Cambridge University Press

0521583748 - Phonology and Language Use

Joan Bybee
Excerpt
More information

1
Language Use as Part of Linguistic Theory

1.1 Substance and Usage in Phonology

This book introduces into the traditional study of phonology the notion
that language use plays a role in shaping the form and content of sound
systems. In particular, the frequency with which individual words or
sequences of words are used and the frequency with which certain pat-
terns recur in a language affects the nature of mental representation
and in some cases the actual phonetic shape of words. It is the goal
of the present work to explore to the extent possible at the present
moment the nature of the relation between the use of linguistic forms
on the one hand, and their storage and processing on the other.
To someone approaching linguistics from other disciplines, it might
seem odd that language use has not been taken into account in for-
mulating theories of language. However, since language is such a
complex phenomenon, it has been necessary to narrow the eld of
study to make it manageable. Thus we commonly separate phonology
from syntax, synchrony from diachrony, child language from adult lan-
guage, and so on, constantly bearing in mind that interactions exist that
will eventually have to be taken into account. We then go on to for-
mulate theories for these domains a theory of syntax, a theory of
phonology, a theory of language acquisition knowing all the while
that the ultimate goal is to encompass all these subelds in one theory
of language.
Early in the twentieth century, a proposal was made to distinguish
the shared knowledge that a community of speakers has from the
actual uses to which that knowledge is put (de Saussure 1916). Many
researchers then focused their attention on the structure of that shared
knowledge (called langue by Saussure and competence by Chomsky

Cambridge University Press www.cambridge.org

Cambridge University Press
0521583748 - Phonology and Language Use
Joan Bybee
Excerpt
More information

2 Language Use as Part of Linguistic Theory

1965) and paid little attention to language use in real time. The focus
on competence, or the structure of language, turned out to be
extremely productive. Structuralism provided linguists with a work-
shop of analytic tools for breaking down the continuous speech stream
into units, and these units into features; structuralism postulated hier-
archical relations among the units and assigned structures to different
levels of grammar, organizing language and the people who study it
into subelds phonology, morphology, syntax, and semantics.
The present work proposes to demonstrate that the focus on struc-
ture needs to be supplemented with a perspective that includes more
than just structure, a view that includes two other important aspects of
the language phenomenon the material content or substance of lan-
guage, and language use. The substance of language refers to the two
polar ends phonetics and semantics that language molds and struc-
tures, the two ends between which language forms the bridge. Lan-
guage use includes not just the processing of language, but all the social
and interactional uses to which language is put. For present purposes,
in the context of phonology, the frequency with which certain words,
phrases, or patterns are used will be shown to have an impact on
phonological structure. I will return to a discussion of these two aspects
of language and the role they play in past and future theories after
describing some recent developments in linguistics and related elds
that suggest a need for an enlarged perspective on language.
In the domain of morphosyntax, a substantial development beyond
structuralism has already taken place. The content of grammatical
categories has been studied as a substantive rather than a structural
matter, for example, in crosslinguistic studies of subject, topic, noun,
verb, tense, aspect (Comrie 1976, 1985, Dahl 1985), mood, and so on.
Also use is being studied as a prime shaper of syntactic structure
(Givn 1979, Haiman 1994, Hopper and Thompson 1984, and others)
and morphological structure (Bybee 1985, Bybee et al. 1994, DuBois
1985). So far, no comparable development has occurred in phonology,
but there are several indicators that it is time to open up the eld to
new questions and new sources of data and explanation.
Despite having looked carefully at matters of structure, having
dened and redened units such as phoneme and morpheme (or for-
mative), having shifted and reshifted levels such as phonemic and mor-
phophonemic, we nd that problems and questions still remain. Units
and levels do not submit to denitions that work for every case. We
still do not have strict denitions of even the most basic units, such as

Cambridge University Press www.cambridge.org

Cambridge University Press
0521583748 - Phonology and Language Use
Joan Bybee
Excerpt
More information

1.1 Substance and Usage in Phonology 3

segment, syllable, morpheme, and word. Instead we nd variation and

gradience commonplace in empirical studies, and we nd phonological
phenomena intimately bound up with lexicon and morphology, syntax,
discourse, and social context.
Developments from outside linguistics also point to a new view of
language. Studies of natural categorization by psychologist Eleanor
Rosch and her colleagues have had an impact on the way that linguists
view categories, including word meaning (Lakoff 1987), grammatical
classes such as gender (Zubin and Kpcke 1981), verb classes (Bybee
and Moder 1983), grammatical functions such as subject and topic, and
phonetic categories (K. Johnson 1997, Miller 1994, and other exem-
plar approaches to phonetic categories). In particular, these studies
show that the way human beings categorize both nonlinguistic and lin-
guistic entities is not by discrete assignments to categories based on the
presence or absence of features, but rather by comparison of features
shared with a central member. All category members need not have
all of the features characterizing the category, but a member is more
central or more marginal depending on the number and nature of
shared features. Moreover, Nosofsky (1988) has shown that the per-
ceived center of a category can shift toward the more frequently expe-
rienced members.
A second development important to linguistic modeling is the
development of computer models that can reproduce apparent
rule-governed behavior as well as probabilistic behavior using
parallel distributed processing (Daugherty and Seidenberg 1994,
Rumelhart and McClelland 1986, and others). In such models, labeled
connectionist models, structures are not given in advance (i.e., innate),
but take their form from the nature of the input, just as neurological
matter is structured by the input it receives. Connectionist models,
then, are quite compatible with usage-based theories of language.
Langacker (1987) and now Ohala and Ohala (1995) argue that storage
of linguistic percepts should be like the storage of other mental
percepts.
Yet a third recent development applicable to a large array of sci-
ences is the study of complex systems and their emergent properties.
The basic idea behind emergence as it will be applicable here is that
certain simple properties of a substantive nature, when applied repeat-
edly, create structure. Lindblom et al. (1984) are, to my knowledge, the
rst to apply the notion of emergent structure in linguistics. They illus-
trate emergence in the following way:

Cambridge University Press www.cambridge.org

Cambridge University Press
0521583748 - Phonology and Language Use
Joan Bybee
Excerpt
More information

4 Language Use as Part of Linguistic Theory

Termites construct nests that are structured in terms of pillars and arches and
that create a sort of air-conditioned environment. The form of these nests
appears to arise as a result of a simple local behavioral pattern which is followed
by each individual insect: the pillars and arches are formed by deposits of gluti-
nous sand avored with pheromone. Pheromone is a chemical substance that is
used in communication within certain insect species. Animals respond to such
stimuli after (tasting or) smelling them. Each termite appears to follow a path of
increasing pheromone density and deposit when the density starts to decrease.
Suppose the termites begin to build on a fairly at surface. In the beginning the
deposits are randomly distributed. A fairly uniform distribution of pheromone
is produced. Somewhat later local peaks have begun to appear serving as stimuli
for further deposits that gradually grow into pillars and walls by iteration of the
same basic stimulus-response process. At points where several such peaks come
close, stimulus conditions are particularly likely to generate responses. Deposits
made near such maxima of stimulation tend to form arches. As termites continue
their local behavior in this manner, the elaborate structure of the nest gradually
emerges. (Lindblom et al. 1984: 185186)

Lindblom et al. point out that the importance of this notion for lin-
guistics is that structure can be explained without attributing a mental
blueprint to the creatures creating the structure that substance and
form are intimately related (see also Hopper 1987, Keller 1994). Note
further that in this example and others of emergence in complex
systems, substance and form are related via the process by which the
structure is created.
If we apply emergence to language, the substance and use interact
to create structure. The substance in question includes both phonetics
and semantics. Phonetic substance has always been included in the eld
of phonology. Only a few phonologists have ever proposed that
phonology is independent of phonetics (see Postal 1968). On the con-
trary, most phonologists see phonetics as motivating phonology (for a
recent statement, see Hayes 1999). They have perhaps not always been
serious enough about pursuing the phonetic facts, however. One promi-
nent feature of generative phonology has been its disdain for the low-
level phonetic properties of speech, properties that presumably border
on performance.
Semantics, on the other hand, has been considered irrelevant to
phonology. This would not seem to be such a serious allegation to level
at phonologists, except that phonological descriptions and theoretical
works are full of references to notions such as morpheme and word
boundaries both of which delimit meaningful units as well as to
specic grammatical categories or specic morphemes. Generative
phonologists and Optimality Theory phonologists have proceeded as

Cambridge University Press www.cambridge.org

Cambridge University Press
0521583748 - Phonology and Language Use
Joan Bybee
Excerpt
More information

1.1 Substance and Usage in Phonology 5

though the content of these categories did not matter. I have shown in
Bybee (1985) that the phonological fusion of morphemes reects their
degree of semantic fusion, and in the chapters of this book, I will
explore further the relation between grammatical and lexical units and
phonological structure. Generative theories have largely neglected
such topics: even though morphological decomposition has played an
important role in the development of generative theories from The
Sound Pattern of English to Lexical Phonology and Optimality Theory,
the semantic derivations that should parallel the phonological ones
have never been attempted.
While substance has found its way into phonology from both the
phonetic and semantic end, use has been systematically excluded from
structuralist theories altogether.As mentioned earlier, distinctions such
as langue versus parole (de Saussure) and competence versus perfor-
mance (Chomsky) were specically designed to set up a mental object
that is separate from the uses to which it is put and to designate the
mental object as the proper domain for linguistics. Of course, there is
some value in distinguishing mental representations from the social
activities upon which they are based, but totally excluding factors of
use from consideration ignores the potential relation between repre-
sentation and use. It is certainly possible that the way language is
used affects the way it is represented cognitively, and thus the way it
is structured.
In fact, a good deal of progress in morphology and syntax has been
made in explaining specic phenomena by making just this assump-
tion. It has been shown that syntactic structures are the result of the
conventionalization of frequently used discourse patterns (e.g., DuBois
1985, Givn 1979), and that grammatical morphemes develop from
lexical morphemes in particular constructions through increases in the
frequency of use and through extension in use to more and more con-
texts (Bybee et al. 1994, Haiman 1994). Greenberg (1966) has demon-
strated that markedness effects are directly related to frequency of use,
with unmarked members of categories being the most frequent, and
Tiersma (1982) has shown that this hypothesis also explains cases of
local markedness in morphology. Psycholinguists have long known that
high-frequency words are accessed faster than low-frequency ones, and
I have argued that high-frequency irregular morphological formations
tend to maintain their irregularities precisely because of their high
frequency (Bybee 1985, Hooper 1976b). In all of these ndings we have
a dynamic aspect language structure is becoming or remaining

Cambridge University Press www.cambridge.org

Cambridge University Press
0521583748 - Phonology and Language Use
Joan Bybee
Excerpt
More information

6 Language Use as Part of Linguistic Theory

because of the way language is used. Thus the emphasis on the static,
synchronic language as the object of study has given way to the view
of language as slowly, gradually, but inexorably mutating under the
dynamic forces of language use.
Very little attention has been given to phonology in this usage-based
approach to language, yet these same ideas can be applied to phono-
logical phenomena with very interesting results. It is the purpose of this
book to explore the phenomena that have traditionally been studied
as phonology, reevaluating structural notions in terms of use and sub-
stance. The successes of structuralism in its various guises are not being
discarded. Rather structural notions will rst be empirically evaluated
to ascertain their viability, then the basis of such notions will be con-
sidered, and the role that substance and especially, use, plays in the phe-
nomena will be discussed. The phenomena discussed here point to a
deep involvement of phonology with lexicon and grammar, and a role
for both token and type frequency in shaping phonological structure.
A dynamic view of language is taken here, one that integrates both
synchronic and diachronic sources of explanation.1

1.2 Some Basic Principles of a Usage-Based Model

The ideas that I will apply to phonology are for the most part already
present in the literature and are now shared by a number of linguists,
phoneticians, and psychologists. A brief statement of these ideas
follows.
1. Experience affects representation. The use of forms and patterns
both in production and perception affects their representation in
memory. High-frequency words and phrases have stronger rep-
resentations in the sense that they are more easily accessed and
less likely to undergo analogical change. Low-frequency words
are more difcult to access and may even become so weak as to
be forgotten. The lexical strength of words may change as they
are used more or less in different contexts. Patterns (represented
as schemas, see below) that apply to more items are also stronger

1
The phonological theory developed here is quite different from Natural Generative
Phonology (NGP) (Hooper 1976a). For while NGP had very concrete lexical represen-
tations, much involvement of morphology and the lexicon with phonology, and the same
view of the relation of synchrony to diachrony, it was a structuralist theory and provided
no means of representing the impact of language use on language structure.

Cambridge University Press www.cambridge.org

Cambridge University Press
0521583748 - Phonology and Language Use
Joan Bybee
Excerpt
More information

1.2 Some Basic Principles of a Usage-Based Model 7

and more accessible, and thus more productive than those apply-
ing to fewer items. This is in contrast to modular approaches in
which representations and rules or constraints are all static and
xed, and in which all rules or representations in the same com-
ponent have the same status (for instance, all being equally acces-
sible no matter how many forms they apply to).
2. Mental representations of linguistic objects have the same
properties as mental representations of other objects. Of course,
this is the simplest assumption we can make that the brain
operates in the same way in different domains. One consequence
of this assumption is that mental representations do not have
predictable properties abstracted away from them, but rather
are rmly based on categorizations of actual tokens. As
Langacker (1987) and Ohala and Ohala (1995) have pointed
out, if predictable properties are taken away from objects,
they become unrecognizable. (See Chapter 2 for further
discussion.)
3. Categorization is based on identity or similarity. Categorization
organizes the storage of phonological percepts. What form this
categorization takes is an interesting question and one that can
be approached through phonetic and psychological experimen-
tation as well as through analogies with ndings in other percep-
tual domains. From structural linguistic analysis we can already
identify many different types of relations among linguistic objects
for example, the relation between two phonetic tokens of the
same word, that between tokens of the same morpheme in dif-
ferent words, and that between two similar phones in different
words in the same or different contexts.
4. Generalizations over forms are not separate from the stored
representation of forms but emerge directly from them. In
Langackers terms, there is no rule/list separation (see Chapter
2). Generalizations over forms are expressed as relations
among forms based on phonetic and/or semantic similarities. New
forms can be produced by reference to existing forms, but most
multimorphemic words are stored whole in the lexicon.
5. Lexical organization provides generalizations and segmentation
at various degrees of abstraction and generality. Units such as
morpheme, segment, or syllable are emergent in the sense that
they arise from the relations of identity and similarity that
organize representations. Since storage in this model is highly

Cambridge University Press www.cambridge.org

Cambridge University Press
0521583748 - Phonology and Language Use
Joan Bybee
Excerpt
More information

8 Language Use as Part of Linguistic Theory

redundant, schemas may describe the same pattern at different

degrees of generality (Langacker 2000).
6. Grammatical knowledge is procedural knowledge. Anderson
(1993) and Boyland (1996) distinguish declarative or proposi-
tional knowledge (e.g., Washington, DC is the capital of the
United States) from procedural knowledge (how to drive a car,
tie your shoelaces, and so on). While linguistic knowledge is in
part declarative (in the sense that we can cite the meanings of
words, for instance), much linguistic knowledge is procedural
(Boyland 1996). A native speaker can form an acceptable sen-
tence quite automatically, yet be unable to explain how this was
done or to list what the properties of an acceptable sentence are.
Thinking of grammatical constructions as procedural units has
profound consequences for our view of phonology. Phonology
then becomes a part of the procedure for producing and decod-
ing constructions, rather than a purely abstract, psychological
system.

1.3 The Creative Role of Repetition

Usage-based functionalism emphasizes language as a conventional-
ized, cultural object. In order to understand the nature of language, we
need to understand what it means for behavior to be conventionalized.
Haiman (1994, 1998) discusses grammar as ritualized behavior and
points to various properties of both ritual and grammar that are the
result of repetition. It is useful here to distinguish between a ritual and
a convention: though both represent repeated behavior, a ritual can
be individual and idiosyncratic, but a convention is agreed upon
socially and evokes a consistent response in other members of a society
(Tomasello et al. 1993). What both concepts have in common is that
their structure is shaped by repetition. The following is a summary of
some aspects of language that are shaped by repetition.
Through repetition we get lexical strength strong, easily accessible
representations, such as a greeting when you see someone you
know or responses such as thank you and youre welcome; that is,
any kind of learned automatic response. It is repetition that ritualizes
these responses and makes them readily available. These are just
extreme examples of a general phenomenon that pervades linguistic
representation repetition leads to strength of representation (Bybee
1985).

Cambridge University Press www.cambridge.org

Cambridge University Press
0521583748 - Phonology and Language Use
Joan Bybee
Excerpt
More information

1.3 The Creative Role of Repetition 9

Repetition also leads to reduction of form. This is true of nonlin-

guistic gestures such as making the sign of the cross. It is true in non-
human rituals: among chimpanzees (according to Plooij 1978, cited in
Haiman 1994) the original gesture of lying down is reduced to just
leaning slightly backwards. And it is true of language in many obvious
cases. Greetings become reduced, (how are you becomes hi), gram-
maticizing phrases with increasing frequency reduce and compress
(going to becomes gonna), and, in less obvious cases, there is a general
frequency effect in reductive sound changes (see Section 1.4).
Repetition also leads to the reduction of meaning. This reduction or
bleaching of meaning can be related to what Haiman calls habituation,
or the loss of impact due to repetition. Habituation is also a general
phenomenon, not restricted to language or to humans. It is a decline
in the tendency to respond to stimuli that have become familiar due to
repeated or persistent exposure (Haiman 1994:7). We recognize habit-
uation in the trivialization by repetition of great music (Beethovens
Fifth Symphony) or great art (Van Goghs sunowers). We also nd it
in language in cases where the emphatic becomes the normal. For
instance, in the French negative construction ne . . . pas, pas, literally
step, was once an emphatic added to the original negative ne, but is
now obligatory and nonemphatic.
Finally, and perhaps most importantly, repetition leads to emanci-
pation. In emancipation, instrumental actions are disassociated from
their original motivation and are free to take on a communicative func-
tion instead. The military salute derives from the more instrumental
gesture used in the Middle Ages when knights in armor greeted one
another. They raised the visor of their helmet to show their faces as an
indication of a peaceful greeting. The armor is gone, the visor is gone,
but a reduced form of the gesture remains, though without its instru-
mental function. It no longer raises the visor, but it has been imbued
instead with the function of communicating respect for the military
hierarchy.
Applications of the principle of emancipation through repetition in
language involve all sorts of cases of conventionalization, and most
commonly, cases in which one communicative function is replaced by
another. For instance, the inquiry into someones current state of being,
how are you, is not just reduced phonologically to hi, but also is eman-
cipated from its original communicative value and now serves simply
as a greeting. (A more conservative function of hi is found in some
dialects of Black English where speakers commonly respond to hi with

Cambridge University Press www.cambridge.org

Cambridge University Press
0521583748 - Phonology and Language Use
Joan Bybee
Excerpt
More information

10 Language Use as Part of Linguistic Theory

ne). Emancipation is also richly illustrated in the process of gram-

maticization during which words lose their categoriality. For instance,
verbs become auxiliaries and sometimes afxes, and also become dis-
associated from their lexical meaning and take on pragmatic or gram-
matical functions, as when be going to loses its motion sense and
becomes a future marker.
Haiman (1994) demonstrates that the development of ritual is a
common process in the animal kingdom, and by no means restricted to
humans, or even primates, as dog and cat owners can attest. He further
argues (Haiman 1998) that ritualization is the basis for the develop-
ment of grammar. The process of grammaticization depends upon rep-
etition and is characterized by the reduction of both meaning and form,
by strong entrenchment of patterns, and by emancipation in the sense
that forms in their grammaticizing constructions often shift from pro-
positional meaning to discourse-oriented functions (Traugott 1989).
Our understanding of the ritualization process can be applied to
syntax, as Haiman has shown, but also to phonology, as we investigate
the role of repetition in the structuring of phonological patterns and
lexical representations.

1.4 Frequency Effects

Much is already known about frequency effects in language, and much
remains to be learned. In this section, I will lay out the basic notions
and terminology that will be taken up again in later chapters.
There are two ways of counting frequency of occurrence that are
applicable to language: token frequency and type frequency. TOKEN
FREQUENCY is the frequency of occurrence of a unit, usually a word,
in running text how often a particular word comes up. Thus broke
(the past tense of break) occurs 66 times per million words in Francis
and Kucera (1982), while the past tense verb damaged occurs 5 times
in the same corpus. In other words, the token frequency of broke is
much higher than that of damaged.
TYPE FREQUENCY refers to the dictionary frequency of a particular
pattern (e.g., a stress pattern, an afx, or a consonant cluster). For
instance, English Past Tense is expressed in several different ways, but
the expression with the highest type frequency is the sufx -ed, as in
damaged, which occurs on thousands of verbs. The pattern found in
broke has a much lower type frequency, occurring with only a handful
of verbs (depending upon how you count them: spoke, wrote, rode,

Cambridge University Press www.cambridge.org