3 Cognitive Processes and Language Use

3.1 Introduction
There is consensus among usage-based researchers that linguistic structure is
shaped by domain-general processes. But what exactly are these processes?
Different researchers have emphasized the importance of very different aspects
of usage and cognition in order to explain how grammar is shaped by perfor-
mance. Some linguists have emphasized the role of discourse (Givón 1979) and
interaction (Auer 2005), others have focused on conceptual factors (Langacker
1987) or iconicity (Haiman 1985), and yet other researchers have claimed that
grammar is shaped by cognitive constraints on sentence processing (Hawkins
2004), language acquisition (Chater and Christiansen 2008) and/or language
production (MacDonald 2013). One factor that has been especially prominent
in recent years is frequency of occurrence (Bybee and Hopper 2001; Ellis 2002;
Behrens and Pfänder 2016). There is ample evidence that frequency is an
important determinant of usage, acquisition and change (for reviews, see
Diessel 2007; Diessel and Hilpert 2016).
The various proposals are not mutually exclusive, but there are so many
different suggestions in the usage-based literature as to how grammar is shaped
by performance that the whole approach has been rightfully criticized for being
arbitrary and ad hoc (Newmeyer 2003). Obviously, what is needed is a more
structured framework that attempts to predict the interaction of the various
cognitive processes.
This chapter provides an overview of the various cognitive processes that are
involved in language use and explains, in general terms, how grammar, usage
and cognition are related. It is argued that language use involves a decision-
making process that is determined by cognitive factors from three general
domains: social cognition, conceptualization and memory. It is the purpose of
the chapter to provide a background and orientation for the analysis of the
emergence of linguistic structure in the grammar network in later chapters of
the book.1

1
Parts of the following discussion are based on Diessel (2017).

23

Downloaded from https://www.cambridge.org/core. Columbia University - Law Library, on 12 Aug 2019 at 15:32:49, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781108671040.003
 24 Foundations

3.2 Linguistic Decisions

Let us begin with some general thoughts on language use or performance.
Language use is a cooperative activity that is driven by interlocutors’ commu-
nicative intentions. The German psychologist Karl Bühler (1934) characterized
language as an “instrument,” Greek “organon,” which speakers use to provide
information, to direct other people’s actions, to ask questions or to express
emotions (Austin 1962). Crucially, there are always multiple ways of expres-
sing a particular communicative intention – of saying more or less the same
thing, e.g., alternative constructions to describe the same scene (1a–b), alter-
native words to designate the same entity (2a–b) and alternative pronunciations
(3a–b).
(1) a. I sent Tom a letter.
b. I sent a letter to Tom.
(2) a. I didn’t see the man.
b. I didn’t see him.
(3) a. They are going to leave.
b. They’re gonna leave.

Of course, alternative does not mean equal. The examples in (a) and (b) are not
equivalent. They differ with regard to style, information structure and concep-
tualization, but they are close enough that speakers can (often) choose between
them. What is more, speakers cannot only choose between alternative means
that are stored in (linguistic) memory, they can also produce novel forms and
meanings. Language is productive in the sense that speakers are able to extend
the use of previously established linguistic patterns within certain limits
(Chapter 7). The productive use of language increases the range of linguistic
means that are potentially available to express a particular communicative
intention.
Given that there are always alternative ways of saying more or less the same
thing, speakers have to make choices, that is, they have to “decide” how to
express a particular intention or meaning. That does not mean, however, that
they consciously consider the various alternatives and then make a deliberate
decision. While there are situations in which language use proceeds in this way,
notably in writing, the decisions speakers make in spontaneous conversation
are often routinized and unconscious. Still, given that there are always alter-
native ways of expressing a particular communicative intention, we will say
that language production involves an (unconscious) decision-making process
that concerns the choice of linguistic means in a particular situation.
A parallel analysis applies to comprehension. Like speakers, listeners have
to make decisions. Every word and every structure has multiple interpretations
that are contingent on the context and listeners’ knowledge. Psycholinguists

Downloaded from https://www.cambridge.org/core. Columbia University - Law Library, on 12 Aug 2019 at 15:32:49, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781108671040.003
 Cognitive Processes and Language Use 25

have characterized sentence comprehension as a “constraint satisfaction pro-

cess” (MacDonald et al. 1994) whereby listeners seek to derive a plausible and
coherent interpretation from (i) the expressions they encounter, (ii) their under-
standing of the current interaction and (iii) general world knowledge
(Trueswell et al. 1994; MacDonald and Seidenberg 2006). If we think of
language comprehension in this way, it involves an (unconscious) decision-
making process, just like language production.
The linguistic decision-making process is key to understanding how gram-
mar, usage and cognition are related. In what follows, I argue that a language
user’s linguistic choices are motivated by competing cognitive processes from
three general domains, namely, the domains of social cognition (§3.3), con-
ceptualization (§3.4) and memory and processing (§3.5). The competing pro-
cesses affect the linguistic decision-making process online, but since recurrent
decisions tend to become automatized, they have long-term effects on language
development in both history and acquisition. If we want to understand how
grammar is shaped by performance, we have to study both the cognitive
processes of language use and their effects on language development.
Note, however, that language use and language development are not only
influenced by cognitive processes but also by social factors. Cognitive linguists
and psychologists tend to focus on the analysis of cognitive processes, but there
is ample evidence that linguistic decisions are also influenced by social factors
such as prestige and group identity (e.g., Labov 1972). Many linguistic features
are associated with particular social groups and social contexts. I gotta go, for
instance, has a different social connotation from I have to leave. Historical
linguists have shown that the sociolinguistic values of linguistic features have a
significant impact on speakers’ choice of linguistic means and their develop-
ment in language history (Weinrich et al. 1968; Labov 1972; Trudgill 1974).
Moreover, social factors are important to understand linguistic conventions.
Since individual speakers seek to speak like their peers, speech communities
develop group-specific patterns of language use that can be seen as linguistic
conventions. There is no doubt that social factors influence language use and
language change, but in this book the focus of analysis is on cognitive pro-
cesses, including cognitive processes of social cognition.

3.3 Social Cognition

Language use is a particular form of social interaction, which crucially relies on
the ability to take another person’s knowledge, intentions and beliefs into
account (Clark 1996). This ability is of fundamental significance in the use of
linguistic symbols and has been characterized as a particular capacity of the
human mind that distinguishes human communication from that of other
species (Tomasello 1999). While other species are able to communicate in

Downloaded from https://www.cambridge.org/core. Columbia University - Law Library, on 12 Aug 2019 at 15:32:49, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781108671040.003
 26 Foundations

one way or another, their ability to understand mental states and linguistic
symbols is limited compared to that of human beings (see Crockford et al. 2012
for a recent discussion).
A basic form of social cognition is “joint attention” (Carpenter et al. 1998;
Tomasello 1999). In order to communicate, speaker and listener must focus their
attention on the same experience, which may involve an object or event in the
surrounding situation or a concept that is evoked by the preceding discourse. In
face-to-face conversation, joint attention is commonly established by nonverbal
means of communication such as eye gaze, head movement and gesture. Of
particular importance is deictic pointing – a communicative device that is
universally available to establish joint attention and that is commonly accom-
panied by demonstratives or spatial deictics (Diessel 2006; Stukenbrock 2015).
Interestingly, the ability to engage in joint attention emerges only gradually
during early childhood (Carpenter et al. 1998; Tomasello 1999). While infants
respond to adults’ communicative behaviors from early on, it takes around nine
months until they begin to follow the eye gaze and head movements of other
people and it usually takes another three months until they begin to produce
their first pointing gestures. Tomasello interprets the emergence of these
behaviors as the first steps of a long-lasting process whereby children gradually
acquire a “theory of mind” (Tomasello 2003: 3).
Joint attention is a basic aspect of social interaction, but in order to communicate,
it is not only important that the speech participants are focused on the same object
or scene, they also have to align their knowledge and beliefs; that is, communica-
tion presupposes that the interlocutors share a “common ground” (Clark and
Brennan 1991). Clark (1996) defines common ground as language users’ aware-
ness of their shared knowledge, which does not only concern information about the
physical speech situation surrounding the interlocutors, but also background infor-
mation about the communicative partner and general world knowledge.
Common ground is a domain-general cognitive phenomenon that is not only
relevant for the use of language but also for other, nonverbal forms of social
interaction. Of particular importance to language are those aspects of common
ground that emanate from discourse. As discourse unfolds, the communicative
partners build up a body of shared representations providing a background and
orientation for the interpretation of elements in the ensuing discourse.
Functional linguists have emphasized the importance of discourse-based infor-
mation for the analysis of various aspects of linguistic structure including
nominal reference (e.g., Chafe 1994), word order (e.g., Givón 1979) and
subordination (e.g., Verhagen 2005).
Common ground provides the basis for what some psychologists (and socio-
linguists) call “audience design” (Clark and Marshall 1981), which is the
process whereby speakers seek to construct a sentence according to what
they think a hearer “needs” in order to understand their communicative

Downloaded from https://www.cambridge.org/core. Columbia University - Law Library, on 12 Aug 2019 at 15:32:49, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781108671040.003
 Cognitive Processes and Language Use 27

intention in a particular situation (see also Horton and Keysar 1996; Arnold
2008). Audience design is of central significance to speakers’ choice of lin-
guistic means. Consider, for instance, the use of referring (nominal) expres-
sions (see 4). In English, speakers can choose between definite and indefinite
NPs, proper names, proximal and distal demonstratives, third-person pronouns
and zero anaphors (in certain constructions):
(4) The man Definite noun phrase
A man Indefinite noun phrase
Peter Proper name
This Proximal demonstrative
That Distal demonstrative
HE Stressed pronoun
He Unstressed pronoun
ø Zero anaphor

The inventory of referring terms is language-particular. Different languages have

different sets of (pro)nominal expressions (Givón 1983; Ariel 1990), but all
languages have multiple types of referring terms so that speakers are forced to
choose between them. Functional linguists have shown that the various types of
referring expressions serve particular pragmatic functions that correlate with
aspects of the linguistic and nonlinguistic context (e.g., givenness, visibility)
(Givón 1983; Chafe 1994). However, from a cognitive perspective we may say
that speakers (often) choose a particular term based on what they think listeners
know and see, and listeners interpret the chosen expressions based on the assump-
tion that speakers construct sentences according to this strategy. In other words, the
choice and interpretation of linguistic expressions is crucially influenced by inter-
locutors’ “assessment of common ground” and the attempt to “tailor” an utterance
according to “hearers’ needs” (Clark and Marshall 1981; but see §3.6).
What is more, common ground and audience design do not only affect speakers’
choice of linguistic expressions, they also influence grammatical development and
grammar evolution. For instance, many languages have information-structure
constructions such as left-dislocation (e.g., That guy over there, he . . .) and cleft-
sentences (e.g., It is John who . . .), which can arguably be seen as grammatical
strategies that have evolved from discourse patterns that were used to establish a
thematic foundation (or common ground) for the interpretation of subsequent
information in the unfolding speech stream (Clark and Brennan 1991: 228).

3.4 Conceptualization
Conceptualization is concerned with the construction of meaning. In (formal)
semantics, linguistic meaning is commonly defined as some kind of correspondence
relation between language and world, or language and thought, but in the usage-

Downloaded from https://www.cambridge.org/core. Columbia University - Law Library, on 12 Aug 2019 at 15:32:49, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781108671040.003
 28 Foundations

based approach, meaning is shaped by conceptualization, which is the cognitive

structuring of experience or semantic content (Langacker 1987, 1991; Talmy 2000;
Croft and Cruise 2004: §3).
Like all other cognitive processes of language use, conceptualization is not
specific to language. In fact, the conceptual approach to semantics is inspired
by general psychological research on vision (see Evans and Green 2006: §3 for
a succinct summary). Pioneering research on conceptualization comes from
gestalt psychology (Koffka 1935), which had a strong impact on cognitive
semantics (Langacker 1991; Talmy 2000; Verhagen 2007). The gestalt psy-
chologists showed that vision involves more than the passive recording of
sensory cues – that visual perception is guided by general cognitive principles
such as the figure–ground distinction and reification (which is the enrichment
of perceptual information through inference).
Inspired by this research, cognitive and usage-based linguists have devel-
oped a conceptual theory of semantics in which the meaning of linguistic
expressions is structured by general processes of conceptualization including
metaphor, metonymy, fictive motion, force dynamics, reification and the
figure–ground segregation (Lakoff and Johnson 1980; Lakoff 1987;
Langacker 1991; Talmy 2000; Coventry and Garrod 2004).
Interestingly, Langacker (1991: 117) argued that there are always multiple
ways of viewing the same experience so that speakers are (often) forced to
decide how to describe and conceptualize a particular object or scene (see also
Croft and Cruse 2004: §3 and Langacker 2008: 43). Consider, for instance, the
use of come and go in the following examples.
(5) She came to school.
(6) She went to school.
Come and go are deictic verbs that can often be used with reference to the same
scene, but they describe the scene from different perspectives, that is, they
evoke different conceptualizations. In the case of come, the conceptual figure is
moving toward the observer, but in the case of go, the figure is moving away
from the observer (Figure 3.1).
Both verbs are interpreted relative to a particular point of reference, the
deictic center, also called the “origo” (Bühler 1934: 107). The deictic center is
the origin of a coordinate system that is usually grounded by the speaker’s body
or location at the time of the utterance, but the deictic center can be shifted from
the speaker to another person or fictive observer. In narrative discourse, for
instance, the deictic center is often located in one of the characters of the
narration who uses deictic expressions in the same way as speakers use them
for spatial orientation in the surrounding situation (see Diessel 2014 for
discussion).

Downloaded from https://www.cambridge.org/core. Columbia University - Law Library, on 12 Aug 2019 at 15:32:49, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781108671040.003
 Cognitive Processes and Language Use 29

come go

Figure 3.1 Conceptualization of come and go

Like words, constructions involve conceptualization. Consider, for instance,

the active–passive alternation in examples (7) and (8).
(7) The man kicked the ball.
(8) The ball was kicked (by the man).
An active sentence construes a scene from the perspective of the agent. In
sentence (7), the agent is the focus of attention and the patient is backgrounded
relative to the agent, but in the passive sentence in (8) it is the other way around.
In this case, the patient serves as figure and the agent is a secondary focal point
(Langacker 1991: 101–148), which can be omitted, but, of course, concep-
tually, the passive construction entails an agent or agentive force. Analyzing
grammatical relations in this way creates an explicit link between argument
structure and general conceptual processes (Chapter 7).
To give one more example, in languages with perfective and imperfective
aspect, action verbs can be construed in two different ways: as ongoing or
imperfective actions (e.g., I am writing a book) and as completed or perfective
actions (e.g., I have written a book). One feature that distinguishes ongoing
from completed events is “conceptual boundedness” (Langacker 1987: 86–87;
1991: 93–95). Ongoing events are temporally unbounded, whereas completed
events are temporally bounded. Of course, every event has a beginning and an
ending, but perfective verb forms construe an event as temporally bounded,
whereas imperfective verb forms present the same event as ongoing and
expansible (Talmy 2000: 50–62).
In general, in the usage-based approach, semantic conventions emerge from
recurrent conceptualizations of the same or similar experiences that become
associated with particular lexemes and constructions. We will consider the
conceptual approach to semantics in detail in Chapter 6. Here we note that
conceptualization is not only the driving force behind the construction of
meaning, it also plays a decisive role in the diachronic development of gram-
mar. For instance, the early stages of grammaticalization are (often) motivated
by general conceptual processes such as metaphor, metonymy and deictic
projection, which lead to the development of grammatical function words
from nouns, verbs and spatial deictics (Heine et al. 1991; Diessel 2012a).

Downloaded from https://www.cambridge.org/core. Columbia University - Law Library, on 12 Aug 2019 at 15:32:49, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781108671040.003
 30 Foundations

Similar conceptual processes occur in first language acquisition (Diessel

2011b, 2012b). We will consider the conceptual processes of language change
and L1 acquisition in later chapters (Chapters 5, 6 and 7) and now turn to
memory-related processes.

3.5 Memory-Related Processes

Functional and cognitive linguistics have always emphasized that linguistic
structure is motivated by semantic and pragmatic aspects of communication
and discourse, but in the recent literature, the focus of analysis has shifted from
communication and meaning to frequency and processing. Frequency and
processing concern the storage, representation and activation of linguistic
information in memory. In the older psychological literature, memory is
often described as some kind of place where information is stored, but in
current cognitive psychology, the term memory subsumes a set of cognitive
processes that concern the processing and organization of knowledge (Cowan
2005; Jonides et al. 2008). In what follows, I briefly consider some of the
memory-related processes that influence speakers’ choice of linguistic means.
All of these processes will be described in more detail in later chapters.

3.5.1 Attention and the Flow of Consciousness

Traditionally, the human memory system is divided into two basic “stores”:
long-term memory, which includes a person’s entire knowledge, and working
memory, which holds the information that is currently activated and processed
(Baddeley 1986). A number of recent studies have argued, however, that
information in working memory cannot really be delineated from information
in the long-term store (MacDonald and Christiansen 2002; Cowan 2005).
Challenging the traditional division between long-term memory and working
memory, these studies argue that memory constitutes a “unitary system” in
which working memory serves as an “attention mechanism” that activates
specific information at a particular point in time. In particular, Cowan (2005)
argued that one should think of memory as an encompassing network with an
inherent focus of attention (see also Oberauer 2002).
Above, we have seen that the creation of joint attention is an important aspect
of communication (§3.3); but the attention mechanism of the memory system is
not only influenced by social interaction but also by memory and sensory
perception, which can interfere with interlocutors’ attempt to coordinate their
attention.
Crucially, at any given moment in time, the focus of attention centers on only
one item, but this item is connected to semiactivated items, which in turn are
linked to other memory traces that are currently not activated but easily

Downloaded from https://www.cambridge.org/core. Columbia University - Law Library, on 12 Aug 2019 at 15:32:49, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781108671040.003
 Cognitive Processes and Language Use 31

Figure 3.2 Long-term memory with moving focus of attention (see Oberauer
2002: 412)

accessible through the activated item (Oberauer 2002; Cowen 2005). On this
view, the traditional notion of working memory corresponds to a cluster of
conceptually related elements with graded activation values in the overall
network of a person’s knowledge.
In accordance with this view, linguistic researchers have argued that the
focus of attention in speaking and listening constitutes an open-ended set of
activated and semiactivated items (Chafe 1994). Crucially, since language
unfolds in time, the focus of attention is moving and the whole cluster of
activated elements is constantly in flux (see Figure 3.2). Chafe refers to this
as the “flow of consciousness” and argues that the moving focus of attention is
an important determinant for language users’ choice of linguistic means. Word
order, for instance, is crucially influenced by the flow of attention, or flow of
consciousness (Chafe 1994: 162–165).

3.5.2 Categorization, Abstraction and Analogy

Categorization is the process whereby a new experience is classified as an
instance of an existing category or schema. Traditionally, categories are defined
by necessary and sufficient features, but in current cognitive psychology,
categories are commonly defined in terms of prototypes and exemplars (see
Murphy 2002: Chapters 3 and 4 for a review).2
Like attention and the flow of consciousness, categorization is a domain-
general process that concerns both linguistic and nonlinguistic concepts
(Lakoff 1987). Crucially, there are always multiple concepts (or categories)
that are potentially available to license (or categorize) a new (linguistic) token.
Consider, for instance, the categorization of speech sounds.
2
Note that while categorization is here subsumed under memory-related processes, it also
involves cognitive processes of conceptualization (Lakoff 1987; Langacker 1991).

Downloaded from https://www.cambridge.org/core. Columbia University - Law Library, on 12 Aug 2019 at 15:32:49, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781108671040.003
 32 Foundations

There is an enormous amount of variation in the phonetic realization of

speech sounds, especially in the domain of vowels. Very often, a vowel
token falls somewhere in between two or more speech-sound categories,
but listeners are forced to categorize (unconsciously) any given phonetic
token as a particular phoneme in order to arrive at a coherent interpretation.
A listener’s choice of category is determined by a number of factors
including the context, frequency and, perhaps most importantly, the simi-
larity between the new token and the phonetic properties of the competing
speech-sound categories. Similarity is a key concept of categorization and
does not only concern the classification of speech sounds, but all linguistic
elements including morphemes, words, phrases and constructions (Chapters
4, 7, 8 and 9).
Closely related to categorization are two other cognitive processes: abstrac-
tion and analogy. By abstraction I mean the process whereby language users
generalize across multiple experiences with overlapping properties and thereby
create a new concept or schema (Anderson 2005: 165–167; see also Langacker
2008: 17, who refers to abstraction as “schematization”). Abstraction plays a
central role in language acquisition, notably in grammar learning (Chapter 4).
The notion of analogy is used in many different ways by different scholars.
In historical linguistics, it is often used as a descriptive term for a certain type of
structural change, notably morphological change (Trask 1996: 105–115), but in
usage-based linguistics, analogy is a domain-general phenomenon that
accounts for an important aspect of linguistic productivity (Bybee and Moder
1983; Barðdal 2008; Behrens 2017). In the usage-based approach, linguistic
productivity is commonly defined as the extension of an existing schema to a
new item (Langacker 2000: 26; Bybee 2010: 94). Two general factors influence
the analogical extension of a constructional schema to novel expressions: (i) the
strength of a particular schema in memory, and (ii) the similarity between
lexical expressions that are licensed by a schema.
We will consider these factors in detail in later chapters (Chapters 4, 7, 8 and 9).
Here we note that there are two different types of similarity that influence analogy
(and also categorization and abstraction): (i) “object similarity,” which refers to
overlapping attributes, and (ii) “structural similarity,” which involves overlapping
structures or relations (Gentner 1983; Holyoak and Thagard 1995). Consider, for
instance, the three sets of circles and squares in Figure 3.3.
As can be seen, set A and set B are structurally similar – they involve similar
arrangements of geometric figures, but C is different. There is no structural
overlap between C and the two other sets, but C includes circles that are similar
in shape, color and size to those in B, that is, C and B exhibit object similarity.
Both types of similarity affect analogy (and categorization and abstraction),
but, as we will see (Chapters 4 and 7), structural similarity is of particular
importance to grammar learning as grammar is concerned with relations.

Downloaded from https://www.cambridge.org/core. Columbia University - Law Library, on 12 Aug 2019 at 15:32:49, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781108671040.003
 Cognitive Processes and Language Use 33

Set B

Set C

Set A

Figure 3.3 Object similarity and structural similarity

3.5.3 Priming
Priming is a well-known recency effect of activation in memory that is related
to analogy. Like analogy, priming is driven by similarity, both object similarity
and structural similarity, which has led some researchers to argue that priming
can be seen as a (particular) form of analogy (Leech et al. 2008; Goldwater et al.
2011). Yet, in contrast to the term analogy, the notion of priming is specifically
used to characterize the role of recency on activation spreading. Elements that
have been recently activated increase the likelihood that the same or related
elements will be (re-)used in the unfolding situation. Two basic types of
(language) priming are commonly distinguished: lexical priming and structural
priming.
Lexical priming refers to the facilitatory (or inhibitory) effect of a lexical
item, the prime, on the activation of a related item, the target. For instance,
people are faster and more accurate in identifying a word such as dog if the
word is preceded by a semantically related item such as cat than if it is preceded
by an unrelated word such as city. There is also evidence that the phonetic
features of a word affect the activation of phonetically related expressions that
rhyme or alliterate with the prime and that repetition speeds up lexical access
and word recognition (Harley 2001: 145–150).
Like lexical priming, structural priming is an implicit memory effect that
concerns the activation of knowledge, but structural priming involves relations
or structures rather than lexemes. If speakers can choose between alternative
constructions, and if one of them has been previously used, speakers are likely
to reuse the construction in the ensuing discourse. For instance, as Bock (1986)
demonstrated in a pioneering study, English speakers are more likely to
describe an act of transfer by the to-dative construction (She gave the book to
John) than the double-object construction (She gave John the book), if they

Downloaded from https://www.cambridge.org/core. Columbia University - Law Library, on 12 Aug 2019 at 15:32:49, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781108671040.003
 34 Foundations

have used a to-dative construction prior to the experimental task (He sent a
picture to his friend). Bock’s study launched a whole new line of research in
psycholinguistics (see Pickering and Ferreira 2008 for a review). There is now
a plethora of results indicating that both lexical priming and structural priming
have a significant impact on language users’ choice of linguistic means and the
development of linguistic knowledge in acquisition (e.g., Chang et al. 2006;
Goldwater et al. 2011; Rowland et al. 2012) and change (e.g., Jäger and
Rosenbach 2008, Pickering and Garrod 2017).

3.5.4 Exemplar Learning and Automatization

Repeated decisions strengthen the representation of linguistic elements in
memory. There is general consensus among usage-based linguists that repeti-
tion or frequency is an important determinant of linguistic knowledge, but the
strengthening effect of frequency can be analyzed from different perspectives
(Diessel 2016).
Some usage-based linguists have argued that exemplar learning provides a
cognitive mechanism that explains the role of frequency in language (Bybee
2006; Goldberg 2006). On this account, every piece of information, that is,
every token, encountered in experience leaves a trace in memory. Over time,
tokens with similar or identical features reinforce each other creating clusters of
overlapping memory traces known as “exemplars.” The whole token cluster
can be interpreted as an emergent category that functions as an “attractor” or
“cognitive reference point” for the classification of novel experiences
(Nosofsky 1988).
Exemplar theory has been especially influential in research on phonetics and
phonology, where speech-sound categories (e.g., vowel phonemes such as /ɛ/
and /ɔ/) emerge from many slightly different phonetic tokens that a language
user encounters in experience (Johnson 1997; Bybee 2001; Pierrehumbert
2001), but parallel analyses have also been proposed for the emergence of
linguistic elements in morphology and syntax (Bybee 1985; Goldberg 1995;
Bod 2009). Specifically, it has been argued that speakers’ knowledge of con-
structional schemas is based on their experience with particular lexical
instances of constructions.
Exemplar theory provides a plausible explanation for certain aspects of
categorization, category learning and the storage of linguistic information
(Chapter 4), but in the current approach, most frequency effects are explained
by automatization, which is immediately related to the network view of lan-
guage (Diessel 2016).
Automatization is a well-known concept of cognitive psychology (Anderson
2005: 99–103) that is closely related to the distinction between “automatic” and
“controlled processes” (Logan 1988; Schneider and Chein 2003). Automatic

Downloaded from https://www.cambridge.org/core. Columbia University - Law Library, on 12 Aug 2019 at 15:32:49, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781108671040.003
 Cognitive Processes and Language Use 35

processes occur without conscious control and effort – they are fast and can be
performed in parallel to other tasks. Using a computer keyboard, for instance, is
an automatic process that most people can perform parallel to other tasks (e.g.,
watching the computer screen). Controlled processes, by contrast, require
attention and monitoring and cannot be so easily combined with other tasks.
Entering new values into an electronic database, for instance, is a controlled
activity that requires attention and monitoring.
Automatization transforms controlled processes into automatic processes
through repetition or practice. This is a very common cognitive phenomenon
that concerns both motor actions and cognitive processes. Automatization
enables people to perform complex activities with little effort (Logan 1988)
but is also a common source for certain types of mistakes, e.g., slips, that occur
for lack of attention or lack of conscious control (Schneider and Chein 2003).
Language use is a highly automated process, which involves the rapid
production of articulatory gestures, the choice of lexemes and constructions
and the monitoring of the addressee. All of this occurs in milliseconds and
would not be possible without automatization (or routinization). In particular,
the motor movements of speech production are highly automated, but auto-
matization also affects the choice of lexical and grammatical means.
If we think of grammar as an encompassing network in which the various
aspects of a speaker’s linguistic knowledge are interconnected by associative
relations, we can define automatization as the process that strengthens the
associations between linguistic elements in memory through repetition. For
instance, if two or more lexemes are frequently used together, they become
associated with each other and develop into a holistic lexical unit (§2.4.2
and §5.2).
Let me emphasize, however, that automatization is not the only factor that
affects the strength of associative connections. As we will see, the various types
of links that constitute the grammar network are subject to different cognitive
processes. While all of them are strengthened by automatization, some are also
influenced by semantic and pragmatic factors. Symbolic links, for instance, are
crucially influenced by relevance. As we will see in Chapter 6, there is evidence
that speakers are able to memorize the meaning of a new form with no or little
repetition if the new expression is relevant to the speaker, suggesting that
relevance (or salience) has a particular impact on symbolic associations.
Automatization has far-reaching consequences for the linguistic decision-
making process and the storage of linguistic information. Linguistic elements
(or linguistic features) that are frequently processed together develop into
cognitive routines that are executed without any further decision once the
routine has started; that is, automatization reduces the number of linguistic
decisions that are involved in the production and comprehension of an utter-
ance (§5.2).

Downloaded from https://www.cambridge.org/core. Columbia University - Law Library, on 12 Aug 2019 at 15:32:49, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781108671040.003
 36 Foundations

3.6 Competing Motivations

Table 3.1 provides an overview of the cognitive processes that have been
considered in this chapter. The various processes concern different aspects of
language use and complement each other but can also be in competition. The
notion of “competing motivations” plays a key role in the usage-based analysis
of language (DuBois 1985; Bates and MacWhinney 1989; MacDonald et al.
1994; see also Diessel 2005). Of particular importance is the competition
between social cognition and memory-related processes.
There is general agreement that communication involves audience design,
but several recent studies have argued that speakers and listeners do not
constantly assess the other person’s mental state (Horton and Keysar 1996;
Pickering and Garrod 2004). More specifically, these studies argue that inter-
locutors often act in a self-oriented way without taking the other person’s
knowledge, beliefs and attention into account (see Arnold 2008 for a review).
For instance, in a series of studies Horton, Keysar and colleagues have
shown that speakers often disregard their listeners’ needs when choosing a
particular referring expression (Horton and Keysar 1996; Keysar et al. 2000;
Horton and Gerrig 2005). In one of their studies, they examined the way
speakers describe a spatial scene that was not visible to the listener under two
different conditions. In condition 1, speakers performed the task at their own
pace, but in condition 2, they were put under time pressure. In accordance with
their hypothesis, Horton and Keysar (1996) found that in condition 1 speakers
carefully considered what listeners could potentially know about the hidden
scene when describing it; but when speakers were put under time pressure, they
often acted in a self-oriented fashion and disregarded their hearers’ needs, as
evidenced by the fact that the chosen expressions were not always informative

Table 3.1 Some domain-general processes of

language use

Social cognition • Joint attention

• Common ground
• Audience design
Conceptualization • Figure–ground segregation
• Metaphor and metonymy
• Deixis and perspective
• Force dynamics
Memory-related processes • Attention and flow of consciousness
• Categorization, abstraction, analogy
• Lexical and structural priming
• Exemplar learning and automatization

Downloaded from https://www.cambridge.org/core. Columbia University - Law Library, on 12 Aug 2019 at 15:32:49, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781108671040.003
 Cognitive Processes and Language Use 37

to the addressee (as they were when speakers performed the task at their own
pace). Parallel results have been found in several other studies on speakers’
choice of linguistic means (Arnold 2008).
In accordance with this research on production, Hanna et al. (2003) observed
that comprehension can proceed in an automatic fashion. Specifically, these
researchers showed that the semantic interpretations listeners assign to linguis-
tic expressions are not always consistent with what they know about speakers’
current mental states, suggesting that comprehension, like production, does not
generally involve a careful assessment of common ground.
Similar results have been reported in linguistic and psycholinguistic research
on phonetic reduction. If listeners are likely to be able to predict the occurrence
of a particular word or utterance from the context, speakers tend to reduce
articulatory effort, but if there are no contextual cues that would allow listeners
to anticipate upcoming expressions, speakers tend to produce them carefully.
There is plentiful evidence that the degree of phonetic reduction correlates with
the predictability of linguistic expressions in a particular context (e.g., Jurafsky
et al. 2001; Bell et al. 2003, 2009; Gahl and Garnsey 2004; Kuperman and
Bresnan 2012; Arnon and Priva 2013). There is, however, also evidence that
phonetic reduction is influenced by production-internal processes, notably by
automatization (Bybee 2001) and priming (Fowler and Housom 1987), and that
reduction processes are not always adjusted to the hearers’ needs (Bard et al.
2000).
In general, there is now a large amount of research indicating that the
linguistic decision-making process is the result of an intricate interplay
between other-oriented and self-oriented processes. In production, we can
distinguish between “hearer-oriented processes” of common ground and audi-
ence design, and “speaker-oriented processes” of memory retrieval, priming
and automatization (Arnold 2008). A similar contrast occurs in comprehension:
very often, listeners’ interpretations of linguistic stimuli are based on their
assessment of common ground, but under time pressure, they may solely rely
on their experience with particular words and constructions without consider-
ing the speaker’s current mental state. More generally, as we will see through-
out this book, while communication needs some audience design, there is
evidence that both speakers and listeners often act in a self-oriented or mechan-
istic way (Pickering and Garrod 2004; Horton and Gerrig 2005; Arnold 2008).

3.7 Acquisition and Change

The linguistic decision-making process has long-term effects on language
development, which can be studied in two different time frames: (i) in ontoge-
netic time (i.e., L1 acquisition) and (ii) in diachronic time (i.e., language
change). There are many parallels between L1 acquisition and language change

Downloaded from https://www.cambridge.org/core. Columbia University - Law Library, on 12 Aug 2019 at 15:32:49, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781108671040.003
 38 Foundations

that have intrigued linguists for a very long time. At the end of the nineteenth
century, many scholars were convinced that the diachronic development of
language is crucially influenced by language acquisition. Henry Sweet, for
instance, argued that sound change can be the result of children’s defective
imitation of adult speakers’ pronunciations, and Max Müller claimed that the
regularization of irregular morphology is caused by children’s errors (see
Diessel 2012b for discussion).
Child-based explanations of language change have been very prominent in
historical linguistics until today (e.g., Lightfoot 1999; Culbertson and Newport
2015); but a number of recent studies have questioned the validity of child-
based theories of language change (Croft 2001: 45; Bybee 2010: 114–119).
Specifically, these studies argue that diachronic innovations do not arise from
errors and misanalyses in child language but from small changes in adult
language use (see also Sankoff and Blondeau 2007). One piece of evidence
for this view comes from the fact that children’s errors do not seem to survive
through adolescence into adulthood. If language acquisition was the source of
language change, one would expect children’s errors to persist through adoles-
cence into adulthood, but this is not the case (see Diessel 2011b, 2012b for
reviews of relevant research).
That does not mean, however, that language acquisition is irrelevant to the
study of language change. That speakers of Present-Day English do not under-
stand the Old English sentence Ne com se here ‘The army did not come’ is, of
course, a consequence of the fact that they have no direct experience with the
words and structural patterns in this sentence, indicating that the transmission
of language from one generation to the next plays an important role in the
diachronic development of language (e.g., Hare and Elman 1995; Kirby 1999).
Moreover, while children’s errors are usually eliminated in the development
from childhood into adolescence, there are striking parallels between L1
acquisition and language change that are interesting from a usage-based per-
spective, as the comparative analysis of language acquisition and language
change can help to better understand the dynamics of the language system. As I
have argued in Diessel (2011b, 2012b), while diachronic innovations are
unlikely to originate from errors in L1 acquisition, the two developments
involve the same (or very similar) cognitive processes, which account for the
many parallels between them. However, since L1 acquisition and language
change occur under very different circumstances, they also differ in important
ways. There are two points to note here.
First, an important aspect of grammar learning is the extraction of structural
patterns from the analysis of lexical sequences with overlapping properties
(Chapter 4). Schema abstraction plays a key role in L1 acquisition. Once the
first constructions are in place, they are often modified and extended under the
influence of new experiences; but the first steps of grammar learning crucially

Downloaded from https://www.cambridge.org/core. Columbia University - Law Library, on 12 Aug 2019 at 15:32:49, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781108671040.003
 Cognitive Processes and Language Use 39

involve the extraction of novel schemas from the analysis of lexical sequences,
which is a relatively rare phenomenon in language change. As we will see
(throughout the book), language change typically involves the modification and
extension of existing schemas rather than the extraction of entirely new ones.
Second, child and adult language are subject to different sociolinguistic
constraints. Young children seek to imitate the speech of adult speakers. The
ambient language provides a model for children’s speech during the preschool
years; but when children get older and reach adolescence, they begin to pay
attention to the speech of their peers. There is copious evidence indicating that
the language of adolescents and adults is influenced by social prestige and
group identity (Labov 1972; Trudgill 1974). In other words, while young
children strive to imitate the language of a few adult speakers, notably the
language of their parents, adults seek to speak in accordance with the linguistic
conventions of particular speech communities.

3.8 Summary
To summarize the discussion in this chapter, I have argued that language use
involves an unconscious decision-making process that is influenced by a wide
range of cognitive processes, which may be divided into three basic types: (i)
processes of social cognition, which concern the interaction between the
speech participants, (ii) processes of conceptualization, which concern the
cognitive structuring of experience, and (iii) memory-related processes,
which concern the storage, retrieval and processing of linguistic information.
The various cognitive processes affect the linguistic decision-making pro-
cess online and have long-term effects on language development. These effects
can be studied in two different time frames: in ontogenetic time (or language
acquisition) and in diachronic time (or language change). Since child and adult
speakers are influenced by the same cognitive processes, there are conspicuous
parallels between acquisition and change, but there are also important differ-
ences between them that reflect the different conditions under which linguistic
structures evolve in L1 acquisition and language change (Diessel 2011b,
2012b). In what follows, we will consider usage-based research on both
acquisition and change in order to better understand how grammar (or linguistic
knowledge in general) is shaped by language use.

Downloaded from https://www.cambridge.org/core. Columbia University - Law Library, on 12 Aug 2019 at 15:32:49, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781108671040.003
Downloaded from https://www.cambridge.org/core. Columbia University - Law Library, on 12 Aug 2019 at 15:32:49, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781108671040.003