpisa Shoshone, word order change must be assumed
for all cases in Hopi.

(451) Hopi (Uto-Aztecan; Kalectaca 1978: 41, Malotki 1979: 123)

a. Pam wuuti kuuyit qa wéhekna.
that woman water NEG spilled
‘That woman didn’t spill the water.’
b. Qa-háqaqw kwii-kwitsi.
NEG-whence RDP-smoke
‘Smoke comes out nowhere.’ (Lit. ‘It smokes from nowhere.’)
c. Qa-hak pu-t aw may ’ta. (NT, John 8.20)
NEG-who he-ACC [Link] hands laid
‘Nobody laid hands on him.’
 252

Another variant of this situation is found in the related language Yaqui

(Lindenfeld 1973: 26-31). Ordinary sentence negation and negative absorption
are quite analogous to Tümpisa Shoshone (with some phonological changes in
indefinites with incorporated indefinites: kaa ‘not’ + habe ‘who’ > ka-abe, kaa +
hita ‘what’ > ka-ita), cf. (452a-c).

(452) Yaqui (Uto-Aztecan; Lindenfeld 1973: 31)

a. Nee kaa Maria-ta bič a-k.
I not Maria-DEP see-REAL
‘I didn’t see Maria.’
b. Ka-abe ne-u nooka-k.
NEG-who me-to speak-REAL
‘Nobody spoke to me.’
c. Ka-ita beete-k.
NEG-what burn-REAL
‘Nothing burned.’

But there is an alternative construction to (452b-c): Instead of the ka-indefinites

with incorporated negation, indefinites marked by huni (‘even’) may be used in
negative sentences. But here the sentence negator kaa must cooccur with the
indefinites. The resulting meaning is more emphatic negation (‘at all’).

(453) Yaqui (Lindenfeld 1973: 31)

a. Habe huni kaa ne-u nooka-k.
who INDEF NEG me-to speak-REAL
‘There is nobody at all who spoke to me.’
b. Hita huni kaa beete-k.
what INDEF NEG burn-REAL
‘Nothing whatsoever burned.’

Yaqui reminds us that the taxonomy of § 8.2.2 refers to individual indefinites,

not to whole languages. Yaqui ka-indefinites belong to type V-NI, while huni-
indefinites, which cannot have arisen by absorption, belong to type NV-NI. I
take this to be confirming evidence for the hypothesis that absorption is the
cause for the V-NI pattern.
Now let us go back to Europe and see whether negative absorption can be
found here, too. The Yaqui case is of course reminiscent of English: English no-
and any-indefinites are quite similar in their behavior to Yaqui ka- and huni-
indefinites. But recall that English has long been an SVO language, so negative
absorption as stated above could not account for English sentences like She saw
nothing. Furthermore, in earlier English the old negative particle ne cooccurred
with the no-indefinites, contrary to what we would expect if it had been
absorbed by them. English, as indeed most other modern European languages,
is adequately accounted for by Jespersen’s Cycle (§ [Link]), and we should not
yield to the temptation of overextending the application of negative absorption.
However, there are two possible cases of negative absorption in Europe,
after all. First, in colloquial Frisian the older nimmen ‘nobody’ (cf. Dutch
niemand, German niemand) has been replaced by the form net ien (lit. ‘not one’),
cf. (454).

(454) Frisian (Bernini and Ramat 1992: 84, 138, 219)

a. Jan yt net fisk.
Jan eats not fish
 253

‘Jan doesn’t eat fish.’

b. Jan het net ien. (older: Jan het nimmen.)
Jan has not one
‘Jan has nobody.’

However, no dramatic change in type results from this replacement because the
old form nimmen belongs to type V-NI (resulting from Jespersen’s Cycle), just
like the new net ien.
The other possible case is Latin. Although Latin shows traces of having
undergone Jespersen’s Cycle (non ‘not’ < ne-oen(um) ‘not-one’, Jespersen 1924:
336), there is no evidence whatsoever that the original negator ne, which was
inherited from Proto-Indo-European, got lost as it did in French, English, or
German. Instead, it seems that it was absorbed into the negative indefinites.
Since the older Latin word order was verb-final, we expect no subject-object
asymmetries in the effect of negative absorption, and indeed none are found.
As predicted by the absorption hypothesis, Latin negative indefinites contain
the element n(e)-.

(455) Origin of Latin negative indefinites

a. nē mo ‘nobody’ < *ne-hemo < *ne-homo, cf. homo ‘man’
b. nihil ‘nothing’ < *ne-hil < *ne hī lum, cf. hī lum ‘thread’
c. nū llus ‘no’ < *ne-ū llus < *ne-oinelos, cf. ū llus ‘any’
d. numquam ‘never’ < *ne-umquam, cf. umquam ‘ever’
e. nusquam ‘nowhere’ < *ne-usquam, cf. usquam ‘anywhere’

Like the changes of Jespersen’s Cycle, negative absorption is motivated

independently (by the Negative First Principle) but leads to the dispreferred V-
NI pattern. The expectation again is that speakers will remedy this state of
affairs at some stage if given enough time. For Arabic and Uto-Aztecan lan-
guages, it is probably too early to expect a new round of changes. But in Latin,
the expectation is borne out fully: According to the standard view (Posner 1984:
1), Proto-Romance belonged to the NV-NI type, i.e. verbal negation was
reintroduced. In the next section (§ 8.2.4), I will discuss the Romance languages
in more detail.

[Link]. The range of functions of negative indefinites of type V-NI. For the
seven negative indefinites of type V-NI in my sample (German, Dutch, English
no-, Swedish ingen, Icelandic enginn, Latin n-, Ossetic ma-/ni-), the following
implication holds:

(456) If a negative indefinite never cooccurs with verbal negation, it has

only the direct-negation function.

The implication appears to hold also for the other languages that were men-
tioned in this section (§ 8.2.3), and I know of no counterexamples. Of course this
universal can be easily motivated functionally: If a V-NI indefinite occurred in
non-negative functions, these would be understood as negative.
However, the reverse does not hold: Some negative indefinites that do not
have functions other than direct negation nevertheless belong to type NV-NI, in
particular negative indefinites in the Slavic and Baltic languages. And the
Georgian negative indefinite, which also does not have other functions, belongs
to type (N)V-NI.
 254

8.2.4. Type (N)V-NI: Variation in cooccurrence with verbal negation

Some negative indefinites sometimes cooccur with verbal negation and

sometimes do not, as illustrated by Portuguese (cf. 457) and New Testament
Greek (cf. 458).

(457) Portuguese
a. Ninguém veio.
nobody came
‘Nobody came.’
b. Não veio ninguém.
NEG came nobody
‘Nobody came.’
(458) New Testament Greek
a. Oudeìs hē mãs emisthō sato. (NT, Mt 20.7)
nobody us hired
‘Nobody hired us.’
b. Egō ou krínō oudéna. (NT, John 8.15)
I NEG judge:1SG nobody:ACC
‘I judge nobody.’

I have found variation of this kind in the following languages: several Romance
languages (Italian, Spanish, Portuguese, colloquial Catalan, cf. Posner 1984),
older Slavic (especially Old Church Slavonic, Old Russian, Old Czech, cf. Kř íž
ková 1968), Albanian, New Testament Greek, non-standard English (Labov
1972), non-standard German (Tanaka 1994: 196) and Georgian. In all these
languages, the negative indefinites that show this variation do not have many
non-negative functions, of course, because otherwise sentences lacking verbal
negation would possibly be ambiguous (cf. the discussion in § 8.1.2).
The following rule seems to capture the distribution of the V-NI pattern and
the NV-NI pattern for all these languages (except for Georgian, for which I do
not know what determines the variation):

(459) Cross-linguistic generalization for V-NI/NV-NI variation

Whenever a negative indefinite may be associated either with the
V-NI pattern or with the NV-NI pattern, the V-NI pattern tends to be
used when the negative indefinite occurs preverbally, and the NV-NI
pattern tends to be used when the negative indefinite occurs
postverbally.

The motivation for this generalization is again straightforward: It is Jespersen’s

Negative First Principle (cf. § [Link]), as noted in Kř íž ková (1968: 31)vii and
Horn (1989: 450). If rule (459) is strictly observed, as it is in some languages, the
result is that the hearer is always informed of the negative character of the
sentence before he or she hears the verb, i.e. rather early in an SVO language.
(Note that the languages with (N)V-NI indefinites almost all have the basic
order SVO.)
But notice that the effect of a rule like (459) would be more easily achieved
by adopting the NV-NI pattern uniformly, as is done in the large majority of
the world’s languages. Against this background, the diachronic development of
Slavic is easily understood: While older Slavic showed variation along the lines
of (850), all modern Slavic languages uniformly show the NV-NI pattern
 255

(Mathesius 1937, Kř íž ková 1968). That is, the change from (N)V-NI to NV-NI
restored the universally preferred state of affairs.
As Kř íž ková (1968) documents, only the NV-NI pattern is attested with
postverbal indefinites in Old Church Slavonic and Old Russian (e.g. 460a). With
preverbal indefinites, both V-NI and NV-NI are possible (Old Church Slavonic
showing a slight preference for NV-NI, Old Russian a slight preference for V-
NI), cf. (460b-c).

(460) Old Russian (Kř íž ková 1968: 24)

a. (NV-NI) i ne idjaš e s nimi nikto ž e
and NEG went with them nobody PT
‘And nobody went with them’
b. (V-NI) Nič ego ž e sja bojat’ bě si, tokmo kresta.
nothing PT self fear demons only cross
‘The demons are afraid of nothing, except the cross.’
c. (NV-NI) jako svoego nikto ž e ne xulit’
because self’s:ACC nobody PT NEG abuses
‘because nobody abuses his own’

Similar tendencies are attested for Old Czech. In the modern languages,
sentences like (460b) are impossible, and only the NV-NI pattern is acceptable.
But when we get to Romance, we seem to be faced with a puzzle: It appears
that the reverse development can be observed, from a uniform NV-NI pattern
to variation along the lines of (459). Such a reverse development would
contradict my thesis that the NV-NI pattern is universally preferred. How can
we understand this change? The facts are as follows (see the summary in
Posner 1984). In the majority of Romance languages, only the NV-NI pattern is
possible (Romanian, standard Catalan, Friulian, Ladin, Romansh, Vegliot,
French with reservations), but in Italian (especially central and southern),
Sardinian, Spanish, and Portuguese, the sentence negator is omitted when the
negative indefinite occurs preverbally. The standard assumption is that this is
an innovation and that Proto-Romance was uniformly NV-NI, not only because
most Romance languages still belong to this type, but mainly because older
texts of (N)V-NI languages have the NV-NI pattern even with preverbal
indefinites, e.g. (461a-b).

(461) a. Old Spanish (Llorens 1929: 86)

Vio que ninguno non pud ffallar.
saw that nobody NEG could speak
‘She saw that nobody could find.’
b. Old Portuguese (Llorens 1929: 89)
Nenhuu o nom podia acordar de seu pensar.
nobody him NEG could wake from his thought
‘Nobody could awaken him from his thoughts.’

Posner (1984) suggests that Spanish and Portuguese may have been influenced
by Italian, and that the Latin model played a role in supporting the V-NI
pattern in Italian (and Sardinian). However, the Latin influence does not, of
course, account for the asymmetry described in (459), which must have a
universal basis.
I would like to propose that the puzzle can be solved in the following way:
The assumption that Proto-Romance was uniformly NV-NI should be given
[Link] The Latin V-NI pattern was replaced by the NV-NI pattern first in the
 256

case of postverbal indefinites, which violate not only the form-meaning

isomorphism, but also the Negative First Principle, and are hence doubly
dispreferred. This change leads to the (N)V-NI pattern as described above.
Central and southern Italian and Sardinian stopped at this stage, while most
other languages extended the NV-NI pattern also to preverbal indefinites.
Before the standard languages were codified, there was a lot of variation, and
in the process of codification, Spanish and Portuguese were influenced by
Italian, and all three languages were to some extent influenced by Latin (cf.
Posner 1984). The change from uniform NV-NI to (N)V-NI that we see in the
texts would then only be a superficial change, resulting from system-external
sociolinguistic factors, and not a deep change in the system. If this
interpretation is correct, my general account can be maintained. Note that a
change from NV-NI to (N)V-NI is predicted not to occur, because it would
mean that a structure that is optimal on two preference parameters (Negative
First and form-meaning isomorphism) is turned into a form that is less
preferred on one of these parameters.
It is instructive to compare the change from Latin (V-NI) to Romance ((N)V-
NI > NV-NI) with the analogous change from standard English to dialectal
English. As documented in Labov (1972), there are many dialects of English
which have not extended the NV-NI pattern to indefinites in subject position
and can be compared to Sardinian or Italian. For example, New York City
White Vernacular allows NV-NI with postverbal no-indefinites, but not with
preverbal no-indefinites:

(462) New York City White Vernacular English (Labov 1972: 785-6)
a. We don’t ever see none of them guys.
b. *Nobody don’t know.

But sentences like (462b) are possible in other white dialects, and quite common
in African-American dialects. But even in African-American English the
tendency of (459) finds a reflection: While the NV-NI pattern is normally
obligatory with no-indefinites, it is always optional if the no-indefinite is in
preverbal position (Labov 1972: 806).

(463) African-American Vernacular English (Labov 1972: 785-6)

a. Nobody don’t know where it’s at.
b. Nobody fights fair.

Thus, different non-standard dialects of English have diverged from the

standard V-NI pattern to various degrees, just like different Romance lan-
guages have diverged from Latin to various degrees. And just like the Latin
influence on Romance was strong, the standard English influence on English
dialects is strong. At the present stage, the most innovative English dialects
seem to be further advanced than Italian (in that they allow (463a)), but none is
as advanced as Romanian or standard Catalan (in that none disallows (463b)).

8.2.5. Type NV-NI: Obligatory cooccurrence with verbal negation

As we saw in § 8.2.2, the type NV-NI, where the indefinite pronoun cooccurs
with verbal negation, is the most widespread type cross-linguistically, as is to
be expected because sentence negation should be expressed on the [Link] When
a negative indefinite pronoun of this type appears to contain a negative marker
 257

(e.g. Russian ni-kto ‘nobody’), Western linguists have traditionally spoken of

'double negation', but I have noted that this is a mistake due to the prejudice of
V-NI languages like Latin that do allow real double negation. This was pointed
out already by van Ginneken (1907), and Mathesius (1937) proposed that we
are instead dealing with a kind of agreement in such cases, or 'negative
concord'. Of course, this case is rather different from standard agreement (e.g.
in gender, number, or case), and probably we would not want to water down
the notion of agreement unnecessarily by subsuming these cases under it.
In any event, there are many negative indefinites that do not contain a
negative marker, and although they are not so prominent in Europe, they seem
to be prevalent outside of Europe. With respect to these indefinites, it seems
best not to ask simply whether they are 'inherently negative', but to ask in
which other functions they are used besides the direct-negation function (as I
proposed in § 8.1). We can then formulate the generalization in (464):

(464) If an indefinite pronoun series that is used in the direct-negation

function is also used in a non-negative function other than
the comparative, question and conditional functions, it occurs in the
NV-NI pattern.

The condition 'other than the comparative, question and conditional functions'
is necessary in order to exclude Romance indefinites like Italian nessuno and
Spanish nada, which are possible in these functions but belong to the (N)V-NI
type. The generalization in (464) captures the intuition that indefinites that are
not 'inherently negative' cannot do without verbal negation and at the same
time avoids the problems associated with the notion 'inherently negative' (cf. §
8.1).
Let us now look at an interesting restriction on negative indefinites that are
also used in non-negative functions. In some SVO languages, such indefinites
cannot be used in subject position, as illustrated in (465-70). This restriction is
rarely found in negation-only indefinites like Russian ni-kto, etc.x

(465) English
a. I didn’t see anybody.
b. *Anybody didn’t come.
(466) Swedish
a. Jag har inte sett någon.
I have not seen anybody
‘I didn’t see anybody.’
b. *Någon har inte kommit.
anybody has not come
‘Nobody came.’
(467) French (Muller 1991: 314)
a. Je n’ ai (pas) vu qui que ce soit.
I NEG have not seen who indef
‘I didn’t see anybody.’
b. *Qui que ce soit n’ est (pas) venu.
who indef NEG is not come
‘Nobody came.’
(468) Swahili
a. Si-ku-ona mtu.
[Link]-see person
‘I did not see anybody.’
 258

b. *Mtu ha-kufa.
person NEG-die
‘Nobody has died.’
(469) Hausa (Ma Newman 1990: 11, 181)
a. Bà mù hàdú dà kóowáa à hányàa bá.
NEG we meet with anyone on road NEG
‘We didn’ meet anyone on the road.’
b. *Kóowáa bài zóo bá.
anyone NEG-he come NEG
‘Nobody came.’
(470) Chinese (Li 1992: 127)
a. Tā bù xǐ huan shénme.
she NEG like what
‘She doesn’t like anything.’
b. *Shénme rén bù xǐ huan tā .
what person NEG like she
‘Nobody likes her.’

In order to express the ideas of the (b) sentences, these languages have to
resort to a different indefinite series (e.g. 471), to an existential construction (e.g.
472), or to a change of the word order (e.g. 473).

(471) a. English
Nobody came.
b. Swedish
Ingen har kommit.
nobody has come
‘Nobody came.’
c. French
Personne n’ est venu.
nobody NEG is come
‘Nobody came.’
(472) a. Swahili
Ha-kuna mtu a-li-ye-tu-ajiri. (NT, Mt 20.7)
NEG-exist person [Link]-hire
‘There’s nobody who hired us.’
b. Hausa
Báà kóowáa wandà yá zóo.
NEG:exist anyone REL 3SG came
‘There’s nobody who came.’
(473) a. Swahili
Ha-kufa mtu.
NEG-die person
‘Nobody died.’
b. French
Il n’ est pas venu qui que ce soit.
it NEG is not come who INDEF
‘Nobody came.’

The reason for the unacceptability of the (b) sentences in (465-70) is the lack of
focusing of the indefinite pronoun: In order to be interpreted negatively, an
indefinite pronoun that also has non-negative uses needs to be focused (cf. § 5.7
on focusing and scalar endpoints), and in many SVO languages only postverbal
constituents can be focused. By contrast, all preverbal constituents can be fo-
 259

cused in verb-final languages, so these languages do not show asymmetries like

those in (465-70).xi
This explanation, which relates the subject-object asymmetries in (465-70) to
focusability rather than to structural relations with the negative element,
receives support from Modern Greek. In this language, postverbal indefinites in
the direct-negation function may be either stressed or unstressed (i.e. focused
or unfocused).

(474) Modern Greek (Veloudis 1982: 179)

a. Dhen érxete kanís IPOPSÍFJOS.
NEG comes any candidate
‘No candidate comes.’
b. Dhen érxete KANÍS ipopsífjos.
NEG comes any candidate
‘Id.’

However, in the pre-verbal position only stressed negative indefinites are

admitted (Veloudis 1982: 194), as shown in (475). Structurally, kanís occurs in
the same position in (475a) and (475b), but it must be focused in order to be
licensed.

(475) a. *Kanís dhen tin ENDHIAFÉRI.

anyone NEG her interests
‘Nobody interests her.’
b. KANÍS dhen tin endhiaféri.
anyone NEG her interests
‘Nobody interests her.’

The tendency for negative indefinites to occur postverbally extends even to

indefinites that are 'inherently negative' and are not categorically excluded
from the subject position. As Labov (1972) notes, in African-American
Vernacular English, sentences like (476a) are possible, but subject-verb inver-
sion as in (476b-c) is also a common option.

(476) African-American Vernacular English (Labov 1972: 786, 811)xii

a. Nobody don’t know where it’s at.
b. Don’t no average motherfucker make no fifty dollars a day.
c. Ain’t no white cop gonna put his hands on me!

However, this tendency to bar from the subject position negative indefinites
that are not 'inherently negative' is not universal – it is only a tendency.
Compare the following example from Hebrew, an SVO language that is
otherwise much like Swahili or Chinese in the relevant aspects.

(477) Hebrew
a. Lo ra?iti iš .
NEG I:saw anybody
‘I didn’t see anybody.’
b. Iš lo ba la-mesiba.
anybody NEG came to:the-party
‘Nobody came to the party.’
 260

Quite analogous are sentences like Personne n’est venu ‘Nobody came’ in older
French, at a time when ne was still the primary negation marker. And compare
sentences like (478) in English, where a lexical scalar endpoint is used to
express negation. In (478b), the subject must bear sentence accent in order to be
interpreted as focused, but otherwise it is as acceptable as (478a) (cf. König
1991: 41).

(478) English
a. He wouldn’t kill A FLY.
b. A ROCKEFELLER could not afford to pay for this.

Thus, the restriction on any- in such environments must be regarded as an

idiosyncratic fact about any- (and cannot be predicted from the low position of
negation, as claimed by Progovac (1994) and Laka (1994)).

8.2.6. Cooccurrence of several negative indefinites in one clause

When several negated indefinites occur in one clause, some languages have
restrictions on certain combinations. In the unmarked case, illustrated by (479-
81), the same negative indefinite series is used for all the indefinites.

(479) Spanish
Nadie hizo nada.
nobody did nothing
‘Nobody did anything.’
(480) Swahili
Ha-wa-kumw-ambia mtu neno. (NT, Mt 16.8)
NEG-they-him-say person word
‘They didn’t say anything to anybody.’
(481) Basque (Saltarelli 1988: 306)
Ez digu i-nor-k e-zer esan-go.
NEG he:it:[Link]:AUX INDEF-who-ERG INDEF-what say-FUT
‘No one will say anything to us.’

But some languages show restrictions on the possible combinations of negative

indefinites, as we will see in this section.

[Link]. Cooccurrence restrictions of V-NI indefinites. In many languages with

V-NI indefinites, only the first indefinite is from the series that is normally used
in the direct-negation function. The other indefinites are from the series that is
used in the indirect-negation function, as exemplified in (482-84).

(482) Latin
Nemini quidquam <*nihil> dixerunt.
to:nobody anything nothing they:said
‘They didn’t say anything to anybody.’
(483) English (standard)
They said nothing to anybody <*nobody>.
(484) standard German
Sie sagten niemandem etwas <*nichts>.
they said nobody something nothing
‘They didn’t say anything to anybody.’
 261

The fact that the other orders are not possible (*Cujquam nihil dixerunt, *They
said anything to nobody) is of course another manifestation of Jespersen’s
Negative First Principle.
The starred variants of (482-84) are impossible only on the intended reading,
but they are grammatical on a different reading in these languages, cor-
responding to the reading of 'real double negation' of (434-35) in § 8.2.2.

(485) English
They said nothing to nobody.
i.e. They said something to everybody.

The fact that this doubly negated reading is available explains the impossibility
of two (or more) V-NI indefinites cooccurring with each other – otherwise
sentence (485) would be ambiguous. Thus, we have the following typological
correlation:

(486) a. If in a language negative pronouns do not cooccur with verbal

negation, they also do not cooccur with each other, and conversely,
b. if negative pronouns do not cooccur with each other, they do not
cooccur with verbal negation, either.

Unfortunately, I know of two counterexamples to the first generalization

(886a), whereas no counterexample to the second generalization (886b) has
come to my attention yet. The first of these counterexamples, Occitan, can
perhaps be explained by its history. In Occitan, sentences like (887a) are
possible, although Occitan belongs to type V-NI (cf. 887b).

(487) Occitan (Bernini and Ramat 1992: 208, 188)

a. Degun vegué ren.
nobody saw anything
nobody saw anything
‘Nobody saw anything.’
b. Ai ges d’ amic.
I:have nothing of friend
‘I have no friend.’

Occitan has recently undergone Jespersen’s Cycle (Bernini and Ramat (1992:
208) note that some dialects like Gascon still retain the preverbal negation),
which means that negative indefinites like degun ‘nobody’ and ren ‘nothing’
have only recently become inherently negative. So probably sentences like I
didn’t say nothing (i.e. ‘I said something’) are not yet possible,xiii so that the
motivation for the impossibility of cooccurrence does not apply.
In contrast to Occitan, the negative indefinites of Latin may be assumed to
have arisen by negative absorption (§ [Link]), which would of course lead to
structures like (482), because the verbal negation can be absorbed only into one
indefinite pronoun. Thus, the different origin of the Latin and Occitan V-NI
pattern also results in different behavior when a clause contains more than one
negated indefinite. The fact that standard English and standard German behave
like Latin although they underwent Jespersen’s Cycle could be attributed to
influence from Latin, or it might be due to the fact that a longer time has passed
since the preverbal negation got lost, so that there is no danger of
misunderstanding sentences like Nobody didn’t come.
 262

However, there is a second counterexample, Ossetic, and I can think of no

way of explaining it away. In Ossetic, negative indefinites are marked by the
prefixes ni- and ma-, which are so similar to the verbal negations næ and ma that
an origin by negative absorption seems plausible. This would also account for
the V-NI pattern, illustrated in (488a). But unfortunately, several negative
indefinites may cooccur in one clause, cf. (488b).

(488) Ossetic (Axvlediani (ed.) 1963: 197)

a. Ma-kædæm acu.
NEG-whither go:IMPV
‘Don’t go anywhere.’
b. Ma-kæmæn ma-cy zæV.
NEG-to:whom NEG-what say:IMPV
‘Don’t say anything to anybody.’

I have no explanation for this unexpected situation, but I think that the
generalization in (486) is still valid as a strong tendency, and that the expla-
nation is also on the right track.

[Link]. Cooccurrence restrictions in NV-NI languages. In some languages that

have several different negative indefinites belonging to type NV-NI, there is a
restriction to the effect that a non-inherently negative indefinite may not
precede an inherently negative indefinite. For example, in Persian either the i-
series (which is used in many other functions, including specific functions) or
the hič -series (which occurs only in negative functions and in questions) may
be used in the direct-negation function, cf. (489). When two negative indefinites
cooccur in one clause, only three of the logically possible four combinations are
acceptable, cf. (490).

Persian
(489) Hič kas(-i)/ kas-i zang na-zad.
INDEF person-INDEF person-INDEF phone NEG-struck
‘Nobody phoned.’
(490) a. Kas-i č iz-i na-š anid.
b. Hič kas hič č iz na-sanid.
c. Hič kas č iz-i na-š anid.
d. *Kas-i hič č iz na-š anid.
‘Nobody heard anything.’

The same restriction is found in Turkish, which borrowed the negative

indefiniteness marker hič from Persian and seems to have borrowed and
preserved the complete pattern.
This restriction, which has not to my knowledge been noted before in the
literature, is similar to the restriction of the preceding subsection (§ [Link]) in
that the 'inherently negative' indefinite must precede the other indefinite,
which is clearly another manifestation of the Negative First Principle. If any of
the four combinations is impossible, the Negative First Principle predicts that
the (d) combination is impossible. But given that (489) is acceptable, why
cannot all four combinations be possible?
I would like to venture the following speculation: Persian *Kas-i hič č iz na-š
anid and its Turkish equivalent are ungrammatical because kas-i cannot be
interpreted as falling in the scope of negation (in this sense there is a similarity
to (465-70) (b) in § 8.2.5). The reason is that the following hič -indefinite, which
 263

is more emphatic than the corresponding hič -less form in (490a), is au-
tomatically focused and 'absorbs the focus', so to speak, so that kas-i can no
longer be the focus (as in 490d).

8.3. Diachronic Sources of Negative Indefinites

Diachronic sources of indefinite pronouns that are used in the direct-negation
function have already been discussed in earlier parts of this work (especially §
7.1, § 8.2.3). In this section, two sources that deserve more discussion will be
treated: negative scalar focus particles (§ 8.3.1) and minimal and maximal unit
expressions (§ 8.3.2). Then § 8.3.3 will briefly summarize the diachronic sources
of negative indefinites, and in § 8.3.4 I discuss and reject the hypothesis that
overtly negative indefinites can be reanalyzed as 'non-negative' indefinites.

8.3.1. Negative scalar focus particles

It is often taken for granted that negative indefinite pronouns like those in (491)
incorporate a negation, so that they are 'inherently negative' not only from a
semantic, but also from a formal point of view.

(491) Russian ni-kto ‘nobody’

Lithuanian nie-kas ‘nobody’
Rumanian nici-odată ‘never’
Classical Greek oud-eís ‘nobody’
Hungarian sen-ki ‘nobody’

However, this way of speaking is misleading at best. In this type of negative

indefinite, which is particularly widespread in Europe, the indefiniteness
marker is not a simple negation marker, but a NEGATIVE SCALAR FOCUS PARTI-
CLE. Thus, I claim that negative indefinites of the type in (491) are more closely
related to the indefinites in § 7.1, which are also formed by means of a scalar
focus particle, than to the negative indefinites in § [Link], which arise by
negative absorption. The important point is that although the negative focus
particles in (491) and (492) are 'negative' in some sense, they are quite
independent of (and sometimes formally unrelated to) the verbal negator which
expresses sentence negation.
The examples in (492) show that indefiniteness markers in negative indef-
inites are often identical to negative focus particles meaning ‘nor, not even’, but
different from the sentence negator associated with the verb.

(492) negative indefinites from negative scalar focus particles

Russian ni-kto ‘nobody’ ni ‘nor, not even’ (< ne + i)
(and other Slavic languages) ne ‘not’
Classical Greek oud-eís ‘nobody’ oudé ‘nor, not even’ (< ou + de)
ou(k) ‘not’
Hungarian sem-mi ‘nothing’ sem ‘nor, not even’ (< is + nem)
nem ‘not’
Lithuanian nie-kas ‘nobody’ ne% ‘nor, not even’ (> nie-)
ne- ‘not’
Albanian as-njeri ‘nobody’ as ‘nor, not even’
s’, nuk ‘not’
 264

Finnish milloin-kaan ‘never’ -(kA)An ‘nor, not even’

ei ‘not’
Spanish ninguno ‘(no)body’
< *nec-unus Latin neque ‘nor, neither’
Latin non ‘not’
Cairene Arabic wala Ìaaga ‘nothing’ wala ‘nor, neither’
ma ...-š ‘not’
Ancash Quechua ni-ima ‘nothing’ ni ‘nor, not even’ (< Spanish)
mana ... -tsu ‘not’
Selkup š i-kuty ‘nobody’ š i ‘nor, not even’ (<Russian)
ašš a ‘not’
Mansi nem-xotti ‘nobody’ nem ‘nor, not even’
at ‘not’
Rumanian nici-un ‘no’ nici ‘nor, not even’
nu ‘not’

These negative scalar focus particles are used just like the scalar additive focus
particles discussed in § 7.1: They convey the meaning that everything that is
higher on the relevant pragmatic scale than their focus value is included, i.e.
negated. In (493) I give some examples where they are used in contexts other
than indefinite pronouns.

(493) a. Russian
Ja ne skaž u ni slova.
I not say [Link] word
‘I won’t say even a word.’
b. Rumanian
Nici mă car unul n-a scă pat.
[Link] at least one NEG-has escaped
‘Not even one escaped.’
c. Cairene Arabic
Ma m¿-ii-š wala waÌda.
NEG with-1SG-NEG even one
‘I haven’t a single one on me.’
d. Finnish
Emme saaneet yhtä-än kalaa.
NEG:1PL receive one-even fish
‘We didn’t receive even a single fish.’

If the focus value is the low endpoint of a pragmatic scale, the resulting
meaning is that of universal negation. Some of the negative indefinites in (492)
are based on the numeral ‘one’, which is of course the low endpoint of the scale
of numbers (Greek oud-eís, Albanian as-njeri, Spanish ning-uno, Rumanian nici-
un). Others are based on generic nouns (Arabic Ìaaga ‘thing’), indefinite
pronouns (Mansi xotti ‘somebody’), or most often bare interrogatives used as
indefinites. The semantic mechanism is always the same (cf. also § 7.1): These
bases are understood as low endpoints of the relevant pragmatic scale, and
together with the sentence negation the meaning of universal negation results.
In most of the cases of (492), the negative indefinite cooccurs with verbal
negation (i.e. belongs to type NV-NI of § 8.2.2).xiv Thus, in these cases there can
be no question of 'negative attraction and incorporation'.
But why do some languages use non-negative scalar focus particles (§ 7.1),
whereas others use negative scalar focus particles to form their negative
 265

indefinites? A first answer is: Because different languages use different focus
particles in contexts like (493), where no indefinite pronouns are involved. An
interesting contrast is provided by the closely related languages Finnish and
Estonian. Finnish has two different focus particles for ‘even’ (-kin) and ‘not
even’ (-kAAn) (cf. 494a-b), whereas Estonian uses the same focus particle -gi/-ki
‘even’ (the cognate of Finnish -kin) in both contexts (cf. 495a-b).

(494) Finnish
a. Se voi maksaa 100-kin markkaa.
that can cost 100-even marks
‘That can cost even 100 marks.’
b. Hän ei edes tahtonut kuulla-kaan mitä minulla olisi ollut sanomista.
he NEG even wanted listen-even what [Link] was been [Link]
‘He didn’t even want to listen to what I had to say.’
(495) Estonian
a. Lapsed-ki teavad seda.
children-even know it
‘Even children know it.’
b. Ta ei julge piuksatada-gi.
he NEG dare [Link]-even
‘He doesn’t dare to even make a sound.’

This difference correlates with a difference in the use of the corresponding

indefinite series: In Finnish, the -kAAn-series is used in the direct-negation
function, whereas Estonian uses its -gi-series there:

(496) Finnish
Hän ei palannut milloin-kaan.
she NEG:3SG returned when-INDEF
‘She never returned.’
(497) Estonian
Ta ei kohanud keda-gi.
she NEG met whom-INDEF
‘She did not meet anybody.’

Focus particles and indefinites based on them also show parallels with re-
spect to cooccurrence with verbal negation. In Greek, neither oudeís (mē deís)
‘nobody’ etc. nor oudé (mē dé) ‘not even’ cooccurred with verbal negation in the
classical period, e.g.

(498) Classical Greek (Herodotus)

a. speírousi dè oudén, all’ apò ktē néō n zō ousi kaì ikhthúō n
they:sow PT nothing but from cattle they:live and fish
‘They sow nothing, but live on cattle and fish.’ (Hdt. 1.216.3)
b. hoi dè dē hirées kuámous oudè horôntes anékhontai
the PT PT priests beans [Link] looking tolerate
‘The priests do not tolerate even to look at beans.’ (Hdt 2.37.5)

Thus, the problem has been pushed back somewhat: A question about
indefinite series has been reduced to a question about focus particles, but no
definitive solution has been found. A detailed investigation of the typology of
scalar focus particles is beyond the scope of this work. But I would like to
present some initial evidence showing that such a typology would apparently
 266

look intriguingly similar to the typology of 'scalar indefinites' (i.e. indefinites

that express a low endpoint on a pragmatic scale) proposed here.
Consider only the systems of Russian, German, and English. In the (a), (b),
(c), and (d) contexts of (499-501), German uses four different focus particles,
Russian uses three different focus particles, and English uses only two different
focus particles.

(499) German
a. Sie hat sogar die Anmerkungen gelesen.
she has even the notes read
b. Wenn sie wenigstens/ auch nur die Schlagzeilen gelesen hat,
if she [Link] [Link] the headlines read has
weiss sie Bescheid
knows she knowledge
c. Sie hat nicht einmal eine Seite gelesen.
she has not even a page read
d. Lies wenigstens/ *auch nur die Sschlagzeilen!
read [Link] [Link] the headlines
(500) Russian
a. Ona proč itala daž e snoski.
she readeven headlines
b. Esli ona proč itala xot’ zagolovki, ona znaet ob ètom.
if she read even headlines she knows about it
c. Ona ne č itala ni stranicy.
she not read [Link] page
d. Proč itaj xot’ zagolovki!
read [Link] headlines
(501) English
a. She even read the notes.
b. If she even/at least read the headlines, she is informed.
c. She hasn’t even read a page.
d. At least read the headlines!

In order to describe the cross-linguistic variation among German, Russian, and

English, one might propose the semantic map in Figure 25.

Figure 25. Semantic map for scalar focus particles

(c) negative
(d) imperative (b) conditional
(a) assertive

The distribution of German, Russian and English focus particles over this map
is shown in Figure 26a-c.

Figure 26.
a. German b. Russian c. English
negative negative negative
imper. conditional imper. conditional imper. conditional
assertive assertive assertive

Although this map is very preliminary and needs a lot of cross-linguistic testing
before it can be considered universally valid, the similarity to the implicational
map for indefinite pronouns is quite striking. The similarity exists both at the
universal level (where Figure 25 shows obvious paralles with the map in Figure
 267

8 in chapter 4), and apparently also at the level of individual languages.

Especially these latter similarities are surprising, because they are not restricted
to cases where an indefinite series is marked by a focus particle (like the
Russian ni-series, whose parallel behavior to the focus particle ni is expected).
In particular, there is an intriguing parallel between English even and any-, and
between at least and some-, between German nicht einmal and the n-series
(niemand, nichts, etc.), and between Russian xot’ and the -nibud’-series. If these
parallels are not accidental (as they may well turn out to be after more
languages are examined), there must be higher-level regularities that have to be
captured somehow. I cannot pursue these matters any further in the present
work and leave them as a challenge for future typological research.

8.3.2. Minimal-unit and maximal-unit expressions

In the preceding section we saw that scalarity is an important starting point for
the creation of negative indefinite pronouns. In addition to scalar focus
particles, in which the semantic components of scalarity and a scalar endpoint
are inherent, languages grammaticalize minimal-unit and maximal-unit
expressions, like ‘single’ or ‘eternity’. These expressions inherently denote an
endpoint, and they give rise to scalar implications when they are used non-
specifically and in the right environment (in particular, in scale-reversing
contexts, cf. § 5.5.3). Such minimal-unit and maximal-unit expressions are in
principle possible in all negative-polarity contexts, but there seems to be a
general tendency for them to become restricted to negative contexts (cf.
Ladusaw 1993).

[Link]. Minimal-unit expressions. The simplest case are minimal-unit nouns

that originally refer to small things that are used metonymically, such as ‘trace’,
‘jot’, ‘bit’, ‘tittle’, ‘atom’. For example, the Irish nouns dath ‘trace’ and dada ‘iota,
jot, tittle’ are commonly used in negative-polarity contexts with the meaning
‘anything’.

(502) Irish (Ó Dónaill (ed.) 1977, de Bhaldraite (ed.) 1959)

a. An bhfuil dada/ a dhath le rá agat?
Q is jot ART trace to say at:you
‘Do you have anything to say?’
b. Má tá a dhath bainne agat,...
if is ART trace milk at:you
‘If you have any milk,...’
c. Má bhíonn dada ag cur ort,...
if is jot on troubling on:you
‘If anything troubles you,...’
d. Ní bhfuaireas dada/ a dhath.
NEG I:got jot ART trace
‘I got nothing.’

In English, sentences like I didn’t get a bit are very emphatic, but in Irish the
above expressions have lost their emphatic flavor.
Such metonymic expressions for humans are also not uncommon, sometimes
containing the attribute ‘living’, e.g. English a living soul, French âme qui vive (lit.
‘a soul that lives’), Irish aon duine beo or aon duine cruthaithe (lit. ‘a living man, a
created man’). Spanish nada ‘nothing’ derives from Latin (res) nata ‘born
(thing)’. Unlike the Irish expressions in (502), Spanish nada has become virtually
 268

restricted to the negative functions (although it is not excluded from the

comparative and question functions yet), e.g. No vi nada ‘I didn’t see anything’,
Nada vi! ‘Nothing I saw!’, ¿Qué viste? — Nada! ‘What did you see? — Nothing!’
Besides such colorful metonymic expressions, the ordinary generic nouns
may also function as minimal-unit expressions and become restricted to
negative-polarity contexts. This has happened, for example, in the cases in
(503).

(503) generic nouns as scalar-endpoint indefinites

French personne ‘anybody < person’ rien ‘anything < thing’
Catalan res ‘anything < thing’ enlloc ‘anywhere < place’
Hebrew iš ‘anybody < man’ davar ‘anything < thing’
Maltese xejn ‘anything < thing’ imkien ‘anywhere < place’
Lezgian sa kas-ni ‘nobody < even one person’ sa zat’-ni ‘nothing < even one thing’

Generic nouns are 'minimal' in a somewhat different sense than the ‘small
things’. In sentences like I will not yield an inch, a scalar-endpoint interpretation
is so readily available because it is so unlikely that the speaker should mean the
literal interpretation. Utterances like I will not yield an inch, I will yield as much as
two centimeters are theoretically possible, but are rarely needed for pragmatic
purposes. By contrast, sentences like I did not do a thing do not describe unlikely
situations, but they are simply uninformative unless they are given a scalar-
endpoint interpretation. Thus, the fact that a scalar-endpoint interpretation is
often conventionalized with them can be attributed to pragmatic strengthening
(cf. Traugott 1988).
In French, Catalan and Hebrew the indefinites in (503) may be used not only
negatively, but also in questions and conditionals. However, in elliptical
contexts the French and the Catalan indefinites all can do without a negative
element, and their use in questions and conditionals appears to be [Link]
In Maltese the only non-negative use is the comparative use, and Lezgian sa
kas-ni and sa zat’-ni are always negative.
A third type of minimal-unit expression is the numeral ‘one’ and its em-
phatic variants like ‘single’. Examples of languages where this has been
conventionalized for negative-polarity environments are given in (504).

(504) ‘one’ as a scalar-endpoint indefinite

English any < Old English ænig, based on an ‘one’ + -ig
Latin ullus‘any’ < un-elos, based on unus ‘one’
Irish aon ‘any’ < ‘one’
Maltese Ìadd ‘anyone’ < ‘one’
Persian hič ‘any-, no-’ < Old Persian aiva ‘one’ + č iy
(emphatic particle)

While the English, Latin and Irish indefinites are also used in questions,
conditionals and in part even in the free-choice function, Maltese Ìadd (like xejn
and imkien of (503)) and Persian hič are practically restricted to the negative
functions.
It should be noted that these three types (‘small things’, generic nouns and
‘one’) are not incompatible with each other and with scalar focus particles, but
can be combined freely. Thus, Lezgian sa zat’-ni ‘nothing’ contains a scalar
focus particle (-ni), the numeral ‘one’ (sa), and a generic noun (zat’ ‘thing’).
 269

[Link]. Maximal-unit expressions. Finally, we have to consider maximal-unit

expressions like ‘in the world’, ‘in eternity’, ‘in my life’. Such expressions are
used as indefinites much more rarely than minimal-unit expressions, but
especially in the ontological category ‘time’ they are attested in several
languages.

(505) ‘(n)ever’ form ‘eternity/life’

Hebrew me-¿olam ‘never (in the past)’ lit. ‘from eternity’
le-¿olam ‘never (in the future’ lit. ‘until eternity’
Cairene Arabic ¿umri ‘never’ lit. ‘my life’
Spanish en mi vida ‘never’ lit. ‘in my life’
Irish go brách ‘never, (for)ever’ lit. ‘until eternity’

Indefinites that make use of the expression ‘in the world’ are rare, and I know
only one case, Irish, where ar bith ‘in the world’ serves as a regular negative-
polarity indefiniteness marker, as illustrated in (506).

(506) Irish
a. An bheiceann tú in áit ar bith é?
Q see you in place on world it
‘Can you see it anywhere?’ (< ‘in a place in the world’)
b. Má tá airgead ar bith agat,...
if is money on world at:you
‘If you have any money,...’ (< ‘money in the world’)
c. Níor labhair sé focal ar bith.
NEG said he word on world
‘He didn’t say a word.’ (< ‘a word in the world’)

Maximal-unit expressions are similar to minimal-unit expressions in that they

are not likely to be given a non-scalar interpretation because otherwise they
would not be relevant. A sentence like In my life I have been to Mekka is
tautological, whereas a sentence like Have you been to Mekka in your life?, with a
scalar interpretation, makes perfect sense.

8.3.3. Summary of diachronic sources

I have identified five main types of sources of negative indefinites:

(i) non-negative scalar focus particles (§ 7.1.1), e.g.

Selkup ämtä kuty ‘nobody’ < ‘even who’
(ii)negative scalar focus particles (§ 8.3.1), e.g.
Hungarian sem-mi ‘nothing’ < ‘not even what’
(iii) diachronic negative absorption (§ [Link]), e.g.
Yaqui ka-abe ‘nobody’ < *kaa habe ‘not anybody’
Latin nemo ‘nobody’ < *ne homo ‘not a man’
(iv) minimal-unit expressions (§ [Link]), e.g.
Irish dada ‘nothing’ < ‘tittle’
French personne ‘nobody’ < ‘person’
Maltese Ìadd ‘nobody’ < ‘one’
(v)maximal-unit expressions (§ [Link]), e.g.
Spanish en mi vida ‘never’ < ‘in my life’
 270

Thus, negative indefinites have rather heterogeneous sources. Unlike free-

choice indefinites, which can be weakened and eventually be extended to all
other functions (§ 5.4.2), negative indefinites which are used only in the
negation functions are never extended to other functions. In the next section,
apparent counterexamples to this generalization are discussed.

8.3.4. From negative to non-negative indefinite?

[Link]. A challenge to unidirectionality. In § 8.3.2. we saw how a non-negative

expression like ‘person’, ‘thing’, may under certain conditions become a
negative indefinite, as shown schematically in (507a). In this section, I will
discuss and reject the claim that the reverse of this process, shown schemat-
ically in (507b), is also a possible diachronic change. I try to show that the
proposed instances of this change can be explained differently, and that the
unidirectionality of the path of change in (507a) can be maintained.

(507) a. NEG V ... (non-NEG) indefinite —> NEG V ... NEG-indefinite

b. NEG V ... NEG-indefinite —> NEG V ... (non-NEG) indefinite

The hypothesis of a change like (507b) has been advanced occasionally for some
indefinites in some Indo-European languages, and most focefully for Semitic
languages by Faber (1988). For Indo-European languages, Brugmann (1911:
351) lists the indefinite pronouns in (508) and notes that they appear to contain
a negative marker (Proto-Indo-European *ne).

(508) Old Irish nech ‘somebody’

Welsh nep ‘anybody’
Old Church Slavonic ně -kŭ to ‘somebody’
Lithuanian nekas ‘something’
nekada ‘sometimes’

Delbrück (1893: 514-18) and Brugmann (1911) suggest that the non-negative
meaning of these indefinites resulted from a reanalysis of a negative indefinite
cooccurring with verbal negation as a non-negative indefinite. Since the two
(Pseudo-English) sentences in (509a-b) are equivalent, the (a) sentence may be
reanalyzed as (b) and vice versa.

(509) a. She not saw anything.

b. She not saw nothing.

The reanalysis from (a) to (b) is historically attested in the paradigm case of
French personne ‘person’ > ‘nobody’. Delbrück’s and Brugmann’s etymologies
for the indefinites in (508) imply the claim that this type of change is not
unidirectional, and that the reverse reanalysis from (b) to (a) is also possible.
This claim is made explicitly by Faber (1988). She discusses indefinite pro-
nouns characterized by the suffix *-ma in various Semitic languages, among
others the forms in (510). As Faber notes, they all have mainly negative-polarity
meaning.
 271

(510) Akkadian mim-ma ‘anything’

Hebrew m´?umå ‘anything’
Phoenician mn-m ‘anything’ (vowels not attested)
Syriac moto:-m ‘ever, always’
Amharic man-´m ‘anybody’
Arabic kayfa-maa ‘however’ (non-specific relative pronoun)

Faber claims that the indefiniteness marker *-ma was originally identical to the
negator *ma ‘not’:

Only after the loss of transparency of -ma in these forms, as the ma negative was
replaced in some or all contexts by other negative markers (*la?, *?al, *?ain, *?i, etc.)
did they acquire negative polarity: as, for example, *mah-ma ceased to be analyzed
as ‘nothing’, in order to avoid misunderstanding, it needed to be supplemented by
another, more transparent, negative, becoming, in effect, ‘any-thing’. (1988: 232).

She notes that if this reconstruction of the diachronic development is correct,

then the reanalysis of (509a) to (509b), well-known from French, 'cannot reflect
a universal, unidirectional process' (1988: 223).

[Link]. Unidirectionality saved. There is a theoretical reason why I am re-

luctant to accept Faber’s conclusion: Many linguistic changes have turned out
to be unidirectional, and the unidirectionality of change is one of the most
important aspects of language change from the point of view of linguistic
theory.
But more importantly, there are no actually attested parallels – all the above
cases were merely reconstructed. It is a general rule of diachronic typology that
reconstructed changes can never disprove a proposed universal – on the
contrary, the burden of proof that the proposed change is possible is on the
reconstructer. But instead of just shrugging off these examples as invalid, I will
present some specific arguments that alternative etymologies are preferable.
First, let us look at Semitic. Here the derivation of the indefiniteness marker
*-ma from ma ‘not’ is very problematic because of the word order: -ma is a
suffix, whereas negative ma always precedes the constituent it negates. In fact, I
know of no language where an indefiniteness marker that derives from an
absorbed negator (§ 8.2.3) or from an overtly negative focus particle (§ 8.3.1)
follows rather than precedes its basis. Second, the evidence from Semitic for a
Proto-Semitic negator *ma is not strong – it occurs mainly in Arabic, where it
may well be a recent innovation. By contrast, an item deriving from *ma(h)
‘what?’ is found in almost all Semitic languages, so it seems promising to ask
whether this could be a source for the indefiniteness marker *-ma. Indeed, there
are parallels for this kinds of indefiniteness marker elsewhere: e.g., Russian koe-
also means ‘which, what’. Note also that the interrogative pronoun ‘what?’ is
used as a general complementizer in some languages (e.g. Russian č to ‘what?;
that’), and that some indefiniteness markers incorporate a general complemen-
tizer (cf. § 6.2.3), which tends to follow the interrogative basis, as Semitic *-ma
does. Thus, while I cannot commit myself to a concrete proposal, it is clear that
Faber’s etymology is by no means the only option.
Of the Indo-European cases in (508), Old Church Slavonic ně -kŭ to
‘somebody’ is better explained as resulting from *ne vě kŭ to ‘I don’t know who’
(cf. § 6.2.1), for purely phonological reasons: The Delbrück-Brugmann
hypothesis cannot account for the long vowel in ně -. The Lithuanian forms like
ne-kas, ne-kada etc. most commonly occur in the reduplicative combinations kas
 272

nekas, kada nekada, etc. It is not clear to me what the function of the negation ne-
is in such combinations, but the highly expressive nature of these indefinites
makes it unlikely that they originated in [Link] Thus, I do not see a
compelling motivation to accept the possibility of a reanalysis along the lines of
(507b), and the universal unidirectionality of (507a) can be maintained.

[Link]. A Romance counterexample? Let us now briefly consider an apparently

attested case where a negative indefinite which resulted from negative
absorption or a negative focus particle is indeed extended to non-negative
functions. This is the non-negative use of some Romance negative indefinites
like Italian nessuno, niente, Spanish ninguno, Catalan ningú. In Italian and
Spanish, they seem to be restricted to comparatives and questions, whereas
Catalan ningú may also be used in conditionals, as illustrated in (511-13).

(511) Spanish
a. A Roberto le gusta la sopa de ajo más que a ningun

to Roberto [Link] pleases the soup of garlic more than to any

otro amigo mío.
other friend mine
‘Roberto likes the garlic soup more than any other friend of mine.’
b. ¿Cuándo ha venido nadie aquí?
when has come anyone here
‘When did anyone come here?’
(512) Italian
Hai visto nessuno?
have:you seen anybody
‘Did you see anybody?’
(513) Catalan (Lleó 1983:309)
Si hagués vingut ningú, t’ haurien avisat.
if had come anybody you [Link] informed
‘If anyone had come, they would have informed you.’

In Latin, n-indefinites were still restricted to the direct-negation function. So

aren’t these cases examples of just the kind of change hypothesized by
Delbrück, Brugmann, and Faber? I do not think so, for the following reason. In
these Romance languages, we are witnessing a merger of two indefinite series
of different origins. On the one hand, there are indefinites of non-negative
origin like Italian mai, Spanish nada, nadie, Catalan res, cap, mai, and on the other
hand, there are negatives inherited from Latin (or late Latin), like Italian
nessuno, Spanish ninguno, etc. These two indefinite series merged into one at
some point. Synchronically, the members of the series in (514) behave alike, and
their different origins are not relevant.

(514) negative and negative-polarity series in Romance

(° = originally negative, † = originally non-negative, p. = postposed)
Italian Spanish Catalan
person: °nessuno †nadie °ningú
thing: °niente †nada †res, †gens
place: — — †enlloc
time: †mai °nunca, †jamás †mai
determiner: °nessuno, †alcuno °ninguno, †alcuno (p.) †cap
 273

The originally non-negative members came to be used as negative indefinites

by the process described in § 8.3.2, but they retained some non-negative uses,
e.g. in questions and conditionals. As they merged with originally negative
indefinites, these non-negative uses were also extended to these new members
of the series. Thus, my claim is that uses as in (511-13) are due to analogical
pressure rather than to any kind of reanalysis.

8.4. Concluding Remarks

In this chapter I treated one functional type of indefinite pronouns, negative
indefinites, in greater detail. Although negative indefinites have received a
great deal of attention in the earlier literature, I hope to have shown that the
overall outlook of this book, in particular the typological implications and the
diachronic perspective, lead to new insights about negative indefinites.
Negative indefinites cannot be divorced from the other functions on the
implicational map that indefinites often have. Since negation is a semantic
property of the whole sentence, it is not easy to pin down indefinites as being
negative or non-negative. We saw this difficulty in the criterion of elliptical
contexts (§ 8.1.2), and it is also evident in the recent debate between Zanuttini
(1991) and Laka (1994) on the nature of Romance n-indefinites, which Zanuttini
regards as 'negative quantifiers' and Laka regards as negative-polarity items.
This controversy is quite parallel to the controversy around any, which some
analysts regard as a universal quantifier and others regard as an existential
quantifier (cf. § 5.2.1). In my view, the search for the true nature of these
indefinites is beside the point. Languages are not constrained by categories
such as 'universal quantifier', 'negative quantifier', or even 'negative polarity'.
They are only constrained by the implicational map, which requires indefinites
to be used only in adjacent functions. Of course, most of the semantic and
syntactic issues discussed in the previous literature remain relevant in one way
or another for setting up and explaining the implicational map, but the nature
of the indefinites is expressed sufficiently by their position on the map. I
invoked a number of additional principles that constrain the negative
interpretation of elliptical indefinites (§ 8.1.2) and the range of functions that
negative indefinites have depending on the cooccurrence with verbal negation
(456, 464, 486), but these principles are only strong tendencies with a
straightforward functional motivation.
The only major parameter of negative indefinite constructions that is not
acounted for by the implicational map is the cooccurrence of negative
indefinites with verbal negation. Here I identified three main types, V-NI,
(N)V-NI, NV-NI, which are not on an equal footing: The NV-NI pattern is
universally preferred because the meaning of negation (predicate denial)
should be isomorphic with its form (verbal negation), and the other two
patterns arise only when other motivations, such as grammaticalization
(reinforcement of negation) and Negative First (negative absorption), come into
play. But even then the preference for NV-NI reasserts itself, with V-NI giving
way to (N)V-NI and finally NV-NI by reintroduction of the verbal negation.
The close parallels of negative indefinites with other emphatic indefinites
were also seen in the conditions of focusing (§ 8.2.5) that some languages
impose, and in the diachronic sources of negative indefinites (§ 8.3), which do
not differ substantially from the kinds of sources that are also found in non-
negative indefinites (except for diachronic negative absorption).
274
 275

i
Kahrel distinguishes a fifth type ('existential'), which does not involve the use of
indefinite pronouns. See § 3.3.2, where such cases are briefly described.
ii
However, Greek is a weak counterexample, because its típota-series comes in two
varieties, a stressed one and an unstressed one. If típota-indefinites are stressed, they are
restricted to the direct-negation function (see (288) above). Since the indefinite is
stressed in elliptical answers, it cannot be mistaken for the unstressed variety.
iii
This is Kahrel’s (1989) term. Bernini and Ramat (1992: 109) call them 'generic
quantifiers', Dahl (1979: 105) just speaks of a 'substituted pronoun'.
iv
But standard French also allows sentences like Je n’ai vu qui que ce soit. ‘I didn’t see
anybody’ (Muller 1991: 256), where the negative interpretation cannot be due to qui
que ce soit ‘anybody’, but must be due to ne.

v
The distribution is as follows:
V-NI indefinites: German, Dutch, English (no-), Swedish (ingen), Icelandic
(enginn),
Latin, Ossetic
(N)V-NI indefinites: Portuguese (n-), Italian, Georgian
NV-NI indefinites: English (any), Swedish (någon), Icelandic (neinn), Portuguese
(qualquer), Catalan, Rumanian, and all others, except for Ossetic,
Maltese and Georgian.
vi
The close connection between the verb and sentence negation is expected if
Aristotle’s and Jespersen’s view of negation as predicate denial is adopted, as argued
extensively in Horn (1989). The Fregean view of negation as a sentence operator
contributes little to our understanding of the form of negation in natural languages.
vii
Kř íž ková (1968) does not cite Jespersen at all, and she does not cite anyone in
connection with this explanation, which suggests that she found this functional
motivation independently.
viii
Posner also hints at this possibility when she says that modern dialectal usage
'appears to have conserved Latin features', and talks about 'a pattern that survived in
certain areas' (1984: 19), although elsewhere she seems committed to the assumption
that Proto-Romance was uniformly NV-NI (1984: 1, 12-13).
ix
Tanaka (1994) puts it somewhat differently: This type is the unmarked case because
both the scope and the focus of negation must be marked. The negative element on the
verb marks the scope, and the negative indefinite marks the focus of negation.
x
A rare case of an indefinite which is limited to negative contexts but also shows the
restriction is Icelandic neinn/neitt:
(i) Ég sá ekki neinn. (ii) *Neinn sá ekki mig.
 276

I saw NEG nobody nobody saw NEG me

‘I saw nobody.’ ‘Nobody saw me.’
xi
Progovac (1994: 35) and Laka (1994: 21) attribute the ungrammaticality of (465b)
(*Anybody didn’t come) to the fact that the negative-polarity item anybody must be
licensed by being c-commanded by negation. They contrast English with
Serbian/Croatian and Basque, where the corresponding sentences are grammatical.
They explain this difference by the position of the negation: In English, negation is
postverbal and therefore does not c-command the subject, whereas in Serbian/Croatian
and Basque it is preverbal and therefore c-commands the subject. This explanation
seems to make the following prediction: Languages with postverbal negation do not
allow negative-polarity items in subject position, whereas languages with preverbal
negation allow them. This prediction is contradicted by Hausa, Swahili, and Chinese,
and is confirmed only by Swedish (French is ambiguous, because it has both preverbal
and postverbal negation). However, strictly speaking Progovac’s and Laka's explanation
fails to generalize to Swedish and Swahili, where någon and mtu (etc.) are clearly not
negative-polarity items. Thus, their approach both makes wrong predictions and is less
general than my explanation.
xii
Labov (1972: 811) notes that negative inversion depends on 'negative transfer' from
the subject indefinite to the verb: Nobody will catch us —> Nobody won’t catch us —>
Won’t nobody catch us, whereas *Will nobody catch us is excluded. This is an arbitrary
restriction in his formal framework, but in the perspective of this work it finds a natural
explanation: The Negative First Principle (cf. § [Link]) makes the latter sentence
dispreferred.
xiii
In French, which is also near the completion of Jespersen’s Cycle (but still retains
preverbal ne in its normative variety), sentences like (i) are possible, although
'l’acceptabilité de telles phrases est généralement difficile' (Muller 1991: 258). But
sentences like (ii), where verbal negation follows the negative indefinite, are still
completely impossible.
(i) Ils n’ont pas rien dit. (ii) *Personne n’est pas venu.
‘They didn’t say nothing.’ ‘Nobody didn’t come.’
xiv
In fact, the only exception is Greek, where oudeís was V-NI in the classical period,
but (N)V-NI in New Testament Greek.
xv
The cases of French personne ‘nobody’ < ‘person’ and rien ‘nothing’ < ‘thing’ are
often explained as resulting from the weakening of the preverbal negator ne, i.e. from
Jespersen’s Cycle (§ [Link]). It is said that the preverbal particle lost its negative force,
which was transferred to personne. This description is not completely wrong, but there
is no direct connection between Jespersen’s Cycle and the change from ‘person’ to
‘nobody’. Jespersen’s Cycle may happen also with other indefinites, and a generic noun
 277

may also become restricted to the negative function if the language does not undergo
Jespersen’s Cycle.
xvi
The Lithuanian indefinites of the type kada ne-kada have an intriguing parallel in
Hindi/Urdu, where koii na koii (lit. ‘somebody not somebody’) means ‘somebody’. But
they also remind one of Polish indefinites like gdzieniegdzie ‘here and there’, which can
hardly be separated from nie-indefinites, and these have a different origin (§ 6.2.1).
 293

Conclusions

This chapter summarizes the main results of this work and looks briefly at
possible further typological connections.

9.1. Summary of the Results of this Work

In this study I looked at indefinite pronouns in the world’s languages with the
goal of discovering cross-linguistic generalizations in this area of grammar and
thereby adding to our knowledge of human language in general. Although
indefinite pronouns are not a very conspicuous part of the grammars of human
languages, their study has wide ramifications in semantics, pragmatics, syntax,
morphology, and by putting special emphasis on the diachronic origin and
development of indefinites, I enlarged the scope of this work further. Since this
study is topic-oriented and encyclopedic in character, it is not easy to
summarize. Nevertheless, in the following sections I will outline the main
findings and their proposed explanations.

9.1.1. Typological generalizations about indefinite pronouns

[Link]. Formal generalizations. Once indefinite pronouns are suitably defined

(§ 2.2), the search for them in different languages shows that most languages
have indefinite pronouns of some kind, and that their shapes are fairly uniform
across languages. In particular, they are almost always of one of two types (§
3.1): either derived from interrogative pronouns by means of an indefiniteness
marker (or identical to interrogatives), or based on generic nouns like ‘person’,
‘thing’. Non-derived indefinites like Dutch iets ‘something’ are very rare.
Whenever indefinites are formally related to another class of words and one of
these classes is formally more complex, the indefinites are formally more
complex.
The indefiniteness marker is usually an uninflected particle (or a sequence of
particles), prefixed or suffixed. Usually an indefiniteness marker characterizes a
whole series of indefinite pronouns. Its position is often outside of case
inflections. Very often indefinite pronouns are formed by reduplication (§ 7.4),
and often the bare interrogatives may also be used as indefinites (§ 7.3). The
origin of the indefiniteness marker is often still transparent, although the
meaning of the indefinite pronoun can never be determined purely
compositionally. The transparency of the markers is due to the generally recent
grammaticalization: Indefinite pronouns have a relatively short lifespan and
are diachronically quite unstable. Closely related languages often show very
different indefinite pronouns and indefinite systems.

[Link]. Functional generalizations. As in other grammatical categories, lan-

guages differ not only with respect to the forms of their indefinite pronouns,
but also with respect to the number and the kinds of functional distinctions that
are expressed by them. Since variation in functional distinctions is not
unlimited, cross-linguistic comparison is possible and fruitful. For indefinite
 294

pronouns, the following distinctions are often associated with cross-linguistic

variation:

(i) known vs. unknown to the speaker (§ 3.2.4)

(ii) specific vs. non-specific (§ 3.2.3, § 5.2.2, § 5.4)
(iii) negative polarity/scale reversal (§3.2.2, § 5.5.4)
(iv) 'emphasis' (scalarity) (§ 5.7.2)
(v) free choice indefinites (§ 3.2.5, § 5.2.1, § 5.5.5)
(vi) direct and indirect negation (§ 3.2.1, chapter 8)
(vii) standard of comparison (§ 3.2.2)

No language has a special indefinite series for all these distinctions – usually an
indefinite series expresses more than one function. Whenever this is the case,
the choice of functions that can be expressed by one series is tightly constrained
by a series of implicational universals that can be summarized as an
implicational map (Figure 27). This map represents the central typological
generalization of this work.

Figure 27. The implicational map of chapter 4

(7)
direct
(4) (6) negation
question indirect
negation
(1) (2) (3)
specific specific irrealis
known unknown non-specific
(5) (8)
conditional comparative
(9)
free choice

When an indefinite series expresses two non-adjacent functions on this map,

the prediction is that it also expresses all the other functions in between these
two functions. Eventually a number of further functional distinctions will have
to be drawn (§ 4.7), but more cross-linguistic evidence is required before they
can be integrated into the map.
In addition, I have explored the connection between these functions and the
additional function of expressing negation in elliptical contexts (§ 8.1.2), and the
relation between negation on the verb and the type of indefinite pronoun used
for the direct-negation function.

[Link]. Diachronic generalizations. While the diachronic development of

indefinite pronouns is not easy to trace, I have looked at the diachronic (i.e.
etymological) sources of indefiniteness markers in a large number of languages.
As in other grammatical categories, the same types of source constructions give
rise to indefinites in language after language. The most important sources are:

(i) I don’t know who (§ 6.2.1)

(ii) who you want (§ 6.2.2)
(iii) whoever it may be (§ 6.2.3)
(iv) no matter who (§ 6.2.4)
 295

(v) even/at least someone (§ 7.1)

(vi) not even someone (§ 8.3.1)

Source constructions (i)-(iv) are subject to formal grammaticalization processes,

and all may undergo semantic change associated with grammaticalization, i.e.
desemanticization or semantic weakening (§ 6.4). Grammaticali-zation theory (§
6.3) predicts that the opposite development, semantic strengthening, is
impossible.
Indefinites can also acquire new meanings by pragmatic strengthening. In
general, specific indefinites acquire appreciative uses (§ [Link]), while non-
specific indefinites acquire depreciative uses (§ [Link]).

9.1.2. Explanations of the generalizations

In contrast to facts of individual languages, typological generalizations im-

mediately lead to the question: Why? While facts of individual languages can
be due to historical accidents and may have no further relevance, cross-lin-
guistic generalizations demand general explanations that can give us valuable
insights into the nature of human language.
Throughout this work, I have sought as general explanations for linguistic
facts as possible, i.e. preferably explanations from outside the linguistic system
proper. The explanation of the implicational map in Figure 27 (§ 5.6) consisted
in a demonstration that those functions that are next to each other on the map
are also functionally the most similar. Thus, it was shown that a correlation
between form and function exists. This involved a discussion of various
theoretical models that have been proposed for describing the functional
distinctions (§ 5.1-5). I concluded that the theories of mental spaces and scalar
implicatures are best suited for the description of the semantic distinctions. In
addition, the significance of focusing and sentence accent for indefinite
pronouns was highlighted (§ 5.7).
Functional explanations are also prominent in the chapter on negative
indefinites: Several facts relating to verbal negation are related to Jespersen’s
functionally-based Negative First principle and the general principle of
isomorphism.
The regularities of diachronic change are explained by the theory of
grammaticalization, which makes strong testable predictions about correlations
of a large number of parameters. I have shown how the evidence of indefinite
pronouns bears on a contentious issue within grammaticalization theory (§
6.4.1).

9.1.3. Remaining puzzles

I have not succeeded in finding explanations for all the generalizations that I
have found. Some puzzles that remain are: What does reduplication express in
reduplicated indefinites? (§ 7.4) What is the mechanism by which bare
interrogatives come to be used as indefinites? (§ 7.3.3) Why are bare in-
terrogatives often restricted to specific uses? (§ 7.3.2) How can the 'multiple
partitive' use of bare interrogatives be understood? Why do many languages
allow bare interrogatives here, even though bare interrogatives cannot be used
as indefinites elsewhere? (§ 7.3.4) Finally, why do indefinites show an areal
distribution of continent size? (This question will be discussed below in § 9.2.3.)
The last four questions have never been asked before to my knowledge, so the
 296

contribution of this work consists in pointing out some puzzles for our science
that future research must address.

9.2. Wider Typological and Areal Connections

9.2.1. Looking for typological correlations

The main synchronic typological generalizations of this work took the form of
universal implications among different functions of indefinite pronouns,
expressed by the implicational map. Now one might ask whether there are
perhaps some wider typological connections between other aspects of the
linguistic system and the types of indefinites that a language has. In other
words, are there other grammatical features that allow us to make predictions
about indefinite pronouns?
The answer seems to be no. Of course, the fact that I have not found any such
wider typological correlations does not prove that none exist. But it seems
unlikely for two other reasons. First, indefinite pronouns show very little
diachronic stability (as I observed in § 2.4.1), so any correlations would not
survive long. And second, there are not many good candidates for features that
might determine the indefinite pronoun system in a language. We would not
expect purely formal grammatical features like word order and
head/dependent marking to influence indefinite pronoun systems because the
most interesting variation among indefinites is in the functional distinctions of
different series of indefinites. It is imaginable that grammatical features having
to do with the semantics and pragmatics of reference might be correlated with
indefinite pronouns, for example the presence of a definite article (e.g., it might
be that languages lacking articles, like Russian, have particularly elaborate
systems of indefinite series to compensate the missing article). I have found no
evidence for any such connection. Other areas where connections might be
sought are the diachronic sources of indefiniteness markers, especially free
relative clauses and scalar focus particles (cf. § 8.3.1). However, these
correlations would be purely diachronic, because any synchronic connection
between the sources and the indefinites is usually lost soon after the indefinites
have been created.
One of the most intriguing questions regarding the formal makeup of in-
definites is the question why some languages use interrogative pronouns and
others use generic nouns as the base of indefinites. One possible explanation
would be diachronic: Languages with generic-noun-based indefinites perhaps
did not develop interrogative-based indefinites because of some properties that
did not allow the corresponding source constructions to be built. However, I
am not aware of any such restrictions – the source constructions that yield
interrogative-based indefinites (chapter 6) seem to be very common and
probably exist in the vast majority of languages. And in any event, many
languages use bare interrogatives as indefinites, and since all languages have
interrogative pronouns (Ultan 1978), any language could use them as
indefinites.
But when we look at the languages of Europe, a pattern seems to emerge.
Western languages such as western Romance, Celtic and the Germanic lan-
guages tend to show generic-noun-based indefinites, or at least mixed systems,
whereas eastern languages such as Slavic, Finno-Ugrian and Turkic languages
as a rule show interrogative-based indefinites. However, Basque is an exception
 297

in western Europe – it has only interrogative-based indefinites, like eastern

European languages. But Basque is also in other respects an atypical western
European language, especially with respect to the well-known word order
parameter. Basque shows a strong tendency for head-final word order (SOV,
genitive-noun, postpositions), whereas other western European languages
generally have head-initial word order (SVO or VSO, noun-genitive,
prepositions). So is there perhaps a correlation between word order type and
indefinite pronoun type after all?
This suspicion is confirmed when we turn to what is known about di-
achronic change in European languages. Classical Greek, Latin and Gothic had
almost only interrogative-based indefinites, but Modern Greek, modern
Romance and modern Germanic have a clear preference for non-interrogative-
based indefinites. In several cases the old interrogative-based indefinites have
been preserved (i.e. no generic nouns are used), but the synchronic connection
with the interrogatives has become obliterated, e.g. Greek típota ‘anything’ (cf. tí
‘what’), Portuguese alguém ‘someone’ (cf. quém ‘who’), Icelandic nokkur
‘anyone’ (cf. hver ‘who’), German etwas, dialectal eppes ‘something’ (cf. was
‘what’). It is also known that the older stages of these languages had much
more head-final word order than the modern stages. Thus, the diachronic
changes fit the putative connection between interrogative-based indefinites and
head-final word order, and between generic-noun-based indefinites and head-
initial order. It is not clear why such a correlation should exist, but in view of
the strong predictive power of word order in other domains of grammar, it is
not unimaginable that such a correction will turn out to be correct. This
hypothesis will be tested on a world-wide sample in the next section.

9.2.2. Indefinite pronoun type and word order in the 100-language sample

Since grammars do not normally give many details on indefinite pronouns, the
amount of information that I could get for the languages in the 100-language
sample is minimal – not more than the type of base of indefinites (generic noun,
interrogative, non-derived), whether indefiniteness markers are suffixed or
prefixed, and the basic word order in the clause, the possessive noun phrase
and the adpositional phrase. The data are listed in Appendix B.
However, this information is sufficient to test the hypothesis formulated in
the preceding section that interrogative-based indefinites are correlated with
head-final word order and that generic-noun-based indefinites are correlated
with head-initial word order. The relevant figures are given in Tables 8-10. (The
columns and rows 'other' include both languages with other types and
languages whose type is unknown. A language may belong to more than one
base type, but each language is assigned to one word order type.)

Table 8. Indefinite pronoun base and clausal word order

clausal word order
base SOV SVO V-initial other total
generic noun 12 17 5 1 35
interrogative pronoun 36 11 5 3 55
other 3 5 2 0 10
total 51 33 12 4 100

Table 9. Indefinite pronoun base and noun phrase word order

 298

noun phrase word order

base GenN NGen other total
generic noun 14 19 2 35
interrogative pronoun 29 13 13 55
other 3 4 3 10
total 46 36 18 100

Table 10. Indefinite pronoun base and adposition phrase word order
adposition phrase word order
base preposition postposition other total
generic noun 13 18 4 35

interrogative pronoun 27 14 14 55
other 2 6 2 10
total 42 38 20 100

It appears from these numbers that a correlation between word order type and
indefinite pronoun type indeed exists, even though it is not very strong: Head-
final word order seems to favor interrogative-based indefinites, whereas head-
initial word order favors generic-noun-based indefinites.
However, on closer examination this correlation turns out to be spurious.
The numbers in Tables 8-10 are not significant because the overall trend is not
replicated within areas. As Dryer (1989) has shown, areal effects may create the
appearance of a typological correlation, especially if the scale of the areal effect
is large (continent-sized). He proposes that a typological correlation is
significant only if it obtains in all the five large areas that he proposes.
By this criterion, the above correlation clearly fails to be significant, because
indefinite pronouns have a strongly areal distribution.

9.2.3. Continent-sized areas of indefinite pronoun type

The geographical map on p. 303 clearly shows what is going on: The indefinite
pronoun type shows a strong areal pattern. With few exceptions, interrogative-
based indefinites occur in the languages of Eurasia, Australia and the Americas,
while generic-noun-based indefinites occur in the languages of Africa and
Oceania. As it happens, the languages of Eurasia, Australia and the Americas
favor head-final word order, while Africa favors head-initial word order, and
this results in the apparent typological correlation. But a closer look
conclusively shows that no such correlation exists, and that the correlation is
purely areal.
First, Eurasian languages have interrogative-based indefinites also if their
word order is head-initial, like many languages in southeastern Asia (in my
sample: Chinese, Thai, Khmer, Xinh Mul, Atayal, Tagalog) and in Europe (in
my sample: Romansh, Upper Sorbian, Ingrian). On the other hand, the
languages of Oceania have generic-noun-based indefinites also if their word
order is head-final (in my sample: Takia, Amele, Kobon, Hua, Haruai). These
patterns are remarkably consistent. Exceptions exist, but do not form a pattern.i
Thus, the large scale of areal distribution of indefinite pronoun types must be
accepted as a fact, although it is a rather curious fact. As Nichols (1992) argues
persuasively, the scale of areal patterning of a feature reflects its diachronic
stability. Phenomena that show areality at the level of the continent can be
 299

assumed to be very old and diachronically stable. This is in striking contrast

with my earlier observation that indefinite pronouns are diachronically quite
unstable. Thus, we certainly would not expect to find the continent-sized areas
that my 100-language sample has demonstrated.
I must leave this puzzle as a challenge for future research.
 300

i
E.g. Amharic and Masalit have interrogative-based indefinites in Africa; Mongolian
and Welsh have generic nouns in Eurasia. But at least the case of Welsh probably
represents a subregularity: In those parts of Eurasia that are closest to Africa, i.e.
southwestern Europe, generic nouns are fairly widespread. This observation is
intriguing, because it implies the question whether this connection can be attributed to
influence from African languages (perhaps an Afro-Asiatic substratum of Celtic?).
 303

Appendix A
The data of the 40-language sample

The distributional schemas of the languages of the 40-language sample were presented in § 4.4.
In this appendix I list the indefinite pronouns of the 40 languages and give examples of the
most important uses of these indefinites.
These data are not complete and conclusive, especially for those languages for which I could
not consult with native speakers (Latin, Catalan, Serbian/Croatian, Yakut, Nanay, Hausa,
Swahili). But also the data of the 33 languages for which I consulted native speakers are
probably deficient in some respects. Clearly, one individual cannot control data from so many
languages in a perfect way. As I observed in § 2.1, typological breadth necessarily implies some
loss of depth in individual languages.
I add this appendix to my work in order to make it easier for the reader to link the abstract
typological generalizations with concrete data.

A.1. German
A.1.1. Inventory. German (Germanic, Indo-European) has three main series of indefinite
pronouns: (i) the irgend-series, (ii) the negative n-series, and (iii) the defective etwas-series.

interrogative etwas-series irgend-series n-series

person: wer jemand irgend-wer, irgend-jemand niemand
thing: was etwas irgend-was, irgend-etwas nichts
place: wo irgend-wo nirgends
time: wann irgend-wann nie
manner: wie irgend-wie (auf keine Weise)
determiner: welcher (ein) irgend-ein, irgend-welche kein

In addition, there is the determiner jeder ‘any, every’ (§ 6.5) and the time adverb je ‘ever’, which
do not belong to any of the series. Also, the bare interrogatives wer, was and wo (and marginally
wann) are used as indefinites in the colloquial language (§ 7.3.1).
A.1.2. Origins. Middle High German had three series, a non-emphatic ete-series (ete-wer
‘someone’, ete-waz ‘something’, ete-wâ ‘somewhere’, etc.; the origin of ete- is unknown), and a
negative-polarity series marked by ie (‘ever’) (ie-man ‘anyone’, cf. man ‘man’; iht ‘anything’ < ie-
wiht ‘ever-thing’; iergen ‘anywhere’ < ie- + hwar-gin ‘where-PT’). This distinction was given up,
and etwas and jemand are now in the same series. A new series was created on the basis of the
particle irgend (< iergen ‘anywhere’), combined with the bare interrogatives or with je-
mand/etwas/ein. The n-series consists of the old negator ne combined with the old ie-series. The
negative determiner kein (< dehhein) was formerly used in all negative-polarity environments
and became restricted to negation only 200 years ago. On the history of jeder, see Kolb (1983).
A.1.3. Distribution. The distribution of the three series is shown in Figure A.1. The data are
from my native speaker knowledge and observations.

Figure A.1. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

The etwas-series is possible in all non-emphatic functions. It can always be replaced by the
irgend-series, except in the specific-known function.

(A.1) specific known/unknown

a. Ich habe etwas <*irgend etwas> verloren. Rate mal, was!
‘I lost something. Guess what!’
b. Ich habe etwas/irgend etwas verloren, aber ich weiss nicht, was.
‘I lost something, but I don’t know what.’
(A.2) irrealis non-specific (imperative, ‘want’)
 304

Sie möchte jemanden/irgend jemanden mit schwarzen Haaren heiraten.

‘She wants to marry someone [non-specific/specific] with black hair.’

In the question/conditional and indirect-negation functions, a possible alternative to

irgendwann is je ‘ever’.

(A.3) question/conditional
a. Hast du (irgend) etwas gehört?
‘Did you hear anything?’
b. Ist sie irgend-wann/je in Chittagong gewesen?
‘Has she ever been to Chittagong?’
c. Wenn (irgend) jemand anruft, sag mir Bescheid!
‘If anyone calls, tell me!’
(A.4) indirect negation
Ich glaube nicht, dass hier je/irgend-wann (irgend) jemand war.
'I don't think that anyone has ever been here.'

In the comparative function, only the irgend-series and je are possible (the etwas-series would
yield a different sense). The irgend-indefinite must be stressed.

(A.5) Joan Baez sang besser als IRGEND JEMAND JE zuvor.

‘Joan Baez sang better than anyone ever before.’

In the free-choice function, the irgend-series may occur, but it must be stressed. However, jeder
‘any, every’ is generally preferred.

(A.6) free choice

Dieses Problem kann IRGEND JEMAND lösen.
‘This problem can be solved by anyone.’

The determiner jeder may be used in the free-choice, comparative and indirect-negation
functions. Jeder is also used as a distributive universal pronoun/determiner, so one might claim
that A.1.7-8 do not belong here. However, its use in the indirect-negation function indicates that
it should also be regarded as an indefinite.

(A.7) free choice

Dieses Problem kann jeder lösen.
‘This problem can be solved by anyone.’
(A.8) comparative
Joan Baez singt besser als jede andere Sängerin.
‘Joan Baez sings better than any other singer.’
(A.9) indirect negation
a. ohne jede/irgend-welche Hilfe
‘without any help’
b. Sie stritt jede/irgend-eine Beteilingung ab.
‘She denied any participation.’

Direct negation is expressed by the n-series, which does not cooccur with verbal negation.

(A.10) direct negation

a. Niemand ist gekommen.
‘Nobody has come.’
b. Ich habe niemandem (irgend) etwas gesagt.
‘I didn’t tell anything to anybody.’

A.2. Dutch
A.2.1. Inventory. Dutch (Indo-European, Germanic) has three series of indefinite pronouns: (i)
the non-emphatic iets-series, (ii) the non-specific dan ook-series, and (iii) the negative niets-series.

interrogative iets-series dan ook-series niets-series

person: wie iemand wie dan ook niemand
 305

thing: wat iets wat dan ook niets

place: waar ergens waar dan ook nergens
time: wanneer ooit wanneer dan ook nooit
manner: hoe hoe dan ook
determiner: welke een ‘one’ welke dan ook geen

A colloquial variant of iets is wat, i.e. the bare interrogative (but the other interrogatives cannot
be used as indefinites). The WH dan ook series has the alternatives WH ook and WH ook maar,
which have the same distribution.
The indefinite determiner enig ‘any’ is the only member of its series.

A.2.2. Origins. The forms of the iets-series originally all contained *ajw ‘ever’ (=German je):
iemand < *ajw + *mann ‘man’; iets < *ajw + wiht ‘thing’ (cf. older German iht); ergens < *ajw +
*hwar-gin ‘where-PT’; ooit < *ajw + jet ‘still’ (cf. English yet). The niets-series has in addition the
negator *ne. Geen goes back to Old Saxon nigên, i.e. nih ‘neither, not even’ + ên ‘one’. The
particle ook is ‘also’, dan is ‘then’. Enig is etymologically identical to English any.

A.2.3. Distribution. The distribution of the three series is shown in Figure A.2. The data are
from Paardekooper (1978, 1979), Geerts et al. (eds.) (1984) (G84), and from native speakers
(Leon Stassen, Kees Hengeveld, Jan Rijkhoff).

Figure A.2. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

The iets-series is used in all non-emphatic functions, from specific-known to indirect negation.

(A.11) specific known/unknown

Er heeft vanmorgen iemand opgebeld. (Raad eens wie./ Ik weet niet meer wie het was.)
‘Someone phoned this morning. (Guess who./ I don’t remember who it was.)’

In non-specific functions, the dan ook-series is a possible alternative.

(A.12) irrealis non-specific (imperative)

Koop iets/wat dan ook voor haar verjaardag.
‘Buy something for her birthday.’
(A.13) question/conditional
a. Zou er nog iemand/wie dan ook komen? (G84: 265)
‘Will anyone else come?’
b. Als je iets/wat dan ook ziet, waarschuw me dan.
‘If you see anything, tell me.’
(A.14) indirect negation (‘without’, ‘difficult’)
a. zonder iemand/wie dan ook
‘without anybody’
b. Het is moeilijk om iets/wat dan ook te zien.
‘It is difficult to see anything.’

In the comparative and the free-choice functions, only the dan ook-series is possible:

(A.15) comparative
De jongen loopt harder dan wie dan ook in zijn klass.
‘The boy runs faster than anyone in his class.’
(A.16) free choice
Je mag wie dan ook uitnodigen.
‘You may invite anyone.’

In the direct-negation function, niets-indefinites are used alone, without verbal negation. As in
German and English, only one negative indefinite per clause is used in the standard language
unless multiple negation is intended.
 306

(A.17) Niemand heeft opgebeld.

‘Nobody called.’

The determiner enig is used in the comparative, indirect-negation, and question/conditional

functions:

(A.18) comparative
Het is nu kouder dan in enige vorige winter. (G84: 275)
‘It is now colder than in any previous winter.’
(A.19) indirect negation
zonder enige help
‘without any help’
(A.20) question/conditional
a. Heeft hij enig succes gehad met al zijn pogingen? (G84: 275)
‘Has he had any success with all his attempts?’
b. Als hij enig succes heeft, dan komt dat door zijn goede kontakten.
‘If he has any success, that is due to his good connections.’

A.3. English
A.3.1. Inventory. English has three main series of indefinite pronouns: the non-emphatic some-
series, (ii) the emphatic any-series, and (iii) the negative no-series. These are formed by
combining the determiners some, any and no with generic nouns or interrogative pronouns.

interrogative some-series any-series no-series

person: who some-body, some-one any-body, any-one no-body, no one
thing: what some-thing any-thing no-thing
place: where some-where any-where no-where
time: when some-time any-time never
manner: how some-how any-how no way
determiner: which some any no

The indefinite ever forms a series of its own. Note that English spelling does not distinguish
between the indefinite determiner sôme [sUm] and the indefinite article sòme [sm]. The latter
only combined with mass nouns and plurals (some sand, some books).

A.3.2. Origins. On the history of English indefinites, see Einenkel (1903). Some (Old English
sum, Gothic sums, cf. Behaghel 1917) is an old indefinite adjective and has Indo-European
cognates meaning ‘one’ (e.g. Latin semel ‘once’). Any (Old English ænig) is derived from Old
English an ‘one’ plus -ig (cf. § [Link]). The uses of sum and ænig in Old English were much like
the modern English uses, so that not much can be said on the diachronic development.
However, the free-choice use of any developed only later (Einenkel 1903).

A.3.3. Distribution. The distribution of the three series is shown in Figure A.3.

Figure A.3. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

No further examples are given here, because they can be found throughout this work (cf. §
4.3.1). There is of course an extensive literature on English indefinites. Monographic treatments
are Sahlin (1979) and Tesch (1990), and the list of papers on English indefinites includes Stoffel
(1899), Bolinger (1960, 1977), Klima (1960), Lakoff (1969), Borkin (1971), Lawler (1971), Horn
(1972), Labov (1972), Warfel (1972), Ferrer (1973), Savin (1974), Fauconnier (1975a), Anthony
(1977), McCawley (1977), Ladusaw (1980), Davison (1980, 1981), Hintikka (1980, 1986), Carlson
(1980, 1981), Léonard (1980, 1983), Linebarger (1980, 1981, 1987), Aldridge (1982, Hirtle (1982,
1988, Strickland (1982, Kadmon and Landman (1993). Diachronic issues are dealt with in
Einenkel (1903), Raumolin-Brunberg (1994), Tottie (1994).
 307

A.4. Swedish
A.4.1. Inventory. Swedish (Germanic, Indo-European) has three series of indefinite pronouns,
only one of which is based on interrogatives: the non-emphatic någon-series, (ii) the negative
ingen-series, and (iii) the free-choice som helst-series.

interrogative någon-series ingen-series som helst-series

person: vem någon ingen vem som helst
thing: vad någon-ting, något ingen-ting, intet vad som helst
place: var någon-stans ingen-stans var som helst
time: när någon gång aldrig när som helst
manner: hur på något vis hur som helst
determiner: vilken någon ingen vilken som helst

The word någonsin ‘ever’ formally belongs to the någon-series, but it it used differently.

A.4.2. Origins. Någon is said to go back to *ne hwait ik hwarir ‘I don’t know who’ (cf. § [Link]).
Ingen goes back to *ain-gi- ‘one-particle’. Någon and ingen are combined with generic nouns
(ting ‘thing’, etc.) Som helst is literally ‘that (is) dearest’, so it represents the type discussed in §
5.3.2. Aldrig ‘never’ is from the dative of alder ‘age, time’ plus the emphatic particle -gi (also
found in ingen).

A.4.3. Distribution. The distribution of the three series is shown in Figure A.4. The data are
from various reference works and a native speaker (Östen Dahl). See also Thelander (1980).

Figure A.4. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

The någon-series is used in all non-emphatic functions and in the negative functions, and also in
the comparative function. WH som helst is a possible alternative only in the comparative
function.

(A.21) specific (known, unknown)

a. Jag har ännu något att tala med honom om.
I have still something to talk with him about
‘I still have something to talk to him about.’
b. Någon, jag minns inte vem, säger...
somebody I remember not who says
‘Somebody, I don’t recall who, says...’
(A.22) irrealis non-specific (future)
Någon gång får han nog erfara.
some time gets he enough experience
‘Sometime he will learn.’
(A.23) question/conditional
a. Har du hört något <*vad som helst>?
have you heard something anything
‘Did you hear anything?’
b. Om du hör någonting <*vad som helst>, väck mig.
if you hear something anything wake me
‘If you hear anything, wake me up.’
(A.24) indirect negation
Jag tror inte att någon <*vem som helst> vet det.
I think not that anyone anyone knows that
‘I don’t think that anyone knows that.’
(A.25) comparative
Den här pojken springer fortare än någon/ vem som helst
this here boy runs faster than anyone who INDEF
annan i klass-en.
 308

other in class-ART
‘This boy runs faster than anyone in his class.’

In the direct-negation function, there is competition between någon (which cooccurs with verbal
negation) and ingen (which does not cooccur with verbal negation). When the negated
indefinite occurs preverbally, ingen must be used (cf. § 8.2.5):

(A.26) direct negation

a. Ingen har sett mig. (*Någon har inte sett mig.)
none has seen me some has not seen me
‘Nobody has seen me.’
b. Jag har aldrig hört någon opera.
I have never heard any opera
‘I never heard any opera.’

When the negated indefinite occurs postverbally, någon is now generally preferred to ingen (cf.
Thelander (1980) for discussion of the various factors contributing to the choice).

(A.27) a. Jag vet inte något. b. Jag vet ingenting.

I know not anything I know nothing
‘I do not know anything.’ ‘I know nothing.’

In the free-choice function, WH som helst is used.

(A.28) free choice

Du kan fråga vem som helst.
you can ask who INDEF
‘You can ask anyone.’

The indefinite någonsin ‘ever’ is only used in negative-polarity environments:

(A.29) a. *Jag har hört det någonsin.

I have heard that ever
‘I ever heard that.’
b. Skivor är dyrare än någonsin.
records are dearer than ever
‘Records are more expensive than ever.’

A.5. Icelandic
A.5.1. Inventory. Icelandic (Germanic, Indo-European) has five main series of indefinite
pronouns, two of which are based on interrogatives: the non-emphatic ein-series, (ii) the
negative-polarity nokkur-series, (iii)-(iv) the negative enginn- and neinn-series, and (v) the free-
choice series marked by sem er.

interrogative ein-series nokkur-series enginn-series neinn-series sem er-series

person: hver ein-hver nokkur enginn neinn hver sem er
thing: hvaD eitt-hvaD nokkuD ekkert neitt hvaD sem er
place: hvar einhvers nokkurs hvergi hvar sem er
staDar staDar
time: hvenær einhvern nokkurn tíma aldrei hvenær sem er
tíma
manner: einhvern engan nein leiD
veginn vegin
determiner: einhver nokkur enginn neinn

A.5.2. Origins. Three series use generic nouns in combination with indefinite determiners for
the adverbial categories place, time and manner. Only in the sem er-series are all indefinites
based on the corresponding interrogatives. Sem is a relative marker, and er is a form of the verb
‘be’ (§ 6.2.3). The ein-series consists of ein- ‘one’ plus hver (§ 7.5.2). For nokkur, going back to Old
Norse nekkver, it has often been claimed that it goes back to *ne wait ik hwarir ‘I don’t know who’
 309

(§ [Link]). The enginn-series was originally marked by the emphatic particle -gi, which is
currently recognizable only in hver-gi: enginn < *ein-gi ‘even one (person)’; ekkert < *eitt-gi [?]
‘even one (thing)’; aldrei < *aldri-gi (aldri, dative of aldr ‘age’) ‘even at age, i.e. at any time’. Neinn
must be from the ancient negator *ne plus einn ‘one’.
A.5.3. Distribution. The distribution of the three series is shown in Figure A.5. The data are
from a native speaker (Halldór SigurDsson). See also Jónsdottir (1991).

Figure A.5. direct

indirect negation
specific specific irrealis question/ negation
known unknown non-specific conditional
comparative
free-choice

The ein-series is used in all non-emphatic functions. However, in the specific-known function a
generic noun would be more natural.

(A.30) specific known

PaD hringdi ein-hver/ maD ur í mig í morgun (en ég segi pér ekki hver).
it rang one-who man in me in morning and I say you not who
‘Someone called me up this morning (but I won’t tell you who).’
(A.31) specific unknown
Strákur-inn sá eitt-hvaD á bak viD tréD (en ég veit ekki hvaD).
boy-the saw one-what on back near tree and I know not what
‘The boy saw something behind the tree (but I don’t know what).’
(A.32) irrealis non-specific (imperative)
Kaupta eitt-hvaD / *nokkuD fyrir mig!
buy one-what anything for me
‘Buy something for me!’

In questions, the nokkur-series is a possible alternative (nokkur is used in questions when a

negative answer is expected). However, nokkur is not used in conditionals, nor is it used in the
irrealis non-specific function (see A.32).

(A.33) question/conditional
a. Sástu nokkuD / eitt-hvaD ?
saw:you anything one-what
‘Did you see anything/something?’
b. Ef ein-hver/ ??nokkur hringir, viltu pá segja aD ég sé ekki heima.
if one-who anyone rings will:you then say that I am not home
‘If anyone calls, please say I’m not at home.’

The nokkur-series is also used in the comparative and indirect-negation functions.

(A.34) comparative
Hann getur hlaupiD hraDar en nokkur/ *hver sem er/ *neinn annar í
he can run faster than anyone who that is noone other in
bekknum.
class
‘He can run faster than anybody else in his class.’
(A.35) indirect negation
Hún neitaDi aD viDurkenna nokkuD .
sherefused to accept anything
‘She refused to accept anything.’

The enginn-series is used in the direct-negation function, without cooccurring verbal negation.

(A.36) direct negation

Enginn sá mig./ PaD sá mig enginn.
nobody saw me it saw me nobody
‘Nobody saw me.’
 310

An alternative to the enginn-series is the neinn-series, which cooccurs with verbal negation.
However, it cannot precede the negation (cf. § 8.2.5, fn. 10), so in subject position only enginn is
possible. Neinn is also used in indirect negation:

(A.37) indirect negation

Ég held ekki aD neinn hafi komiD.
I think not that nobody has come
‘I don’t think that anybody has come.’

In the free-choice function, the sem er-series is used.

(A.38) free choice

Pú mátt heimsækja okkur hvenær sem er.
you may visit us when that is
‘You may visit us anytime.’

A.6. Latin
A.6.1. Inventory. Latin (Italic, Indo-European) has a rich system of five major series of in-
definite pronouns, four of which are mostly derived from interrogatives: (i) the non-emphatic
ali-series, (ii) the negative-polarity series marked by -quam, (iii-iv) the two free-choice series
marked by -vis and -libet. The negative n-series (v) is not derived from interrogatives. The bare
interrogatives are also commonly used as indefinites. In the following table, the -libet-series
is omitted because it does not differ from the -vis-series.

interrogative ali-series -quam-series n-series -vis-series

person: quis ali-quis quis-quam nemo qui-vis
thing: quid ali-quid quid-quam nihil quid-vis
place: ubi ali-cubi usquam nusquam ubi-vis
time: quando ali-quando umquam numquam —
determiner: qui ali-qui ullus nullus qui
dual determiner: uter — — neuter uter-vis

In addition, there is the specific-known determiner qui-dam ‘a certain’ (also quid-dam

‘something’, qui-dam ‘somebody’); an old-fashioned negative-polarity series marked by -
piam (quis-piam ‘anybody’, quid-piam ‘anything’, uspiam ‘anywhere’); and two series which are
most commonly used as relative pronouns in non-specific free relatives, but may also be used
as free-choice indefinites: the -cumque-series (qui-cumque ‘whoever; anyone’, ubi-cumque
‘wherever; anywhere’, etc.), and a reduplicated series (quis-quis ‘whoever; anyone’, ubi-ubi
‘wherever; anywhere’, etc.).
A.6.2. Origins. The etymologies of ali-, -dam, -piam, and -quam are uncertain. The suffix -vis is
identical to vis ‘you want’, and -libet is identical to libet ‘it pleases’. The suffix -cumque seems to
go back to *quom-que ‘anytime, ever’. The negative indefinites consist of the old negator ne plus
a general noun (*ne-homo ‘not man’ > nemo; ne hilum ‘not string’ > nihil), or plus
usquam/umquam/ullus of the -quam-series (cf. § [Link]). The origin of the roots us- and um- is also
unclear, but ullus can be traced back to an extension of unus ‘one’ (un-elos > un-lus > ullus) (cf. §
[Link]).
A.6.3. Distribution. The distribution of the five major series and the -dam-indefinites is shown
in Figure A.6. The data are from Kühner and Stegmann (1914, I.1: 633ff.), Orlandini (1981, 1983),
Hahn (1933), Serbat (1985), Mellet (1992, 1994). (For convenience, many of the examples are
from the New Testament.)

Figure A.6. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

The indefinite qui-dam is used when the referent is known to the speaker, whereas the ali-series
is used when it is not known. Some counterexamples to this generalization can apparently
 311

found in Latin texts, but I know of no better description of the distinction between ali- and -dam.
In any event, -dam is absolutely never used non-specifically.

(A.39) specific known

Magister, vidimus quem-dam in nomine tuo ejicientem daemonia.
Master we:saw whom-INDEF in name your [Link] devils
‘Master, we saw someone casting out devils in thy name.’ (NT, Mk 9.38)
(A.40) specific unknown
Tetigit me ali-quis. (NT, Lk 8.46)
touched me INDEF-who
‘Somebody hath touched me (for I perceive that virtue is gone out of me).’

The ali-series is also used non-specifically in irrealis-contexts and in questions and conditionals.

(A.41) irrealis non-specific

At ille intendebat in eos, sperans se ali-quid accepturum ab eis
but that [Link] in them hoping self INDEF-what accept:FUT from them
‘And he gave heed unto them, expecting to receive something of them.’ (NT, Act. 3.5)

However, in questions and conditionals the bare interrogatives are more common. But bare
interrogatives can be used as indefinites only when they are enclitic upon an element (such as si
‘if’, num ‘question particle’) early in the sentences as shown in (A.42-43) (a). When an indefinite
in a question or conditional cannot be enclitic, the ali-series is used (A.42-43 (b)).

(A.42) question
a. Num quid simile populus Romanus audierat? (Cic. Lael. 41)
Q what similar people Roman heard
‘Had the Roman people heard anything similar?’
b. Licet mihi loqui ali-quid ad te? (NT, Acts 21.37)
[Link] to:me to:speak INDEF-what to you
‘May I say something to you?’
(A.43) conditional
a. Si quid petieritis me in nomine meo, hoc faciam. (NT, John 14.14)
if what you:ask me in name my this I:will:do
‘If ye shall ask anything in my name, I will do it.’
b. Et si quid ali-quem defraudavi, reddo quadruplum. (NT, Lk 19.8)
and if what INDEF-whom I:deceived I:return fourfold
‘And if I have taken anything from any man by false accusation, I restore him
fourfold.’

The -quam-series is used in all negative-polarity contexts, except direct negation.

(A.44) question/conditional
a. Aut quomodo potest quis-quam intrare in domum fortis...
or how can who-INDEF enter in house of:strongman
nisi prius alligaverit fortem? (NT, Mt 12.29)
unless before he:will:bind strongman
‘Or else how can anyone enter into a strong man’s house...except he first bind the
strong man.’
b. Si quis-quam est timidus in magnis periculosis-que rebus, is ego sum.
if who-INDEF is timid in big dangerous-and things this I am
‘If anyone is timid in important and dangerous situations, it is me.’
(Cic. Fam. 6.14.1)
(A.45) comparative
Ac videte quanto taetrior hic tyrannus Syracusanis fuerit quam
and you:see [Link] more:horrible this tyrant for:Syracusans was than
quis-quam superiorum. (Cic. Verr. 4.123)
who-INDEF of:[Link]
‘And you see how much more horrible this tyrant was for the Syracusans than any
of the earlier ones.’
(A.46) indirect negation
Et non sinebat ut quis-quam transferret vas per templum.
and NEG admitted that who-INDEFtransferred vessel through temple
 312

‘And he would not admit that any man should carry any vessel through the
temple.’ (NT, Mk 11.16)

The n-indefinites occur only in the function of direct negation. As in standard English, there is
no verbal negation in such sentences. When more than one indefinite is in the scope of
negation, only the first is from the n-series; the others are from the -quam-series (again, as in
standard English).

(A.647) direct negation

a. Nemo potest duobus dominis servire. (NT, Mt 6.24)
nobody can two masters serve
‘Nobody can serve two masters.’
b. Nemini quid-quam dixerunt. (NT, Mk 16.8)
to:nobody what-INDEF they:said
‘They did not say anything to anybody.’

Here are some examples of the free-choice function, where the -vis-series (A.48a), the -libet-
series (A.48b), the -cumque-series, and the reduplicated interrogatives can be used.

(A.48) free choice

a. Mihi utrum-vis satis est. (Cic. Off. 3.33)
to:me which-INDEF enough is
‘Either is sufficient for me.’
b. Utrum-libet elige; alterum incredibile est, alterum nefarium. (Quint. 81)
which-INDEF choose one incredible is other sinful
‘Choose either: one is incredible, the other sinful.’

A.7. Portuguese
A.7.1. Inventory. Portuguese (Romance, Indo-European) has three major series of indefinite
pronouns: (i) the non-emphatic alg-series, (ii) the non-specific qualquer-series, and (iii) the
negative n-series.

interrogative alg-series qualquer-series n-series

person: quem alguém qualquer pessoa ninguém
thing: que alguma coisa, algo qualquer coisa nada
place: onde algures, em algum lugar em qualquer lugar nenhures
time: quando alguma vez em qualquer altura nunca, jamais
manner: como de algum modo de qualquer modo
determiner: que, qual algum qualquer nenhum

A.7.2. Origins. The alg-series goes back to the Latin ali-series (cf. A.6) (algum < *alic-unus, alguém
< *ali-quem, algo < *ali-quod, algures < alicubi). Qualquer is from qual ‘which’ plus quer ‘wants’ (§
6.2.2).

A.7.3. Distribution. The distribution of the three series is shown in Figure A.7. The data are
from my own observations and from a native speaker (Lourenço C. Finatti). (My data are
mostly from Brazilian Portuguese. It is possible that European Portuguese differs from this in
some respects.)

Figure A.7. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

The alg-series is used in all non-emphatic functions. In the specific-known function, only the
alg-series is possible, but in other functions the qualquer-series is a possible alternative.

(A.49) specific known

O João telefonou e disse alguma coisa/ *qualquer coisa –
 313

ART João phoned and said some thing any thing

adivinhe o que!
guess what
‘João called and said something – guess what!’
(A.50) specific unknown
a. Senti que alguém me observava.
I:felt that someone me observed
‘I felt that someone was observing me.’
b. Ivan disse qualquer coisa em russo que não entendi.
Ivan said any thing in Russian that not I:understood
‘Ivan said something in Russian that I did not understand.’
(A.51) irrealis non-specific (possibility)
Alguém/ qualquer pessoa pode passar aí e apanhar a caixa.
someone any person can pass here and take the box
‘Someone [non-specific] can come along here and take the box.’
(A.52) question/conditional
a. Você ouve alguma coisa/ qualquer coisa?
you hear some thing any thing
‘Can you hear anything?’
b. Se você ouvir alguma coisa/ qualquer coisa, acorde-me.
fi you hear some thing any thing wake-me
‘If you hear anything, wake me up.’

In the comparative and free-choice functions, only the qualquer-series is used.

(A.53) comparative
Eu amo minha filha mais do que qualquer outra pessoa.
I love my daughter more than any other person
‘I love my daughter more than anyone else.’
(A.54) free choice
Pelo Bábel ele é capaz de fazer qualquer coisa.
for:ART Babel he is capable of doing any thing
‘For Babel he is capable of doing anything.’

The qualquer-series may also be used in the negation functions (cooccurring with verbal
negation in the direct-negation function). However, the n-series is more common in both
functions. The n-series also cooccurs with verbal negation, except when an n-indefinite pre-
cedes the verb.

(A.55) indirect negation

Liliana perguntou isso sem encenar qualquer/ nenhum charme sensual.
Liliana asked that without showing any no charm sensual
‘Liliana asked this without showing any sensual charm.’
(A.56) direct negation
a. Não contém qualquer sentimento nobre.
not contains any feeling noble
‘It does not contain any noble feeling.’
b. Ninguém veio./ Não veio ninguém.
nobody came not came nobody
‘Nobody came.’

A.8. Catalan
A.8.1. Inventory. Catalan (Indo-European, Romance) has three rather heterogeneous series of
indefinite pronouns which are synchronically unrelated to interrogatives: (i) the non-negative-
polarity algun-series, (ii) the negative-polarity cap-series, and (iii) the qualsevol-series.

interrogative algun-series cap-series qualsevol-series

person: qui algú ningú qualsevol
thing: què alguna cosa res, gens qualsevol cosa
place: on en algun lloc enlloc
 314

time: quan alguna vegada mai, en ma vida

determiner: quin algun cap qualsevol

A.8.2. Origins. The algun-series is based on the determiner algun plus a general noun (cosa
‘thing’, lloc ‘place’, etc.). Algun is from late Latin alicunus ‘some(one)’ (cf. French aucun, Latin
aliquis). Ningú is from Latin nec unus ‘not even one’ (cf. § 8.3.1); res is from Latin res ‘thing’; gens
is from Latin gens ‘people’; enlloc is from in loco ‘at a place’; mai is from magis ‘more’; en ma vida
is literally ‘in my life’; cap is said to be from caput ‘head’, but the semantic development is
obscure. Qualsevol (cf. Italian qualsivoglia) is from Latin qualis ‘which, what kind’ plus vol-
‘want’.

A.8.3. Distribution. The distribution of the three series is shown in Figure A.8. The data are
from Lleó (1983) (L83), Solà (1973) (S73), Espósito (1988) (E88), and Hualde (1992) (H92).

Figure A.8. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

The algun-series is used in non-negative polarity contexts.

(A.57) specific
Alguna vegada he vist algú en algun lloc. (S73: 92)
some time I:have seen someone in some place
‘I once saw somebody somehwere.’
(A.58) irrealis non-specific (future)
Algun dia ho entendràs.
some day it you’[Link]
‘Some day you’ll understand.’

In negative-polarity contexts, the cap-series is used, but except in the direct-negation function,
the algun-series is a possible alternative.

(A.59) question/conditional
a. Hi ha res/ alguna cosa de nou? (E88: 105)
there exists anything some thing of new
‘Is there anything/something new?’
b. Si hi hagués ningú/ algú de Barcelona, li podríem demanar.
if there existed anyone someone from Barcelona him we:coulf ask
‘If there were anyone/someone from Barcelona, we could ask him.’ (H92: 163)
(A.60) indirect negation
Nega que hagui arrivat ningú/ algú. (E88: 105)
denies that have arrived anyone someone
‘She denies that anyone has arrived.’

In the direct-negation function, only the cap-series may be used. When the indefinite follows the
verb, it always cooccurs with verbal negation. When the indefinite precedes it, the prescriptive
norm also requires verbal negation, but in the colloquial language, verbal negation is always
absent (cf. Solà (1973), Espósito (1988) for discussion).

(A.61) direct negation

a. No he vist mai ningú enlloc. (S73: 93)
not I:have seen ever anyone anywhere
‘I have never seen anybody anywhere.’
b. Ningú (no) ens ha vist. (S73: 93)
nobody not us has seen
‘Nobody has seen us.’

In the comparative function, both the cap-series and the qualsevol-series are possible (José
Hualde, p.c.).
 315

(A.62) Aqui el clima es mes agradable que

here the climate is more pleasant than
en qualsevol altre lloc/ enlloc d’ Europa.
in any other place anywhere of Europe
‘Here the climate is more pleasant than anywhere else in Europe.’

An example for qualsevol in the free-choice function is the following:

(A.63) free choice

La negació expressada per l’adverbi no pot recaure no sobre tota la proposició...sinó sobre
una altra qualsevol de las sevas parts. (S73: 94-5)
‘The negation expressed by the adverb no can have scope not over the whole
sentence, but over any other of its parts.’

A.9. French
A.9.1. Inventory. French (Romance, Indo-European) has five major series of indefinite
pronouns: (i) the non-emphatic quelque-series, (ii) the negative-polarity series of personne etc.,
(iii) the free-choice series marked by n’importe, and (iv-v) the two emphatic series marked by
que ce soit and -conque.

interrogative quelque-series personne-series -conque-series

person: qui quelqu’un personne qui-conque
thing: quoi quelque chose rien
place: où quelque part nulle part
time: quand (en quelque temps) jamais
manner: comment (en quelque sorte) aucunement
determiner: quel quelque aucun quel-conque

The n’importe-series and the que ce soit series are regularly derived from the interrogatives
(n’importe qui, n’importe quoi, etc., and qui que ce soit, quoi que ce soit, etc., but note that the
determiner here is quelque ... que ce soit). In addition, there is a marginal je ne sais-series (cf. §
[Link]).
A.9.2. Origins. On the origin of the determiner quelque, see Foulet (1919) (que is the general
subordinator, § [Link]). Personne and rien go back to earlier generic nouns (Latin persona
‘person’, rem ‘thing(ACC)’). Aucun is from late Latin alicunus (cf. Portuguese algum, Catalan
algú). Nul continues the Latin negative pronoun nullus. Jamais is from jam ‘already, anymore’ +
magis ‘more’. N’importe literally means ‘it doesn’t matter’ (§ 6.2.4). The indefiniteness marker -
conque seems to come from qu’onques ‘that ever’ (§ [Link]), although some have claimed that it
goes back directly to Latin -cumque (Foulet 1919). The marker que ce soit literally means ‘that it
be’ (§ [Link]).

A.9.3. Distribution. The distribution of the three series is shown in Figure A.9. The data are
from Gaatone 1971 (G71), Culioli (1983), and my own observations.

Figure A.9. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

The quelque-series is used in all non-emphatic functions.

(A.64) specific known/unknown

a. Quelqu’un m’attend. C’est lui que j’épouserai.
‘Someone is waiting for me. It’s him that I will marry.’
b. Quelqu’un est à la porte. Je ne sais pas qui.
‘Somebody is at the door. I don’t know who.’
(A.65) irrealis-non-specific (imperative, ‘want’)
Elle veut épouser quelqu’un aux cheveux noirs.
‘She wants to marry someone with black hair.’
 316

However, as Culioli (1983) notes, the determiner quelque behaves somewhat differently from
quelqu’un etc. in that it can only be used non-specifically.
In the question and conditional functions, the personne- and que ce soit-series are also possi-
ble, but only in rhetorical questions that imply a negative answer.

(A.66) question
a. As-tu écouté quelque chose?
‘Have you heard anything?’
b. Si l’art n’était vraiment qu’un prolongement de la vie, valait-il de lui
rien sacrifier? (G71: 169)
‘If art were really only a continuation of life, would it be worth
sacrificing anything for it?’
c. Comment peut-on en conclure quoi que ce soit sur l’ordre de la pensée?
‘How can one conclude anything from it about the order of thoughts?’
(A.67) conditional
a. Je vous rends responsable si rien/quelque chose s’ébruite dans la presse.
‘I hold you responsible if anything transpires in the media.’
b. Éveillez-moi s’il arrive quoi que ce soit.
‘Wake me up if anything happens.’

The personne-series is most often used in the direct-negation function, cooccurring with the
preverbal negative particle ne, but not with the postverbal negative particle pas. Furthermore, it
is used in the indirect-negation and comparative functions. In all these functions, the que ce soit-
series is a possible alternative.

(A.68) direct negation

a. Je ne vois rien.
‘I cannot see anything.’
b. Personne n’ jamais dit rien.
‘Nobody ever said anything.’
c. Je n’ai vu qui que ce soit. (Muller 1991: 256)
‘I haven’t seen anybody.’
(A.69) indirect negation
a. Je doute que personne y réussisse.
‘I doubt that anybody will succeed in it.’
b. J’hésite à rien écrire sur cela/à écrire quoi que ce soit sur cela.
‘I hesitate to write anything about that.’
(A.70) comparative
a. Il parle mieux qu’aucun orateur.
‘He speaks better than any orator.’
b. Je préfère de beaucoup fermer les yeux que lui reprocher quoi que ce
soit qui puisse vous ennuyer.
‘I much prefer to close my eyes than to blame him for anything that could
annoy you.’

The que ce soit-series is also used in the free-choice function.

(A.71) free choice

Vous pouvez dire quoi que ce soit.
‘You can say anything.’

The range of functions of the -conque-series (i.e. quiconque and quelconque) is very similar to that
of the que ce soit-series.

(A.72) a. Demandez à quiconque des assistents! (free choice)

‘Ask any of the assistants!’
b. Je le sais mieux que quiconque. (comparative)
‘I know it better than anyone.’
c. Il n’est pas permis de parler de ceci à quiconque. (indirect negation)
‘It isn’t possible to talk about this to anyobody.’
d. Si quiconque insiste pour me parler, dites qu’on m’écrive. (conditional)
‘If anyone insists on talking to me, tell them to write to me.’
 317

The n’importe-series is used in the free-choice and comparative functions.

(A.73) a. Tu peux dire n’importe quoi.

‘You can say anything.’
b. Le général avait compris assez tôt que cette guerre, plus que n’importe quelle guerre
dans le passé, était une guerre de réserves d’hommes et de matériel.
‘The general had understood pretty soon that this war, more than any war in
the past, was a war of human and material resources.’

A.10. Italian
A.10.1. Inventory. Italian (Indo-European, Romance) has a rather heterogeneous system,
consisting of three major series. Of these only one, the free-choice series in -unque, is formally
homogeneous, but it is quite incomplete. The other two are the negative nessuno-series and the
general qualche-series.

interrogative qualche-series nessuno-series -unque-series

person: chi qualcuno nessuno chiunque
thing: che qualche cosa, niente, nulla —
qualcosa
place: dove in qualche luogo in nessun luogo dovunque
time: quando qualche volta <mai> —
manner: come in qualche modo —
determiner: quale qualche nessuno qualunque

There is an important determiner qualsiasi ‘any’ which forms a series of its own. Other in-
definites are archaic and therefore marginal: two forms in -chessia (chicchessia ‘anyone’,
checchessia ‘anything’), the determiners qualsivoglia ‘any’ and veruno ‘any’, as well as the three
negative indefinites from the stem alcun- (alcuno ‘nobody’, alcunché ‘nothing’, alcuno ‘no’).

A.10.2. Origins. The etymology of Italian qualche is analogous to French quelque. Nessuno is from
the negation nec plus uno ‘one’. Nulla is from Latin nulla (res) ‘no (thing)’, and niente is probably
from nec gentem ‘not people’ (cf. the discussion in Zanuttini (1987)). Mai is from Latin magis
‘more’. The suffix -unque is identical to old Italian unque ‘ever’ (< Latin unquam ‘ever’). The
forms in -chessia are analogous to the French forms in que ce soit.

A.10.3. Distribution. The distribution of the three series is shown in Figure A.10. The data are
from various reference works (e.g. Brunet 1981), Longobardi (1989), and native speakers (Paolo
Ramat, Davide Ricca). See also Zanuttini (1987, 1991).

Figure A.10. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

The most general series is the qualche-series.

(A.74) specific known

Qualcuno è venuto – indovina chi!
‘Someone has come – guess who!’
(A.75) specific unknown
Non trovo la penna, eppure in qualche parte l’avrò messa.
‘I can’t find the pen, and yet I must have put it somewhere.’
(A.76) irrealis non-specific (imperative)
Compra qualcosa per me.
‘Buy something for your niece.’
 318

In the question (but not the conditional!) function and in indirect-negation functions, the
nessuno-series is also possible.

(A.77) question
Vedi qualcosa/niente?
‘Can you see anything?’
(A.78) conditional
Se senti qualcosa <*niente>, svegliami.
‘If you hear anything, wake me up.’
(A.79) indirect negation
Non è necessario che venga nessuno/che qualcuno venga.
‘It is not necessary that anyone come.’

The -unque-series is restricted to the free-choice and comparative functions.

(A.80) free choice

Puoi andare dovunque.
‘You can go anywhere.’
(A.81) comparative
Christie ha scritto più romanzi che chiunque altro in questo secolo.
‘Christie has written more novels than anyone else in this century.’

In the direct-negation function, only the nessuno-series is possible. When the negative pronoun
follows the verb, it cooccurs with verbal negation (except in colloquial northern Italian), but
when it precedes the verb, verbal negation is not possible.

(A.82) direct negation

a. Non ho veduto nulla.
‘I have not seen anything.’
b. Nessuno (*non) è venuto.
‘Nobody has come.’
c. Nessun professore ha scritto mai nessun libro.
‘No professor has ever written any book.’

Alcuno/alcunché may replace nessuno/niente only in the direct/indirect negation function, not in
the question function.

A.11. Romanian
A.11.1. Inventory. Romanian (Indo-European, Romance) has four major series of indefinite
pronouns, three of them derived from interrogatives: (i) the non-emphatic -va-series, (ii) the
free-choice ori-series, (iii) the negative ni-series, and (iv) the oare-series.

interrogative -va-series ori-series oare-series ni-series

person: cine cine-va ori-cine oare-cine nimeni
thing: ce ce-va ori-ce oare-ce nimic
place: unde unde-va ori-unde oare-unde nică ieri
time: cînd cînd-va ori-cînd oare-cînd niciodată , nici-cînd
manner: cum cum-va ori-cum oare-cum nici-cum
amount: cît cît-va ori-cît oare-cît
determiner: care care-va ori-care oare-care nici un

In addition, there is the indefinite determiner vre-un ‘any’ (Graur 1939).

A.11.2. Origins. The particle ori also means ‘or’ and derives from Latin velis ‘you want
(subjunctive)’. The particle oare also occurs as a question particle and as a disjunctive particle
(‘be it ... be it’); it derives from late Latin *volet ‘wants’ (classical vult) > voare > oare. The suffix -
va has the same origin (voare > vare > va), or perhaps it derives from vrea, which also means
‘want’. Some of the members of the ni-series are from nici ‘neither, not even’ plus interrogative
or general noun (dată ‘time’), but nimeni goes back directly to Latin nemo (Accusative neminem).
Nimic is from nec mica ‘not even a bit’.
 319

A.11.3. Distribution. The distribution of the three series is shown in Figure A.11. The data are
from Manoliu-Manea (1966), Berea Gă geanu (1980), Graur (1939), various reference works, the
New Testament, and a native speaker (Beatrice Primus).

Figure A.11. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

The -va-series is used in non-emphatic functions.

(A.83) specific (known, unknown)

a. Am cumpa*rat ce-va, ghici ce!
I:have bought what-INDEF guess what
'I have bought something, guess what!'
b. Cine-va te-a ca*utat la telefon, dar nu i-am înt3eles numele.
who-INDEF you-has asked on phone but not him-have understood name
'Somebody asked for you on the phone, but I didn't understand the name.'
(A.84) irrealis non-specific (‘want’)
As3 vrea ce-va mîncare s3i bă utură .
INTJ want what-INDEF [Link] and drink
‘I want something to eat and to drink.’
(A.85) question/conditional
a. Ai fost vreodata* /cînd-va la Chis3ina*u?
have been ever when-INDEF at Chis3ina*u
'Have you ever been to Chis3ina*u?'
b. Dacă auzi ce-va/ *ori-ce, trezes3te-ma*.
if [Link] what-INDEF wake-me
‘If you hear anything, wake me up.’

The ori-series is used in the comparative and free-choice functions.

(A.86) comparative
Niculina fuge mai repede decît ori-ce/ *vre-o fata* din clasa*.
Niculina runs more fast thanINDEF-what INDEF-one girl in class
'Niculina runs faster than any girl in her class.'
(A.87) free choice
Se gă seste în ori-ce pră vă lie.
REFL gets in INDEF-what shop
‘You can get it in any shop.’

The ni-series is used for direct and indirect negation. In the direct-negation function, it cooccurs
with verbal negation.

(A.88) indirect negation

Nu mai îndră zneau să -l întrebe nimic. (NT, Lk 20.40)
NEG more [Link] SBJV-him ask nothing
‘They did not dare to ask him anything anymore.’
(A.89) direct negation
Nu ne angajează nimeni. (NT, Mt 20.7)
NEG us hired nobody
‘Nobody has hired us.’

The determiner vre-un is used in questions/conditionals, and in indirect negation (cf. Graur
(1939) for discussion).

(A.90) question
Ai vre-un prieten care te-ar putea ajuta?
[Link] INDEF-one friend who you-would [Link] help
‘Do you have any friend who could help you?’
 320

(A.11.8) indirect negation

Nu este probabil să aibă loc nici o/ vre-o nouă
NEG is probably SBJV have place NEG one INDEF-one new
întrevedere cu d. Molotov. (Graur 1939: 174)
interview with Mr. Molotov
‘It’s unlikely that there will be any new interview with Mr. Molotov.’

A.12. Modern Greek

A.12.1. Inventory. Modern Greek (Greek, Indo-European) has three main series of indefinite
pronouns. One of these is based on interrogative pronouns (the specific ka-series), one is based
on relative pronouns (the free-choice series marked by -dhípote), and one is formally very
heterogeneous (the non-specific típota-series).

interrogative ka-series típota-series -dhípote-series

person: pjos ká-pjos kanénas, kanís opjos-dhípote
thing: ti ká-ti típota oti-dhípote
place: pu ká-pu puthená opu-dhípote
time: póte ká-pote poté opote-dhípote
manner: pos ká-pos (me kanénan trópo)

A.12.2. Origins. The prefix ka- comes from kan ‘even, at least’ (< kaì eán ‘even if’). The típota-
series is quite diverse not only synchronically, but also in terms of its origins: Kanénas consists
of kan plus énas ‘one’; the suffix -pota in típota probably goes back to poté ‘ever, never’; poté
survives from Ancient Greek, where indefinites were identical to interrogatives except that they
were unstressed or had final stress. Likewise, puthená is said to go back to póthen ‘whence’. The
-dhípote-series is not originally part of the Demotic (i.e. vernacular) language – it came into the
modern standard language from the older archaizing Katharévousa variety and can effectively
be regarded as a borrowing from Ancient Greek dē pote (consisting of the ' emphatic' particle
dē plus poté ‘ever’).

A.12.3. Distribution. The distribution of the three series is shown in Figure A.12. The data are
from my own observations, Veloudis (1982), Dhelverúdhi (1989), Giannakidou (1993, 1994a, b,
c), Tsimpli and Roussou (1993) and from native speakers (Anastasia Christofidou, Soteria
Svorou, Axel Theofilidis, Artemis Alexiadou).

Figure A.12. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

The ka-series is used in all non-emphatic functions (except apparently ‘specific known’).
However, in the non-specific functions (irrealis, question, conditional) the típota-series is
preferred.

(A.91) specific known/unknown

a. *Ká-pjos tilefónise. Mándepse pjos!
INDEF-who phoned guess:IMPV who
‘Someone called. Guess who!’
b. Ká-pjos tilefónise. Dhen kséro pjos.
INDEF-who phoned not I:know who
‘Someone called. I don’t know who.’
(A.92) irrealis non-specific (imperative, ‘want’)
a. Fére típota/ ká-ti na fáme!
bring:IMPV anything INDEF-what SBJV we:eat
‘Bring something to eat!’
b. Théli na pandreftí kanénan/ ká-pjon pu na milái ghaliká.
wants SBJV she:marry anyone INDEF-whom who SBJV speak French
‘She wants to marry someone [non-specific] who speaks French.’
(A.93) question/conditional
 321

a. Ídhes típota/ ká-ti?

you:saw anything INDEF-what
‘Did you see anything/something?’
b. An dhís típota/ ká-ti, pes mu.
if you:see anything INDEF-what say:IMPV me
‘If you see anything/something, tell me.’

The típota-series is also used in the negation functions, cooccurring with verbal negation in the
case of direct negation. See § 5.7.3 and § 8.2.5 for the role of stress.

(A.94) direct negation

Kanís dhen mu ípe típota.
anyone not me said anything
‘Nobody told me anything.’
(A.95) indirect negation
Dhen nomízo óti írthe kanís.
not I:think that came anyone
‘I don’t think that anyone came.’

The -dhípote-series is used in the free-choice function and in the comparative function.

(A.96) free choice

Opjos-dhípote borí na lísi aftó to próvlima.
who-INDEF can SBJV solve this the problem
‘Anyone can solve this problem.’
(A.97) comparative
To aghóri borí na tréksi ghrighorótera apó opjon-dhípote sto sxolío tu.
theboy can SBJV he:run faster from whom-INDEF in:the school his
‘The boy can run faster than anyone in his school.’

The -dhípote-series is also possible in the conditional function with an emphatic value, but not in
the question function.

(A.98) question/conditional
a. *Ídhes oti-dhípote?
you:saw what-INDEF
‘Did you see anything at all?’
b. An dhis oti-dhípote, pes mu.
if you:see what-INDEF say:IMPV me
‘If you see anything at all, tell me.’

A.13. Bulgarian
A.13.1. Inventory. Bulgarian (Indo-European, Slavic) has three major series of indefinite
pronouns, all of them derived from interrogatives: (i) the general nja-series, (ii) the negative ni-
series, and (iii) the free choice series in -to i da e.

interrogative nja-series ni-series -to i da e-series

person: koj nja-koj ni-koj koj to i da e
thing: š to ne-š to ni-š to š to to i da e
property: kakâv nja-kakâv ni-kakâv kakâv to i da e
place: kâde nja-kâde ni-kâde kâde to i da e
time: koga nja-koga ni-koga koga to i da e
manner: kak nja-kak ni-kak kak to i da e
amount: kolko nja-kolko ni-kolko kolko to i da e

The nja-series has an alternative form where the particle si follows the base (njakoj si, neš to si,
etc.). The -to i da e-series has the more emphatic alternative form -to i da bilo (kojto i da bilo, etc.).
Marginal series are the ne znam si-series (ne znam si koj ‘someone’, lit. ‘I don’t know who’), the -
gode-series (koj-gode ‘anyone’, etc.), and the -da e-series (koj da e, etc.).
A.13.2. Origins. The nja- and ni-series go back to Old Church Slavonic ně - (cf. § 6.2.1) and ni-
(cf. § 8.3.1). The -to i da e-series originates in a parametric concessive conditional clause (cf. Kojto
 322

i da e, mož e da vleze. ‘Whoever she is, she may come in.’). The particle si is perhaps cognate with
Polish -ś (cf. A.15) .
A.13.3. Distribution. The distribution of the three series is shown in Figure A.13. The data are
from Blaž ev (1980), Donč eva (1970) (D70), Dogramadž ieva (1974), Kirova (1986), Mostovska
(1988), Nicolova (1972, 1978, 1986), Guentchéva (1981), and from native speakers (Tania Kuteva,
Pavlina Vlajkova). See also Paš ov (1965), Xaralampiev (1977).

Figure A.13. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

The nja-series can be specific or non-specific:

(A.99) specific unknown

Govorix vč era s nja-koga v avtobusa.
I:talked yesterday with INDEF-whom in the:bus
‘Yesterday I talked to someone on the bus.’
(A.100) specific known
Nja-koj se obadi, otgatni koj.
INDEF-who REFL called guess:IMPV who
‘Somebody has called, guess who.’
(A.101) irrealis non-specific (‘want’)
Kogato edin č ovek iska da vârš i ne-š to – daj mu rabota.
when one person wants SBJV he:do INDEF-what give:IMPV him work
‘When a person wants to do something – give him or her work.’ (D70: 445)

In conditionals, but not in questions, the more emphatic -to i da e-series may alternatively be
used.

(A.102) question/conditional
a. Vidja li ne-š to <*kakvo-to i da e>?
you:saw Q INDEF-what what-INDEF
‘Did you see anything (at all)?’
b. Ako vidiš ne-š to/ kakvo-to i da e, kaž i mi.
if you:see INDEF-what what-INDEF tell:IMPV me
‘If you see anything/anything at all, tell me.’

The -to i da e-series is also used in the free-choice, comparative, and indirect-negation functions.

(A.103) free choice

Mož eš da vzemeš koj-to i da e moliv.
you:can SBJV you:take which-INDEF pencil
‘You can take any pencil.’
(A.104) comparative
Na mene poveč e, otkolkoto i na kogo-to i da bilo
to me more than even to whom-INDEF
drug mi e č už d pokaznijat patos. (D70: 449)
other me is alien ostentatious pathos
‘Ostentatious pathos is more alien to me than to anyone else.’
(A.105) indirect negation
Momč e-to go napravi bez kakva-to i da e/ ni-kakva pomoš t.
boy-the it did without which-INDEF NEG-which help
‘The boy did it without any help.’

The ni-series is used in the direct-negation function (cf. A.106), and also in some cases in the
indirect-negation function (cf. A.105).

(A.106) direct negation

Ni-koj ne mi kaza ni-š to.
NEG-who not me told NEG-what
 323

‘Nobody told me anything.’

A.14. Serbian/Croatian
A.14.1. Inventory. Serbian/Croatian (Slavic, Indo-European) has four major series of indefinite
pronouns, all of which are derived from interrogatives: (i) the non-emphatic ne-series, (ii) the
negative-polarity series marked by i-, (iii) the free-choice and negative polarity series marked
by bilo, and (iv) the negative ni-series.

interrogative ne-series i-series ni-series bilo-series

person: ko ne-ko i-tko ni-(t)ko ko bilo
thing: š to ne-š to i-š ta ni-š to š to bilo
property: kakav ne-kakav i-kakav ni-kakav kakav bilo
place: gdje ne-gdje i-gdje ni-gdje gdje bilo
time: kada ne-kada, ne-gda i-kad(a), i-gda ni-kada kada bilo
manner: kako ne-kako kako ni-kako kako bilo
determiner: koji ne-koji, ne-ki koji ni-koji koji bilo

A.14.2. Origins. The prefix ne- has the same origin as Russian ne-, Bulgarian nja- and Polish nie-
(§ 6.2.1). The prefix i- is from the conjunction and focus particle i ‘and; also; even’ (§ 7.1.1). For
ni-, see § 8.3.1 (cf. Russian ni-). The particle bilo is the perfect tense of ‘be’, so the bilo-series is an
example of an ‘it-may-be’ indefinite (§ 6.2.3).
A.14.3. Distribution. The distribution of the three series is shown in Figure A.14. The data are
from Progovac (1988, 1990, 1991, 1992, 1994) and reference works.

Figure A.14. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

The ne-series is used in the specific and irrealis non-specific functions.

(A.107) specific
Mira voli ne-koga. (P91: 569)
Mira loves INDEF-whom
‘Mira loves someone.’
(A.108) irrealis non-specific
Daj mi ne-š to!
give me INDEF-what
‘Give me something!’

The i-series is used in negative-polarity contexts, i.e. in the question/conditional, indirect-

negation and comparative functions. In all these functions, the bilo-series is a possible
alternative.

(A.109) question/conditional
a. Da li je i-ko/ bilo ko doš ao? (P90: 130-1)
Q Q AUX INDEF-who INDEF who come
‘Did anyone come?’
b. Ako je i-ko/ bilo ko doš ao, poč n-i-te! (P90: 130)
if AUX INDEF-who INDEF who come start-IMPV-2PL
‘If anyone has come, start!’
c. Jesi li ne-koga naš ao?
[Link] Q INDEF-whom found
‘Did you find anyone?’
(A.110) indirect negation
Milan ne tvrdi da je i-ko/ bilo ko/ *ni-ko doš ao.
Milan NEG claims that AUX INDEF-who INDEF who INDEF-who come
‘Milan does not claim that anyone has come.’ (P90:130-1)
(A.111) comparative (cf. § 4.7.1)
 324

Marija je pametnija {nego i-ko/ od bilo koga} u razredu.

Marija is smarter than INDEF-who from INDEF who:GEN in class
‘Marija is smarter than anyone in her class.’

The bilo-series is also used in the free-choice function.

(A.112) free choice

Bilo ko mož e istuć i malo dete. (P90: 130)
INDEF who can beat small child
‘Anyone can beat a small child.’

In the direct-negation function, only the ni-series is used, cooccurring with verbal negation.

(A.113) direct negation

Milan ne voli ni-koga <*i-koga>. (P91: 568)
Milan NEG loves INDEF-whom INDEF-whom
‘Milan does not love anyone.’

A.15. Polish
A.15.1. Inventory. Polish (Indo-European, Slavic) has three major series of indefinite pronouns,
all derived from interrogatives: (i) the general ś -series, (ii) the free-choice series in -kolwiek, and
(iii) the negative ni-series. The marginal nie-series is also shown in the following table:

interrogative ś -series -kolwiek-series ni-series nie-series

person: kto kto-ś kto-kolwiek ni-kto —
thing: co co-ś co-kolwiek nic nie-co
quality: jaki jaki-ś jaki-kolwiek ni-jaki nie-jaki
place: gdzie gdzie-ś gdzie-kolwiek ni-gdzie —
time: kiedy kiedy-ś kiedy-kolwiek ni-gdy nie-kiedy
manner: jak jako-ś jak-kolwiek ni-jak (niejako)
determiner: który który-ś który-kolwiek z%aden nie-który

The three major series are quite regular, except for a few phonological alternations (jak : jako-ś ,
kiedy : ni-gdy) and the negative determiner z%aden (from a different root). The nie-series is
incomplete. A variant of it is the X-nie-X-series: gdzie-nie-gdzie ‘here and there’, kiedy nie-kiedy
‘once in a while’, co nie-co ‘a little something’ (§ [Link]). There are three further marginal free-
choice series, formed with byle (byle kto ‘anyone’, etc.), lada (lada kto ‘anyone’), and (nie)
ba4dź (kto (nie) ba4dź ‘anyone’, etc.) . Finally, the bare interrogatives are commonly used as
indefinites in the colloquial language (cf. § 7.3.2).
A.15.2. Origins. The suffix -ś (< -si, cf. Czech -si) perhaps goes back to *sit ‘(it) be’ (cf. § 6.2.3).
The origin of the suffix -kolwiek (cf. Slovak -kol’vek, Czech -koli(v(ě k)), Sorbian -kuli) is not clear
(cf. Cieś likowa (1965: 80-3) for some discussion), but a plausible possibility is that koli is
identical to the older interrogative-indefinite pronominal adverb koli ‘when?; ever’, and -wiek is
derived from wiek ‘age’. The prefix nie- is identical to Russian ne-, Bulgarian nja- (cf. § 6.2.1), and
(nie) ba4dź is identical to Russian -nibud’. Byle (also ‘if only’) seems to be connected with the
modal hypothetical particle by (originally a subjunctive form of być ‘be’), and the origin of lada
is unknown.
A.15.3. Distribution. The distribution of the three series is shown in Figure A.15. The data are
from Grzegorczykowa (1972a) (G72), various reference works, and a native speaker (Thomas
Bak). See also Grzegorczykowa (1972b) and Mostovska (1988).

Figure A.15. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

The ś -series is used in all functions from ‘specific-known’ to ‘indirect negation’.

 325

(A.114) specific known/unknown

Kto-ś do ciebie zadzwonil, [zgadnij kto/ ale nie zrozumialam nazwisko].
who-INDEF to you called guess who but not I:understood name
‘Someone called you, [guess who/but I didn’t understand the name].’
(A.115) irrealis non-specific (possibility)
Zawsze moz5na sie czego-ś nauczyć .
always [Link] REFL what-INDEF teach
‘One can always learn something.’

In the question/conditional and in the indirect-negation functions, the -kolwiek-series is a

possible more emphatic alternative (except in questions, where it sounds a bit odd).

(A.116) question/conditional
a. Widzisz kogo-ś <?kogo-kolwiek>?
you:see whom-INDEF whom-INDEF
‘Can you see anybody?’
b. Jezeli co-ś / co-kolwiek zobaczysz, odrazu mnie obudź .
if what-INDEF what-INDEF you:see immediately me wake:IMPV
‘If you see anything, wake me up immediately.’
(A.117) indirect negation
Nie sadze4, ze kto-ś / kto-kolwiek przyszedl.
not I:think that who-INDEF who-INDEF came
‘I don’t think that anybody came.’

The -kolwiek-series is most typically used in the free-choice function, as well as the comparative
function.

(A.118) free choice

Moz%esz mnie odwiedzić o której-kolwiek godzinie.
you:can me visit at which-INDEF hour
‘You can visit me at any time.’
(A.119) Graz%yna napisala wie4cej ksia4z%ek niz% jaki-kolwiek inny autor.
Graz%yna wrote more books than which-INDEF other author
‘Grazyna has written more books than any other author.’

The ni-series occurs only in the direct-negation function, cooccurring with verbal negation.

(A.120) Ni-kt ni-c nie wie.

NEG-who NEG-what not knows
‘Nobody knows anything.’

A.16. Russian
A.16.1. Inventory. Russian has seven major series of indefinite pronouns: (i-ii) the specific to-
and koe-series, (iii)-(iv) then non-specific -nibud’- and -libo-series, (v) the negative ni-series, (vi)
the negative-polarity series marked by WH by to ni bylo, and (vii) the free-choice series marked
by ugodno. In the following table, the by to ni bylo-series is omitted for lack of space.

interrogative -to-series -nibud’-series -libo-series ni-series koe-series

person: kto kto-to kto-nibud’ kto-libo ni-kto koe-kto
thing: č to č to-to č to-nibud’ č to-libo ni-č to koe-č to
place: gde gde-to gde-nibud’ gde-libo ni-gde koe-gde
time: kogda kogda-to kogda-nibud’ kogda-libo ni-kogda koe-kogda
manner: kak kak-to kak-nibud’ kak-libo ni-kak koe-kak
amount: skol’ko skol’ko-to skol’ko-nibud’ — ni-skol’ko —
determiner: kakoj kakoj-to kakoj-nibud’ kakoj-libo ni-kakoj koe-kakoj

In addition, there are some forms of the marginal ne-series (ne-kto ‘somebody’, ne-č to
‘something’, ne-kogda ‘once’, ne-skol’ko ‘several’), and some little grammaticalized series (xot’ kto
‘anyone’; Bog vest’ kto ‘God knows who’). The determiner ljuboj ‘any’ is the only member of its
series.
 326

A.16.2. Origins. On the diachrony of Russian indefinite pronouns, see especially Malovickij
(1971: 60-111). The origin of -to is unclear, but it is probably identical with to ‘now’ (< Old
Russian tŭ , ≠ to ‘that’). The suffix -nibud’ goes back to ni budi ‘it may be’ (§ [Link]), and -libo
(Old Russian ljubo) comes from ljubo ‘dear, pleasant’ (§ [Link]). Ni means ‘neither, not even’ (§
8.3.1). Koe is the neuter form of koj ‘which?’ – why it is used as an indefiniteness marker is not
clear to me. Ugodno is from ugodno ‘suitable, pleasing’ (§ [Link]), and by to ni bylo is a more
modern form of ni budi (§ [Link]).
A.16.3. Distribution. The distribution of the three series is shown in Figure A.16. The data are
from my own observations, from native speakers (especially Maria Koptjevskaja-Tamm, Vera
Podlesskaya, Elena Kalinina), and from the huge literature on Russian indefinites. Some refer-
ences are Galkina-Fedoruk (1963), Veyrenc (1964), Seliverstova (1964, 1988), Rybák (1965, 1975),
Dahl (1970), Malovickij (1971), Ruž ič ka (1973), Paduč eva (1974, 1985), Ward (1977), Fontaine
(1978), Ponomareff (1978), Kobozeva (1981), Markowich and Paillard (1981), Giusti (1982),
Paillard (1983, 1984), Nikolaeva (1983, 1985), Bonnot-Saoulski (1983), Kreisberg (1986),
Kuz’mina (1989).

Figure A.16. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

The -to-series is mainly used specifically, when the referent is not known to the speaker.

(A.121) specific unknown

Kto-to postuč al v dver’.
‘Somebody knocked at the door.’

In the irrealis-non-specific and question/conditional functions, normally the -nibud’-series is

used, for which the -libo-series is a more formal alternative.

(A.122) irrealis non-specific (imperative, future)

a. Spojte nam kakoj-nibud’/kakoj-libo romans.
‘Sing some romance for us.’
b. My vstretimsja gde-nibud’/gde-libo.
‘We’ll meet somewhere.’
(A.123) question/conditional
a. Zvonil li mne kto-nibud’/kto-libo?
‘Did anyone call me?’
b. Esli č to-nibud’/č to-libo sluč itsja, ja skaž u mame.
‘If anything happens, I’ll tell mom.’

The -to-series is also not excluded from the above contexts, though it is associated with a
different pragmatics in questions and conditionals (§ 4.7.3). In the irrealis contexts, -to-
indefinites show a strong tendency to be interpreted specifically if such an interpretation is
possible, cf. (A.124). However, a non-specific reading seems to be possible as well, although the
-nibud’-series is clearly preferred in this case.

(A.124) irrealis context: specific/non-specific

a. Ona xoč et vyjti zamuž za kogo-to iz Ameriki.
‘She wants to marry someone [specific] from America.’
b. Ona xoč et vyjti zamuž za kogo-nibud’/kogo-to iz Ameriki.
‘She wants to marry someone [non-specific] from America.’

The -libo-series has a wider application than the -nibud’-series: it is also used in the indirect-
negation and comparative functions, where -nibud’ is not possible. However, -libo can be
replaced by by to ni bylo in these cases.

(A.125) comparative
Zdes’ prijatnee ž it’ č em gde-libo/gde by to ni bylo v mire.
‘Here it is more pleasant to live than anywhere in the world.’
(A.126) indirect negation
 327

a. Otsutstvujut kakie-libo/kakie by to ni bylo ukazanija o sposobe ix primenenija.

‘Any indications about the mode of their use are lacking.’
b. bez kakoj-libo/kakoj by to ni bylo pomošč i
‘without any help’

The by to ni bylo-series is also possible as an emphatic variant in conditionals, but not in

questions.

(A.127) a. Esli ty uslyš iš ’ č to by to ni bylo, razbudi menja.

‘If you hear anything (at all), wake me up.’
b. *Slyš ala li ty č to by to ni bylo?
‘Did you hear anything (at all)?’

In the free-choice function, either the determiner ljuboj or the ugodno-series is used. Marginally
the by to ni bylo-series is also possible.

(A.128) free choice

Ty mož eš ’ kupit’ ljubuju/kakuju ugodno/?kakuju by to ni bylo knigu.
‘You may buy any book.’

In the direct-negation function, only the ni-series is possible.

(A.129) direct negation

Nikogda ja ne zabudu tebja.
‘I will never forget you.’

The koe-series may be used when the referent is specific and the speaker knows its identity, as
in (A.130) (cf. § 3.2.4).

(A.130) specific known

Nam nado pogovorit’ s toboj koe o č em.
‘You and I have to talk about something.’

A.17. Lithuanian
A.17.1. Inventory. Lithuanian (Indo-European, Baltic) has five major series of indefinite
pronouns, all of them derived from interrogatives: (i) the specific kaž -series, (ii) the non-specific
nors-series, (iii) the free-choice series in bet-, (iv) the negative nie-series, and (v) the kai-series.

interrogative kaž -series nors-series bet-series nie-series kai-series

person/thing: kas kaž -kas kas nors bet kas nie-kas kai kas
property: koks kaž -koks koks nors bet koks kai koks
place: kur kaž -kur kur nors bet kur nie-kur kai kur
time: kada kaž -kada kada nors bet kada nie-kada kai kada
manner: kaip kaž -kaip kaip nors bet kaip nie-kaip kai kaip
amount: kiek kaž -kiek kiek nors bet kiek
determiner: kuris kaž -kuris kuris nors bet kuris kai kuris

There is also a marginal X-ne-X-series (kas-ne-kas ‘someone’, kuris-ne-kuris ‘some’, cf. §8.3.4), and
the determiner joks ‘any’, which forms a 'series' of its own.

A.17.2. Origins. The prefix kaž - has the variant kaž in (kaž in kas, etc.), which shows that it goes
back to kas ž ino k- ‘who knows wh-’ (cf. § 6.2.1). The indefiniteness marker nors is identical to
nors ‘although; if only, at least’, and it goes back to a converb form (noris) of the verb nore%ti
‘want’. The marker bet is identical to bet ‘but, nevertheless’, but there may also be a connection
to bent ‘at least, even’. The prefix nie- is a phonological variant of ne ‘neither, not even’. The
marker kai is identical to kai ‘when’.

A.17.3. Distribution. The distribution of the three series is shown in Figure A.17. The data are
from Pilka (1984) (P84), Ambrazas (1985), and native speakers (Emma Geniuš iene%, Greta
Lemanaite%) .
 328

Figure A.17. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

The kaž -series is primarily used in the specific-unknown function, but it is also possible in the
question/conditional function. In non-specific functions, the nors-series is generally used, see §
3.2.3, examples (69), (71), (75), (77), (79), (82). In questions (cf. ex. 88) and conditionals, the nors-
series is normal:

(A.131) conditional
Jei tu ka nors/ kaž -ka matai, pasaky-k man.
if you what INDEF INDEF-what see tell-IMPV to:me
‘If you see anything/something, tell me.’

In the indirect-negation function, only the nors-series is possible.

(A.132) indirect negation

Ne-langva rasti ka4 nors kas pade%tu4 jums.
NEG-possible find whom INDEF who help you
‘It’s impossible to find anyone who would help you.’

The direct-negation function is generally expressed by the nie-series, which cooccurs with
verbal negation.

(A.133) direct negation

Nie-kas man nie-ko ne-saki.
NEG-who to:me NEG-what NEG-told
‘Nobody told me anything.’

The free-choice and comparative functions are expressed by the bet-series. In the comparative
function, the nors-series is also possible.

(A.134) free choice

Bet kas gali paaiš kinti š ia problema.
INDEF who can solve this problem
‘Anybody can solve this problem.’
(A.135) comparative
Aš esme suprantu geriau negu kas nors/ bet kas kitas. (P84: 146)
I essence understand better than who INDEF INDEF who other
‘I understand the essence better than anyone else.’

The single determiner joks is used in both the direct-negation and the indirect-negation
functions:

(A.136) direct negation

Ž irardo ne-atskleidž ia jokiu nauju savo talento pusiu. (P84: 87)
Girardeaux NEG-shows any new her talent’s sides
‘Girardeaux does not show any new sides of her talent.’
(A.137) indirect negation
Dauguma meteoru pasirodo nakti be jokiu/ kokiu nors
majority of:meteors appear at:night without any which INDEF
desningumu. (P84: 87)
regularities
‘The majority of meteors appear at night without any regularities.’

The kai-series indicates that the referent is known to the speaker, much like the Russian koe-
series (see ex. 100 in § 3.2.4).
 329

A.18. Latvian
A.18.1. Inventory. Latvian (Indo-European, Baltic) has three major series of indefinite
pronouns, all derived from interrogatives: (i) the general kaut-series, (ii) the negative ne-series,
and (iii) the free-choice jeb-series.

interrogative kaut-series ne-series jeb-series

person: kas kaut kas, kā ds ne-viens jeb-kā ds
thing: kas kaut kas ne-kas jeb-kas
place: kur kaut kur ne-kur jeb-kur
time: kad kaut kad ne-kad jeb-kad
manner: kā kaut kā ne-kā
determiner: kā ds kaut kā ds ne-kā ds jeb-kā ds
kurš jeb-kurš

There is also a marginal X ne-X-series (kur ne-kur ‘here and there’, kā ds ne-kā ds ‘some(one) or
other’), and the bare interrogatives may also be used as indefinites (especially kā ds in the
meaning ‘somebody’).
A.18.2. Origins. The particle kaut also means ‘at least, even’, cf. § 7.1. The particle jeb also means
‘or’, cf. § 7.2. The prefix ne- seems to be identical to the verbal prefix ne- ‘not’.
A.18.3. Distribution. The distribution of the three series is shown in Figure A.18. The data are
all from reference works and from a native speaker (Ivonna Wagner).

Figure A.18. direct

question indirect negation
specific irrealis negation
non-specific
conditional comparative
free-choice

The kaut-series is used in all non-emphatic functions:

(A.138) specific
Vinš kaut kur aizgā jis.
he INDEF where left
‘He went somewhere.’
(A.139) irrealis non-specific (imperative)
Nopē re kaut ko.
buy INDEF what
‘Buy something.’
(A.140) question/conditional
a. Vai tu gatavojies kaut kur braukt?
Q you you:are:prepared INDEF where to:go
‘Are you going to go anywhere?’
b. Ja kā ds atnā k(tu), saki, lai vinš pagaida.
if who arrives say:IMPV let he wait
‘If anyone comes, say that he should wait.’
c. Ja kaut kur <*jeb-kur> ejat, tad pasakiet man.
if INDEF where INDEF-where you:go then tell:IMPVme
‘If you go anywhere, tell me.’

The jeb-series is used in the free-choice, comparative, and indirect-negation functions.

(A.141) free choice

Varat nā kt jeb-kurā laikā .
you:can to:come INDEF-which time:LOC
‘You can come at any time.’
(A.142) comparative
Tas vinu iepriecinā ja vairā k nekā jeb-kas cits.
that him delighted more than INDEF-what other
‘It made him more glad than anything else.’
(A.143) indirect negation (cf. § 4.7.2)
a. bez jeb-kā dā m pā rmainā m
 330

without INDEF-which changes

‘without any changes’
b. Es nedoma@ju, ka ka@ ds atna@ks.
I [Link] that who came
'I don't think that anybody came.'

In the direct-negation function, the ne-series cooccurs with verbal negation.

(A.144) direct negation

Vinu ne-kas ne-interesē .
him NEG-what NEG-interests
‘Nothing interests him.’

A.19. Irish
A.19.1. Inventory. Irish has three series of indefinite pronouns, all of them derived from generic
nouns: (i) the non-emphatic éigin-series, (ii) the negative-polarity series marked by aon, and (iii)
the emphatic ar bith-series.

interrogative éigin-series aon-series ar bith-series

person: cé duine (éigin), neach aon duine duine ar bith
thing: cad, cén rud (éigin), ní aon rud/ní rud ar bith
place: cá, cár in áit éigin, in aon bhall, in áit ar bith
i mball éigin in aon áit
time: cathain
manner: conas ar chaoi éigin, in aon chor ar chor/dhóigh ar bith,
ar dhóigh éigin ar chaoi ar bith
determiner: cé éigin aon ... ar bith

In addition, there is one isolated indefinite pronoun dada 'anything', and several expressions
for 'ever': riamh, go brách (referring to past events), choíche, go deo (referring to future events).
Irish indefinites show a very low degree of grammaticalization, and perhaps only two in-
definite determiners (éigin, aon) and one indefinitizing prepositional phrase (ar bith)should be
i
recognized. However, the above paradigm makes it easier to see the parallels with other
languages (where some slots are also occupied by little grammaticalized expressions).

A.19.2. Origins. Aon is originally ‘one’, ar bith is ‘in the world’, dada is 'jot, tittle'.

A.19.3. Distribution. The distribution of the three series is shown in Figure A.19. The data are
from Ó Dónaill (ed.) (1977) (D77), De Bhaldraithe (ed.) (1959) (B59), and a native speaker
(Dónall P. Ó Baoill).

Figure A.19. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

The éigin-series is used in specific and irrealis non-specific functions (perhaps there is a
preference to omit éigin and use the bare generic nouns when the referent is known to the
speaker):

(A.145) specific known/unknown

a. Tá rud agam le rá leat. (NT, Lk 7.40)
is thing on:me for telling to:you
‘I have something to tell you.’
b. Dúirt duine éigin liom é. (B59: 683)
told person certain to:me he
‘Somebody told me.’
(A.146) irrealis non-specific (modal, imperative)
 331

a. Caithfear teacht ar réiteach éigin eile. (B59: 682)

must:FUT coming on solution certain other
‘Some other solution will have to be found.’
b. Abair rud éigin. (B59: 683)
say:IMPV thing certain
‘Say something.’

In all negative-polarity functions, the aon-series or the ar bith-series may be used:

(A.147) questions
a. An bhfuil aon airgead/ airgead ar bith agat?
Q is any money money on world at:you
‘Do you have any money?’
b. An bhfuil dada le rá agat? (D77: 361)
Q is anything for telling at:you
‘Do you have anything to say?’
c. An bhfeiceann tú duine ar bith ansin thall?
Q see you person on world there over
‘Do you see anyone over there?’
(A.148) conditionals
a. Má thagann aon duine acu ... (B59: 28)
if comes any person at:them
‘If any of them should come ...’
b. Má thagann sé go deo/ choíche ... (B59: 229)
if comes he ever
‘If he ever comes ...’
c. Má chuir sé fearg ort ar dhóigh ar bith ... (B59: 28)
if puts he anger on:you on manner on world
‘If he has in any way offended you ...’
(A.149) indirect negation
a. Ní mheasaim gur tháinig aon duine/ duine ar bith acu. (B59: 28)
NEG I:think that arrived any person man on world at:them
‘I don’t think any of them has arrived.’
b. Níorbh áilleis go mbeadh fios ag aon duine.
NEG:was desire with:him that be knowledge at any person
‘He didn’t want anyone to know it.’ (NT, Mk 9.30)
(A.150) comparative
a. Tháinig an bháisteach níos tréine ná riamh. (B59: 229)
arrived the rain more power than ever
‘It started to rain, faster than ever.’
b. chomh maith le aon duine beo (B59: 28)
as good with any man living
‘as good as any man’

In the direct-negation function, the indefinites cooccur with verbal negation.

(A.151) direct negation

a. Níor labhair aon duine/ duine ar bith liom. (B59: 477)
NEG spoke any person person on world with:me
‘Nobody spoke to me.’
b. Ní dúirt siad aon ní/ dada/ rud ar bith le haon duine.
NEG said they any thing anything thing on world to any person
‘They didn’t say anything to anybody.’
c. Ní labhair sé riamh air. (D77: 997)
NEG spoke he ever on:it
‘He never spoke of it.’

In the free-choice function, too, either the aon-series or the ar bith-series is used:

(A.152) a. Tar lá ar bith/ aon lá (a thogrós tú). (B59: 28)

come day on world any day that like you
‘Come any day (you like).’
b. Íosfaidh sé rud ar bith/ aon rud/ *dada. (B59: 28)
 332

eats:HAB he thing on world any thing anything

‘He eats anything.’

The indefinite dada and the various worlds for 'ever' are only used in the negative-polarity
functions. They are not possible in the free-choice function.

A.20. Ossetic
A.20.1. Inventory. Ossetic (Indo-European, Iranian) has five major series of indefinite pro-
nouns, all of them based on interrogative pronouns: (i) the specific -dær-series, (ii) the non-
specific is-series, (iii)-(iv) the two negative series in ni- and ma-, and (v) the free-choice series in
-dæriddær.

interrogative -dær-series is-series ni-series ma-series -dæriddær-series

person: č i (obl. kæj) č i-dær is-č i ni-č i ma-č i č i-dæriddær
thing: cy (obl. cæj) cy-dær is-ty ni-cy ma-cy cy-dæriddær
place: kæm kæm-dær is-kæm ni-kæm ma-kæm kæm-dæriddær
time: kæd kæd-dær is-kæd ni-kæd ma-kæd kæd-dæriddær
manner: kuyd kuyd-dær is-kuyd ni-kuyd ma-kuyd kuyd-dæriddær
determiner: kæcy kæcy-dær is-kæcy ni-kæcy ma-kæcy kæcy-dæriddær

There is another free-choice series marked by -fændy which s eems to be equivalent to the -
dæriddær-series: č i-fændy ‘anyone’, cy-fændy ‘anything’, etc.
A.20.2. Origins. The suffix -dær is identical to the focus particle dær ‘also, even’. the prefix is- is
originally the 3rd person singular of ‘be’. I have no information on the origin of -
dæriddær. The two negative prefixes are related to the negative particles næ ‘not’ and ma ‘not’.
A.20.3. Distribution. The distribution of the three series is shown in Figure A.20. The data are
from Kulaev (1958) (K58), Axvlediani (1963), Bagaev (1965), and a native speaker (Zarema
Xubecova) .

Figure A.20. direct

question indirect negation
specific irrealis negation
non-specific
conditional comparative
free-choice

The -dær-series is used only in specific functions:

(A.153) specific
Cy-dær mæ qygdær-y. (K58: 52)
what-INDEF me bother-PRES.3SG
‘Something bothers me.’

The is-series is used in non-specific functions, in irrealis contexts, in questions and conditionals,
and in indirect-negation contexts.

(A.154) irrealis non-specific (imperative)

a. Mænæm dær ma is-ty ratt.
I:DAT also PT INDEF-what give(IMPV)
‘Give me something, too.’
(A.155) question/conditional
a. Dy is-kæj uyn-ys?
you INDEF-who see-PRES.2SG
‘Do you see anybody?’
b. Dy is-ty kuy ssar-i-s, uæd mænæm dær æxsyzgon uaid.
you INDEF-what if find-SBJV-2SG then I:DAT also glad will:be
‘If you find anything, I’ll be glad.’
(A.156) indirect negation
Æz dyzærdyg kæn-yn, is-č i ærbacæu-dzæn ævi næ.
I doubt do-PRES.1SG INDEF-who come-FUT(3SG) or not
‘I doubt that anyone will come.’
 333

The -dæriddær-series is used in the free-choice and comparative functions.

(A.157) free choice

Aj xuymætæg xyncinag u, č i-dæriddær dær æj baxync-dzæn.
this simple problem is who-INDEF also it solve-FUT(3SG)
‘This is a simple problem, anyone will solve it.’
(A.158) comparative
Uyj jæ k’lasy kæmæj-dæriddær zonddž yn-dær u.
he his class who:ABL-INDEF smart-COMPR is
‘He is smarter than anyone in his class.’

In the direct-negation function, the ni-series or the ma-series is used. The choice depends on the
modality of the clause, like the choice between the corresponding verbal negators næ and ma.
Næ is used in indicative clauses, and ma is used in imperative and subjunctive clauses. Ossetic
is like Latin, standard English and standard German in that negative pronouns do not cooccur
with verbal negation. (However, unlike Latin, English and German, Ossetic allows more than
one negative pronoun in one clause, see § [Link].)

(A.159) direct negation

a. Æz uym ni-kæj uyd-t-on.
I there NEG-who see-PAST-1SG
‘I didn’t see anyone there.’
b. Ma-kædæm acu.
NEG-where go(IMPV)
‘Don’t go anywhere.’

A.21. Persian
A.21.1. Inventory. Persian (Indo-European, Iranian) has only two series of indefinite pronouns:
(i) the generic-noun-based -i-series, and (ii) the negative hič -series.

interrogative i-series hič -series

person: ki kas-i hič kas(-i)
thing: č e č iz-i hič č iz(-i)
place: koǰ â ǰ â-yi hič ǰ â(-yi)
time: key vaqt-i hič vaqt
manner: hič gune
determiner: kodâm yek X-i hič (yek)

The indefinites of the i-series may also be preceded by yek ‘one’ (yek kas-i, yek č iz-i, yek ǰ â-yi,
etc.).
A.21.2. Origins. The i-series is nothing but generic nouns (‘person’, ‘thing’, ‘place’, etc.) suffixed
with the indefinite article -i. The determiner yek is identical to yek ‘one’. The hič -series is also
based on generic nouns. The particle hič originally meant ‘anybody’ and appears to go back to
Old Persian aiva ‘one’ plus č iy (an emphatic particle cognate with Sanskrit cit, cf. A.22.2).
A.21.3. Distribution. The distribution of the two series is shown in Figure A.21. The data are
from reference works and from a native speaker (Parviz Rostampour) .

Figure A.21. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

The i-series can be used in all functions except for the comparative and free-choice functions.

(A.160) specific known/unknown

Diruz kas-i-râ dar otobus did-am.
yesterday person-INDEF-ACC on bus see:PAST-1SG
‘Yesterday I saw someone [known/unknown] on the bus.’
 334

(A.161) irrealis non-specific (imperative)

Yek vaqt-i be xâne-ye mâ bey-âyid.
one time-INDEF to house-LK we SBJV-come:IMPV
‘Come to our house sometime.’

In the question function, but not in the conditional function, hič is also possible (yielding an
emphatic reading).

(A.162) question/conditional
a. Âyâ (hič ) č iz-i mi-š anav-i?
Q INDEF thing-INDEF IMPF-hear-2SG
‘Can you hear anything?’
b. Agar (*hič ) č iz-i š anid-i, be man be-gu.
if INDEF thing-INDEF hear-2SG to me SBJV-say
‘If you hear anything, tell me.’
(A.163) indirect negation
Fekr na-mi-kon-am (ke) Sorâyâ (hič ) č iz-i š anide bâš ad.
thinking NEG-IMPF-do-1SG that Soraya INDEF thing-INDEF heard has
‘I don’t think that Soraya heard anything.’

While hič in the question and indirect-negation functions is possible, but not preferred, it is
quite normal in the direct-negation function. However, not even here it is required. (For the
case of two negated indefinites in one clause, see § [Link].)

(A.164) direct negation

Hič kas(-i)/ kas-i zang na-zad.
INDEF person-INDEF person-INDEF ring NEG-hit
‘Nobody called.’

For the free-choice and the comparative series, har ‘every’ has to be used:

(A.165) Har kas(-i) mi-tavânad in mas?ale-râ hall konad.

every person-INDEF IMPF-can this problem-ACC solving do
‘Anyone can solve this problem.’
(A.166) Pesar az har kas-i tondtar mi-davad.
boy than every person-INDEF faster IMPF-runs
‘The boy runs faster than anyone.’

A.22. Hindi/Urdu
A.22.1. Inventory. Hindi/Urdu (Indo-European, Indic) has only two major series of indefinite
pronouns: (i) the koii-series, which is morphologically similar to, but not synchronically
derivable from, the interrogative pronouns; (ii) the bhii-series, derived from the koii-series.

interrogative koii-series bhii-series

person: kaun (obl. kis) koii (obl. kisii) koii bhii
thing: kyaa kuch kuch bhii
place: kahã kahĩ kahĩ bhii
time: kab kabhii kabhii bhii

Besides these, there is a peripheral X na X series (koii na koii ‘someone or other’, kuch na kuch
‘something or other’), cf. § [Link].
A.22.2. Origins. The indefiniteness marker bhii is identical to the focus particle bhii ‘also, even’.
The koii-series is etymologically derived from the interrogative series: koii < Sanskrit ko ’pi,
sandhi form of kah5 api (kah5 ‘who’, api ‘particle’), kuch < Sanskrit kim%cit (kim% ‘what’, cit, a
particle) (Beames 1875, vol. II: §72).
A.22.3. Distribution. The distribution of the three series is shown in Figure A.22. The data are
from Bhatia (1978) (B78), Davison (1978a) (D78), (1978b), Mahajan (1990), and a native speaker
(Sangeeta Sharma). See also Dayal (1995) on bhii.

Figure A.22. direct

 335

question indirect negation

specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

The koii-series may be used in all functions except the comparative and the free-choice
functions. However, in the direct-negation function the bhii-series seems to be preferred.

(A.167) specific known/unknown

Kisii-ne <*bhii> fon kiy-aa thaa,
someone-ERG INDEF phone do-PFV was
— par mãĩ tumhẽ nahĩ bataaũ gii, kis-ne.
but I you NEG I:will:tell who-ERG
— par mujhe nahĩ maaluum, kis-ne.
but I:DAT NEGknown who-ERG
‘Someone has phoned, but I won’t tell you who/but I don't know who.’
(A.168) indirect negation
Mujhe nahĩ lag-taa hai ki kal koii (bhii) aaege. (D78: 37)
I:DAT NEG seem-IMPF is that tomorrow someone INDEF he:will:come
‘It doesn’t seem to me that anyone will come tomorrow.’
(A.169) direct negation
a. Ghar mẽ koii (bhii) nahĩ hai. (B78: 69)
house in someone INDEF NEG is
‘No one (at all) is at home.’
b. Kisii-ne kuch nahĩ dekh-aa. (D78: 29)
someone-ERG something NEG see-PFV
‘Nobody saw anything.’ (Or: ‘Someone saw nothing.’)

In irrealis contexts, the bhii-series is always non-specific.

(A.170) irrealis non-specific (‘want’, imperative)

a. Wah kisii-ko/ kisii-ko bhii fon kar-naa caah-tii hai.
she someone-DAT someone-DAT INDEF phone do-INF want-IMPF is
‘She wants to phone someone (specific/non-specific).’
b. Kisii-ko (bhii) fon kar-o!
someone-DAT INDEF phone do-IMPV
‘Phone someone!’

In questions and conditionals, the bhii-series differs from the koii-series in being more emphatic
(‘any ... at all’).

(A.171) question/conditional
a. Kyaa aap-ne kisii-ko/ kisii-ko bhii dekh-aa? (D78: 27-8)
Q you-ERG someone-DAT someone INDEF see-PFV
‘Did you see somebody/anybody?’
b. Agar koii (bhii) fon kare, mujhe bataanaa.
if someone INDEF phone calls I:DAT tell
‘If anybody (at all) calls, tell me.’

The free-choice and comparative functions are only expressed by the bhii-series:

(A.172) comparative
Yeh larkaa apne vidyaalay ke kisii bhii larke se zyaada tez
this boy his class of someone INDEF boy from more fast
daur-taa hai.
run-IMPF is
‘This boy runs faster than anyone in his class.’
(A.173) free choice
Ghar mẽ koii bhii aa sak-taa hai. (B78:70)
house in someone INDEF come can-IMPF is
‘Anyone can come into the house.’
 336

A.23. Turkish
A.23.1. Inventory. Turkish (Turkic) has three major series of indefinite pronouns, all of them
based on generic nouns and bir ‘one’: (i) the non-emphatic bir-series, (ii) the negative hiç-series,
and (iii) the free-choice series marked by herhangi:

interrogative bir-series hiç-series herhangi-series

person: kim biri(si) hiçbiri, hiç kimse herhangi biri
thing: ne bir s3ey hiç bir s3ey herhangi bir s3ey
place: nerede bir yerde hiç bir yerde herhangi bir yerde
time: ne zaman bir zaman hiç bir zaman herhangi bir zaman
manner: nasıl bir s3ekilde hiç bir s3ekilde herhangi bir s3ekilde
determiner: hangi bir hiç bir herhangi bir

In addition, there is the isolated indefinite (bir) kimse ‘someone, anyone’, which can only be
used non-specifically.
A.23.2. Origins. Bir is identical to the numeral ‘one’. Some of the combinations of bir plus
generic nouns are strongly lexicalized; bir s3ey is felt as a single word (cf. § 1.2.2). The indef-
initeness marker hiç was borrowed from Persian (cf. A.21). Herhangi consists of her ‘every’ (also
borrowed from Persian) and hangi ‘which?’. Kimse seems to be derived from kim ‘who’ and the
conditional suffix -se.
A.23.3. Distribution. The distribution of the three series is shown in Figure A.23. The data are
from Xovratovič (1989) and from a native speaker (Fethi Inan) .

Figure A.23. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

In the specific-known function, only the bir-series is possible:

(A.174) specific-known
San-a biri(si) telefon et-ti. Bil, bakalım, kim.
you-DAT one phone do-PAST know we’[Link] who
‘Someone phoned you. Guess who.’

In all the functions from specific-unknown to direct negation, either the bir-series or the
herhangi-series are possible.

(A.175) specific-unknown
San-a biri(si)/ herhangi biri telefon et-ti. Kim ol-duǧ -u-nu
you-DAT one any one phone do-PAST who be-REL-3SG-ACC
bil-m-iyor-um.
know-NEG-IMPF-1SG
‘Someone phoned you. I don’t know who.’
(A.176) irrealis non-specific (imperative)
Herhangi bir yer-e/ bir yer-e telefon et!
any one place-DAT one place-DAT phone do(IMPV)
‘Call somewhere (or other)!
(A.177) question/conditional
a. Herhangi bir s3 ey/ bir s3 ey/ hiç bir s3 ey duy-du-n mu?
any one thing one thing INDEF one thing hear-PAST-2SG Q
‘Did you hear anything?’
b. Herhangi biri/ biri telefon ed-er-se, ban-a hemen söyle.
any one one phone do-AOR-COND I-DAT [Link] say(IMPV)
‘If anyone calls, tell me immediately.’

In the question function (cf. A.177a) and in the negation functions, the hiç-series is another
possible alternative.
 337

(A.178) indirect negation

Kimse-nin/ herhangi biri-nin/ hiç kimse-nin gel-me-si
anyone-GEN any one-GEN INDEF anyone-GEN come-MASD-3SG
gerekli deǧ il.
necessary [Link]
‘It is not necessary for anyone to come.’

In the direct-negation function, the hiç-indefinites are the most usual form. They always cooccur
with verbal negation. (For the case of two negated indefinites in one clause, see § [Link].)

(A.179) direct negation

Hiç bir s3 ey/ herhangi bir s3 ey/ bir s3 ey gör-me-di-m.
INDEF one thing any one thing one thing see-NEG-PAST-1SG
‘I didn’t see anything.’

In the free-choice and comparative functions, only the herhangi-series is possible:

(A.180) free choice

Herhangi biri bu problem-i çöz-ebil-ir.
any one this problem-ACC solve-ABIL-AOR(3SG)
‘Anyone can solve this problem.’
(A.181) comparative
Murat sınıf-ın-da-ki herhangi birin-den daha hızlı kos3-ar.
Murat class-3SG-LOC-ATTR any one-ABL more fast run-AOR
‘Murat runs faster than anyone in his class.’

A.24. Kazakh
A.24.1. Inventory. Kazakh (Turkic) has four major series of indefinite pronouns, three of which
are based on interrogatives: (i) the non-emphatic bir-series, (ii) the specific älde-series, (iii) the
negative eš -series, and (iv) the free-choice series marked by bolsa da.

interrogative bir-series älde-series eš -series bolsa da-series

person: kim birew älde-kim eš -kim kim bolsa da
thing: ne birnärse, älde-ne eš -närse ne bolsa da
birdeme eš -teme
place: qajda bir ž erde älde-qajda eš -qajda qajda bolsa da
time: qaš an bir kezde älde-qaš an eš -qaš an qaš an bolsa da
manner: qalaj qalaj bolsa da
determiner: qandaj älde-qandaj eš -qandaj qandaj bolsa da
älde-bir eš -bir

A.24.2. Origins. The bir-series is based on bir ‘one’ and generic nouns (except for birew and
birnärse, birdeme, whose origin is not clear to me). The prefix eš - is etymologically identical to
Turkish hiç and goes back to Persian hič (Section 21) . The origin of the prefix älde- is obscure,
but a connection with Tatar and Bashkir ällä ‘possibly; question particle; indefiniteness marker’
seems possible.
A.24.3. Distribution. The distribution of the four series is shown in Figure A.24. The data are
from a native speaker (Kusain Rsaldinov) and from Sauranbaev (ed.) (1954) (S54).

Figure A.24. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

The älde-series is used only in specific functions.

(A.182) specific
 338

a. Älde-qajdan maš ina-nyN gudag-y estil-di. (S54: 151)

INDEF-where car-GEN honk-3SG sound-PAST
‘Somewhere the honk of a car sounded.’
b. Ol älde-qandaj dälel-der keltir-di. (S54:319)
he INDEF-which argument-PL adduce-PAST(3SG)
‘He adduced some arguments or other.’

The bir-series is used in specific functions, both known and unknown.

(A.183) specific known/unknown

Birew telefon soq-ty — byraq kim ekenin bil-me-j-min.
someone phone hit-PAST but who [?] know-NEG-PRES-1SG
— kim ekenin tapš y.
who [?] guess(IMPV)
‘Someone called — but I don’t know who.’
— guess who.’

It is also used non-specifically in irrealis contexts, in questions and conditionals, and in indirect-
negation contexts.

(A.184) irrealis non-specific (imperative, hortative)

Bir ž er-de kezde-ser-miz.
one place-INESS meet-FUT-1PL
‘Let’s meet somewhere.’
(A.185) question/conditional
a. Birew <*älde-kim> kel-di me?
someone INDEF-who come-PAST(3SG) Q
‘Did anyone come?’
b. Birew <*älde-kim> kel-se, ma-Van ajt.
someone INDEF-who come-COND(3SG) I-DAT tell(IMPV)
‘If anyone comes, tell me.’
(A.186) indirect negation
Birew kel-di dep sen-be-j-min.
someone come-PAST(3SG) that think-NEG-PRES-1SG
‘I don’t think that anyone came.’

In the direct-negation function, only the eš -series is possible, cooccurring with verbal negation.

(A.187) direct negation

a. Eš -kim eš -närse de-me-di.
NEG-who NEG-anything say-NEG-PAST(3SG)
‘Nobody said anything.’

The bolsa da-series is used in the free-choice function, and also (though much less commonly
than the bir-series) in questions and conditionals. (I lack data for the comparative function.)

(A.188) free choice

SeniN kim-di bolsa da š aqyru-Na bolady.
you who-ACC INDEF invite-MASD may
‘You may invite anyone.’
(A.189) Bir waqyt-ta/ qaš an bolsa da seniN Moskva-da bol-ǧ a-nyN bar ma?
one time-INESS when INDEF you Moscow-INESS be-PAST-2SG is Q
‘Have you ever been to Moscow?’

A.25. Yakut
A.25.1. Inventory. Yakut (Turkic) has four major series of indefinite pronouns, all of which are
derived from interrogatives: (i) the specific ere-series, (ii) the non-specific eme-series, (iii) the
negative da-series, and (iv) the free-choice series marked by baVarar.

interrogative ere-series eme-series da-series baVarar-series

 339

person: kim kim ere kim eme kim da kim baVarar

thing: tuox tuox ere tuox eme tuox da tuox baVarar
place: xanna xanna ere xanna eme xanna da xanna baVarar
time: xahan xahan ere xahan eme xahan da xahan baVarar
manner: xajdax xajdax ere — xajdax da xajdax baVarar
determiner: xannyk xannyk ere — xannyk da xannyk baVarar
xaja xaja ere xaja eme xaja da xaja baVarar

A.25.2. Origins. The particle ere is also used with the meaning ‘or’ (cf. § 7.2). The particle eme is
also used with the meaning ‘at least’ (cf. § 7.1). The particle da is also used with the meaning
‘and, also, even’ (cf. § 7.1). The indefiniteness marker baVarar is derived from baVar ‘want’.

A.25.3. Distribution. The distribution of the four series is shown in Figure A.25. The data are
from Afanas’ev and Xaritonov (1968) (AX68) and from Ubrjatova (ed.) (1982) (U82).

Figure A.25. direct

question indirect negation
specific irrealis negation
non-specific
conditional comparative
free-choice

The ere-series is restricted to specific functions.

(A.190) specific
Kim ere eji-exe kiir-e syld’y-byt-a. (AX68: 250)
who INDEF you-DAT enter-CONV go-PERF-3SG
‘Someone has come to you.’

The eme-series is used in non-specific functions in irrealis contexts, questions and conditionals.

(A.191) irrealis non-specific (future)

Xojut kim-inen eme tug-u eme yyt-ya-m. (U82: 200)
afterwards who-INSTR INDEF what-ACC INDEF send-FUT-1SG
‘Afterwards I’ll send something with someone.’
(A.192) question/conditional
a. En xahan eme Moskva-Va syld’y-byt-yN duo? (AX68: 229)
you when INDEF Moscow-DAT go-PERF-2SG Q
‘Have you ever been to Moscow?’
b. Kim eme kel-leVine, mii-gin üle-t-igerbar-byt-a die.
who INDEF come-COND I-ACC work-3SG-DAT go-PERF-3SG say(IMPV)
‘If anyone comes, say that I went to work.’ (AX68: 250)

The da-series is used in the direct-negation function (cooccurring with verbal negation), the
comparative function and the free-choice function. Unfortunately, I lack data on the indirect-
negation function, but my implicational map predicts that the da-series is also possible in this
function.

(A.193) direct negation

Kini-ni tuox da interiehirge-ppet. (AX68: 336)
he-ACC what INDEF interest-NEG.3SG
‘Nothing interests him.’
(A.194) comparative
AjylVa kiergel-in kim-ten da syanalyy-r. (U82: 209)
nature beauty-3SG:ACC who-ABL INDEF appreciate-3SG
‘He appreciates nature’s beauty more than anyone.’

The baVarar-series may also be used in the free-choice function.

(A.195) free choice

Kini-tten tug-u baV arar küüt-üö-xxe söp. (U82: 207)
 340

he-ABL what-ACC INDEF expect-MASD-DAT possible

‘One can expect anything from him.’

A.26. Hungarian
A.26.1. Inventory. Hungarian (Finno-Ugrian) has four major series of indefinite pronouns, all of
them derived from interrogatives: (i) the general vala-series, (ii) the negative sem-series, and
(iii-iv) the two free-choice series in akár- and bár-.

interrogative vala-series sem-series akár-series bár-series

person: ki vala-ki sen-ki akár-ki bár-ki
thing: mi vala-mi sem-mi akár-mi bár-mi
property: milyen vala-milyen sem-milyen akár-milyen bár-milyen
place: hol vala-hol se-hol akár-hol bár-hol
time: mikor vala-mikor sem-mikor akár-mikor bár-mikor
manner: hogy(an) vala-hogy(an) se-hogy(an) akár-hogy(an) bár-hogy(an)
amount: hány — se-hány akár-hány
mennyi — se-mennyi akár-mennyi bár-mennyi
determiner: mely(ik) mely(ik) — akár-mely(ik) bár-mely(ik)

There is also a marginal né-series (né-mi ‘something, a little’, né-hol ‘in some places, here and
there’, né-hány ‘some, a few’, né-melyik ‘some’).
A.26.2. Origins. The indefiniteness prefix vala- is probably related to van/vol- ‘be’. The prefix
akár- is identical to akár ‘or; at least’, related to akar ‘want’. The prefix bár is identical to bár
‘although; if only’. The prefix sem- (se-) is identical to sem ‘not either, neither’ (< is ‘too’ plus nem
‘not’). The prefix né- was borrowed from Slavic (cf. Russian ne- Polish nie-, Bulgarian nja-).
A.26.3. Distribution. The distribution of the three series is shown in Figure A.26. The data are
from Stephanides (1983, 1985) (S85), Hunyadi (1987), dictionaries, the New Testament, and a
native speaker (Edith Moravcsik). (See also Szarvas (1892), Beke (1914).)

Figure A.26. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

The vala-series is used in all functions from ‘specific’ to ‘indirect negation’.

(A.196) specific unknown

Vala-hol már tolál-koz-t-unk.
INDEF-where already meet-REFL-PAST-1PL
‘We have met somewhere before.’
(A.197) specific known
Mester, lát-ánk vala-ki-t, a ki a te nev-ed-del
master see-1PL INDEF-who-ACC who ART you name-2SG-INSTR
ördög-ök-et üz. (NT, Mk 9.38)
devil-PL-ACC drive
‘Master, we saw one casting out devils in thy name.’
(A.198) irrealis non-specific
Majd csak meglesz-ünk vala-hogy.
later only manage-1PL INDEF-how
‘We’ll manage somehow.’

The sem-series can only be used in the direct-negation function:

(A.199) direct negation

a. Nem lát-t-am sem-mi-t <*akár-mi-t, *bár-mi-t>.
NEG see-PAST-1SG INDEF-what-ACC
‘I did not see anything.’
b. Sen-ki-nek sem-mi-t sem szólának. (NT, Mk 16.8)
INDEF-who-DAT INDEF-what-ACC NEG say:3PL
 341

‘Neither said they any thing to any man.’

The akár- and bár-series are used mainly in the free-choice function, but also in the comparative
function, the indirect-negation function, and with an emphatic value in the conditional
function. However, these series cannot be used in questions.

(A.200) free choice

Akár-ki tanul-hat-ott.
INDEF-who learn-POT-PAST(3SG)
‘Anybody could learn.’
(A.201) comparative
Itt jobb a klíma, mint akár-hol/bár-hol Európá-ban.
here better ART climate than INDEF-where Europe-INESS
‘Here the climate is better than anywhere in Europe.’
(A.202) indirect negation
Nem hisz-em, hogy vala-ki/ bár-ki/akár-ki lát-t-a volna.
NEG think-1SG that INDEF-who INDEF-who see-PAST-3SG SBJV
‘I don’t think that anybody has seen it.’
(A.203) conditional
Ha bár-mi-t/akár-mi-t/vala-mi-t hal-asz, ébressz fel.
if INDEF-what-ACC hear-2SG wake up
‘If you hear anything (at all), wake me up.’
(A.204) question
Hall-ott-ál vala-mi-t <*akár-mi-t, *bár-mi-t>?
hear-PAST-2SG INDEF-what-ACC INDEF-what-ACC INDEF-what-ACC
‘Did you hear anything?’

A.27. Finnish
A.27.1. Inventory. Finnish (Finno-Ugrian) has three main series of indefinite pronouns, two of
which are derived from interrogatives: (i) the mainly negative -kaan-series and (ii) the free-
choice series marked by hyvänsä. The non-emphatic -kin-series (iii), by contrast, is based on
relative pronouns.

interrogative series series series

person: kuka jo-ku (jo-n-ku-n) kuka-an kuka hyvänsä
thing: mikä, mitä jo-kin, jota-kin mikä-än mikä hyvänsä
place: missä jossa-kin missä-än missä hyvänsä
time: milloin jolloin-kin milloin-kaan milloin hyvänsä
manner: miten, kuinka joten-kin miten-kaan miten hyvänsä
dual determiner: kumpi (‘which of two’) jompi-kumpi kumpi-kaan kumpi hyvänsä

Note that the members joku and jompikumpi of the -kin-series do not contain the suffix -kin, but
are compounds of the roots jo- and ku- and inflect doubly: joku, Partitive jotakuta, Inessive
jossakussa, etc.; jompikumpi, Partitive jompaakumpaa, etc. An equivalent of hyvänsä is tahansa. In
addition, there is eräs 'a certain'.
A.27.2. Origins. The suffix -kin also means ‘also, even’. The suffix -kaan also means ‘not either;
not even’. Hyvänsä is somehow related to hyvä ‘good’, and tahansa is somehow related to tahtoa
‘want’.
A.27.3. Distribution. The distribution of the three series is shown in Figure A.27. The data are
from native speakers (Juhani Rudanko, Angela Bartens, Hannu Tommola). See also Karttunen
and Peters (1980).

Figure A.27. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

The -kin-series is used in non-emphatic functions, but not in the specific-known function, where
eräs ‘a certain’ must be used.
 342

(A.205) specific known

a. *Joku soitt-i. Arvaa kuka (se oli)?
someone call-PAST(3SG) guess:IMPV who it was
‘Someone called. Guess who (it was).’
b. Eräs ystäväni kerto-i sen.
certain friend tell-PAST(3SG) it
'A certain friend told him.'
(A.206) specific unknown
Joku soitt-i, mutta en saa-nut
someone call-PAST(3SG) but NEG:1SG [Link]
nime-stä selvää.
name-ELAT clear:PRTV
‘Someone called, but I didn’t understand the name.’
(A.207) irrealis non-specific (‘want’)
Hän haluaa mennä naimisiin jonkun kanssa, jolla on tumma tukka.
she wants go marriage someone with who-on is dark hair

‘She wants to marry someone [non-specific] with black hair.’

In questions, the -kaan-series is a possible alternative (cf. Schmid 1980: 139), but the hyvänsä-
series is impossible. In conditionals, the hyvänsä-series is a possible (more emphatic) alternative,
but the -kaan-series is impossible.

(A.208) question
Soitt-i-ko joku/ kuka-an/ *kuka hyvänsä?
call-PAST(3SG)-Q someone who-INDEF who INDEF
‘Did someone/anyone call?’
(A.209) conditional
Jos joku/ kuka hyvänsä/ *kuka-an soittaa, sano minulle.
if someone who INDEF who-INDEF calls tell(IMPV) I:on
‘If someone/anyone at all calls, tell me.’

The -kaan-series is used in the negation functions and in the comparative function. In the direct-
negation function, it cooccurs with verbal negation.

(A.210) direct negation

Kuka-an ei kerto-nut minulle mitä-än.
who-INDEF NEG:3SG [Link] I:on when-INDEF
‘Nobody told me anything.’
(A.211) indirect negation
Epäilen että poika varast-i mitä-än.
I:doubt that boy steal-PAST(3SG) what-INDEF
‘I doubt that the boy stole anything.’
(A.212) comparative
Ilmasto on täällä mielyttävä-mpi kuin missä-än muualla Euroopa-ssa.
climate is here pleasant-COMPR than where-INDEF else Europe-INESS
‘Here the climate is more pleasant than anywhere in Europe.’

The hyvänsä- (or tahansa-) series is used mainly in the free-choice function.

(A.213) free choice

Kuka tahansa osaa natkaista tämän ongelman.
who INDEF can solve this problem
‘Anyone can solve this problem.’

Since the hyvänsä-series also occurs in the conditional function (cf. A.209), the implicational map
predicts that it can be used in the comparative function. This is indeed the case, but only in the
standard of equative comparison:

(A.214) comparative
Täällä on yhtä mukavaa kuin missä tahansa muualla.
here is as pleasant as where INDEF else
 343

‘It is as pleasant here as anywhere else.’

In the standard of comparison of inequality, the -kaan-series is used, cf. (A.212). In my data
Finnish is the only case of a language where comparison of equality behaves differently from
comparison of inequality. Maybe this case shows that eventually the map will have to be
refined, with the equative comparative being further away from the negation functions.

A.28. Nanay

A. 28.1. Inventory. Nanay (Manchu-Tungusic) has three major series of indefinite pronouns:
the specific -nuu-series, (ii) the non-specific -daa-series, and (iii) the free-choice series marked by
-daa xem aja. The data are given in § 4.3.3.

A.29. Lezgian
A.29.1. Inventory. Lezgian (Nakh-Daghestanian) has three major series of indefinite pronouns,
two of which are derived from interrogatives: (i) the general sa X jat’ani-series, and (ii) the free-
choice x̂ ajit’ani-series. The third series is the negative sa X-ni-series.

interrogative jat’ani-series x̂ ajit’ani-series sa X-ni-series

person: wuž /ni- sa wuž jat’ani wuž x̂ ajit’ani sa kas-ni
thing: wuč /kü- sa wuč jat’ani wuč x̂ ajit’ani sa zat’-ni
property: hix̂ tin sa hix̂ tin jat’ani hix̂ tin ... x̂ ajit’ani —
place: hina sa hina jat’ani hina x̂ ajit’ani sana-ni
time: mus sa mus jat’ani mus x̂ ajit’ani sadra-ni
manner: hik’ sa hik’ jat’ani hik’ x̂ ajit’ani sak’-ni
determiner: hi sa hi jat’ani hi ... x̂ ajit’ani sa...-ni

The sa in the sa X jat’ani-series and in sa kas-ni/sa zat’-ni may optionally be absent. Perhaps also a
special specific sa-series (sa kas ‘somebody’, sa zat’ ‘something’, sadra ‘once’, etc.) must be
recognized.
A.29.2. Origins. Sa is identical to the numeral sa ‘one’. Both jat’ani and x̂ ajit’ani consist of ‘be’ (ja
‘be (standard copula)’, x̂ un ‘become; be’), the conditional suffix -t’a, and the focus marker -ni
‘also, even’. Kas is ‘person’, zat’ is ‘thing’.
A.29.3. Distribution. The distribution of the three series is shown in Figure A.29. The data are
from Haspelmath (1993a: 194-9).

Figure A.29. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

The sa X jat’ani is used for all functions from specific-known to question/conditional:

(A.215) specific known/unknown

a. Za wa-z sa wuč jat’ani luhu-da.
I:ERG you-DAT one what INDEF say-FUT
‘I’ll tell you something.’
b. Gila č ar-ar wuč iz jat’ani xükwe-zma-č .
now letter-PL why INDEF [Link]-NEG
‘Now the letters don’t come anymore for some reason.’
(A.216) irrealis non-specific (imperative)
Sa ni-z jat’ani ewer ce!
one who-DAT INDEF call give:IMPV
‘Call someone!’
(A.217) question/conditional
a. Wa-z sa wuč jat’ani aku-r-t’a, za-z lah!
you-DAT one what:ABS INDEF [Link]-COND I-DAT say:IMPV
 344

‘If you see anything, tell me.’

b. Sa wuž jat’ani ata-nwa-ni?
one who:ABS INDEF come-PERF-Q
‘Has anyone come?’

The x̂ ajit’ani-series is used in the free-choice and comparative functions.

(A.218) free choice

Ada-qh galaz kwe-kaj x̂ ajit’ani sühbet iji-z ž e-da.
she-POSTESS with what-SUBEL INDEF talk do-INF can-FUT
‘With her one can talk about anything.’
(A.219) comparative
Jusuf.a ne-laj x̂ ajit’ani qhsan-diz mani-jar luhu-zwa.
Jusuf(ERG) who-SUPREL INDEF good-ADV song-PL say-IMPF
‘Jusuf sings better than anyone.’

The sa X-ni-series is used in the indirect-negation and direct-negation functions (cooccurring

with verbal negation in the latter).

(A.220) direct negation

Däwe mad sadra-ni sa kasdi-z-ni q’ismet ta-x̂ u-raj.
war again once-even one person-DAT-even destiny NEG-be-OPT
‘May war never again be anyone’s destiny.’
(A.221) indirect negation
Č a-z sada-z-ni hič sa č ’aw.u-z-ni däwe x̂ a-na
we-DAT one-DAT-even ever one time-DAT-even war [Link]
k’an-da-č .
want-FUT-NEG
‘None of us wants there ever to be war.’

A.30. Maltese
A.30.1. Inventory. Maltese (Afro-Asiatic, Arabic) has two series of indefinite pronouns which
are not related to interrogatives: (i) the non-negative xi-series, and (ii) the negative ebda-series.

interrogative xi-series ebda-series

person min xi Ìadd Ìadd
thing xi xi Ìag· a xejn
place fejn xi mkien imkien
time meta xi darba qatt
manner kif b’xi mod (u manjiera) bl’ ebda mod
determiner liema xi ebda

There is also a marginal free-choice determiner kwalunkwe ‘any’ (a recent loan from Italian
qualunque).
A.30.2. Origins. The determiner xi goes back to Classical Arabic š ay? ‘thing, something’. Some
of the second elements in the xi-series are synchronically existing generic nouns (Ìaga ‘thing’,
darba ‘time, i.e. French fois’, mod ‘manner’). The root Ìadd goes back to Classical Arabic ?aÌad
‘one’ (related to waaÌid, Maltese wieÌed ‘one’), and mkien goes back to makaan ‘place’. Qatt goes
back to Arabic qattu ‘never’, and ebda goes back to ?abadan ‘(n)ever’, the adverbail accusative of
?abad ‘eternity’.
A.30.3. Distribution. The distribution of the two series is shown in Figure A.30. The data are
from Haspelmath and Caruana (to appear).

Figure A.30. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice
 345

The xi-series is used in non-negative functions.

(A.222) Xi Ì add qed jaÌdem fil-g· nien.

INDEF anyone PROG he:work:IMPF in:the-garden
‘Somebody is working in the garden.’
(A.223) irrealis non-specific (future)
Xi darba is-sena d-dieÌla mmur l-Afrika.
INDEF time the-year the-entering I:go:IMPF the-Africa
‘I’ll go to Africa sometime next year.’
(A.224) question/conditional
a. Jekk tara xi Ì ag· a <*xejn>, gÌidli.
if you:see:IMPF INDEF thing anything tell:me:IMPV
‘If you see anything, tell me.’
b. Rajt xi Ì ag· a <*xejn>?
you:see:PERF INDEF thing anything
‘Did you see anything?’

The ebda-series is used in the negative and comparative functions.

(A.225) Comparative
Dan huwa l-isbaÌ inkwatru li Ì add qatt ping· a.
this it the-beautiful:COMP picture that anyone ever he:paint:PERF
‘This is the most beautiful painting that anyone has ever painted.’
(A.226) Indirect negation
It-tifla gÌamlitha mingÌajr ebda gÌajnuna.
the-girl she:do:it:PERF without any help
‘The girl did it without any help.’
(A.227) Direct negation (see also § 8.2.1)
It-tifla ma rat xejn.
the-girl NEG she:see:PERF anything
‘The girl didn’t see anything.’

The free-choice function is not expressed by indefinite pronouns. Either universal pronouns are
used (kulÌadd ‘everyone’), or circumlocutions like min trid ‘whoever you want’ (cf. § 3.3.3-4). The
use of kwalunkwe ‘any’ is marginal.

A.31. Hebrew
A.31.1. Inventory. (Modern) Hebrew (Afro-Asiatic, Semitic) has three major series of indefinite
pronouns: (i) the -š ehu-series, which is derived from interrogative pronouns, (ii) the negative-
polarity iš -series (iš ‘man’, davar ‘thing’), (iii) the kol-series, and (iv) the negative af/š um-series.

interrogative -š ehu-series iš -series kol-series af/š um-series

person: mi mi-š ehu/-š ehi iš af eÌad
thing: ma ma-š ehu davar š um davar
place: eyfo eyfo-š ehu kol makom be-š um/af makom
time: matay pa¿am, matay-š ehu kol zman af pa¿am, le¿olam
manner: ex ex-š ehu
determiner: eyze/eyzo eyze-š ehu kol af, š um

Besides š um davar ‘nothing’, there are also two non-derived expressions: klum ‘nothing’, and
me?uma ‘nothing’, which behave in the same way. The kol-series and the iš -series almost occur
in complementary distribution, so they could perhaps be regarded as forming one formally
heterogenous series.
A.31.2. Origins. Two of the three series are formally rather heterogeneous. Pronouns from the
negative af/š um-series are generally made up of one of the determiners af and š um plus a
generic noun (eÌad ‘one’, davar ‘thing’, makom ‘place’, pa¿am ‘time, fois’), but le¿olam is literally ‘in
eternity’. Two of the pronouns of the iš -series are originally general nouns (iš ‘man’, davar
 346

‘thing’), and kol ‘any’ is originally ‘all’. The suffix of the -š ehu-series stems from a parametric
concessive conditional clause (cf. mi-š ehu yihiye ‘whoever he is...’; cf. § 6.2.3).
A.31.3. Distribution. The distribution of the three series is shown in Figure A.31. The data are
from Glinert (1982) (G82), (1989) (G89), and a native speaker (David Gil) .

Figure A.31. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

The -š ehu-series is normal for the specific functions and up to the indirect-negation function.

(A.228) specific
Ra?i-ti mi-š ehu <*iš >.
saw-1SG who-INDEF person
‘I saw somebody.’
(A.229) irrealis non-specific
Ti-r?e ma-š ehu <*davar>.
2sg-will:see what-INDEF thing
‘You will see something.’

In the question, conditional and indirect-negation functions, the -š ehu-series and the iš -series
overlap. The iš -series is rather formal or literary – the colloquial language prefers the -š ehu-
series.

(A.230) conditional
a. Im ti-r?e ma-š ehu, t-agid li. (informal)
if 2SG-will:see what-INDEF 2SG-will:tell to:me
b. Im ti-r?e davar, haged li. (formal)
if 2SG-will:see thing tell:IMPV to:me
‘If you see anything, tell me.’
(A.231) question
Ha?im amr-u zot le-iš / -mi-š ehu? (G82: 463)
Q said-3PL that to-person who-INDEF
‘Did they tell this to anyone?’
ii
(A.232) indirect negation
Ani lo Ìoš evet š e-nir?a iš / mi-š ehu ba-reÌov.
I NEG think:F that-was:seem(3SG) person who-INDEF in:the-street
‘I don’t think that anyone was seen in the street.’

When the negation is in the same clause, either the iš -series or the af/š um-series may be used:

(A.233) direct negation

a. Lo ra?i-ti iš / afeÌ ad.
NEG saw-1SG person INDEF one
‘I did not see anybody.’
b. Lo ra?i-ti kol/ š um student.
NEG saw-1SG any any student
‘I did not see any student.’
c. Af eÌ ad/ iš lo ba. (G82: 4 34,)
INDEF one person NEG came(3SG)
‘Nobody came.’

For the free-choice function, kol ‘every, any’ or kol...š ehu is used:

(A.234) a. Ti-rš om kol ktovet š ehi. (G82: 445)

2SG-will:note any address INDEF
‘Note down any address.’
b. Ani muxan la-gur be-dira kol-š ehi. (G89)
I willing to-live in-apartment any-INDEF
 347

‘I am willing to live in any apartment.’

A.32. Hausa
A.32.1. Inventory. Hausa (Chadic, Afro-Asiatic) has two series of indefinite pronouns, (i) the
emphatic koo-series which is based on interrogative pronouns, and (ii) the non-emphatic wani-
series, which consists of the determiner wani plus generic nouns.

interrogative koo-series wani-series

person: wàa, wàanee koo-waa wani/wata
thing: mèe, mèenee koo-mee wani àbù (‘some thing’)
place: ìnaa koo-’ìnaa wani wurii (‘some place’)
time: yàushè/yàushe koo-yàushè, -yàushe wani lookàcii (‘some time’)
manner: yàayàa koo-yàayàa
determiner: wànè/wàcè koo-wànè/-wàcè wani/wata

A.32.2. Origins. The particle koo has a plethora of other uses, described in Meyers (1974): ‘or’,
question particle, ‘whether’, ‘even’, ‘even though’. In all likelihood, its use as a scalar additive
focus particle (‘even’) is responsible for its use as indefiniteness marker. The determiner wani is
from the same pronominal root wa- as the interrogatives wàa, wànè (cf. Attouman and Caron
1984: 12-13), but it need not be interrogative-based because this root is also used for
demonstratives (wannàn ‘this’) and relatives (wandà ‘(the one) which’).
A.32.3. Distribution. The distribution of the three series is shown in Figure A.32. The data are
from Ma Newman (1990) (M90), Attouman and Caron (1984) (AC84), Abraham (1946) (A46),
and Marit Lobben (p.c.).

Figure A.32. direct

question indirect negation
specific irrealis negation
non-specific
conditional comparative
free-choice

The wani-series is used in all non-emphatic functions.

(A.235) specific
Wani àbù yaa sàamee shi. (M90: 252)
some thing 3SG:PAST happen him
‘Something has happened to him.’
(A.236) irrealis non-specific (future)
Wata raanaa sáà daawoo. (M90: 252)
some day 3PL:FUT return
‘Some day they’ll return.’
(A.237) question/conditional
a. Kaa ga wani mùtûm à oofìs? (M90: 252)
2SG:PAST see some man in office
‘Did you see someone at the office?’
b. Ìdan kaa jíì wani àbù, kà taashè ni.
if 2SG:PERF hear some thing 2SG:SBJV wake me
‘If you hear anything, wake me up.’

The koo-series is used in the comparative and free-choice functions.

(A.238) comparative
Yaa san tà fìye dà koo-waa. (M90: 11)
3SG:PAST know her better than INDEF-who
‘He knows her better than anyone else.’
(A.239) free choice
Anàa saamùn-sà koo-’ìnaa. (M90: 12)
one:PRES get-3SG INDEF-where
‘You can get it anywhere.’
 348

It is also used in the direct-negation function (cooccurring with verbal negation), but only after
the negation, i.e. in non-subject position. (Unfortunately, I lack data on the indirect-negation
function. When the negated indefinite is in subject position, a paraphrase must be used (cf. §
8.2.5).

(A.240) direct negation

Bà mù hàdu dà koo-waa à hanyàa ba. (M90: 11)
NEG 1PL:PAST meet to INDEF-who on road NEG
‘We didn’t meet anyone on the road.’

The koo-series is apparently also used in questions. Abraham (1946) has examples like the
following:

(A.241) a. Koo-mèe ya batà? (A46: 535)

INDEF-what 3SG:PAST lost
‘Is anything lost?’
b. Koo-wàa ya zoo? (A46: 541)
INDEF-who 3SG:PAST come
‘Has anyone come?’

In these examples, koo-mèe and koo-wàa have a low tone on the root syllable, and it appears that
this tonal difference corresponds to a functional difference. The high-tone counterpart of
(A.241a) has a different meaning.

(A.242) Koo-mee ya batà?

INDEF-what 3SG:PAST lost
‘Is everything lost?’

Unfortunately, my data do not allow me to come to any clear conclusions with respect to this
question, but these data are reminiscent of the accentual contrasts observed in § 5.7.3.

A.33. Swahili
A.33.1. Inventory. Swahili (Niger-Congo, Bantu) has only two major series of indefinite
pronouns: (i) a series consisting of generic nouns, and (ii) a free-choice and negative series
consisting of the determiner CL-o CL-ote (where CL stands for a gender prefix).

interrogative generic-noun-series CL-o CL-ote-series

person: nani mtu (‘person’) mtu ye yote
thing: nini neno, kitu (‘thing’) kitu cho chote, neno lo lote
place: wapi mahali (‘place’) mahali po pote
time: lini wakati (‘time’) wakati wo wote
manner: -je kwa njia yo yote, vyo vyote
determiner: -pi -moja (‘one’) CL-o CL-ote

It is quite conceivable that the generic nouns have not yet been grammaticalized as pronouns
yet. I describe them as an indefinite series for the sake of comparability with the other
languages.
In the CL-o CL-ote-series, the generic noun may also be omitted when the gender and the
context make it clear which generic noun is to be understood. Thus, po pote is common for
mahali po pote (because this gender is restricted to a few nouns), ye yote, cho chote and lo lote are
possible variants, but *wo wote ‘ever’ is not possible because this gender is too large.
Another indefinite determiner is fulani ‘a certain’ (from Arabic fulaan).
A.33.2. Origins. All I can say about CL-o CL-ote is that it is based in CL-ote, which means ‘all,
whole’ (e.g. daftari yote ‘the whole notebook’, wanafunzi wote ‘all the pupils’).
A.33.3. Distribution. The distribution of the two series is shown in Figure A.33. The data are
from various reference works and from the New Testament.

Figure A.33. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
 349

conditional comparative
free-choice

Generic nouns can be used in all functions except the free-choice and the comparative
functions.

(A.243) specific
a. Mtu a-li-ni-gusa. (NT, Lk 8.46)
man 3SG-PAST-me-touch
‘Somebody has touched me.’
b. A-na neno la ku-mw-arifu.
3SG-has word of INF-him-tell
‘She has something to tell him.’
(A.244) irrealis non-specific (subjunctive complement)
Yesu a-li-mw-ambia... kwamba a-wa-pe maskini kitu.
Jesus 3SG-PAST-him-tell that 3SG-them-give:SBJV poor thing
‘Jesus told him to give something to the poor.’ (NT, John 13.29)
(A.245) question/conditional
a. Wa-ona kitu? (NT, Mk 8.23)
2SG-see thing
‘Do you see anything?’
b. Kama mtu a-ki-wa-ambia neno, seme-ni, Bwana a-na haja na-o.

if man 3SG-if-you-tell word say-PL Lord 3SG-has need with-3PL

‘AIf anyone says anything to you, say, the Lord needs them.’ (NT, Mt 21.3)
(A.246) indirect negation
Tokea hapo ha-i-ja-siki-wa ya kuwa mtu a-me-ya-fumbua macho ya kipofu.
since then NEG-hear-PASS that man 3SG-PERF-them-open eyes of blind
‘Since the world began was it not heard that any man opened the eyes of one that
was born blind.’ (NT, John 9.32)

In the direct-negation function, bare generic nouns can only be used postverbally (A.247). If the
indefinite is the subject, a different strategy has to be resorted to (§ 8.2.5).

(A.247) direct negation

Wala ha-wa-kumw-ambia mtu neno. (NT, Mk 16.8)
and NEG-them-him-say man thing
‘And they did not say anything to anybody.’

The CL-o CL-ote-series can be used in the free-choice and negative functions, as well as in
questions (at least rhetorical questions) and conditionals, where it seems to have emphatic
value. I have no data for the comparative function, but I predict that CL-o CL-ote is possible there
as well.

(A.248) free choice

Daktari yule a-ta-ponya magonjwa yo yote.
doctor that 3SG-FUT-cure disease any
‘That doctor will cure any disease whatsoever.’
(A.249) direct negation
Ninyi ha-m-jui neno lo lote. (NT, John 11.49)
you NEG-it-know word any
‘You don’t know anything.’
(A.250) question/conditional
a. A-li-mw-ambia malaika yupi wakati wo wote, Ndi-we mwana-ngu?
3SG-past-him-say angel which time any you-are son-1SG
‘To which of the angels did he ever say, You are my son?’ (NT, Heb 1.5)
b. M-ki-ni-omba neno lo lote kwa jina langu, ni-ta-li-fanya.
you-COND-me-ask word any by name mine 1SG-FUT-it-do
‘If you ask anything in my name, I will do it.’ (NT, John 14.14)

A.34. Georgian
 350

A.34.1. Inventory. Georgian (Kartvelian) has five series of indefinite pronouns, all of them
derived from interrogatives: (i) the specific -Vac-series, (ii) the non-specific -me-series, and (iii)-
(iv) the three negative series marked by ar(a)-, ver(a)- and nur(a)-. (Below only the ara-series is
given.)

interrogative -Vac-series -me-series ara-series

person: vin vi-Vac vin-me ara-vin
thing: ra ra-Vac ra-me ara-peri, ara-ra
place: sad(a) sad-Vac sad-me ar-sad
time: rodis rodis-Vac rodis-me ara-sodes
manner: rogor rogor-Vac rogor-me
determiner: romeli romeli-Vac romeli-me ara-vitari

A.34.2. Origins. The origin of -me is unknown. The suffix -Vac consists of -Va ‘only’ and -c(a)
‘also’. The prefixes ara-, vera- and nura- derive from the negative particles ar, ver and nu ‘not’.

A.34.3. Distribution. The distribution of the four series is shown in Figure A.34. The data are
from Tschenkéli 1958 (T58), Vogt 1971 (V71), Revaz Tchantouria, Winfried Boeder (p.c.) and
dictionaries (cf. also Apridonidze 1972, Topuria 1925).

Figure A.34. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

The -Vac-series is restricted to specific functions.

(A.251) specific known/unknown

a. Es c’ignisad-V ac v-iš ove. (V71: 47)
this book where-INDEF 1SG-found
‘I found this book somewhere [I could say where].’
b. Movida vi-V ac <*vin-me> rusi. (V71: 47)
came who-INDEF who-INDEFRussian
‘Some Russian person has come [I don’t know him/her].’

In all non-emphatic non-specific functions, including indirect negation and comparative, the -
me-series is used.

(A.252) irrealis non-specific (imperative, modal, future, distributive)

a. DauJaxet vi-s-me!
call:IMPV:PL who-DAT-INDEF
‘Call somebody!’
b. Es c’igni sad-me unda v-iš ovo. (V71: 47)
this book where-INDEF necessary 1SG-find:SBJV
‘I have to find this book somewhere.’
(A.253) question/conditional
a. Mosula vin-me <*vi-V ac>? (V71: 47)
came who-INDEF who-INDEF
‘Has anyone come?’
b. Tu vin-me movides, utxari.
if who-INDEF comes tell:IMPV
‘If anyone comes, tell (me).’
(A.254) indirect negation
Vera-prit verdaitanxmes rom rogor-me sad-me sc’avla
NEG-way NEG persuade:him that how-INDEF where-INDEF study
ganegrJo.
he:continue:SBJV
‘They couldn’t persuade him to continue his studies anywhere in any way.’
 351

(A.255) comparative
Man kartuli ena uk'et icis vidre sxva vin-me-m.
he:ERG Georgian language better knows than other who-INDEF-ERG
'He knows Georgian better than anyone else.'

In the direct-negation function, one of the three negative series (ara-, vera-, nura-) is used,
normally together with verbal negation (though this may optionally be omitted). Ar is the
unmarked negator, ver contains an additional semantic component of possibility, and nu is used
in prohibitive sentences.

(A.256) a. Me ara-sodes ara-vis-gan ara-pers ar mi-v-iVeb. (T58: 223)

I NEG-when NEG-who-from NEG-thing NEG PREV-1SG-get
‘I’ll never get anything from anybody.’
b. Mas ara-vis-tvis utkvams.
he NEG-who-for has:told
‘He hasn’t told anybody.’
c. Nura-vi-s nu et’q’vi! (V71: 48)
NEG-who-DAT NEG tell
‘Don’t tell anybody!’

The free-choice sense is expressed by means of the adjective nebismieri 'arbitrary, any', which
can hardly be considered an indefinite pronoun. Other possibilities are the universal quantifier
q’vela ‘every’ and the free relative clause vinc ginda 'whoever you want'.

A.35. Kannada
A.35.1. Inventory. Kannada (Dravidian) has three series of indefinite pronouns, all of them
derived from interrogatives: (i) the specific -oo-series, (ii) the non-specific -aadaruu-series, and
(iii) the negative and free-choice -uu-series.

interrogative -oo-series -aadaruu-series -uu-series

person: yaaru yaar-oo yaar-aadaruu yaar-uu
thing: eenu een-oo een-aadaruu een-uu
place: yalli, elli elliy-oo elliy-aadaruu elliy-uu
time: yaavaaga yaavaagal-oo yaavaagal-aadaruu yaavaagal-uu
manner: heege heeg-oo heeg-aadaruu heeg-uu
amount: yas5t5u, es5t5u es5t5-oo es5t5-aadaruu es5t5-uu
determiner: yaava-nu/-l5u yaava-n-oo/-l5-oo yaavanaadaruu yaava X-uu

A.35.2. Origins. The suffix -oo is identical to -oo ‘or; question marker’. The suffix -aadaruu also
has the meaning ‘even, German auch nur’ and consists of aa-d- ‘be(come)’, -are ‘if’ and -uu (i.e.,
the original meaning is ‘even if it be(come)’). The suffix -uu is identical to -uu ‘also, even’.

A.35.3. Distribution. The distribution of the three series is shown in Figure A.35. The data are
from Bhat (1981) (B81), Sridhar (1990) (S90), and native speakers (Anilkumar Belvadi, D.N.S.
Bhat) .

Figure A.35. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

The -oo-series is only used in the specific-unknown function – it cannot be used if the speaker
knows the identity of the referent.

(A.257) specific unknown

Yaar-oo <*yaar-uu, *yaar-aadaruu> bandaru; (*yaaru uuhisi.) (B81: 7)
who-INDEF came who guess
‘Someone <*anyone> came; (*guess who).’
 352

The non-specific -aadaruu-series is used in irrealis and question/conditional functions.

(A.258) irrealis non-specific (imperative)

Ellig-aadaruu <*ellig-oo> hoogu. (B81: 6)
where-INDEF go
‘Go somewhere.’
(A.259) question/conditional
a. Raamu ellig-aadaruu <*ellig-oo, *ellig-uu> hoodan-oo? (B81: 6)
Ramu where-INDEF go-Q
‘Did Ramu go anywhere?’
b. Yaar-aadaruu <*yaar-oo, *yaar-uu> band-are heel5-utt-eene. (B81: 7)
who-INDEF come-COND tell-FUT-1SG
‘I’ll tell you if anyone comes.’

The -uu-series is used in the free-choice function, in the direct- and indirect-negation functions
(cooccurring with verbal negation in the latter), and in the comparative function (although ella
‘all’ is preferred there).

(A.260) free choice

Raamu ellig-uu hoodaanu. (B81: 4)
Ramu where-INDEF [Link]
‘Ramu may go anywhere.’
(A.261) direct negation
Illige yaar-uu baral-illa. (S90: 256)
hither who-INDEF come-NEG
‘No one came here.’
(A.262) indirect negation
Yaar-uu baruvudu avaś ya illa.
who-INDEF come:MASD necessary NEG
‘It is not necessary for anyone to come.’
(A.263) comparative
Illi have beere yaava kad5e-g-intal-uu cann-aag-ide.
here climate other which place-DAT-than-INDEF good-become-is
‘Here the climate is better than in any other place.’

(Better: beer-ell-aa kad5e-g-intal-uu ‘than in all other places’.)

A.36. Chinese
A.36.1. Inventory. Mandarin Chinese (Sino-Tibetan) makes extensive use of the bare
interrogatives as indefinites. In addition, two series are derived by the words dō u and yě , and
the determiner rènhé ‘any’ can be considered a series of its own.

interrogative dō u-series yě -series

person: shéi shéi dō u shéi yě
thing: shénme shénme dō u shénme yě
place: nǎ r, shénme dìfang shénme dìfang dō u shénme dìfang yě
time: shénme shíhou
manner: zě nme
determiner: ně i ně i ... dō u

Generic nouns (e.g. ren ‘person; someone’) are also commonly used as indefinites.

A.36.2. Origins. Yě and dō u mean ‘even, also; every, all’. Rènhé is composed of the old
interrogative hé ‘what?’ plus rèn ‘allow; appoint’. Perhaps the current meaning ‘any’ arose from
an original meaning ‘choose what’ > ‘anything’ (‘choose’ is similar to ‘appoint’).

A.36.3. Distribution. The distribution of the three series is shown in Figure A.36. The data are
from Li (1992) (L92), Li and Thompson (1981) (LT81), Tsai 1990 (T90), reference works , and
native speakers (Fengxiang Li, Wenfang Zhang) .
 353

Figure A.36. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

The bare interrogatives are used in all non-specific non-emphatic functions.

(A.264) irrealis non-specific (imperative)

Chī diǎ n shénme zài zǒ u ba! (L92: 152)
eat [Link] what then go PT
‘Please eat a little something before you leave.’

As shown in detail in Li (1992), bare interrogativs are also used in contexts of uncertainty and
inference.

(A.265) a. Tā hǎ oxiàng xǐ huan shénme. (L92: 131)

he seem like what
‘He seems to like something.’
b. Tā kàndao (le) shénme le. (L92: 133)
he see PERF what perf
‘He must have seen something.’

Examples of the question/conditional functions and the direct negation functions are given
below (unfortunately, I lack data for the indirect negation function).

(A.266) question/conditional
a. Shéi xǐ huan tā ma? (L92: 128)
who like him Q
‘Does anyone like him?’
b. Nǐ yàoshì yào shénme dehuà, èn língr zhòu xíngle.
you if want what press
‘If you want something, ring the bell.’

In the direct-negation function, only shénme is perfect, shéi being less acceptable, and nage (N)
‘which (N)’ is unacceptable (Li 1992: 150):

(A.267) direct negation

a. Tā bù wèi shénme rén zuò shì. (L92:127)
he not for what man do thing
‘He does not work for anyone.’
b. Tā bù xǐ huan shénme.
he not like what
‘He does not like anything.’
c. ?Tā bù xǐ huan shéi.
he not like who
‘He does not like anyone.’
d. *Tā bù xǐ huan nǎ ge (rén).
he not like which man
‘He does not like anyone.’

However, bare interrogatives are not admissible in specific contexts that lack an element of
iii
uncertainty. In such cases, generic nouns are used.

(A.268) specific (known/unknown)

Yǒ u rén dǎ diànhuà le.
exist man hit phone PERF
(Wo bù zhī dao shì shéi./ Cā i yī cā i shì shéi.)
I not know is who guess one guess is who
‘Someone called. (I don’t know who./Guess who.)’
 354

The dō u- and yě -series are used interchangeably in the direct-negation function.

(A.269) direct negation

a. Wǒ shéi-yě bù rènshi. (LT81: 529)
I who-INDEF not know
‘I don’t know anyone.’
b. Tā shéi-dō u bù xìnren. (LT81: 528)
she who-INDEF not trust
‘She doesn’t trust anyone.’
c. Wǒ shénme dìfang dō u/yě bù qù.
I which place INDEF not? go?
‘I’m not going anywhere.’
d. Shèi-dō u bù cō ngmíng.
who-INDEF not intelligent
‘No one is intellingent.’

Two things are remarkable about these forms. First, they only occur preverbally, which Li and
Thompson (1981: 530) explain by the fact that dō u and yē are adverbs and hence must occur
preverbally. If they are adverbs and follow their own syntactic rules, it may seem doubtful
whether the combinations shéi-yě , shénme-dō u etc. can be regarded as single constituents at all,
iv
and hence whether yě and dō u can be considered as indefiniteness markers. This is done here
because they never non-adjacent to interrogative pronouns (i.e. *Tā dō u bu xìnren shéi is not a
possible alternative to (A.36.6b)).
Second, shéi-dō u etc. are also used as universal quantifiers (e.g. wǒ shéi-dō u xǐ huan ‘I like
everyone’), so sentences (A.36.6a-d) should perhaps not be regarded as instances of indefinite
pronouns, but as a rare type of replacement strategy: Instead of saying ‘She doesn’t trust
anybody/She trusts nobody’, Chinese says ‘She doesn’t trust everybody’, with wide scope of
‘everybody’ (i.e. for every person x, she doesn’t trust x). I have to leave this question open here.

The indefinite determiner rènhé is used (mainly in conjunction with the particle/adverb dō
u) to render the free-choice and comparative functions.

(A.270) free choice

Rènhé shíhou nǐ dō u kě yǐ lái.
any time you INDEF can come
‘You can come anytime.’
(A.271) comparative
Tā pǎ o de bǐ rènhé yī ge nánshē ng dō u kuài.
he run PT than any one boy INDEF fast
‘He runs faster than any boy.’

Finally, the determiner rènhé may also be used in the direct-negation and indirect-negation
functions.

(A.272) direct negation

Tā bù néng táo dào rènhé dìfang qu.
he NEG can escape to any place go
‘He cannot escape anywhere.’
(A.273) indirect negation
Wō bù xiǎ ngxin rènhé rén lái le.
I not think any man come PERF
‘I don’t think that anyone came.’

These generalizations should be approached with some caution – my data on Chinese are less
complete than I would have wished.

A.37. Ancash Quechua

A.37.1. Inventory. Ancash Quechua uses only the bare interrogatives and one interrogative-
derived series, the non-specific -pis-series.
 355

interrogative -pis-series
person: pi pi-pis
thing: ima ima-pis
place: may may-pis
time: imay imay-pis
manner: imanaw imanaw-pis

A.37.2. Origins. The suffix -pis also means ‘also, even’.

A.37.3. Distribution. The distribution of the three series is shown in Figure A.37. The data are
from a native speaker (Hernán Aguilár) (cf. also Parker 1976: 74-5). The data from Weber (1989),
where a closely related dialect is described, are also taken into account.

Figure A.37. direct

question indirect negation
specific irrealis negation
unknown non-specific
conditional comparative
free-choice

The bare interrogative pronouns are used in the specific-unknown function.

(A.274) specific unknown

a. Ima-ta-chi wambra yurapa waqta-n-chaw riqa-rqa-n.
what-ACC-VAL boy tree behind-3SG-LOC see-PAST-3SG
‘The boy saw something behind the tree.’
b. Pi-wan-chi qanyan awtobus-chaw parla-rqu-u.
who-COMIT-VAL yesterday bus-LOC talk-PAST-1SG
‘I talked to someone on the bus yesterday.’

The -pis-series is used in all non-specific functions.

(A.275) irrealis non-specific (imperative, future)

a. Mana musya-pti-iki-qa pi-ta-pis tapuku-y.
not know-CONV-2SG-TOP who-ACC-INDEF ask-IMPV
‘If you don’t know, ask somebody.’
b. Imay-pis Afrika-man-mi aywa-shaq.
when-INDEF Africa-ALL-VAL go-FUT.1SG
‘I’ll go to Africa sometime next year.’
(A.276) question/conditional
a. Ima-ta-pis rika-rqu-nki-ku?
what-ACC-INDEF see-PAST-2SG-Q
‘Did you see anything?’
b. Ima-ta-pis rika-rnin-qa willa-ma-y.
what-ACC-INDEF see-2SG-TOP tell-1SG-IMPV
‘If you see anything, tell me.’
(A.277) comparative
Ollqu wambra-m pi-piq-pis ras ayqiqninmi.
male boy-VAL who-ABL-INDEF fast runs
‘The boy can run faster than anyone.’
(A.278) free choice
Pi-pis kay problema-ta-qa atinman-mi.
who-INDEF this problem-ACC-TOP solves
‘Anyone can solve this problem.’

The -pis-series is also used in the negative functions. The scope of negation is indicated by the
discontinuous negators mana/ama/ni ... -tsu.

(A.279) direct negation

a. Ama ima-pis ni-y-chu. (Weber 1989: 21)
not what-INDEF say-IMPV.2SG-NEG
‘Don’t say anything!’
b. Pi-pis ni ima-ta-pis ni-ma-rqa-n-tsu.
 356

who-INDEF [Link] what-ACC-INDEF tell-1SG-PAST-3SG-NEG

‘Nobody told me anything.’
(A.280) indirect negation
Mana muna-shkaa-chu pi-pis musyanan-ta. (Weber 1989: 340)
not want-PERF.1SG-NEG who-INDEF know-ACC
‘I didn’t want anyone to know.’

A.38. Japanese
A.38.1. Inventory. Japanese has three series of indefinite pronouns, all of them derived from
interrogatives: (i) the non-negative ka-series, (ii) the negative mo-series, and (iii) the free-choice
-demo-series.

interrogative ka-series mo-series demo-series

person: dare dare-ka dare-mo dare-demo
thing: nani nani-ka nani-mo nani-demo
place: doko doko-ka doko-mo doko-demo
time: itu itu-ka itu-demo
manner: doo doo-ka doo-demo
amount: ikura ikura-ka ikura-demo
ikutu ikutu-ka ikutu-demo
determiner: dore dore-ka dore-mo dore-demo

A.38.2. Origins. The suffix -ka is formally identical to ka ‘or; question particle’ (cf. Tamba-Mecz
1984). The suffix -mo is formally identical to -mo ‘also’. The suffix -demo is is from de mo ‘even if
it is’ (demo is also a focus particle ‘even’).

A.38.3. Distribution. The distribution of the three series is shown in Figure A.38. The data are
from McGloin (1976) (M76), Ohno (1984) (O84), and native speakers (Masa Koizumi, Yoshiko
Ono) , and various reference works. See also Yamada (1993), Chang and Labrune (1994).

Figure A.38. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

The ka-series is used in non-negative functions.

(A.281) specific known/unknown

Dare-ka kara denwa at-ta kedo,
who-INDEF from phone be-PAST though
‘Someone called,’
— dare kara da ka wakara-na-i.
who from be:PRES Q know-NEG-PRES
‘I don’t know who.’
— dare kara da ka ate-te goran.
who from be:PRES Q [Link]-CONV try:IMPV
‘guess who!’
(A.282) irrealis (hortative, distributive)
a. Dare-ka ni ki-ite mi-masyoo.
who-INDEF DAT ask-CONV try-POL:HORT
‘Let’s ask somebody.’
b. Minna-ga nani-ka-o tabe-ta. (O84: 238)
every-NOM what-INDEF-ACC eat-PAST
‘Everybody ate something (non-specific).’
(A.283) question/conditional
a. Rusutyuu-ni dare-ka <*dare-mo> ki-masi-ta ka?
[Link]-DAT who-INDEF who-INDEF come-POL-PAST Q
‘Did somebody/anybody come while I was gone?’ (M76: 409)
 357

b. Nani-ka <*nani-mo> tabe-tara, eiga e ture-te it-te

what-INDEF what-INDEF eat-COND movie to take-CONV go-CONV

age-na-i yo. (M76: 415)
give-NEG-PRES ASS
‘If you eat something/anything, I won’t take you to the movie.’

The -mo-series is used in the negative functions and in the comparative function. In the case of
direct negation, it cooccurs with verbal negation.

(A.284) comparative
Kono syoonen-wa kono kurasu-no dare-yori-mo hayaku hasir-u.
this boy-TOP this class-GEN who-from-INDEF fast run-PRES
‘This boy can run faster than anyone in his class.’
(A.285) direct negation
Dare-mo kanojo-o aisi-te i-na-i.
who-INDEF she-ACC love-CONV DUR-NEG-PRES
‘Nobody loves her.’
(A.286) indirect negation
Kyoo-wa dare-mo ku-ru to omow-ana-i. (M76: 405)
today-TOP who-INDEF come-PRES SBOR think-NEG-PRES
‘I don’t think anybody is coming today.’

The demo-series is used in the free-choice function.

(A.287) free choice

Dare-demo deki-masu.
who-INDEF possible-POL-PRES
‘Anybody can do it.’

A.39. Korean
A.39.1. Inventory. Korean has six series of indefinite pronouns, three of which are derived from
interrogatives: (i) the general -nka-series, (ii-v) the four free-choice series in -na and -tunci,
which may be derived either from interrogatives or from general nouns, in which latter case
they are preceded by the determiner amu, and (vi) the negative -to-series, which is also derived
from amu + general noun. In addition, bare interrogatives are also commonly used as indefinite
pronouns.

interrogative -nka-series -na-series -tunci-series amu X-to-series

person: nwukwu nwukwu-nka nwukwu-na nwukwu-tunci amu (salam)-to

thing: mues mue-nka mues-ina mues-itunci amu kes-to
property: etten etten ...-na etten ...-tunci
place: eti eti-sunka eti-na eti-tunci amu kos-to
time: encey encey-nka encey-na encey-tunci amu ttay-to
determiner: enu enu ...-na enu ...-tunci amu X-to

For reasons of space, the amu X-na-series and the amu X-tunci-series are omitted from the above
table – they are completely analogous to the amu X-to series. There is also an incomplete -to-
series derived from interrogatives (nwukwu-to ‘nobody’, but *mues-to ‘nothing’).

A.39.2. Origins. The suffix -nka seems to come from i- ‘be’ + -n (present tense) + -ka ‘question
particle’. The suffix -na is also used with the meanings ‘or; for example; approximately’; it
seems to come from the Adversative mood (marked by -(u)na) of i- ‘be’. The Adversative is also
the mood used in parametric concessive conditional clauses, so nwukwu-na would come from
‘whoever it may be’ (cf. § 6.2.3). The suffix -tunci (also used in the sense ‘or’) has a similar
origin: -tunci is a verbal suffix that can also be used in parametric concessive conditional
clauses. The suffix -to also means ‘also’.

A.39.3. Distribution. The distribution of the series is shown in Figure A.39. Since the indefinites
derived frominterrogatives do not seem to differ in their distribution from the indefinites based
 358

on amu X-, the two types are not distinguished. The data are from Martin and Lee (1986)
(ML69), Chang and Labrune (1994) as well as from native speakers (Hee-Rahk Chae, Kyu-Ryun
Choi, and Du-Pyo Hong). See also Lee (1983).

Figure A.39. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

The bare interrogatives are the most commonly used indefinites. The -nka-indefinites are more
formal, less colloquial.

(A.288) specific known/unknown

a. Nwu’-ka/ nwukwu-inka-ka kel-ess-ta.
who-NOM who-INDEF-NOM call-PAST-DECL
‘Somebody called. (I didn’t catch her name./Guess who.’)
b. Ku sonyen-un mues-ul/ mues-inka-lul po-ass-ta.
this boy-NOM what-ACC what-INDEF-ACC see-PAST-DECL
‘The boy saw something.’
(A.289) irrealis non-specific (imperative, hortative)
a. Molu-myen, nwukwu-eykey/ nwukwu-nka-eykey mul-ela.
ignorant-CONV who-DAT who-INDEF-DAT ask-IMPV
‘If you don’t know, ask somebody.’
b. Eti ka-psita. (ML69: 411)
where go-HORT
‘Let’s go somewhere.’
(A.290) question/conditional
a. Ku sonye-kamues-ul/ mue-nka-lul <*mues-ina/ *mues-itunci>
this girl-TOP what-ACC what-INDEF-ACC what-INDEF what-INDEF
chacu-myen, kunye emeni-eykey malha-lkesi-ta.
find-CONV she mother-DAT tell-FUT-DECL
‘If the girl finds anything, she will tell her mother.’
b. Chelswu mues-ul/ mue-nka-lul <*mues-ina *mues-itunci>
Chelswu what-ACC what-INDEF-ACC what-INDEF what-INDEF
ilk-ess-ni?
read-PAST-Q
‘Did Chelswu read anything?’

The -na- and -tunci-series are used interchangeably in the free-choice function.

(A.291) Nwukwu-na/ nwukwu-tunci/ amu-na/ amu-tunci i muncey-lul

who-INDEF who-INDEF any-INDEF any-INDEF this problem-ACC

phul swu-iss-ta.
solve do-can-DECL
‘Anybody can solve this problem.’

The -to-series is used in the indirect and direct negation functions. In the direct-negation
function, it cooccurs with verbal negation. In the indirect-negation function, the bare
interrogatives and the -nka-series are also possible.

(A.292) indirect negation

Nwukwu-to/ nwukwu-lul/ nwukwu-nka-lul chacki-ka pulkanung ha-ta.
who-INDEF who-ACC who-INDEF-ACC find-TOP impossible be-DECL
‘It’s impossible to find anyone.’
(A.293) direct negation
a. Ku sonye-nun amu-to/ nwukwu-to an po-ass-ta.
this girl-TOP any-INDEF who-INDEF NEG see-PAST-DECL
‘The girl did not see anybody.’
b. Amu-to amu kes-to malha-ci anha-ss-ta.
any-INDEF any thing-INDEF say-CONV NEG-PAST-DECL
 359

‘Nobody said anything.’

Furthermore, the -to-series is used in the standard of comparison:

(A.294) comparative
Yeki-ka enu kos pota to te alumtap-ta.
here-TOP which place than INDEF more beautiful-DECL
‘Here it is more beautiful than anywhere.’

A.40. Basque
A.40.1. Inventory. Basque has four major series of indefinites, all of them derived from in-
terrogatives: (i) the non-emphatic bait-series, (ii) the negative-polarity series in i-, and (iii-iv)
the two free-choice series in edo- and -nahi.

interrogative -bait-series i-series edo-series nahi-series

person: nor nor-bait i-nor edo-nor nor-nahi
thing: zer zer-bait e-zer edo-zer zer-nahi
place: non non-bait i-non edo-non non-nahi
time: noiz noiz-bait i-noiz edo-noiz noiz-nahi
manner: nola nola-bait i-nola edo-nola nola-nahi
determiner: zein — — edo-zein zein-nahi

In addition, non-standard dialects have an X edo X-series (nor edo nor ‘someone’, zer edo zer
‘something’), and a neh-series (nehor ‘anyone’, nehon ‘anywhere’, nehoiz ‘ever’). When used in
their negative function, the pronouns of the i-series may optionally be followed by ere (‘even’).

A.40.2. Origins. The suffix -bait is formally identical to the verbal prefix bait- ‘because’, but its
relation to it is obscure. Other potential connections include ba- ‘if’, and bai ‘yes; and’. The
prefix edo- is identical to edo ‘or, probably’, and the suffix -nahi is identical to the stem nahi
‘want’.

A.40.3. Distribution. The distribution of the three series is shown in Figure A.40. The data are
from Saltarelli 1988 (S88), Aulestia 1989 (A89), and native speakers (Karmele Rotaetxe, José
Hualde).

Figure A.40. direct

question indirect negation
specific specific irrealis negation
known unknown non-specific
conditional comparative
free-choice

The bait-series is used in non-negative polarity environments.

(A.295) specific unknown

Nor-bait-i utzi nion argazki-makina eta ez naiz oroi-tzen
who-INDEF-DAT lend I:it:[Link] photo-machine and NEG I:it remember-HAB
nor-i. (S88: ex. 1070)
who-DAT
‘I lent the camera to someone and I do not remember to whom.’
(A.296) irrealis non-specific (intention, hypothetical)
a. Nora-bait joa-te-ko gogoa dut, baina ez dakit nora.
where-INDEF go-HAB-REL thought I:have:it but NEG I:know:it where
‘I have an urge to go somewhere but I do not know where.’ (S88: ex. 976b)
b. Pozik hartu-ko nuke nor-bait-ek laguntzea. (S88: ex. 158)
gladly accept-FUT I:would who-INDEF-ERG helping
‘I would gladly accept someone helping me.’

In the specific-known function, the bait-series is apparently not possble, and the numeral bat
'one' is used:
 360

(A.297) specific known

Kanpoa-n zengoze-n-en-ean baten batek dei egin zizun.
away-LOC be-2SG-REL-LOC one:GEN one:ERG call make he:you:has
Esan nor.
say who
'Someone called while you were away. Guess who it was.'

The i-series is used in negative-polarity contexts: questions/conditionals, comparatives, direct

and indirect negation. In the direct negation function, it cooccurs with verbal negation.

(A.298) question/conditional
a. E-zer <??zer-bait> jan nahi al duzu? (A89: 206)
INDEF-what what-INDEF eat want Q you:it
‘Would you like to eat something?’
b. E-zer <??zer-bait> ikusten ba-duzu, esaidazu.
INDEF-what what-INDEF see COND-you:have tell:you:me
'If you see anything, tell me.'
(A.299) comparative
Donostia-ko eguraldia i-non/ edo-non baino
Donostia-ATTR weather INDEF-where INDEF-where than
atsegin-ago-a da.
pleasant-COMPR-ART is
The weather in San Sebastián is more pleasant than anywhere else.'
(A.300) indirect negation
Ez dut uste [gizon hori i-noiz ikusi dud-anik. (S88:ex. 178)
NEG I:it believe [man that INDEF-when see I:him]-SBOR
‘I don’t believe that I’ve ever seen that man.’
(A.301) direct negation
I-nor-i i-noiz ez diozu deusik eman. (A89: 305)
INDEF-who-DAT INDEF-when NEG you:it:[Link] anything give
‘You haven’t ever given anyone anything.’

The edo- and the -nahi-series are used in the free-choice and indirect-negation functions.

(A.302) free choice

a. Galde-tzen duzuna edo-zein liburu-tan aurki dezakezu. (S88: ex. 895)
ask-HAB you:it INDEF-which book-LOC find you:can:it
‘You can find what you are asking about in any book.’
b. Nor-nahi-k eduki-ko du hik baino botere gehiago. (A89: 432)
who-INDEF-ERG have-FUT he:it you:ERG than power more
‘Anyone will have more power than you.’
 361

i
But note that expressions such as aon duine acu ‘anyone of them’ (A.149a) are
possible, which suggests that the phrase aon duine does have pronominal status.
Partitive phrases are generally not possible with true nouns (*‘a person of them’).
ii
The situation in indirect-negation contexts seems to be complicated, with different
negators behaving differently, see the discussion in Glinert (1982: § 6).
iii
Thus, Li 1992:127 stars sentence (i):

(i) *Tā xǐ huan shénme.

he like what
‘He likes something.’

iv
Notice also that yē and dō u can occur only once per sentence. Thus, the second
negative indefinite in (i) must be a bare interrogative.

(i) Shéi-yě méi tingjian shénme.

who-INDEF NEG:PFV hear what
‘No one heard anything.’
 402

Appendix B:
The data of the 100-language sample
1. The sample. The languages of the 100-language sample are listed in Table 2
(for the selection of the sample from the world's languages, see § [Link]). The
numbers in parentheses after genetic groupings are the numbers of languages
from each grouping that the sample should contain according to the selection
procedure. A number of the form '(n>m)' means that the number should have
been n, but due to readjustment (resulting from the lack of data for certain
groupings) the actual number is m.
The geographic distribution of the languages of the 40-language sample and
the 100-language sample is shown on the map on p. 413.

Table 2. The languages of the 100-language sample

————————————————————————————————
I. Afro-Asiatic (6>7)
Ancient Egyptian (1): Ancient Egyptian
Berber (1): Kabyle
Chadic (1>2):
Biu-Mandara Margi
West Hausa
Cushitic (1): Harar Oromo
Semitic (1>2):
West Amharic
East Akkadian
Omotic (1>0)
II. Altaic (2>3)
Altaic proper (1>2):
Mongolian-Tungusic: Khalkha Mongolian
Turkic: Bashkir
Korean-Japanese (1): Ainu
III. Amerind (17>19)
Northern Amerind (5>7)
Almosan-Keresiouan (1):
Keresiouan:
Siouan-Yuchi Lakhota
Iroquoian Cayuga
Penutian (2)
Gulf: Koasati
Mexican: Huastec
Oregon: Takelma
Hokan (2)
Seri-Yuman: Maricopa
Northern: Eastern Pomo
Central Amerind (2)
Uto-Aztecan: Classical Nahuatl
Oto-Manguean: Mixtec
Chibchan-Paezan (2)
Chibchan (1): Sumu
Paezan (1): Paez
Andean (1>2):
Aymaran: Jaqaru
Southern: Mapuche
Equatorial-Tucanoan (4>2)
Macro-Tucanoan (2>0)
Equatorial (2)
Kariri-Tupi: Guarani
Macro-Arawakan: Goajiro
Ge-Pano-Carib (3>4)
Macro-Carib (1>2):
 403

Peba-Yaguan: Yagua
Carib: Panare
Ge-Pano (2):
Macro-Panoan (1>0):
Macro-Ge (1>2):
Unclassified: Iatê
Ge-Kaingang: Kaingang

IV. Australian (7)

Unclassified: Mangarayi
Maran: Warndarang
Bunaban: Gooniyandi
Pama-Nyungan (1>4):
Dyirbalic: Dyirbal
Kalkatungic: Kalkatungu
Wiradhuric: Yuwaalaraay
South-West: Djaru
V. Austric (15>17)
Miao-Yao (1): Hmong Njua
Austro-Asiatic (2>3)
Mon-Khmer (1>2)
North: Xinh Mul
East: Khmer
Munda (1): Santali
Austro-Tai (12>13)
Daic (1): Thai
Austronesian (11>12)
Atayalic (1): Atayal
Tsouic (1>0)
Paiwanic (1): Paiwan
Malayo-Polynesian (8>10)
Western Malayo-Polynesian (4>5)
Meso-Philippine Tagalog
Celebes Bugis
Borneo Malagasy
Sundic Indonesian
Unclassified Palauan
Central-Eastern Malayo-Polynesian (4>5)
Central Malayo-Polynesian (1>0)
Eastern Malayo-Polynesian (3>5)
S Halmahera-NW New Guinea (1>0)
Oceanic (2>5)
Siassi Takia
Milne Bay-Central Province Kilivila
Southern New Hebrides Kwamera
Remote Oceanic
Micronesian Yapese
Central Pacific Samoan
VI. Caucasian (1>2):
Nakh-Daghestanian Lak
Abkhaz-Adyghean Abkhaz
VII. Chukchi-Kamchatkan (1): Koryak
VIII. Elamo-Dravidian (1): Telugu
IX. Eskimo-Aleut (1): W. Greenlandic Eskimo
X. Indo-Hittite (4>5)
Anatolian (1): Hittite
Indo-European (3>4)
Indo-Iranian: Pashto
Celtic: Welsh
Italic: Romantsch
Balto-Slavic: Upper Sorbian
 404

XI. Indo-Pacific (12>4)

Trans-New Guinea
Main section
Central and Western
East New Guinea Highlands
Kalam Kobon
East-Central Hua
Madang-Adelbert Range
Madang Amele
Sepik-Ramu Haruai
XII. Khoisan (1>0)
XIII. Na-Dene (1): Slave
XIV. Niger-Kordofanian (11>12)
Kordofanian (1>0)
Niger-Congo (10>12)
Mande (1>2)
Unclassified: Bobo
Northern-Western: Bambara
Niger-Congo proper (9>10)
West Atlantic (1): Fula
Central Niger-Congo (8>9)
North Central Niger-Congo (4>3)
Kru (1>0)
Dogon (1): Dogon
Gur=Voltaic (1): Koromfe
Adamawa-Ubangian (1): Gbeya
South Central Niger-Congo (4>6)
Ijo-Defaka (1>0):
Western (1): Ewe
Eastern (2>5)
Yoruba-Northern Akoko Yoruba
Lower Cross Ibibio
Benue-Zambesi
Nyima
Wel
Bantoid
Broad Bantu Babungo
Narrow Bantu
Northwest Bantu Ntomba
Central Bantu Xhosa
XV. Nilo-Saharan (5>6)
Songhai Songhai
Saharan Kanuri
Maban Masalit
East Sudanic (1>2)
Eastern Dongolawi
Nilotic Lango
Central SudanicLogbara
XVI. Sino-Tibetan (3>4)
Sinitic (1): Chinese
Tibeto-Karen (2>3)
Karen (1>0)
Tibeto-Burman (1>3)
Tibetic Newari
Burmic
Kuki-Naga Manipuri
Burmese-Lolo Burmese
XVII. Uralic-Yukaghir (1>2)
Yukaghir: Yukaghir
Uralic: Ingrian
XVIII. Yeniseyan (1): Ket
XIX. Pidgins and Creoles (2):
Pacific Tok Pisin
Indian Ocean Seychelles Creole
XX. Isolates:
 405

Sumerian (1): Sumerian

Nahali (1): Nahali
Burushaski (1): Burushaski
Basque (1): Basque
Nivkh (1): Nivkh
Hurrian (1>0)
Meroitic (1>0)
Etruscan (1>0)
———————————————————————————————
2. The data. The following pages contain the data of the 100-language sample. The
following abbreviations are used:
Q interrogative-based indefinite N north
G generic-noun-based indefinite S south
sf suffixed indefiniteness marker W west
pf prefixed indefiniteness marker E east
rd reduplication PNG Papua New Guinea

‘sth’ ‘something’
‘sb’ ‘somebody’

The bibliographical references to the reference works consulted for the 100-
language sample are not integrated into the main bibliography and follow
immediately after the data.
 406

indefinite example
type
1. Abkhaz N a-J°´ ‘someone’, a-k’´ ‘something’
2. Ainu Q-sf, Q/rd nen ‘who’ nen-ka ‘sb’
3. Akkadian Q-sf minû ‘what’ min-ma ‘sth’
4. Amele G dana oso ‘sb’ (‘a man’)
5. Amharic Q-sf m´n ‘what’ m´n-´m ‘sth’
6. Atayal Q ima? ‘who; anyone’ nanu? ‘what; anything’
7. Bambara G fèn ‘thing; sth’
8. Bashkir Q-sf, pf-Q kem ‘who’ kem-dyr ‘sb’
9. Basque Q-sf, pf-Q nor ‘who’ nor-bait ‘sb’
10. Bobo G sǒ n ‘person; sb’
11. Bugis Q-sf, Q/rd niga ‘who’ niga-re ‘sb’
12. Burmese N, Q-sf badhu ‘who’ badhu-m´shou ‘anybody’
13. Burushaski Q men ‘who; sb’, bo ‘what; sth’
14. Cayuga Q-sf so: ‘who’ so-kwa ‘sb’
15. Chinese Q, Q-sf shenme ‘what; sth’ shenme-yě ‘anything’
16. Djaru Q nyamba ‘what; sth’
17. Dogon G kidé ‘thing; something’
18. Dongolawi G hāǧ a ‘thing; sth’, wEr ‘one; sb’
19. Dyirbal Q wanya ‘who; sb’
20. Ancient Egyptian G s ‘person; sb’ ht ‘things; sth’
21. Ewe G ame ‘person; sb’
22. Fula G huunde ‘thing; sth’
23. Gbeya G mO ‘thing; sth’
24. Goajiro Q kasa’ ‘what; sth’
25. Gooniyandi Q, Q-sf ngoorndoo ‘who; sb’ ngoorndoo-ngaddaya ‘sb’
26. W. Greenlandic Q-sf suna ‘what’ suna-luunniit ‘anything’
27. Guarani G, Q ava ‘who; person; sb’, mba-e ‘what; thing; sth’
28. Haruai G ap=y ‘sth’ (‘thing-INDEF’); nöbö=y ‘sb’ (‘man-INDEF’)
29. Hausa G, pf-Q koo-mee ‘anything’; wani abu ‘sth’ (cf. A.32)
30. Hittite Q-sf kuiš ‘who’ kuiš -ki ‘sb’
31. Hmong Njua Q leej twg ‘who; sb’
32. Hua G vi-ka ‘someone’ (man-2SG)
33. Huastec Q, Q-sf hit ‘who’ hita? ‘sb’
34. Iatê Q-sf tô ‘what’ to:-ná ‘sth’
35. Ibibio G áwó ‘person; sb’, mkpO ‘thing; sth’
36. Babungo G w´ mO’ ‘sb’ (‘person some’)
37. Indonesian G Q-sf sese-orang ‘sb’ (‘one person’); siapa-pun ‘anyone’
38. Ingrian Q-sf ken ‘who’ ken-ikke ‘sb’
39. Jaqaru Q-sf qachi ‘who’ qach-psa ‘sb’
40. Kabyle G lxelq ‘person; sb’
41. Kaingang Q ne ‘what; sth’
42. Kalkatungu N minaNara ‘sth’
43. Kanuri G, Q-sf aví ‘what’ aví-yaye ‘anything’
44. Ket pf-Q anæ ‘who’ tam anæ ‘sb’
45. Khmer Q qw´y ‘what; sth’
46. Kilivila G, pf-Q te-tala (‘man-one’) ‘sb’
47. Koasati Q ná:si ‘what; sth’
48. Kobon G bI ap (‘a man’) ‘sb’
49. Koromfe G a fo ‘person; sb’
50. Koryak Q-sf meki ‘who’ meki-N´n ‘b’
 407

S, O, V N, G Pr/Po location source

1. SOV GN Po NW Georgia Hewitt 1979:158

2. SOV GN Po Hokkaido (Japan) Refsing 1986:103-109
3. SOV NG Pr Mesopotamia Ryckmans 1960:35
4. SOV GN Po New Guinea Roberts 1987:209
5. SOV GN Pr/Po NC Ethiopia Cohen 1936:125-26
6. VSO NG Pr N Taiwan Egerod 1980
7. SOV GN Po/Pr SW Mali Bailleul 1981
8. SOV GN Po SW Urals (C Russia) Juldaš ev 1966:182
9. SOV GN Po NE Spain, SW France Saltarelli 1988
10. SOV GN Po W Burkina Faso Le Bris and Prost 1981:42
11. VSO NG Pr SW Sulawesi (Indonesia) Sirk 1979
12. SOV GN Po SWC Burma Richter 1983
13. SOV GN Po N Jammu and Kashmir (India) Berger 1974:26
14. ? ? ? New York State, S Ontario Hans-Jürgen Sasse, p.c.
15. SVO GN Pr China Li and Th. 1981, Li 1992
16. SVO/SOV ? ? NE Western Australia Tsunoda 1981
17. SOV GN Po SC Mali Vladimir Plungian, p.c.
18. SOV GN Po N Sudan Armbruster 1960:176-77
19. OSV NG – NW Queensland (Australia) Dixon 1972:265
20. VSO NG Pr Egypt Gardiner 1957:79-80
21. SVO GN Po/Pr S Benin, S Togo, SE Ghana Westermann 1907:62
22. SVO NG Pr W Africa Jungraithmayr & A. ’89:119
23. SVO NG Pr W C. Afr. Republic, NW Zaire Samarin 1966
24. VO ? Pr NE Venezuela, N Colombia Olza Zubiri and J. 1986:92
25. SOV(?) NG Po N Western Australia McGregor 1990
26. SOV GN Po W Greenland Fortescue 1984:254-56
27. SVO ? Po E Paraguay Ortiz Mayans 1980
28. SOV GN Po NE Papua New Guinea, Madang Prov. Bernard Comrie, p.c.
29. SVO NG Pr NW Nigeria, S Niger Kraft & Kirk-Greene 1973
30. SOV GN Po ancient Anatolia Friedrich 1940
31. SVO ? Pr S China, N Vietnam, Laos Harriehausen 1990:132
32. SOV GN Po Eastern Highlands (PNG) Haiman 1991
33. VSO GN(?) Pr E San Luis Potosí, N Veracruz (Mex.) McQuown 1984
34. OV ? Pr Pernambuco (Brazil) Lapenda 1968
35. SVO NG Pr SE Nigeria Kaufman 1972
36. SVO NG Pr NW Cameroon Schaub 1985:197
37. SVO NG Pr Indonesia Kähler 1965
38. SVO ? Po/Pr St. Petersburg region (W Russia) Laanest 1966:109
39. SOV? GN Po SC Peru Hardman 1983:133
40. VSO NG Pr N Algeria Dallet 1985
41. SOV? ? ? S Brazil (Paraná, Sta Catarina, R.G. do Sul) Wiesemann 1972
42. SOV GN ? W Queensland (Australia) Blake 1979:105
43. SOV NG Pr NE Nigeria Lukas 1937
44. SOV GN Po Yenisey river (Siberia) Krejnovič 1968b:462
45. SVO NG Pr Cambodia Huffman 1967:153-56
46. SVO GN/NG Pr Trobriand Islands (SE PNG) Senft 1986:67
47. SOV GN Po SW Louisiana (USA) Kimball 1991:423-428
48. SOV GN Po C PNG (W Highlands & Madang Pr.) Davies 1981:156
49. SVO GN Po/Pr SC Burkina Faso Rennison 1986
50. SVO ? Po Kamchatka peninsula (Russian Far E) Ž ukova 1968:273
 408

indefinite example
type
51. Kwamera G nari riti (‘thing one’) ‘sth’
52. Lak Q-sf cu ‘who’ cu-unugu ‘anyone’
53. Lakhota Q, Q-sf tuwe ‘who; sb’ tuwe-hci ‘anyone’
54. Lango G Nàt ‘person; sb’
55. Logbara G àfa ‘thing; sth’
56. Malagasy G pf-Q/red olona ‘person; sb’ iza ‘who’, na iza na iza ‘anyone’
57. Mangarayi Q Niñja ‘who; anybody’
58. Manipuri Q-sf k´dayd´ ‘where’ k´dayd´-no ‘somewhere’
59. Mapuche Q, Q-sf chem ‘what; sth’ chem-rume ‘anything’
60. Margi G pf-Q(-sf) mdù ‘person; sb’ mì ‘what’, kó mì(-yé) ‘anything’
61. Maricopa Q mki ‘who; sb’ kawish ‘what; sth’
62. Masalit Q Ngaw ‘who; sb’
63. Mixtec N, G ñiyIvI ‘people; sb’, sa ‘sth’
64. Khalkha Mongolian G neg xün (‘one person’) ‘sb’
65. Nahali Q-sf nā ni ‘who’ nā ni-kā ‘anyone’
66. Classical Nahuatl Q-sf ā c ‘who’ ac-ah ‘sb’
67. Newari Q su ‘who; anyone’
68. Nivkh Q-sf aN ‘who’ aN-lu, aN-laq ‘sb’
69. Ntomba G moto Ǒ mO (‘person one’) ‘sb’
70. Harar Oromo G namá ‘person; sb’
71. Paez Q-sf kim ‘who’ kim-pa ‘sb’
72. Paiwan G, Q-sf tsautsau ‘person;sb’, a-nema ‘what’, nema-nga ‘thing; sth’
73. Palauan G chad ‘person; sb’
74. Panare Q ?
75. Pashto Q cok ‘who; sb’
76. Eastern Pomo Q č hi ‘how; somehow’
77. Romantsch pf-Q tgi ‘who’ insa-tgi ‘somebody’
78. Samoan G se tangata ‘sb’ (‘any man’); se mea ‘sth’ (‘any thing’)
79. Santali Q ?
80. Seychelles Creole G keksoz ‘thing; sth’
81. Slave G t’asíi ‘thing; sth’, léé ‘sb’ (lée ‘one’)
82. Songhai G boro ‘person; sb’
83. Upper Sorbian pf-Q, Q-sf š to ‘what’, ně -š to ‘sth’, š to-ž kuli ‘anything’
84. Sumerian Q-sf ana ‘what’ [Link] ‘anything’ (< ana-me)
85. Sumu G di (as) ‘sth’ (thing (one))
86. Tagalog Q-sf sino ‘what’ sinu-man ‘anything’ (§2.3.2)
87. Takelma Q nekc ‘who, sb’
88. Takia G tamol ‘man; sb’; mel ‘thing; sth’
89. Telugu Q-sf evaru ‘who’ evar-oo ‘sb’
90. Thai Q khray ‘who; anyone’
91. Tok Pisin G wanpela man (‘one person’) ‘sb’
92. Warndarang Q, pf-Q -ngaNa ‘what; sth’; ra-ngaNa ‘sth’
93. Welsh G peth ‘thing’, rhyw-beth ‘sth’
94. Yagua Q táá(ra) ‘what?’, tara ‘sth’; chíí(ra) ‘who’, tíí ‘sb’
95. Yapese N, G beaq ‘sb’; ba-nean ‘sth’ (ba ‘a’, nean ‘thing’)
96. Yoruba G ohun ‘thing; sth’
97. Yukaghir pf-Q kin ‘who’ me-kin ‘sb’
98. Yuwaalaraay Q-sf Na:n- ‘who’ Na:n-du-wa: (ERG) ‘somebody’
99. Xinh Mul Q nôh ‘who; sb’
100. Xhosa G umntu ‘man; sb’, into ‘thing; sth’
 409

S, O, V N, G Pr/Po location source

51. SVO NG Pr Tanna Island (S Vanuatu) Lindstrom 1986:183

52. SOV GN Po SC Daghestan (S Russia) Murkelinskij 1967:498
53. SOV GN Po N and S Dakota (USA) Schmidt 1980:128
54. SVO NG Pr N Uganda Noonan 1992:166
55. SVO GN/NG Po NE Zaire Crazzolara 1960:65
56. VOS NG Pr Madagascar Korneev 1966
57. OVS NG/GN Pr NC Northern Territory (Australia) Merlan 1982:118-19
58. SOV GN ? Manipur (E India) Bhat 1993+:22
59. SVO GN Po C Chile de Moesbach 1963:60
60. SVO GN Pr NE Nigeria Hoffmann 1963
61. SOV GN Po C Arizona (near Phoenix) (SE USA) Gordon 1986:61-65
62. SOV NG Po W Sudan, E Chad Edgar 1989:60
63. VSO NG Pr SW Oaxaca (SC Mexico) Pensinger 1974
64. SOV GN Po Mongolia Luvsanvandan 1982
65. SOV GN Po W Madhya Pradesh (C India) Kuiper 1962:30-31
66. SOV GN Po Mexico Andrews 1975:175
67. SOV GN Po C and W Nepal Korolev 1989:62
68. SOV GN Po Lower Amur, N Sakhalin (E Russia) Panfilov 1968:419
69. SVO NG Pr W Zaire (Lake Tumba) Mamet 1955
70. SOV NG Po NE Ethiopia Owens 1985:192
71. SOV GN Po SW Colombia Jung 1989:218, 306
72. VSO ? ? S Taiwan Ferrell 1982
73. SVO NG Pr Palau Islands (W Micronesia) Josephs 1975:459
74. OVS ? ? Venezuela Thomas Payne, p.c.
75. SOV GN Pr/Po NE Pakistan, Afghanistan
76. SOV GN N California (E USA) McLendon 1975:130
77. SVO NG Pr Graubünden (E Switzerland) Gartner 1883:105-106
78. VSO NG Pr Samoa Mosel & Hovdhaugen 1993
79. SOV NG Po E Bihar (India) Vermeer 1969
80. SVO NG Pr Seychelles Islands D’Offay & Lionnet 1982:162
81. SOV GN Po N Alberta, NE B. C. (E Canada) Rice 1989:259
82. SVO GN Po C Mali, SW Niger Prost 1956
83. SVO NG Pr E Saxony (=Lusatia) (E Germany) Fasske 1981:591-605
84. SOV NG/GN Po Mesopotamia Thomsen 1984
85. ? ? ? N Nicaragua, E Honduras von Houwald 1980
86. VOS NG Pr S Luzon, N Mindoro (N Philippines) Schachter and Otanes 1972
87. SOV ? Pr/Po Oregon (NE USA) Sapir 1922:§105
88. SOV GN Po Karkar Island, N PNG Malcolm Ross, Bruce Waters, p.c.
89. SOV GN Po Tamil Nadu (SE India), N Sri Lanka Bhat 1989
90. SVO NG Pr/Po S Thailand Hudak 1987:770
91. SVO NG Pr Papua New Guinea Mihalic 1971
92. SVO ? ? SE Arnhemland (N Australia) Heath 1980:88
93. VSO NG Pr Wales (United Kingdom) S. Williams 1980
94. VSO GN Po NE Peru Payne and Payne 1990, p.c.
95. SVO NG Pr Yap Island (Micronesia) Jensen 1977:182
96. SVO NG Pr SW Nigeria Schmid 1980
97. SOV GN Po Kolyma river (Russian Far E) Krejnovič 1968a:448
98. SOV NG ? S Queensland, N N.S.W. (Australia) C. Williams 1980
99. SVO NG Pr Vietnam Pogibenko and Buj 1990:59
100. SVO NG Pr E South Africa Fischer et al. 1985
 R 410

References for the 100-language sample

Andrews, J. Richard (1975), Introduction to Classical Nahuatl (Austin: University of Texas Press).
Armbruster, Charles Hubert (1960), Dongolese Nubian: a Grammar (Cambridge: Cambridge
University Press).
Bailleul, Charles (1981), Petit dictionnaire bambara-français, français-bambara (Averbury Publishing
Company, England).
Berger, Hermann (1974), Das Yasin-Burushaski (Werchikwar): Grammatik, Texte, Wörterbuch
(Wiesbaden: Harrassowitz).
Bhat (1989), (1993) —> general references
Blake, Barry (1979), A Kalkatungu grammar (Pacific Linguistics, B, 37.) (Canberra: Australian
National University).
Cohen, Marcel (1936), Traité de langue amharique (Abyssinie) (Paris: Institut d’Ethnologie).
Crazzolara, J. P. (1960), A study of the Logbara (Ma’di) language: grammar and vocabulary (London:
Oxford University Press).
Dallet, J. M. (1985.),Dictionnaire francais-kabyle (Paris: SELAF).
Davies, John (1981), Kobon. (Lingua Descriptive Series, 3.) (Amsterdam: North Holland).
de Moesbach, Ernesto Wilhelm (1963) Idioma Mapuche (Las Casas: Imp. San Francisco).
Dixon 1972 —> general references
D’Offay, Danielle and Lionnet, Guy (1982), Diksyonner kreol-franse. (Kreolische Bibliothek, 3.)
(Hamburg: Buske).
Edgar, John (1989), A Masalit grammar (Berlin: Reimer).
Egerod, Søren (1980), Atayal-English dictionary. 2 vols. (London and Malmö: Curzon Press).
Fasske, Helmut (1981), Grammatik der obersorbischen Schriftsprache der Gegenwart: Morphologie.
(Bautzen: Domowina).
Ferrell, Raleigh (1982), Paiwan dictionary (Pacific Linguistics, C, 73.) (Canberra: Australian
National University).
Fischer, A. et al. (1985), English-Xhosa dictionary (Cape Town: Oxford University Press).
Fortescue, Michael (1984), West Greenlandic. (Croom Helm Decriptive Grammars) (London:
Croom Helm).
Friedrich, Johannes (1940), Hethitisches Elementarbuch. 1. Teil: Kurzgefasste Grammatik.
(Heidelberg: Winter).
Gardiner, Alan (1957), Egyptian grammar. Rev. 3rd ed. (London: Oxford University Press).
Gartner, Theodor (1883), Raetoromanische Grammatik (Heilbronn: Henninger).
Gordon, Lynn (1986), Maricopa Morphology and Syntax (University of California Publications in
Linguistics, 108.) (Berkeley: University of California Press).
Haiman, John (1991), Hua Dictionary (Wiesbaden: Harrassowitz).
Hardman, Martha J. (1983), Jaqaru: compendio de estructura fonológica y morfológica (Lima:
Instituto de Estudios Peruanos).
Harriehausen, Bettina (1990), Hmong Njua. (Linguistische Arbeiten, 245.) (Tübingen: Niemeyer).
Heath, Jeffrey (1980), Basic Materials in Warndarang: Grammar, texts and dictionary (Pacific
Linguistics, B, 72.) (Canberra: Australian National University).
Hewitt, B. George (1979), Abkhaz (Lingua Descriptive Series, 2.) (Amsterdam: North Holland).
Hoffmann, Carl (1963), A Grammar of the Margi Language. London: Oxford University Press.
Hudak, Thomas John (1987), 'Thai', in: Comrie, Bernard (ed.) (1987), The World’s Major
Languages (London: Croom Helm), 757-75.
Huffman 1967 —> general references
Jensen, John Thayer (1977), Yapese reference grammar (Honolulu: The University Press of
Hawaii).
Josephs, Lewis S. (1975.), Palauan Reference Grammar (Honolulu: University Press of Hawaii).
Juldaš ev, A.A. (1966), 'Baš kirskij jazyk', in Baskakov, N. A. (ed.), Jazyki narodov SSSR. Vol. 2:
Tjurkskie jazyki (Moscow: Nauka), 173-193.
Jung, Ingrid (1989), Grammatik des Paez: Ein Abriss. Dissertation, Universität Osnabrück.
Jungraithmayr, Hermann and Abu-Manga, Al-Amin (1989), Einführung in die Ful-Sprache
(Berlin: Reimer).
Kähler, Hans (1965), Grammatik der Bahasa Indonesia (Wiesbaden: Harrassowitz).
Kaufman, Elaine Marlowe (1972), Ibibio Dictionary (Leiden: African Studies Center).
Kimball, Geoffrey D. (1991), Koasati Grammar (Lincoln and London: University of Nebraska
Press).
Korneev, L.A. (1966), Mal’gaš sko-russkij slovar’ (Moscow: Sovetskaja Ènciklopedija).
Korolev 1989 —> general references
Kraft, Charles H. and Kirk-Greene, A.H.M. (1973), Hausa. (Teach Yourself Books) (Sevenoaks,
Kent: Hodder & Stoughton).
Krejnovič , E. A. (1968a), 'Jukagirskij jazyk', in Skorik (ed.) (1968), 435-452.
Krejnovič , E. A. (1968b), 'Ketskij jazyk', in Skorik (ed.) (1968), 453-473.
 R 411

Kuiper, F. B. J. (1962), Nahali: a comparative study. (Mededelingen der Koninklijke Nederlandse

Akademie van Wetenschappen, Afd. Letterkunde, Nieuwe Reeks, Deel 25, No. 5)
(Amsterdam: Noord-Hollandsche Uitgevers Maatschappij).
Laanest, A. (1966),'Iž orskij jazyk', in: Lytkin, V.I. and Majtinskaja, K.E. (eds.) (1966), Jazyki
narodov SSSR. Vol. 3: Finno-ugorskie i samodijskie jazyki. (Moscow: Nauka), 102-117.
Lapenda, Geraldo (196),. Estrutura da língua iatê falada pelos índios Fulniôs em Pernambuco (Recife:
Universidade Federal de Pernambuco).
Le Bris, Pierre and Prost, André (1981), Dictionnaire bobo-français.
Li 1992 —> general references
Li, Charles N. and Thompson, Sandra A. (1981), Mandarin Chinese: a Functional Reference
Grammar (Berkeley: University of California Press).
Lindstrom, Lamont (1986), Kwamera dictionary (Pacific Linguistics) (Canberra: Australian
National University).
Lukas, Johannes (1937), A study of the Kanuri language (London: Oxford University Press).
Luvsanvandan, Š . (ed.) 1982), Oros mongol tol’ (Ulan Bator: Ulsyn xevleliin gazar).
Lyman, Thomas Amis (1974) Dictionary of Mong Njua (The Hague: Mouton).
Mamet, M. (1955), La langue ntomba (Tervuren: Musée Royal du Congo Belge).
McGregor, William (1990), A Functional Grammar of Gooniyandi. (Studies in Language
Companion Series, 22.) (Amsterdam: Benjamins).
McLendon, Sally (1975), A grammar of Eastern Pomo. (University of California Publications in
Linguistics, 74.) Berkeley: University of California Press.
McQuown, Norman A. (1984), 'A sketch of San Luis Potosí Huastec', in: Edmonson, Munro S.
(ed.) 1984. Supplement to the Handbook of Middle American Indians, vol. II: Linguistics. Austin:
University of Texas Press, 83-146.
Merlan, Francesca (1982), Mangarayi. (LDS, 4.) Amsterdam: North Holland.
Mihalic, F. (1971), The Jacaranda dictionary and grammar of Melanesian Pidgin. Milton, Q. etc.: The
Jacaranda Press.
Mosel, Ulrike and Hovdhaugen, Even (1993), Samoan reference grammar. Oslo: Scandinavian
University Press.
Murkelinskij, G.B. (1967), 'Lakskij jazyk', in: Bokarev, E.A. and Lomtatidze, K.V. (eds.) (1967),
Jazyki narodov SSSR, Vol. 4: Iberijsko-kavkazskie jazyki (Moskva: Nauka), 488-507.
Noonan, Michael (1992), A grammar of Lango. (Mouton Grammar Library, 7.) (Berlin: Mouton de
Gruyter).
Olza Zubiri and Jusayú, Miguel Angel (1986), Gramática de la lengua guajira (San Cristóbal,
Venezuela: Universidad Católica del Tachira).
Ortiz Mayans, António (1980), Nuevo diccionário español-guaraní, guaraní-español. 11th ed.
(Buenos Aires: Editorial Universitária de Buenos Aires).
Owens, Jonathan (1985), A grammar of Harar Oromo (northeastern Ethiopia) (Cushitic language
studies, 4.) (Hamburg: Buske).
Panfilov, V.Z. (1968), 'Nivxskij jazyk', in Skorik (ed.) (1968), 408-434.
Payne, Doris L. and Payne, Thomas E. (1990), 'Yagua', in Derbyshire, Desmond C. and Pullum,
Geoffrey K. (eds.), Handbook of Amazonian Languages. Vol. 2. (Berlin: Mouton de Gruyter),
249-474.
Pensinger, Brenda J (1974), Diccionário mixteco-español, español-mixteco (México: Instituto
Lingüístico de Verano).
Pogibenko, T. G. and Buj Kxang Txe (1990), Jazyk ksingmul: lingvistič eskij oč erk (Materialy
sovetsko-v’etnamskoj lingvistič eskoj èkspedicii 1979 goda.) (Moskva: Nauka, Glavnaja
redakcija vostoč noj literatury)
Prost, R. P. A. (1956), La langue sonay et ses dialectes (Dakar: IFAN).
Refsing 1986 —> general references
Rennison, John (1986), Dictionnaire koromfe (Hamburg: Buske).
Rice, Keren (1989), A grammar of Slave. (Mouton Grammar Library, 5.) (Berlin: Mouton de
Gruyter).
Richter, Eberhardt (1983), Lehrbuch des modernen Burmesisch (Leipzig: Enzyklopädie).
Roberts, John (1987), Amele (Croom Helm Descriptive Grammars) (London: Croom Helm).
Ryckmans, G. (1960), Grammaire accadienne (Louvain: Institut Orientaliste, Université de
Louvain).
Saltarelli 1988: —> general references
Samarin, William J. (1966), The Gbeya language: grammar, texts, and vocabularies. (University of
California Publications in Linguistics, 44.) (Berkeley: University of California Press).
Sapir, Edward (1922), 'The Takelma language of southwestern Oregon', in: Boas, Franz (ed.),
(1922) Handbook of American Indian languages, Part 2. (Washington: Government Printing
Office), 1-296.
Schachter and Otanes 1972 —> general references
Schaub, Willi (1985), Babungo (Croom Helm Descriptive Grammars) (London: Croom Helm).
 R 412

Schmid 1980 —> general references

Senft, Gunter (1986), Kilivila: The Language of the Trobriand Islanders (Mouton Grammar Library,
3.) (Berlin: Mouton de Gruyter).
Sirk 1979 —> general references
Skorik, P. Ja. (ed.) (1968), Jazyki narodov SSSR, Vol. 5: Mongol’skie, tunguso-man’čž urskie i
paleoaziatskie jazyki (Leningrad: Nauka).
Thomsen, Marie-Louise (1984) The Sumerian language: an introduction to its history and
grammatical structure (Copenhagen: Akademisk Forlag).
Tsunoda, Tasaku (1981), The Djaru language of Kimberley, Western Australia. (Pacific Linguistics,
B, 78.) (Canberra: Australian National University).
Vermeer, Hans J. (1969), Untersuchungen zum Bau Zentral-Süd-Asiatischer Sprachen (Heidelberg:
Gross).
von Houwald, Götz (1980), Diccionário Español-Sumu/Sumu-Español. (Managua).
Westermann, Diedrich (1907), Grammatik der Ewe-Sprache (Berlin: Reimer).
Wiesemann, Ursula (1972), Die phonologische und grammatische Struktur der Kaingáng-Sprache
(The Hague: Mouton).
Williams, C. J (1980), A grammar of Yuwaalaraay (Pacific Linguistics, B, 74.) (Canberra: Australian
National University).
Williams, Stephen J. (1980), A Welsh grammar (Cardiff: University of Wales Press).
Ž ukova, A. N. (1968), 'Korjakskij jazyk', in Skorik (ed.) (1968), 271-293.
 413

References

ABRAHAM, ROY C. (1946), Dictionary of the Hausa Language (London: University

of London Press).
AFANAS’EV, P. S. and XARITONOV, L. N. (1968), Russko-jakutskij slovar’ (Moscow:
Sovetskaja Ènciklopedija).
ALDRIDGE, MAURICE V. (1982), English Quantifiers: A Study of Quantifying
Expressions in Linguistic Science and Modern English Usage (Amersham:
Avebury).
ALI, LATIF H. (1970), 'Some aspects of negation in English and Baghdad Arabic',
Archivum Linguisticum 1: 67-83.
— (1972), 'Observations on the preverb negative particle ma in Baghdad Arabic',
Studia Linguistica 26(2): 48-60.
ALVRE, PAUL (1983), 'Az uráli nyelvek névmásainak igei komponense',
Nyelvtudományi Közlemények 85(2): 291-5.
AMBRĀ ZAS, VYTAUTAS (ed.) (1985), Grammatika litovskogo jazyka (Vilnius:
Mokslas).
ANDERSON, LLOYD (1982), 'The ‘Perfect’ as a universal and as a language
particular category', in Paul Hopper (ed.), Tense and Aspect: Between Semantics
and Pragmatics. (Typological Studies in Language, 1.) (Amsterdam:
Benjamins), 227-64.
ANTHONY, MICHAEL N. (1977), 'Some remarks on any', Forum Linguisticum 2:
15-32.
AOUN, JOSEPH (1986), Generalized Binding (Dordrecht: Foris).
APRIDONIDZE, Š UKIA (1972), 'Drois k’at’egoria rogorc sit’qvata maorganizebeli
c’inadadebaš i', Tbilisis saxelmc’ipo universit’et’is š romebi B 5 (148): 137-45.
ATTOUMAN, MAHAMAN BACHIR and CARON, BERNARD (1984), 'Extraction et
opération de quantification-qualification en haoussa: les emplois de wání', in
Culioli (ed.) (1984): 9-32.
AULESTIA, GORKA (1989), Basque-English Dictionary (Reno: University of Nevada
Press).
AXVLEDIANI, G. S. (ed.) (1963), Grammatika osetinskogo jazyka. Tom 1: Fonetika i
morfologija (Ordž onikidze: NII pri SM Severo-osetinskoj ASSR).
BACH, EMMON (1968), 'Nouns and noun phrases', in Emmon Bach and Robert
T. Harms (eds.), Universals in Linguistic Theory (New York: Holt, Rinehart,
and Winston), 91-124.
BACH, EMMON and JELINEK, ELOISE and KRATZER, ANGELIKA and PARTEE,
BARBARA H. (eds.) (1995), Quantification in Natural Languages (Studies in
Linguistics and Philosophy, 54.) (Dordrecht: Kluwer).
BAGAEV, NIKOLAJ K. (1965), Sovremennyj osetinskij jazyk. Č . 1. Fonetika i
morfologija (Ordž onikidze: Severo-osetinskoe kniž noe izdatel’stvo).
BAKER, C.L. (1970), 'Double negatives', Linguistic Inquiry 1: 169-86.
BAKKER, EGBERT J. (1988), Linguistics and Formulas in Homer (Amsterdam:
Benjamins).
BARR, DONALD (1988), 'The functions of reduplication in Da’a', in Hein
Steinhauer (ed.) Papers in Western Austronesian Linguistics, IV. (Pacific
Linguistics A, 79.) (Canberra: Australian National University), 1-9.
BAXTER, ALAN N. (1988), A Grammar of Kristang (Malacca Creole Portuguese).
(Pacific Linguistics, B, 95.) (Canberra: Australian National University).
BEAMES, J. (1875), A Comparative Grammar of Modern Aryan Languages (Reprinted
by Munshiram Manoharlal, Delhi 1966).
 414

BEHAGHEL, OTTO (1917), 'Die Indefinitpronomina hwas und sums', Beiträge zur
Geschichte der deutschen Sprache und Literatur 42: 158-60.
BEKE, Ö. (1913-14), 'Das finnisch-ugrische Indefinitum', Keleti Szemle 14: 130-78.
— (1914), 'A határozatlan névmás', Magyar Nyelvör 43: 70-81.
— (1932), 'Über den verallgemeinernden Gebrauch des finnisch-ugrischen
Interrogativums', Ungarische Jahrbücher 12: 352-3.
BEREA GĂ GEANU, ELENA (1980), 'Pronumele nedefinite compuse cu oare- în
limba română literară contemporană ', Limba Română 29(5): 451-5.
BERCHEM, JÖRG (1991), Referenzgrammatik des Somali (Cologne: Omimee).
BERNINI, GIULIANO and RAMAT, PAOLO (1992), La frase negativa nelle lingue
d’Europa (Bologna: Il Mulino).
BHAT, D. N. S. (1981), 'Physical identification in Kannada', Studies in the
Linguistic Sciences 11(2): 1-8.
— (1989), 'An introduction to Indian grammars. Part 2: Wh-words', A report
being submitted to the University Grants Commission, Mysore.
— (1993), 'Indefiniteness and indeterminacy', manuscript, Mysore, Central
Institute of Indian Languages.
BHATIA, TEJ KRISHAN (1978), 'A syntactic and semantic description of negation
in South Asian languages', Ph.D. dissertation, University of Illinois at
Urbana-Champaign.
BLAŽ EV, B. (1980), 'Različ ie v upotreblenii nekotoryx mestoimenij v russkom i
bolgarskom jazykax', Bâlgarska rusistika 1980(1)
BOLINGER, DWIGHT (1960), 'Linguistic science and linguistic engineering', Word
16: 374-91.
— (1977), 'Any and some', in Dwight Bolinger, Meaning and Form (English
Language Series; 11) (London and New York: Longman), 21-36.
BOLLÉE, ANNEGRET (1978), 'Reduplikation und Iteration in den romanischen
Sprachen', Archiv für das Studium der neueren Sprachen und Literaturen 215(2):
318-36.
BONNOT-SAOULSKI, CHRISTINE (1983), 'L’Étude des indéfinis dans une théorie
de l’énonciation', in Paul Garde (ed.), IIIe colloque de linguistique russe.
(Bibliothèque russe de l’Institut d’Études slaves, 65.) (Paris: Institut d’Études
Slaves), 11-24.
BORETZKY, NORBERT and IGLA, BIRGIT (1991), Morphologische Entlehnung in den
Romani-Dialekten (Arbeitspapiere des Projektes 'Prinzipien des
Sprachwandels', 4.) (Essen: Universität-GH Essen).
BORKIN, ANN (1971), 'Polarity items in questions', Chicago Linguistic Society 7:
53-62.
BOSQUE, IGNACIO (1980), Sobre la negación (Madrid: Cátedra).
BRAUNE, WILHELM (1961), Althochdeutsche Grammatik (10th edition, revised by
Walther Mitzka). (Tübingen: Niemeyer).
BRIGHT, WILLIAM (ed.) (1992), International Encyclopedia of Linguistics (New
York: Oxford University Press).
BROWN, R. (1985), 'Term operators', in Machtelt Bolkestein and Casper de Groot
and J. Lachlan Mackenzie (eds.), Predications and Terms in Functional
Grammar (Dordrecht: Foris), 127-45.
BRUGMANN, KARL (1911), Vergleichende Laut-, Stammbildungs- und Flexionslehre
nebst Lehre vom Gebrauch der Wortformen der indogermanischen Sprachen (Bd. 2,
Teil 2, 2. Bearbeitung. Strassburg: Trübner). (Part of Brugmann and Delbrück
1893-1916)
— (1916), Grundriss der vergleichenden Grammatik der indogermanischen Sprachen
(Bd. 2, Teil 3, 2. Strassburg: Trübner). (Part of Brugmann and Delbrück 1893-
1916)
 415

— and DELBRÜCK, BERTHOLD (1893-1916), Grundriss der vergleichenden

Grammatik der indogermanischen Sprachen (Strassburg: Trübner).
BRUNET, JACQUELINE (1981), Grammaire critique de l’italien (Université de Paris
VIII-Vincennes).
BYBEE, JOAN L. (1985), Morphology: A Study of the Relation Between Meaning and
Form (Typological Studies in Language, 9.) (Amsterdam: Benjamins).
— (1988), 'Semantic substance vs. contrast in the development of grammatical
meaning', Berkeley Linguistics Society 14: 247-64.
— and PAGLIUCA, WILLIAM (1987), 'The evolution of future meaning', in Anna
Giacalone Ramat and Onofrio Carruba and Giuliano Bernini (eds.), Papers
from the 7th International Conference on Historical Linguistics. (Current Issues in
Linguistic Theory, 48.) (Amsterdam: Benjamins), 109-22.
— and PAGLIUCA, WILLIAM and PERKINS, REVERE D. (1991), 'Back to the future',
in Traugott and Heine (eds.) (1991), vol. 2: 17-58.
— and PERKINS, REVERE D. and PAGLIUCA, WILLIAM (1994) The Evolution of
Grammar: Tense, Aspect, and Modality in the Languages of the World (Chicago:
The University of Chicago Press).
CARDEN, GUY (1976), English Quantifiers: Logical Structure and Linguistic
Variation (New York: Academic Press).
CARLSON, GREG (1980), 'Polarity any is existential', Linguistic Inquiry 11(4): 799-
804.
— (1981), 'Distribution of free-choice any', Chicago Linguistic Society 17: 8-23.
CHANG, IN-BONG and LABRUNE, LAURENCE (1994), 'Les «indéfinis» en coréen et
en japonais", Faits de langues 4 (L'indéfini): 73-80.
CHOMSKY, NOAM (1970), 'Deep structure, surface structure, and semantic
interpretation', in Roman Jakobson and Shigeo Kawamoto (eds.), Studies in
General and Oriental Linguistics Presented to Shiro Hattori on the Occasion of His
Sixtieth Birthday. Tokyo: TEC.
— (1981), Lectures on Government and Binding (Dordrecht: Foris).
CIESLIKOWA, ALEKSANDRA (1965), 'Partykula -kolwiek w historii i dialektach
jezyka polskiego', Prace Jezykoznawcze 15: 45-84. (Zeszyty Naukowe
Uniwersytetu Jagiellonskiego, CXIV)
COMRIE, BERNARD (1976), Aspect (Cambridge: Cambridge University Press).
— (1981), Language Universals and Linguistic Typology. Oxford: Blackwell.
COYAUD, MAURICE (1972), Linguistique et documentation: Les articulations logiques
du discours (Paris: Librairie Larousse).
COYAUD, MAURICE and AÏT HAMOU, KALED (1971), 'Un universal dans les
quantificateurs indéfinis', Al-lisā niyyā t 2: 5-20.
— (1972), 'Les quantificateurs dans les langues naturelles', in Luigi Heilmann
(ed.), Proceedings of the Eleventh International Congress of Linguists, vol. II.
Bologna: Il Mulino, 601-18.
— (1976), 'Indéfinis et interrogatifs', Semantikos 1(3): 83-8.
CROFT, WILLIAM (1983), 'Quantifier scope ambiguity and definiteness', Berkeley
Linguistic Society 9: 25-36.
— (1985), 'Determiners and specification', in Jerry Hobbs et al. (eds.),
Commonsense Summer: Final Report (CSLI Report, 35.) Stanford: Center for the
Study of Language and Information.
— (1990), Typology and Universals (Cambridge: Cambridge University Press).
— (1991), Syntactic Categories and Grammatical Relations: The Cognitive
Organization of Information (Chicago: University of Chicago Press).
— and SHYLDKROFT, HAVA BAT-ZEEV and KEMMER, SUZANNE (1987),
'Diachronic semantic processes in the middle voice', in Anna Giacalone
Ramat and Onofrio Carruba and Giuliano Bernini (eds.), Papers from the 7th
 416

International Conference on Historical Linguistics. (Current Issues in Linguistic

Theory, 48.) (Amsterdam: Benjamins), 179-192.
CULIOLI, ANTOINE (1983), 'A propos de quelque', in Fisher and Franckel (eds.),
21-29.
— (ed.) (1980), Opérations de détermination: Théorie et description. Vol. 1. (Paris:
Département de Recherches Linguistiques, Université Paris 7).
— (ed.) (1984), Opérations de détermination: Théorie et description. Vol. 2. (Paris:
Département de Recherches Linguistiques, Université Paris 7).
DAHL, ÖSTEN (1970), 'Some notes on indefinites', Language 46(1): 33-41.
— (1979), 'Typology of sentence negation', Linguistics 17: 79-106.
— (1985), Tense and Aspect Systems (Oxford: Blackwell).
DAUSENDSCHÖN-GAY, ULRICH (1977), «Ah oui? tu connais pas un tel?»
Textlinguistische Untersuchungen zum französischen Indefinitartikel (Forum
Linguisticum, 17.) (Frankfurt/Main: Lang).
DAVISON, ALICE (1978a), 'Negative scope and rules of conversation: evidence
from an SOV language', in Peter Cole (ed.), Pragmatics (Syntax and
Semantics, 9.) (New York: Academic Press), 22-45.
— (1978b), 'Negative scope and its consequences for word order', in Donald M.
Lance and Daniel E. Gulstad (eds.), Papers from the 1977 Mid-America
Linguistics Conference (Columbia, Missouri: University of Missouri), 179-85.
— (1980), 'Any as universal or existential?', in Johan van der Auwera (ed.), The
Semantics of Determiners (London: Croom Helm; Baltimore: University Park
Press), 11-40.
— (1981), 'Contextual effects on ‘generic’ indefinites: cross-linguistic arguments
for pragmatic factors', in David Sankoff and Henrietta Cedergren (eds.),
Variation Omnibus (Carbondale and Edmonton, Alberta: Linguistic Research,
Inc.), 359-68. (Also in Studies in the Linguistic Sciences 9(2) (1979): 55-66.)
DAYAL, VENEETA (1995), 'Quantification in correlatives', in Bach et al. (eds.)
(1995), 179-205.
DAYLEY, JON P. (1989), Tümpisa (Panamint) Shoshone grammar (University of
California Publications in Linguistics, 115.) (Berkeley: University of
California Press).
DE BHALDRAITHE, TOMÁS (ed.) (1959), English-Irish dictionary (Dublin: Oifig an
tSoláthair).
DELBRÜCK, BERTHOLD (1893-1900) Vergleichende Syntax der indogermanischen
Sprachen, in Karl Brugmann and Berthold Delbrück, Grundriss der
vergleichenden Grammatik der indogermanischen Sprachen. Band 3, 4, 5.
(Strassburg: Trübner).
DHELVERÚDHI, RÉA (1989): 'I dhiaforés xrísis tú kanís: mja prospáthja erminías',
in Studies in Greek linguistics: Proceedings of the 9th Annual Meeting of the
Department of Linguistics (Thessaloniki: Kyriakidis), 407-26.
DIXON, R. M. W. (1972), The Dyirbal Language of North Queensland (Cambridge:
Cambridge University Press).
— (1980), The languages of Australia (Cambridge: Cambridge University Press).
— (1988), A grammar of Boumaa Fijian (Chicago: University of Chicago Press).
DOGRAMADŽ IEVA, E. (1974), 'Sintaktič na xarakteristika na sâč etanijata ot tipa
kojto i da e v sâvremennija bâlgarski kniž oven ezik', in V pamet na prof. d-r. St.
Stojkov: Ezikovedski izsledovanija (Sofija: BAN-Institut za bâlgarski ezik), 417-
23.
DONALDSON, TAMSIN (1980), Ngiyambaa: the Language of the Wangaaybuwan
(Cambridge: Cambridge University Press).
 417

DONČ EVA, LILJANA (1970), 'Ruskite neopredelitelni mestoimenija i nareč ija ot

tipa -to, -nibud’(libo), ... ni bylo i texnite ekvivalenti v bâlgarski ezik', Bâlgarski
ezik 1970(5): 442-51.
DRYER, MATTHEW (1989), 'Large linguistic areas and language sampling',
Studies in Language 13: 257-92.
DURIE, MARK (1985), A Grammar of Acehnese: on the Basis of a Dialect of North Aceh
(Dordrecht: Foris).
EDMONDSON, J. A. (1981), 'Affectivity and gradient scope', Chicago Linguistic
Society 17: 38-44.
— (1983), 'Polarized auxiliaries', in Frank Heny and Barry Richards (eds.),
Linguistic Categories: Auxiliaries and Related Puzzles, vol. I: Categories
(Dordrecht: Reidel), 49-68.
EINENKEL, EUGEN (1903), 'Das englische Indefinitum', Anglia 26: 461-572.
ENÇ, MÜRVET (1991), 'The semantics of specificity', Linguistic Inquiry 22(1): 1-25.
ENGEL, ULRICH (1988), Deutsche Grammatik (Heidelberg: Groos).
ESPINAL, M. TERESA (1991), 'Disjunct constituents', Language 67(4): 726-62.
ESPÓSITO, ANTHONY P. (1988), 'Thematization of negative polarity terms in
Catalan, Aragonese and Castilian: a diachronic and comparative perspective',
in Thomas J. Walsh (ed.), Georgetown University Round Table on Languages and
Linguistics 1988. (Washington, D.C.: Georgetown University Press), 102-11.
ETZRODT, WILLY (1909), Die Syntax der unbestimmten Fürwörter PERSONNE und
MÊME. Dissertation Göttingen.
FABER, ALICE (1988), 'Indefinite pronouns in early Semitic', in Yoël L.
Arbeitman (ed.), FUCUS: A Semitic/Afrasian gathering in remembrance of Albert
Ehrmann. (Current Issues in Linguistic Theory, 58.) (Amsterdam: Benjamins),
221-38.
FAUCONNIER, GILLES (1975a), 'Pragmatic scales and logical structures', Linguistic
Inquiry 6: 353-75.
— (1975b), 'Polarity and the Scale Principle', Chicago Linguistic Society 11: 188-99.
— (1977), 'Polarité syntaxique et sémantique', Lingvisticae Investigationes 1: 1-37.
— (1978), 'Is there a linguistic level of logical representation?' Theoretical
Linguistics 5.1.
— (1979), 'Implication reversal in a natural language', in F. Guenthner and S. J.
Schmidt (eds.), Formal Semantics and Pragmatics for Natural Languages
(Dordrecht: Reidel), 289-301.
— (1980), 'Pragmatic entailment and questions', in John R. Searle and Ferenc
Kiefer and Manfred Bierwisch (eds.), Speech Act Theory and Pragmatics.
(Synthese Language Library, 10.) (Dordrecht: Reidel), 57-69.
— (1985), Mental Spaces: Aspects of Meaning Construction in Natural Language
(Cambridge, Mass.: MIT Press).
— (1987), 'Quantification, roles, and domains', in Umberto Eco et al. (eds.),
Meaning and Mental Representation (Bloomington: Indiana University Press),
61-80.
FEOKTISTOV, A. P. (1966), 'Mokš anskij jazyk', in V. I. Lytkin and K. E.
Majtinskaja (eds.), Jazyki narodov SSSR. Tom 3: Finno-ugorskie i samodijskie
jazyki. (Moscow: Nauka), 199-220.
FERGUSON, CHARLES A. (1970), 'Grammatical categories in data collecton',
Working Papers in Language Universals 4. (Stanford: Stanford University).
FERRER, CHRISTIAN (1973), 'Remarques sur le fonctionnement de any en anglais',
Les Langues Modernes 67: 175-90.
FIJN VAN DRAAT, P. (1898), 'A remarkable application of the word any', Englische
Studien 24: 152-5.
 418

FISHER, SOPHIE and FRANCKEL, JEAN-JACQUES (eds.) (1983), Linguistique,

énonciation: aspects et détermination (Connaissance et langage, 7.) (Paris:
Éditions de l’École des Hautes Études en Sciences Sociales).
FODOR, JANET DEAN (1970), The linguistic description of opaque contexts (New
York: Garland, 1979). [Ph.D. dissertation, MIT, 1970]
— (1977), Semantics: Theories of meaning in generative grammar (Hassocks, Sussex:
Harvester Press).
FOLEY, WILLIAM A. (1991), The Yimas language of New Guinea (Stanford: Stanford
University Press).
— and VAN VALIN, ROBERT D. (1984), Functional syntax and universal grammar
(Cambridge: Cambridge University Press).
FONTAINE, J. (1978), 'Sur les rapports syntaxiques de l’aspect verbal et des
pronoms indéfinis', Revue des Études Slaves 51: 97-106.
FOULET, LUCIEN (1919), 'Études de syntaxe française, I: quelque', Romania 45
(1918-19): 220-49.
FREI, HENRI (1940), Interrogatif et indéfini (Paris: P. Geuthner).
GAATONE, DAVID (1971), Étude descriptive du système de la négation en français
contemporain (Publications Romanes et Françaises, 114.) (Genève: Droz).
GALKINA-FEDORUK, E. M. (1963), Vyraž enie neopredelennosti v russkom jazyke
neopredelennymi mestoimenijami i nareč ijami (Moscow: Moskovskij
gosudarstvennyj universitet).
GALMICHE, M. (1983), 'Les ambigüités référentielles ou les pièges de la
référence', in G. Kleiber and M. Riegel (eds.), Grammaire et référence (Langue
française, 57.) (Paris: Larousse).
GAO, MOBO C. F. (1994), ‘Dou as a wide scope universal quantifier’, Australian
Journal of Linguistics 14: 39-62.
GARDINER, ALAN H. (1932), The Theory of Speech and Language (Oxford)
GEERTS, G. and HAESERYN, W. and DE ROOIJ, J. and VAN DEN TOORN, M. C.
(eds.) (1984), Algemene nederlands spraakkunst (Groningen: Wolters-
Noordhoff).
GENIUŠ IENE, EMMA (1987), The Typology of Reflexives. (Empirical Approaches to
Language Typology, 2.) (Berlin: Mouton de Gruyter).
GESSNER, E. (1895), 'Das spanische indefinite Pronomen', Zeitschrift für
Romanische Philologie 19: 153-69.
GIANNAKIDOU, ANASTASIA (1993), 'KANIS/kanis: A case of polarity sensitivity in
Modern Greek'. Studies in Greek Linguistics 14: 130-44.
— (1994a), 'On the semantic licensing of polarity items', Studies in Greek
Linguistics 15
— (1994b), 'The semantic licensing of negative polarity items and the Modern
Greek subjunctive', To appear in: Language and Cognition 4, Yearbook of the
Research Group for Theoretical and Applied Linguistics of the University of
Groningen.
— (1994c), 'Habituality, veridicality and polarity items', Ms., University of
Groningen.
GIL, DAVID (1991), Universal Quantifiers: a Typological Study (EUROTYP Working
Papers, VII/12) Strasbourg: European Science Foundation.
— (1993), 'Conjunctive operators: areal phenomenon or semantic universal', in
David Gil (ed.), Studies in Number and Quantification (EUROTYP Working
Papers VII/19.) Konstanz: University of Konstanz, 18-47.
GIUSTI, FRANCESCA (1982), 'Nereferentnye pokazateli imeni naricatel’nogo',
Russian Linguistics 7(1): 3-19.
 419

GIVÓN, TALMY (1973), 'Opacity and reference in language: An inquiry into the
role of modalities', in John P. Kimball (ed.), Syntax and Semantics, Vol. 2.
(New York: Academic Press), 95-121.
— (1975), 'Serial verbs and syntactic change: Niger-Congo', in Charles Li (ed.),
Word Order and Word Order Change (Austin: University of Texas Press), 47-
112.
— (1978), 'Definiteness and referentiality', in Greenberg (ed.) (1978): 291-330.
— (1981), 'On the development of the numeral ‘one’ as an indefinite marker',
Folia Linguistica Historica 2(1): 35-53.
— (1984), Syntax: A Functional-Typological Introduction. Vol. I (Amsterdam:
Benjamins).
— (1990), Syntax: A Functional-Typological Introduction. Vol. II (Amsterdam:
Benjamins).
GLINERT, LEWIS H. (1982), 'Negative and non-assertive in contemporary
Hebrew', Bulletin of the School of Oriental and African Studies 45: 434-70.
— (1989), The Grammar of Modern Hebrew (Cambridge: Cambridge University
Press).
G∫USZKOWSKA, JADWIGA (1985), Zaimki nieokreslone we wspóΩczesnym
biaΩoruskim je4zyku literackim (Monografie Slawistyczne, 45) (WrocΩaw etc.:
Wydawnictwo Polskiej Akademii Nauk).
GONDA, J. (1949), 'The functions of word duplication in Indonesian languages',
Lingua 2: 170-97.
— (1954-55), 'Notes on the Indo-European kui- and kuo-pronouns', Lingua 4: 241-
85.
GRAUR, AL (1939), 'Vre-o en fonction négative', Bulletin linguistique 7: 173-6.
GREENBERG, JOSEPH H. (1963), 'Some universals of grmmar with particular
reference to the order of meaningful elements', in Joseph H. Greenberg (ed.)
Universals of language (Cambridge/MA: MIT Press), 73-113.
— (ed.) (1978), Universals of Human Language. Vol. 4. Syntax. (Stanford: Stanford
University Press).
GREIMAS, A. J. (1963), 'Comment définir les indéfinis? (Essai de description
sémantique)' Études de Linguistique Appliquée, n°é, pp. 110-125.
GREVISSE, MAURICE (1986), Le bon usage: grammaire française. 12th edition,
revised by André Goosse (Paris and Louvain-la-Neuve: Duculot).
GRZEGORCZYKOWA, RENATA (1972a), 'Funkcje semantyczne tzw. zaimków
nieokreslonych', Prace Filologiczne 22: 63-83.
— (1972b), 'Wykladniki kwantyfikacji w polskim zdaniu', Z polskich studiów
slawistycznych, seria 4, Jezykoznawstwo.
GUENTCHÉVA, ZLATKA (with DESCLÉS, JEAN-PIERRE) (1981), 'Recherches sur les
valeurs des indéfinis njakoj et njakakâv en bulgare', Revue des Études slaves
53(3): 403-26. (Also in Culioli (ed.) (1980): 239-86.)
GUREVIČ , V. V. (1983), 'O semantike neopredelennosti', Filologič eskie nauki 1:
54-60.
HAEGEMAN, LILIANE (1991), 'Enkele opmerkingen over de analyse van eentwa
en het Westvlaams van Guido Gezelle', Taal en Tongval 43: 159-168.
HAGÈGE, CLAUDE (1993), The Language Builder (Amsterdam: Benjamins).
HAHN, E. ADELAIDE (1933), 'Light from Hittite on Latin indefinites', Transactions
and Proceedings of the American Philological Association 64: 28-40.
HAIMAN, JOHN (1974), 'Concessives, conditionals and verbs of volition',
Foundations of Language 11: 341-59.
— (1980), 'The iconicity of grammar: isomorphism and motivation', Language
56(3): 565-89.
 420

HASEGAWA, NOBUKO (1991), 'Affirmative polarity items and negation in

Japanese', in Carol Georgopoulos and Roberta Ishihara (eds.), Interdisciplinary
Approaches to Language: Essays in Honor of S.-Y. Kuroda. (Dordrecht: Kluwer),
271-85.
HASPELMATH, MARTIN (1990), 'The grammaticization of passive morphology',
Studies in Language 14(1): 25-71.
— (1991), 'Zur Grammatikalisierung von Indefinitpronomina', in Norbert
Boretzky and Werner Enninger and Benedikt Jeßing and Thomas Stolz (eds.),
Sprachwandel und seine Prinzipien: Beiträge zum 8. Bochum-Essener Kolloquium
vom 19.10. - 21.10. 1990 an der Ruhruniversität Bochum. (Bochum-Essener
Beiträge zur Sprachwandelforschung, 14.) (Bochum: Universitätsverlag
Brockmeyer), 103-25.
— (1993a), A Grammar of Lezgian. (Mouton Grammar Library, 9.) (Berlin:
Mouton de Gruyter).
— (1993b), 'The diachronic externalization of inflection', Linguistics 31(2): 279-
309.
— (1994) "Implicational universals in the distribution of indefinite pronouns."
Sprachtypologie und Universalienforschung 47(3): 156-81.
— (1995), 'Diachronic sources of 'all' and 'every' ', in Bach et al. (eds.) (1995), 363-
82.
— and CARUANA, JOSEPHINE (to appear), 'Indefinite pronouns in Maltese', to
appear in EUROTYP Working Papers, VII, and Rivista di Linguistica
— and KÖNIG, EKKEHARD (to appear) 'Concessive conditionals in the languages
of Europe', to appear in Johan van der Auwera (ed.) Adverbial constructions in
the languages of Europe (Berlin: Mouton de Gruyter).
HAWKINS, JOHN A. (1978), Definiteness and Indefiniteness: A Study in Reference and
Grammaticality Prediction (London: Croom Helm).
— (1983), Word order universals (New York: Academic Press).
HEINE, BERND and CLAUDI, ULRIKE and HÜNNEMEYER, FRIEDERIKE (1991),
Grammaticalization: a Conceptual Framework (Chicago: The University of
Chicago Press).
HERCUS, L. A. (1982), The Bagandi Language (Pacific Linguistics, B, 67.)
(Canberra: Australian National University).
HERINGER, JAMES T. (1969), 'Indefinite noun phrases and referential opacity',
Chicago Linguistic Society 5: 89-97.
HINTIKKA, JAAKKO (1980), 'On the any-thesis and the methodology of
linguistics', Linguistics and Philosophy 4: 101-22.
— (1986), 'The semantics of a certain', Linguistic Inquiry 17: 331-6.
HIRTLE, WALTER H. (1982), 'Can any express any quantity?' LACUS Forum 9:
336-43.
— (1988), 'Some and any: Exploring the system', Linguistics 26: 443-77.
HOEKSEMA, JACK (1983), 'Negative polarity and the comparative', Natural
Language and Linguistic Theory 1(3): 403-34.
— (1994) 'On the grammaticalization of negative polarity items', Berkeley
Linguistics Society 20
HOPPER, PAUL J. AND TRAUGOTT, ELIZABETH CLOSS (1993), Grammaticalization
(Cambridge: Cambridge University Press).
HORN, LAURENCE R. (1972), 'On the semantic properties of logical operators in
English', Ph. D. dissertation, UCLA. (Distributed by IULC, 1976.)
— (1989), A Natural History of Negation (Chicago: The University of Chicago
Press).
HUALDE, JOSÉ I. (1992) Catalan (London: Routledge).
 421

HUFFMAN, FRANKLIN E. (1967), 'An outline of Cambodian grammar', Ph.D.

dissertation, Cornell University.
HUMBOLDT, WILHELM VON (1827), 'Ueber den Dualis', reprinted in Wilhelm von
Humboldt (1985), Über die Sprache, ed. by Jürgen Trabant (Munich:
Deutscher Taschenbuchverlag), 104-128.
HUNYADI, LÁSZLÓ (1981), 'Remarks on the syntax and semantics of topic and
focus in Hungarian', Acta Linguistica Academiae Scientiarum Hungaricae 31:
107-36.
— (1987), 'On the interpretation of the Hungarian quantifiers mind 'every' and
akár 'any' ', Acta Linguistica Academiae Scientiarum Hungaricae 37: 125-36.
IOUP, GEORGETTE (1977), 'Specificity and the interpretation of quantifiers',
Linguistics and Philosophy 1: 233-45.
JACKENDOFF, RAY S. (1972a), Semantic Interpretation in Generative Grammar
(Cambridge, Mass.: MIT Press).
— (1972b), 'Any vs. Every', Linguistic Inquiry 3: 119-20.
— (1983), Semantics and Cognition (Cambridge, Mass.: The MIT Press).
JAKOBSON, ROMAN (1936), 'Beitrag zur allgemeinen Kasuslehre: Gesamt-
bedeutungen der russischen Kasus', Travaux du Cercle Linguistique de Prague
6: 240-88. (Also in Jakobson (1971), 23-71)
— (1957), 'Shifters, verbal categories, and the Russian verb', in Jakobson (1971),
130-47.
— (1971), Selected Writings. Vol. II: Word and Language (The Hague: Mouton).
JESPERSEN, OTTO (1917), Negation in English and Other Languages (Copenhagen:
A. F. Høst).
— (1924), The Philosophy of Grammar (London: George Allen and Unwin Ltd.).
JÓNSDÓTTIR, M. (1991), 'On indefinite pronouns in Icelandic", in Halldór Á.
SigurDsson (ed.), Papers from the Twelfth Scandinavian Conference of Linguistics
(Reykjavík: Institute of Linguistics, University of Iceland), 153-70.
KADMON, NIRIT and LANDMAN, FRED (1993), 'Any', Linguistics and Philosophy
16(4): 353-422.
KAHREL, PETER (1989), 'A typology of term negation', Ms., University of
Amsterdam.
KALECTACA, MILO (1978), Lessons in Hopi (Tucson: The University of Arizona
Press).
KAMP, HANS and REYLE, UWE (1993), From Discourse to Logic. 2 vols. (Dordrecht:
Kluwer).
KANG, MYUNG-YOON (1988), 'Topics in Korean syntax: phrase structure,
variable binding and movement', Ph.D. dissertation, MIT.
KARCEVSKI, SERGE (1969), 'Sur la parataxe et la syntaxe en russe', in R. Godel
(ed.), A Geneva school reader (Bloomington: Indiana University Press).
KARTTUNEN, LAURI (1976), 'Discourse referents', in James D. McCawley (ed.),
Notes from the linguistic underground (Syntax and Semantics, 7.) (New York:
Academic Press), 363-85.
— and PETERS, STANLEY (1980), 'Interrogative quantifiers', in Christian Rohrer
(ed.), Time, tense and quantifiers (Tübingen: Niemeyer), 181-205.
KEMMER, SUZANNE (1993), The middle voice. (Typological Studies in Language,
23.) (Amsterdam: Benjamins).
KIMBALL, GEOFFREY D. (1991), Koasati grammar (Lincoln: University of Nebraska
Press).
KIROVA, TANJA (1986), 'Osnovni funkcii na ruskite i bâlgarskite otricatelni
mestoimenija ot stilistič no-pragmatič no glediš te', Sâpostavitelno Ezikoznanie
11(3): 14-22.
 422

KLIMA, EDWARD S. (1964), 'Negation in English', in Jerry A. Fodor and Jerrold

J. Katz (eds.), The structure of language (Englewood Cliffs, N.J.: Prentice-Hall),
246-323.
KOBOZEVA, IRINA M. (1981), 'Opyt pragmatič eskogo analiza to- i nibud’-
mestoimenij', Izvestija AN SSSR, Serija literatury i jazyka 40(2): 165-72.
KOLB, HERBERT (1983), 'Jeder: Eine sprachgeschichtliche Untersuchung',
Sprachwissenschaft 8: 48-76.
KÖNIG, EKKEHARD (1985), 'Conditionals, concessive conditionals and
concessives: areas of contrast, overlap and neutralization', in Elizabeth C.
Traugott et al. (eds.), On conditionals (Cambridge: Cambridge University
Press), 229-46.
— (1988) 'Concessive connectives and concessive sentences: cross-linguistic
regularities and pragmatic principles', in John A. Hawkins (ed.), Explaining
language universals (Oxford: Blackwell), 145-66.
— (1991), The meaning of focus particles: a comparative perspective (London:
Routledge).
— (1992), 'From discourse to syntax: the case of concessive conditionals', in
Rosemarie Tracy (ed.), Who climbs the grammar-tree (Linguistische Arbeiten,
281.) (Tübingen: Niemeyer), 423-33.
— (1995), 'Emphatische Reflexiva, Reflexivität und Fokuspartikeln', DFG Project
Proposal, Free University of Berlin.
KOPTJEVSKAJA-TAMM, MARIA (1993), Nominalizations (London: Routledge).
KOROLEV, N. I. (1989), Nevarskij jazyk (Moscow: Nauka, Glavnaja redakcija
vostoč noj literatury).
KREISBERG, ANITA (1986), 'Certains cas d’emploi obligatoire des adjectifs
indéfinis en russe', in Maurice Comtet (ed.), IVe colloque de linguistique russe
(Bibliothèque russe de l’Institut d’Études slaves, 77.) (Paris: Institut d’Études
slaves).
KRIFKA, MANFRED (1992), 'Some remarks on polarity items', in: Zaefferer,
Dietmar (ed.), Semantic universals and universal semantics (Berlin: Foris), 150-
89.
KŘ ÍŽ KOVÁ, HELENA (1968), 'K voprosu o tak nazyvaemoj dvojnoj negacii v
slavjanskix jazykax', Slavia 37(1): 21-39.
— (1971), 'Systém neurč ity!ch zájmen v souč asny!ch slovansky!ch jazycích',
Slavia 1971: 342-70.
— (1974), 'Neurč itá zájmena typu kdo v souč asny!ch slovansky!ch jazycích',
Jazykovedné stúdie 12: 88-95.
KÜHNER, RAPHAEL and STEGMANN, CARL (1914), Ausführliche Grammatik der
lateinischen Sprache. 2. Teil: Satzlehre. Vol. 1-2. (Hannover: Hahnsche
Buchhandlung).
KULAEV, NIKOLAJ X. (1958), Mestoimenija v sovremennom literaturnom osetinskom
jazyke (Ordž onikidze: Severo-osetinskoe kniž noe izdatel’stvo).
KUZ’MINA, S. M. (1989), 'Semantika i stilistika neopredelennyx mestoimenij', in
D. N. Š melev (ed.), Grammatič eskie issledovanija: Funkcional’no-stilistič eskij
aspekt (Moscow: Nauka), 158-231.
LABOV, WILLIAM (1972), 'Negative attraction and negative concord in English
grammar', Language 48: 773-818.
LADD, D. ROBERT (1980), The Structure of Intonational Meaning (Bloomington:
Indiana University Press).
LADUSAW, WILLIAM A. (1980), Polarity Sensitivity as Inherent Scope Relations
(New York: Garland).
 423

— (1993), 'Negation, indefinites, and the Jespersen Cycle', Berkeley Linguistics

Society 19: 437-46.
LAKA, ITZIAR (1994), On the Syntax of Negation. New York: Garland.
LAKOFF, GEORGE (1974), 'Syntactic amalgams', Chicago Linguistic Society 10: 321-
44.
LAKOFF, ROBIN (1969), 'Some reasons why there can’t be any some-any rule',
Language 45: 608-15.
LANGACKER, RONALD (1992), Foundations of Cognitive Grammar. Vol. II.
(Stanford: Stanford University Press).
LAVRIC, EVA (1990), Missverstehen verstehen: Opake Kontexte und Ambiguitäten bei
indefiniten und definiten Nominalphrasen (Grazer Linguistische Monographien;
7) (Graz: Institut für Sprachwissenschaft der Universität Graz).
LAWLER, JOHN M. (1971), 'any Questions?' Chicago Linguistic Society 7: 163-80.
LEE, YAE-SHEIK (1994), ‘Negative-polarity items and the semantics of the
particles -to and -na in Korean’, Kansas Working Papers in Linguistics 19(1): 89-
123.
LEE, YOUNG-HEON (1983), 'Yanghwasa amwu-uy uymipwunsek' [A semantic
analysis of the quantifier amwu 'any'], The Linguistic Journal of Korean 8: 71-87.
LEES, R. B. (1960), 'Review of Dwight Bolinger, Interrogative Structures in
American English', Word 16: 119-25.
LE GOFFIC, PIERRE (1994), 'Indéfinis, interrogatifs, relatifs (termes en Qu-):
parcours avec ou sans issue", Faits de langues 4 (L'indéfini): 31-40.
LEGRAND, JEAN E. (1974), 'AND or OR; some SOMEs and all ANYs', Chicago
Linguistics Society 10: 390-401.
LEHMANN, CHRISTIAN (1982a), Thoughts on Grammaticalization: a Programmatic
Sketch. Vol. I. (akup, 48.) (Cologne: Institut für Sprachwissenschaft der
Universität zu Köln).
— (1982b), 'Directions for interlinear morphemic translations', Folia Linguistica
16: 193-224.
— (1984), Der Relativsatz. (Language Universals Series, 3.) (Tübingen: Narr).
— (1985), 'Grammaticalization: synchronic variation and diachronic change',
Lingua e Stile 20: 308-18.
LÉONARD, ANNE-MARIE (1980), 'A propos de quelques indéfinis en anglais', in
Culioli (ed.) (1980), 99-153.
— (1983), 'Étude différentielle de quelques indéfinis anglais', in Fisher and
Franckel (eds.) (1983), 45-80.
LEVIN, JU. I. (1973), 'O semantike mestoimenij', in Andrej A. Zaliznjak (ed.),
Problemy grammatič eskogo modelirovanija (Moscow: Nauka), 108-21.
LEVINSON, STEPHEN C. (1983), Pragmatics (Cambridge: Cambridge University
Press).
— (1991), 'Pragmatic reduction of the Binding Conditions revisited', Journal of
Linguistics 27(1): 107-61.
LI, CHARLES N. and THOMPSON, SANDRA A. (1981), Mandarin Chinese: A
Functional Reference Grammar (Berkeley: University of California Press).
LI, YEN-HUI AUDREY (1992), 'Indefinite wh in Mandarin Chinese', Journal of East
Asian Linguistics 1(2): 125-55.
LINDENFELD, JACQUELINE (1973), Yaqui Syntax (University of California
Publications in Linguistics, 76.) (Berkeley: University of California Press).
LINEBARGER, MARCIA C. (1980), 'Polarity any as an existential quantifier',
Chicago Linguistic Society 16: 211-18.
— (1981), The Grammar of Negative Polarity (Bloomington/IN: Indiana
University Linguistics Club).
 424

— (1987), 'Negative polarity and grammatical representation', Linguistics and

Philosophy 10: 325-87.
— (1991), 'Negative polarity as linguistic evidence", Chicago Linguistic Society 27,
part 2: Parasession on negation, 165-88.
LLEÓ, CONXITA (1983), 'Mots suposadament negatius: necessitat d’un
tractament semàntico-pragmàtic', in Estudis de llengua i literatura catalanes
oferts a R. Aramon i Serra en el seu stantè aniversari. Vol. 3. (Estudis
Universitaris Catalans, 25.) (Barcelona: Curial Edicions Catalanes), 295-330.
LLORENS, E. L. (1929), La negación en español antiguo (Madrid: Molina).
LOMBARD, ALF (1938-39), 'Une classe spéciale de termes indéfinis dans les
langues romanes', Studia Neophilologica 11: 186-209.
— (1947-48), 'A propos de quienquiera', Studia Neophilologica 20: 21-36.
LONGOBARDI, GIUSEPPE (1988), 'I quantificatori', in Lorenzo Renzi (ed.), Grande
grammatica italiana di consultazione. Vol. I. La frase. I sintagmi nominale e
proposizionale (Bologna: Il Mulino).
LÜDTKE, HELMUT (1980), 'Auf dem Wege zu einer Theorie des Sprachwandels',
in Helmut Lüdtke (ed.), Kommunikationstheoretische Grundlagen des
Sprachwandels (Berlin: de Gruyter), 182-252.
LYONS, JOHN (1977), Semantics. Vol. 1-2. (Cambridge: Cambridge University
Press).
MACIEV, AXMAT GEXAEVIČ (1961) , Č eč ensko-russkij slovar’. (Moscow:
Gosudarstvennoe izdatel’stvo inostrannyx i nacional’nyx slovarej).
MAHAJAN, ANOOP KUMAR (1990), 'LF [Logical Form] conditions on negative
polarity licensing', Lingua 80(4): 333-48.
MAJTINSKAJA, KLARA E. (1969), Mestoimenija v jazykax raznyx sistem (Moscow:
Nauka).
MALOTKI, EKKEHART (1979), Hopi-Raum: Eine sprachwissenschaftliche Analyse der
Raumvorstellungen in der Hopi-Sprache (Tübingen: Niemeyer).
MALOVICKIJ, L. JA. (1971), 'Voprosy istorii predmetno-lič nyx mestoimenij
(mestoimenija kto-. č to-osnov)', in Uč enye zapiski leningradskogo
gosudarstvennogo pedagogič eskogo instituta im. Gercena, Vol. 517, 3-130.
MA NEWMAN, ROXANA (1990), An English-Hausa Dictionary (New Haven: Yale
University Press).
MANOLIU-MANEA, MARIA (1966), 'Le pronom indéfini dans le dacoroumain
contemporain standard: Essai de classification sémantique', Revue roumaine de
linguistique 11(5): 441-68.
— (1977), 'How can somebody become nobody? Perceptive strategy and
Romance negation', Revue Roumaine de Linguistique 22: 13-19.
MARKOWICH, D. and PAILLARD, DENIS (1981), 'A propos de quelques indéfinis
en russe', in Culioli (ed.) (1980).
MARTIN, SAMUEL E. and LEE, YOUNG-SOOK C. (1986), Beginning Korean
(Rutland/VT: Tuttle).
MATHESIUS, VILÉM (1937), 'Double negation and grammatical concord', in
Mélanges de linguistique et de philologie offerts à Jacq. van Ginneken à l’occasion du
soixantième anniversaire de sa naissance (Paris: Klincksieck), 79-83.
MAY, ROBERT (1985), Logical Form (Cambridge/MA: MIT Press).
MCCAWLEY, JAMES D. (1977), 'Lexicographic notes on English quantifiers',
Chicago Linguistic Society 13: 372-83.
— (1981), Everything that Linguists Always Wanted to Know about Logic but were
Afraid to Ask (Chicago: The University of Chicago Press).
— (1988), The Syntactic Phenomena of English. 2 vols. (Chicago: University of
Chicago Press).
 425

MCGLOIN, NAOMI HANAOKA (1976), 'Negation', in Masayoshi Shibatani (ed.),

Japanese generative grammar (Syntax and Semantics, 5.) (New York: Academic
Press), 371-419.
MEIER, H. (1950), 'Indefinita vom Typus Span. cualquiera, It. qualsivoglia',
Romanische Forschungen 62: 385-401.
MELLET, SYLVIE (1992), 'Opérations de détermination: remarques sur deux
définis latins, quis et aliquis", Bulletin de la Société de linguistique de Paris 87:
147-59.
— (1994), 'A propos de quelques indéfinis latins: Classes d'occurrences et
construction référentielle", Faits de langues 4 (L'indéfini): 49-56.
MEUNIER, FR. (1875)), 'Sur le passage du sens interrogatif au sens affirmatif',
Mémoires de la Société de Linguistique de Paris 2: 246-60.
MEYERS, LAURA F. (1974), 'The Hausa particle ko: an uncertainty modality', in
Sandra A. Thompson and Carol Lord (eds.), Approaches to the Lexicon (UCLA
Papers in Syntax, 6.) (Los Angeles: UCLA), 130-46.
MIKLOSICH, FRANZ (1886), Etymologisches Wörterbuch der slavischen Sprachen
(Wien: Braunmüller).
MILNER, JEAN-CLAUDE (1979), 'Le système de la négation en français et l’opacité
du sujet', Langue française 44: 110-63.
MITCHELL, T. F. (1962), Colloquial Arabic (Teach Yourself Books)
(Sevenoaks/Kent: Hodder and Stoughton).
MORAVCSIK, EDITH A. (1969), 'Determination', Working Papers in Language
Universals. Vol. 1 (Stanford: Stanford University), 64-98a.
— (1978), 'Reduplicative constructions', in Greenberg, Joseph H. (ed.) Universals
of Human Language. Vol. 3: Word Structure (Stanford: Stanford University
Press).
MOSTOVSKA, MARIOLA (1988), 'Neopredelitelni mestoimenija v bâlgarskija i
polskija ezik', Sâpostavitelno ezikoznanie 13(4-5): 73-79.
MULLER, CLAUDE (1991), La négation en français: syntaxe, sémantique et éléments de
comparaison avec les autres langues romanes (Geneva: Droz).
MUSIĆ (1938) , 'Les pronoms indéfinis dans les langues serbo-croate, latine et
grecque', Bulletin International de l’Académie de Zagreb 10: 27ff.
NEDJALKOV, VLADIMIR P. (ed.) (1988), Typology of resultative constructions
(Typological Studies in Language, 12.) (Amsterdam: Benjamins).
NICHOLS, JOHANNA (1986), 'The bottom line: Chinese Pidgin Russian', in
Wallace Chafe and Johanna Nichols (eds.), Evidentiality: the Coding of
Epistemology in Language (Norwood/NJ: Ablex).
— (1992), Linguistic Diversity in Space and Time (Chicago: The University of
Chicago Press).
— (1994), 'Chechen', in Rieks Smeets (ed.), The Indigenous Languages of the
Caucasus, vol. 4: North East Caucasian Languages (Delmar/NY: Caravan), 1-77.
NICOLOVA, RUSELINA (1972), 'Zur Semantik der Pronomina (Die
Indefinitpronomina im Bulgarischen)', Zeitschrift für Slawistik 1972(1): 89-97.
— (1978), 'Za znač enieto na bâlgarskite neopredelitelni mestoimenija', in
Bâlgarska morfologija — Imena (Sofia), 435-44.
— (1986), Bâlgarskite mestoimenija (Sofia: Nauka i izkustvo).
NIKOLAEVA, TAT’JANA M. (1983), 'Funkcional’naja nagruzka neopredelennyx
mestoimenij v russkom jazyke i tipologija situacij', Izvestija AN SSSR, Serija
literatury i jazyka 42(4): 342-52.
— (1985), Funkcii č astic v vyskazyvanii (Moscow: Nauka).
NOONAN, MICHAEL (1992), A Grammar of Lango (Mouton Grammar Library, 7.)
(Berlin: Mouton de Gruyter).
 426

Ó DÓNAILL, NIALL (ed.) (1977), Foclóir gaeilga-béarla (Dublin: Oifig an

tSoláthair).
OHLHOFF, KARL (1912), Die Syntax der unbestimmten Fürwörter rien, néant,
quelque, chose und quelque chose. Dissertation Göttingen.
OHNO, YUTAKA (1984), 'Remarks on quantification in Japanese', in Descriptive
and Applied Linguistics (Bulletin of the ICU Summer Institute in Linguistics) 18:
233-244. (Tokyo: International Christian University).
ONENKO, SULUNGU N. (1980), Nanajsko-russkij slovar’ (Moscow: Russkij jazyk).
— (1986), Russko-nanajskij slovar’ (Moscow: Russkij jazyk).
ORLANDINI, ANNA (1981), 'Semantica e pragmatica dei pronomi indefiniti
latini', Lingua e Stile 16(2): 215-34.
— (1983), 'Une analyse sémantique et pragmatique des pronoms indéfinis en
latin', in Harm Pinkster (ed.), Latin Linguistics and Linguistic Theory:
Proceedings of the 1st International Colloquium on Latin Linguistics, Amsterdam,
April 1981. (Studies in Language Companion Series, 12.) (Amsterdam:
Benjamins), 229-40.
PAARDEKOOPER, P. C. (1978), 'Wie dan ook enz., vooral als ww-patroondeel', De
Nieuwe Taalgids 71: 569-81.
— (1979), 'Ook maar iémand', De Nieuwe Taalgids 72: 429-48.
PADUČ EVA, ELENA V. (1974), O semantike sintaksisa: materialy k transformacionnoj
grammatike russkogo jazyka (Moscow: Nauka).
— (1985), Vyskazyvanie i ego sootnesennost’ s dejstvitel’nost’ju (Moscow: Nauka).
PAILLARD, DENIS (1983), 'Détermination et opérations énonciatives: le
fonctionnement de certains indéfinis en russe', in Fisher and Franckel (eds.)
(1983), 31-44.
— (1984), Énonciation et détermination en russe contemporain. (Bibliothèque russe
de l’Institut d’Études slaves, 72.) (Paris: Institut d’Études slaves).
PÁLFY, M. (1982), 'Les pronoms indéfinis en français et en hongrois: quelques
problèmes d’emploi', Annales Universitatis Scientiarum Budapestinensis de
Rolando Eötvös Nominatae, Sectio philologica moderna 13: 27-31.
PANTHER, KLAUS-UWE (1977), Studien zur semantischen Struktur englischer
Quantoren unter besonderer Berücksichtigung des Präsuppositionenproblems
(Frankfurt/Main: Peter Lang).
PARKER, GARY J. (1976), Gramática quechua Ancash-Huailas (Lima: Ministério de
Educación).
PARŠ IN, P. B. (1987), 'Le lexème xot’: structure sémantique de l’unité
linguistique xot’ et pragmatique de la concession dans le dialogue déontique',
in Les particules énonciatives en russe contemporain. Vol. 2: 69-91.
— (1988), 'Ustupka i antiustupka v deontič eskom dialoge (funkcionirovanie
leksemy xot’)', in N. D. Arutjunova, (ed.), Referencija i problemy
tekstoobrazovanija (Moscow: Nauka), 146-68.
PARTEE, BARBARA (1972), 'Opacity, coreference and pronouns', in Donald
Davidson and Gilbert Harman (eds.), Semantics of natural language
(Dordrecht: Reidel), 415-41.
PAŠ OV, PETR (1965), 'Osoben vid neopredelitelni mestoimenija v severoiztoč ni
bâlgarski govori (na-koj, na-kakvo, na-kâde, na-kak i.t.n.)', Bâlgarski ezik 1965(4-
5): 362-70.
PAUL, HERMANN (1916-1920), Deutsche Grammatik. Bd. 1-5. (Halle a. S.:
Niemeyer).
— (1920), Prinzipien der Sprachgeschichte (Halle a. S.: Niemeyer).
PERKINS, REVERE D. (1980), 'The evolution of culture and grammar', Ph.D.
dissertation, SUNY at Buffalo.
 427

PILKA, ALFONSAS (1984), 'Neopredelennye determinativy litovskogo jazyka (v

sopostavlenii s anglijskim)', Candidate's dissertation, Vilnius State
University, Vilnius.
PONOMAREFF, SHIRLEY H. T. (1978), 'A semantic feature study of Contemporary
Standard Russian determinative and indefinite pronominals', Ph.D.
dissertation, University of California, Los Angeles.
POSNER, REBECCA (1984), 'Double negatives, negative polarity and negative
incorporation in Romance: a historical and comparative view', Transactions of
the Philological Society 1984: 1-26.
POSTMA, GERTJAN (1994), ‘The indefinite reading of WH’, in Reineke Bok-
Bennema and Crit Cremers (eds.), Linguistics in the Netherlands 1994
(Amsterdam: Benjamins), 187-198.
POTT, AUGUST FRIEDRICH (1862), Doppelung (Reduplikation, Gemination) als eines
der wichtigsten Bildungsmittel der Sprache, beleuchtet aus allen Sprachen aller
Weltteile (Lemgo and Detmold).
PROGOVAC, LJILJANA (1988), 'A binding approach to polarity sensitivity', Ph.D.
dissertation, University of Southern California.
— (1990), 'Free-choice bilo in Serbo-Croatian: existential or universal?' Linguistic
Inquiry 21(1): 130-35.
— (1991), 'Polarity in Serbo-Croatian: Anaphoric NPIs and pronominal PPIs',
Linguistic Inquiry 22(3): 567-72.
— (1992a), 'Negative polarity: a semantico-syntactic approach', Lingua 86: 271-
99.
— (1992b), 'Non-negative polarity licensing must involve COMP', Linguistic
Inquiry 23(2): 341-7.
— (1993), 'Negative polarity and binding', Linguistics and Philosophy 16(2): 149-
180.
— (1994), Negative and Positive Polarity: a Binding Approach (Cambridge Studies
in Linguistics, 68.) Cambridge: Cambridge University Press.
QUINE, WILLARD VAN ORMAN (1956), 'Quantifiers and propositional attitudes',
Journal of Philosophy 53(5)
— (1960), Word and Object (Cambridge/MA: MIT Press).
QUIRK, RANDOLPH and GREENBAUM, SIDNEY and LEECH, GEOFFREY and
SVARTVIK, JAN (1985), A Comprehensive Grammar of the English Language
(London: Longman).
RAUMOLIN-BRUNBERG, HELENA (1994), 'The development of the compound
pronouns in -body and -one in Early Modern English", in: Kastovsky, Dieter
(ed.), Studies in Early Modern English (Berlin: Mouton de Gruyter), 301-24.
REFSING, KIRSTEN (1986), The Ainu Language: the Morphology and Syntax of the
Shizunai Dialect (Aarhus: Århus University Press).
REICHENBACH, HANS (1947), Elements of Symbolic Logic. (Reprint 1980, New
York: Dover).
RIJKHOFF, JAN and BAKKER, DIK and HENGEVELD, KEES and KAHREL, PETER
(1993), 'A method of language sampling', Studies in Language 17(1): 169-203.
RISSANEN, MATTI (1987), 'Old English indefinite pronouns meaning ‘some’ and
‘any’, with special reference to hw-forms', in Leena Kahlas-Tarkka (ed.),
Neophilologica Fennica: Société Néophilologique 100 ans. (Mémoires de la Société
Néophilologique de Helsinki, 45.) (Helsinki: Societé Néophilologique).
ROHRER, CHRISTIAN (1973), 'On the relation between disjunction and existential
quantification', in Maurice Gross and Morris Halle and Marcel-Paul
Schützenberger (eds.), The Formal Analysis of Natural Languages: Proceedings of
the First International Conference (The Hague: Mouton), 224-32.
ROSS, JOHN ROBERT (1969), 'Guess who', Chicago Linguistic Society 5: 252-78.
 428

RUHLEN, MERRITT (1987), A Guide to the World’s Languages, vol. 1: Classification

(Stanford: Stanford University Press).
°
RUŽ IČ KA, R. (1973), 'Kto-to und kto-nibud’', Zeitschrift für Slawistik 18: 705-36.
RYBÁK, J. (1965), 'Charakter opozicie neurč itych zámen typu kto-nibud’/kto-to',
Č eskoslovenská Rusistika 10: 196-201.
— (1975), 'Kategória č isla a neurč ité zámena', Jazykovedny č asopis 26(2): 150-7.
SAHLIN, ELISABETH (1979), Some and Any in Spoken and Written English (Studia
Anglistica Uppsaliensia, 38) (Stockholm: Almqvist and Wiksell).
SALTARELLI, MARIO (1988), Basque. (Routledge Descriptive Grammars) London:
Routledge.
SANNIKOV, VLADIMIR Z. (1989), Russkie soč initel’nye konstrukcii: semantika,
pragmatika, sintaksis (Moscow: Nauka).
Š APIRO, MOISEJ A. et al. (1984), Russko-evrejskij (idiš ) slovar’ (Moscow: Russkij
jazyk).
SAURANBAEV, N. T. (1954), Russko-kazaxskij slovar’ (Moscow: Gosudarstvennoe
izdatel’stvo inostrannyx i nacional’nyx slovarej).
SAVIN, HARRIS B. (1974), 'Every any means every', in Problèmes actuels en
psycholinguistique (Colloques internationaux du Centre National de la
Recherche Scientifique, 206.) (Paris: Éditions du CNRS), 327-35.
SCHACHTER, PAUL and OTANES, FE T. (1972), Tagalog Reference Grammar
(Berkeley: University of California Press).
SCHMERLING, SUSAN F. (1971), 'A note on negative polarity', Papers in Linguistics
4(1): 200-6.
SCHMID, MAUREEN ALICIA (1980), 'Cooccurrence restrictions in negative,
interrogative, and conditional clauses: a cross-linguistic study', Ph.D.
dissertation, State University of New York at Buffalo.
SCHWYZER, EDUARD (1953), Griechische Grammatik, vols. 1-2. (Handbuch der
Altertumswissenschaft, II.1) (Munich: C.H. Beck).
SELIVERSTOVA, OL’GA N. (1964), 'Opyt semantič eskogo analiza slov tipa vse i
tipa kto-nibud’', Voprosy Jazykoznanija 13(4): 80-90.
— (1988), Mestoimenija v jazyke i reč i (Moscow: Nauka).
SENFT, GUNTER (1986), Kilivila: the Language of the Trobriand Islanders (Mouton
Grammar Library, 3.) (Berlin: Mouton de Gruyter).
SERBAT, GUY (1985), 'Erat Pipa quaedam', Revue des études latines 62: 344-56.
SERIANNI, LUCA (1988), Grammatica italiana: Italiano comune e lingua letteraria
(Turin: UTET).
SHNUKAL, ANNA (1988), Broken: an Introduction to the Creole Language of Torres
Strait (Pacific Linguistics C, 107.) (Canberra: Australian National University).
SIRK, V. H. (1979), La langue bugis (Paris: Association Archipel).
SKVORCOV, M. I. (ed.) (1985), Č uvaš sko-russkij slovar’ (Moscow: Russkij jazyk).
SOLÀ, JOAN (1973), 'La negació', in Joan Solà, Estudis de sintaxi catalana. Vol. 2.
(Barcelona: Ediciones 52), 87-118.
SRIDHAR, S. N. (1990), Kannada. (Routledge Descriptive Grammars) (London:
Routledge).
STASSEN, LEON (1985), Comparison and Universal Grammar (Oxford: Blackwell).
STEEVER, SANFORD B. (1988), The Serial Verb Formation in the Dravidian Languages
(Delhi: Motilal Banarsidass).
STEPHANIDES, ÉVA H. (1983), 'A contrastive analysis of some and any in English
and their Hungarian equivalents', in Sándor Rot (ed.), Languages in Function:
Materials of the 13th Annual Conference of the SLE Held in Budapest 3.-6.9. 1980.
(Budapest), 283-90.
 429

— (1985), 'A contrastive analysis of English some and any and their Hungarian
equivalents', Papers and Studies in Contrastive Linguistics 19: 65-83.
STOFFEL, C. (1899), 'The intensive and depreciative functions of any', Englische
Studien 26: 144-50.
STRICKLAND, MARTHA (1982), 'À propos de any et la valeur ‘n’importe quel’ en
anglais', Bulletin de linguistique appliquée et générale 9: 17-48. (Travaux du
Département de Linguistique Générale et Appliquée, Université de
Besançon)
SUBBARAO, K. V. and ARORA, HARBIR (1988), 'On extreme convergence: the case
of Dakkhini Hindi-Urdu', Indian Linguistics 49: 92-108.
SUGAMOTO, NOBUKO (1989), 'Pronominality: A noun-pronoun continuum', in
Roberta Corrigan and Fred Eckman and Michael Noonan (eds.), Linguistic
Categorization (Amsterdam Studies in the Theory and History of Linguistic
Science, 61.) (Amsterdam: Benjamins), 267-91.
Š VEDOVA, NATALIJA JU. (ed.) (1980), Russkaja grammatika. Vol. 1-2. (Moscow:
Nauka).
SVOROU, SOTERIA (1994), The Grammar of Space. (Typological Studies in
Language, 25.) (Amsterdam: Benjamins).
SZARVAS, G. (1892), 'A határozatlan névmások és határozók', Magyar Nyelvör 21:
145-8.
TANAKA, SHIN (1994), 'Die Mehrfachnegation: Ein Sprachgut der
Unraffinierten?', in Dieter W. Halwachs and Irmgard Stütz (eds.), Sprache –
Sprechen – Handeln: Akten des 28. Linguistischen Kolloquiums, Graz 1993, vol. 1
(Linguistische Arbeiten, 320.) (Tübingen: Niemeyer), 191-8.
TESCH, FELICITAS (1990), Die Indefinitpronomina some und any im authentischen
englischen Sprachgebrauch und in Lehrwerken: Eine empirische Untersuchung
(Tübingen: Narr).
THELANDER, M. (1980), 'Någonting om bl. a. ingenting', Nysvenska studier 59-60
(1979-80): 323-40.
TOPOLINSKA, Z. (1972), 'Semantyczne cechy dystynktywne zaimków
slowianskich', Jazykovedny casopis 23(2).
TOPURIA, VARLAM (1925), 'Ormagi uarq’opa kartulš i', [Double negation in
Georgian] C’elic’deuli I-II: 74-116.
TOTTIE, GUNNEL (1994), 'Any as an indefinite determiner in non-assertive
clauses: evidence from Present-day and Early Modern English", in:
Kastovsky, Dieter (ed.), Studies in Early Modern English (Berlin: Mouton de
Gruyter), 413-27.
TRAUGOTT, ELIZABETH C. (1988), 'Pragmatic strengthening and
grammaticalization', Berkeley Linguistics Society 14: 406-16.
TRAUGOTT, ELIZABETH C. and HEINE, BERND (eds.) (1991), Approaches to
Grammaticalization. 2 vols. (Typological Studies in Language, 19: 1-2.)
(Amsterdam: Benjamins).
TSAI, CHING-YUAN (1990), 'Question words as quantifiers in Chinese', Journal of
Chinese Linguistics 18(1): 125-57. (Also in Working Papers in Linguistics 20(2):
35-59. Honolulu: Department of Linguistics, University of Hawaii.)
TSIMPLI, IANTHI-MARIA and ROUSSOU, ANNA (1993), 'Polarity items in Modern
Greek: their distribution and interpretation', University College of London
Working Papers in Linguistics 5: 129-59.
TSCHENKÉLI, KITA (1958), Einführung in die georgische Sprache. Bd. 1:
Theoretischer Teil. (Zürich: Amirani).
UBRJATOVA, E. I. (ed.) (1982), Grammatika sovremennogo jakutskogo literaturnogo
jazyka. Vol. 1. (Moscow: Nauka).
 430

ULTAN, RUSSELL (1978), 'Some general characteristics of interrogative systems',

in Greenberg (ed.) (1978): 211-48.
VACHEK, JOSEF (1947), 'Obecný zápor v anglič tině a v č eš tině ." Facultas
philosophica universitatis Carolinae Pragensis, Práce z vě decký ch ústavů , 51: 7-
72.
– (1962), 'On the interplay of external and internal factors in the development of
language", Lingua 11: 433-48.
VAN DER WOUDEN, TON (1994), Negative Contexts (Groningen Dissertation in
Linguistics, 12.) Groningen: University of Groningen.
VAN VALIN, ROBERT D. (1990a), 'Layered syntax in Role and Reference
Grammar', in Machtelt Bolkestein and Jan Nuyts and Co Vet (eds.), Layers
and Levels of Representation in Language Theory: a Functional View (Amsterdam:
Benjamins), 193-231.
— (1990b), 'Functionalism, anaphora and syntax', Studies in Language 14(1), 169-
219.
VASMER, MAX (1953-58), Russisches etymologisches Wörterbuch, Vols. I-III.
(Heidelberg: Winter).
VATER, HEINZ (1975), 'Pro-Formen des Deutschen', in Michael Schecker and
Peter Wunderli (eds), Textgrammatik (Tübingen: Niemeyer), 20-42.
VELOUDIS, JOANNIS (1982), 'Negation in Modern Greek', Ph.D. dissertation,
University of Reading.
VENDLER, ZENO (1967), 'Each and every, any and all', in Zeno Vendler, Linguistics
in Philosophy. Ithaca: Cornell University Press, 70-96.
VENNEMANN, THEO (1988), 'Language change as language improvement', in
Vincenzo Orioles (ed.), Modelli esplicativi della diacronia linguistica (Pisa:
Giardini), 11-35.
VERMEYLEN, ALPHONSE (1984), 'Réflexions sur des ‘négations implicites’ en
espagnol', Cahiers de l’Institut de linguistique de Louvain 10(1-3): 305-19.
VEYRENC, JACQUES (1964), 'Kto-nibud’ et kto-libo formes concurrentes?' Revue des
études slaves 40: 224-33.
— (1975), 'L’indéfini du virtuel comme base de l’interrogatif?' Annales de
l’Institut français de Zagreb, 3ème série, No. 1.
VOEGELIN, C. F. and VOEGELIN, F. M. (1977), Classification and Index of the World’s
Languages (New York: Elsevier).
VOGT, HANS (1971), Grammaire de la langue géorgienne (Oslo:
Universitetsforlaget).
VON BERGEN, ANKE and VON BERGEN, KARL (1993), Negative Polarität im
Englischen. (Tübinger Beiträge zur Linguistik, 379.) (Tübingen: Narr).
VON BREMEN, KLAUS (1983), Question-words: A Study in the Syntax of
Relativization, Free Relatives, Pseudo-Cleft Sentences and Certain Indefinite
Pronouns. Ph.D. dissertation, University of Stockholm.
— (1986), 'Le problème des forclusifs romans', Lingvisticae Investigationes 10(2):
223-65.
WARD, DENIS (1977), 'On indefinite pronouns in Russian', Slavonic and East
European Review 55(4): 444-69.
WARFEL, SAM L. (1972), 'Some, reference, and description', in J. H. Battle and
John Schweitzer (eds.), Mid-America Linguistics Conference Papers.
(Stillwater/OK: Oklahoma State University), 41-9.
WEBER, DAVID JOHN (1989), A Grammar of Huallaga (Huánuco) Quechua.
(University of California Publications in Linguistics, 112.) (Berkeley:
University of California Press).
WIERZBICKA, ANNA (1980a), Lingua Mentalis: The Semantics of Natural Language
(Sydney: Academic Press).
 431

— (1980b), 'The ignorative: the semantics of speech acts', in Wierzbicka (1980a):

287-345. (Revised version of an article originally published in International
Journal of Slavic Linguistics 2.2-3 (1977): 251-313.)
XARALAMPIEV, I. (1977), 'Starinni neopredelitelni mestoimenni i adverbialni
formi v edin jugo-zapaden govor', Bâlgarski ezik 27(2): 130-32.
XOVRATOVIČ , V. V. (1989), 'Semantika neopredelennyx mestoimenij (na
materiale tureckogo jazyka)', Sovetskaja Tjurkologija 1989(5): 26-37.
YAMADA, SAE (1993), 'La négation et l'ordre de la polarisation: Expressions
négatives et NPIs du japonais', in: André Crochetière et al. (eds.), Actes du
XVe Congrès International des Linguistes, Québec, vol. 3. (Sainte-Foy: Presses de
l'Université Laval), 131-4.
ZANUTTINI, RAFFAELLA (1987), 'Negazione e concordanza negativa in italiano e
in piemontese', Rivista di Grammatica Generativa 12: 153-72.
— (1991), 'Syntactic properties of sentential negation: a comparative study of
Romance languages', Ph.D. dissertation, University of Pennsylvania.