DOCUMENT RESUME

ED 302 075 FL 017 698

AUTHOR Bates, Elizabeth; MacWhinney, Brian

TITLE What Is Functionalism?
PUB DATE Aug 88
NOTE 17p.; In: Papers and Reports on Child Language
Development, Volume 27; see FL 017 572.
PUB TYPE Reports - Evaluative/Feasibility (142) -- Viewpoints
(120) -- Speeches/Conference Papers (150)

EDRS PRICE MFO1 /PCO1 Plus Postage.

DESCRIPTORS *Child Language; *Language Acquisition; Language
Processing; Language Research; Linguistic Theory
IDENTIFIERS *Functionalism; Grammatical Mapping

ABSTRACT
A defense of functionalism in linguistics, and more
specifically the competition model of linguistic performance,
examines six misconceptions about the functionalist approach.
Functionalism is defined as the belief that the forms of natural
languages-are created, governed, constrained, acquired, and used for
communicative functions. Functionalism is viewed as the natural
alternative to language theories that postulate the separation of
structure and function or describe structural facts without reference
to communicative goals or the capabilities of human information
processing. These six misconceptions are addressed, then replaced
with more viable functionalist accounts: (1) grammars reflect the
interaction between cognitive content and cognitive processes; (2)
symbolic and indexical relations exist between form and function; (3)
mappings between form and function are many-to-many; (4) grammatical
mappings are inherently probabilistic; (5) functionalism is
biologically plausible; and (6) functionalist claims are made at
different levels. (MSE)

***********************i***********************************************
Reproductions supplied by EDRS are the best that can be made
from the original document.
************************************************W**********************
 PRCLD 27 (1988)
Panel "PERMISSION TO REPRODUCE THIS
MATERIAL HAS BEEN GRANTED BY

Cl
WHAT IS FUNCTIONALISM?
N-
C) Elizabeth Bates* Brian MacWhinney TO THE EDUCATIONAL RESOURCES
INFORMATION CENTER (ERIC)."
e\J UC San Diego
C) Carnegie-Mellon
teN
CI For the last fifteen years, we have been involved in collaborative
research on language acquisition in children and language processing in
adults, across a range of structurally and functionally distinct language
types (Bates and MacWhinney, 1979, 1982, 1987, in press; MacWhinney,
1987; MacWhinney and Bates, in press). We have brought those findings
together within a framework for the study of linguistic performance called
the Competition Model, a model that is in turn inspired by a broader
approach to the study of language called functionalism, defined as the
belief that "the forms of natural languages are created, governed,
constrained, acquired and used in the service of communicative functions"
(Bates and MacWhinney, 1982). So defined, functionalism is the natural
alternative to theories of language that postulate a severe separate
between structure and function, and/or theories that attempt to describe
and explain structural facts sui generis, without reference to the
constraints on form that are imposed by the goals of communication and
the capabilities and limitations of human information processing.

Although this definition'seems sensible enough as stated, it has

become sadly clear to us over the years that the term "functionalism" does
not communicate very well on its own. It means different things to
different people, and worst of all, there seems to be a Straw Man
Functionalism out there in the hustings that causes trouble wherever we
go. In this short paper, we would like to compare and contrast the
principles of Straw Man Functionalism with an approach that is (we
believe) much more reasonable and much more likely to succeed. The
Straw Man theory can be summarized with the following six beliefs:

(1) Grammar is a direct reflection of meaning. That is, we

can explain all universal and particular aspects of grammar by
uncovering the meanings they convey.

'Portions of this,paper are taken from E. Bates & B. MacWhinney (in press), "Functionalism and the
Competition Model". In B. MacWhinney & E. Bates (Eds.), The cross-linguistic study of sentence
processing. New York: Cambridge University Press.
u.s. DEPARTMENT OF EDUCATION
Once of Educational Research and Improvement
EDUCATIONAL RESOURCES INFORMATION
CENTER (ERIC)
137 t/his document has been reproduced as
received from me person or organization
originating IL
O Minor changes have been made to improve
reproduction quality
2 Poi fltS 01 vieW or opm,onsstatedin th docu-
ment do not necessarily represent official
OERI pos.fion or policy
 138

(2) Grammar is iconic. That is, grammatical devices "look

like" their meanings.

(3) Mappings from meaning to grammar are one to one.

For every meaning there is one and only one expressive
device, and for every device there is one and only one
associated interpretation.

(4) Mappings from meaning to grammar are deterministic.

If the meaning conditions associated with a given grammatical
form are met, the form lel always be used.

(5) Functionalism is anti-nativist. Grammars are a cultural

invention, and biological principles are irrelevant to their
description and explanation.

(6) Functionalism is anti-linguistic. Functionalist theories

of human performance will ultimately replace linguistic
theory altogether. We will bury Chomsky and all the other
generative grammarians with him!

In fact, we do not believe that any of the above six statements are true,
and we have never espoused them ourselves. So let us go through these six
Straw Beliefs one at a time, and replace each one with a more viable
functionalist account.

(1) Grammars reflect the interaction between cognitive

content and cognitive processes. We believe that grammars carry
out important communicative work. Like individual lexical items, specific
grammatical devices (ordering constraints, bound and free morphemes,
suprasegmental cues) are associated with meanings and/or communicative
goals. But the association is rarely direct. We think it more useful to
think of language as a complex, multivectorial problem space. Many
different meanings are competing for expression in a linear (i.e.
time-delimited) channel. The limits imposed by human information
processing (limits of perception, articulation, learning and memory) may

1
0
 139

ultimately prove more important than meaning itself in elucidating why

grammars come to look the way they do.

In the Competition Model, we have borrowed the term cue validity

to refer to the information value of a given lexical or grammatical device
for any particular meaning or function. The term comes from Gestalt
psychology, where it was broadly used to refer to the information
structure of some aspect of the environment for any goal or condition that
is of interest to the organism. In an ideal world, an ideal animal would
behave in perfect accordance with cue validity. But we do not live in an
ideal world, and we are not ideal animals. The relationship between
meaning and form in language cannot be perfect, because of all the
constraints imposed by our information processing system. Our
experiments to date have shown that cue validity strongly determines
the order of acquisiL 1 of cues by children, and the weights that adult
speakers attach to th. same cues during sentence interpretation. Cue
validity also plays a major role in the sentence comprehension and
production profiles displayed by brain-damaged adults suffering from
severe forms of aphasia. However, there are still many systematic
exceptions to this principle. We have been able to account for most of
these exceptions by invoking principles of cue cost, i.e. the information
processing costs associated with the real-time use any given lexical or
grammatical cue. For example, cues that are equally informative can vary
in their perceivability (e.g. Hungarian accusative case suffixes that follow
a strong vowel, compared with the same suffix following a final
consonant). This factor will influence that degree to which adults "trust"
this particular cue to meaning, the age at which children come to rely on
the cue, and the degree of resistance to impairment associated with this
particular cue in sentence-processing by brain damaged adults. Similarly,
cues can vary in the demands they place on memory: "local" cues that can
be used as soon as they are encountered (e.g. a nominative case suffix)
seem to have an advantage over "long distance" cues that require storage
and comparison across a set of discontinuous elements (e.g. subject-verb
agreement), even though the two sets of grammatical devices may both
point strongly toward the same meaning (e.g. the actor role in a transitive
action). A full account of how grammars come to look the way they do,
how and when they are acquired by children, will require an analysis of the
complex interplay between meaning (cue validity) and information
 140

processing (cue cost). Grammars represent a compromise among these

forces, and for this reason, the communicative function of a given
grammatical form may be quite opaque.

(2) Symbolic and indexical relations between form and

function. Linguistic forms rarely, if ever, resemble their meanings.
There are of course a few examples of words that "sound like" the things
they stand for (e.g. Bang!), but these are few and far between. It is even
more difficult to think of grammatical devices that bear a literal physical
resemblance to their meanings. There is of course the apocryphal claim
that natural languages prefer basic words orders in which the subject
precedes the verb because human beings "naturally" tend to perceive actors
before they perceive their actions. This claim is silly enough that it is not
worth pursuing. But if grammars do not "look like" their meanings, then
what kind of natural cause-and-effect relationship could be said to hold
between form and function?

C. S. Peirce (1932) has provided an analysis of sign-referent

relations that may be as useful in-the-study of grammar as it is in the
study of single signs. Icons are signs that come to stand for their
referents because of a literal physical resemblance (e.g. a stylized picture
of a cigarette to indicate a smoking zone). Indices are another class of
"natural" signs that come to stand for their referents not because of a
physical resemblance, but because their participation in the same event
(e.g. contiguity rather than similarity). For example, smoke can serve as
an index to fire because the two are commonly associated in real life.
Symbols are signs that bear no natural relation to their referents (neither
iconic nor indexical); instead, they carry meaning only because of an
arbitrary convention, an agreement that was reached by a particularly
community of users. As Langacker (1987) has pointed out, most lexical and
grammatical signs bear a symbolic relationship to their meanings.
Grammatical devices exist in order to carry out communicative work, but
the work they do does not determine their form. However, in the domain of
grammar there may well be many cases of indexical causality if we keep in
mind that grammars are jointly caused by cognitive content and cognitive
processing.

5
 141

To offer just one example, consider the relative clause. This device
is typically used to identify referents in discourse (e.g. "The man that sold
me the car", as opposed to some other man), a functional motive
constitutes in itself only a form of symbolic determinism. However, the
functions served by a relative clause can also help to determine its shape.
Bindings between a referent and its modifier are easier to make if the two
are in close proximity. Hence the function of referent-identification is
best served if the relative clause is placed near its governing noun phrase,
where other modifiers are located. However, this solution usually poses
another problem: the relative clause must interrupt a main clause. Such
interruption is costly for two reasons. First, because relative clauses are
longer than most modifiers, the main clause has to be held open.for a
rather long time. Second, because relative clauses resemble main clauses
in many respects, there is a potential for confusion (e.g. which verb goes
with which noun). In principle, this problem could be solved by placing a
warning signal at the beginning of a sentence to indicate that "a relative
clause will be placed within the following sentence at some point; you
guess which point". Although this is a logical possibility, it should be
obvious why it would not work very well. It makes much more sense to
place the marker at the point of interruption, to keep the listener from
chasing down some garden path and to help him/her construct and attach
the clause right where it belongs (i.e. near the element that it modifies).
Finally, insofar as an interruption is already placing quite a burden on the
processor, the interruption-marking device had best be kept short and
sweet. Hence the functions of the relative clause have an effect not only
on the existence of certain devices (symbolic determinism), but also on
their position and overall shape (indexical determinism). In neither case
is it reasonable to say that the resulting grammatical device "looks like"
its meaning!

(3) Mappings between for:n and function are many-to-many.

Grammars can be viewed as a class of solutions to the problem of mapping
non-linear meanings onto a highly-constrained linear medium. The
universal and culture-specific contents of cognition interact with
universal constra;nts on information processing, creating a complex
multivectorial problem space with a finite number of solutions. Natural
languages exhaust the set of possible solutions to this mapping problem,
and because these solutions represent many competing forces, they
 142

invariably involve many-to-many mappings between form and function (c.f.

Karmiloff-Smith, 1979), with correlated meanings riding piggy-back on
correlated bits of grammar. No single meaning (however abstract) can be
allowed a grammatical monopoly.

The many-to-many nature of grammatical mapping is both a cause

and a result of the instability inherent in linguistic systems. In fact,
there may be no stable, perfect pathway through the linguistic problem
space. As Slobin (1982) has pointed out, many processing constraints
stand in direct competition; hence stability in one area may create
instability in another. From the listener's point of view, a given linguistic
marker will signal its meaning most efficiently if it is consistent, salient
and unique. But from the speaker's point of view, the same linguistic
device has to be easy to retrieve and produce. Hence the clear and
perceivable markers that evolve for comprehension are often subject to
erosion in the service of rapid and efficient speech output. Faced with
these competing demands, languages have been known to cycle back and
forth across the course of history, from one set of solutions to another
Hence we must view grammars as a set of partial solutions to the mapping
problem, each representing one pathway through the constraints imposed
by cognitive content and cognitive processing. No solution is perfect, and
each one is constantly subject to change; but every grammar used by a
community of human adults and acquired by their children has to meet
certain some implicit but implacable limits of tolerance.

(4) Grammatical mappings are inherently probabilistic.

Languages differ qualitatively, in the presence or absence of certain
linguistic devices (e.g. word order constraints, case-marking), but they
also differ quantitatively, in the extent to which the "same" linguistic
device is used at all and in the range of functional roles that the "same"
linguistic device has come to serve.

We have given a number of examples of quantitative differences

between languages throughout our work (see especially papers in
MacWhinney and Bates, in press). One particularly important example has
to do with the relative strength of word order versus subject-verb
agreement as cues to sentence meaning. In English, word order is rigidly
preserved; in almost all structures (we will consider a few exceptions

7
 143

later), the order that is preserved is Subject-Verb-Object or SVO. In

Italian, word order can be varied extensively for pragmatic purposes -- a
fact that comes -s something of a surprise to those who believe that such
pragmatic word order variation occurs only in case-inflected languages
(i.e. languages with markers on the noun to indicate "who did what to
whom"). The following list (from Bates and MacWhinney, in press)
illustrates some possible variations in the order of major constituents in
Italian, in a hypothetical restaurant conversation. This short conversation
(a fake, but quite plausible according to our Italian informants) contains
all possible orders of Subject, Verb and Object.

1. SVO: lo mangerei un primo. (I would eat a first course).

2. OSV: La pastasciutta Franco la prende sempre qui.

(Pasta Franco it orders always here).

3. VSO: Al lora, mangio anche io la pastasciutta. (Well

then, am eating also I pasta).

4. VOS: Ha consigliato la lasagna qui Franco, no? (Has

recommended the lasagna here Franco, no?).

5. OVS: No, la lasagna l'ha consigliata Elizabeth. (No, the

lasagna it has recommended Elizabeth).

6. SOV: Al lora, io gli spaghetti prendo. (In that case, I the

spaghetti am having).

Some of these require particular intonation patterns to sound exactly

right, and some are definitely better with particular grammatical markers
like the object clitic. But all these orders can be found in a large enough
sample of free speech, and all of them occur at some point in the input
received by Italian children (Bates, 1976).

At one level, this discourse serves merely to illustrate a

well-known qualitative difference between languages: Italian has word
order options that do not exist in English at all. However, this qualitative
variation also has quantitative implications. We have now demonstrated in

8
 144

several different experiments that Italian listeners "trust" word order

even good old-fashioned Subject-Verb-Object order -- less than their
English counterparts. Given a sentence like "The pencil hits the cow",
English listeners from ages 2 to 80 have a strong tendency to pick the
pencil as the agent/subject. Given the Italian equivalent ("La matita
colpisce la vacca"), Italians are much more likely to choose the cow as the
agent/subject. Hence a qualitative difference in the availability of word
order types has a quantitative effect even on that subset of grammatical
structures that both languages share (e.g. SVO order).

Most of our joint research to date has concontrated on sentence

comprehension. But we have also uncovered some interesting quantitative
differences in the domain of sentence production. For example, Bates and
Devescovi (in press) have described some robust differences between
Italian and English in the use of relative clauses. The structural options
available in the two languages are the same, at least for the set of
structures studied by these investigators. In both languages, it is
perfectly grammatical to describe a picture of a monkey eating a banana by
saying either "A monkey is eating a banana" or "There is a monkey that is
eating a banana". However, English speakers typically use the first option;
by contrast, Italian speakers describing exactly the same pictures, under
the same conditions, are three to five times more likely to produce a
relative clause. This cross-linguistic difference in relative clause use is
already well-established in children by the age of three, and it tends to
persist even in elderly patients who have suffered left-hemisphere
damage. How can we capture a quantitative difference between two
structures that are equally grammatical from a traditional grammatical
perspective? To be sure, there are some differences between the two
languages in the range of functions that control these particular forms. In
particular, Italians appear to use the relative clause as a kind of topic
marker. But in addition to (and perhaps because of) these differences in
function, there are also clear processing differences between English and
Italian in the "accessibility" of the relative clause. We have uncovered
similar statistical differences between Italian and English children in
rates of article omission (greater in English children well before the age
of 3), and in rates of subject omission (with much higher rates of subject
omission in Italian children even in the stage of first word combinations
-- Bates, 1976).. Some of these differences (e.g. subject omission) are

9
 145

treated in current linguistic theory in terms of a discrete set of rules or

parameters; others (e.g. article omission) receive no treatment in current
linguistic theory at all. We think that these early differences in
performance can only be captured by assuming that very small children are
sensitive to statistical as well as structural facts about the language they
are trying to acquire. Function and frequency co-determine the selection
of grammatical forms in sentence production, in language use by adults and
in language acquisition by children.

Physicists have made their peace with the counter-intuitive

predictions of quantum mechanics, and they now accept the premise that
the position of a subatomic particle may be unknowable in the absolute.
Uncertainty lies at the core of the universe; it is not just a byproduct of
our imperfect measures. We argue that the human language processor is
also probabilistic at its core. In the Competition Model, the adult
speaker's knowledge of his native language is represented in a
probabilistic form, and probabilities play a fundamental role in the process
of language acquisition. The difference between obligatory rules and
statistical tendencies is simply a matter of degree. This does not mean
that we ignore the powerful laws that separate one language from another.
After all, the values "0" and "1" do exist even in a probabilistic system, and
an adult native speaker may thus come to know with some certainty that a
particular structure is impossible in his or her language. The difference
between our characterization of adult knowledge (i.e. "competence to
perform") and the characterizations offered in most competence models
lies in our ability to capture the many values that fall between 0 and 1.
We describe linguistic representations in terms of a complex set of
weighted form-function mappings, a dynamic knowledge base that is
constantly subject to change.

. In a sense, language acquisition can thus be viewed as a process of

meaning driven distributional analysis, similar in spirit to the approach
outlined some time ago by Maratsos (1982). However, the Competition
Model also furnishes some non-linear principles that permit us to capture
sudden phase transitions, U-shaped functions, and the effect of rare events
-- all the phenomena that forced psychologists to abandon the
simple linear associative models of American Behaviorism. Many of these
discoveries within our model have fallen out of two approaches to the

10
 146

quantification and formalization of language learning: (a) mathematical

modelling of the effects of cues on choice behavior in sentence
comprehension (McDonald, 1986; McDonald and MacWhinney, in
press), and (b) computer simulations of the learning process (Taraban,
McDonald and MacWhinney, in press). For example, we have discovered that
cue validity can be operationalized in two ways: overall cue validity (the
proportion of all the cases in which an interpretation must be made in
which a given cue is available and leads to a correct interpretation), and
conflict validity (the proportion of cases in which one cue competes with
another in which the cue in question "wins"). Both these metrics can be
calculated objectively from texts of real speech, and used to predict the
choice behavior of children and adults in sentence comprehension
experiments. Interestingly, we have discovered that overall cue validity
drives the early stages of language acquisition; conflict validity (affected
primarily by rare cases, particularly those that are encountered in complex
discourse) drives the late stages of learning in older children and adults.
With these two statistical principles, we can capture abrupt changes in
sentence processing strategies that occur as late as 7 -10 years of age.

Although the Competition Model has been developed on independent

grounds (to deal with facts of acquisition and processing across different
natural languages), the model in its current form has a great deal in
common with a recent movement that is alternatively referred to as
connectionism, neural modelling and/or parallel distributed processing
(e.g. Rumelhart, McClelland and the PDP Research Group, 1986; Elman,
1988). It remains to be seen how strong that relationship will be, but we
are at least convinced that the tools we share will prove to be
exceptionally important in the next era of language acquisition research.
Cognitive psychology has proceeded for more than thirty years without an
adequate model of learning. Unfortunately, research in language
acquisition has done the same. The new focus on learning in "brain-like
systems" is a healthy one, whatever its limits may prove to be. And the
new tools (i.e. mathematical modelling, multivariate statistics, computer
simulation) are bound to lead to progress. Natural languages are so
complex that "eyeball analysis" alone can only take us so far -- probably no
farther than we have come to date.

11
 147

(5) Functionalism is biologically plausible. The innateness

issue is one of the major sources of anger and misunderstanding in the
field of psycholinguistics. We think that much of this misunderstanding
comes from a failure to distinguish between innateness and
domain-specificity. The innateness issue has to do with the extent to
which human language is determined by the unique biological heritage of
our species. But this biological heritage may include many capacities that
are not unique to language itself: our large and facile brain, our particular
social organization, our protracted infancy, and a variety of unknown
factors that may contribute in indirect but very important ways to the
problem of mapping universal meanings onto a limited channel, and to the
particular solutions that we have found to that problem. Hence the human
capacity for language could be both innate and species-specific, and yet
involve no mechanisms that evolved specifically and uniquely for language
itself. Language could be a new machine constructed entirely out of old
parts (Bates, 1979). The universal properties of gramml may be
indirectly innate, based on interactions among innate categories and
processes that are not specific to language. In other words, we believe in
the innateness of language, but we are skeptical about the degree of
domain-specificity that is required to account for the structure and
acquisition of natural languages.

(6) Functionalist claims are made at different levels.

Functionalist theories of performance are not in direct competition with
any linguistic theory. Different kinds of functionalist claims require
different kinds of evidence. This is a point that we have tried to make in
several places (notably Bates and MacWhinney, 1982; Bates and
MacWhinney, 1987 and in press), but it is sufficiently important that we
think it deserves reiterating here. We distinguish four different levels of
functionalist claims, ordered from weakest to strongest (in the sense that
claims at the higher levels presuppose that claims at the lower levels are
true).

Level 1 focusses on the role of cognitive and communicative functions

in the evolution of language proper, and the h:,tory of individual languages.
Claims at Level 1 constitute a kind of linguistic Darwinism, i.e. arguments
that functional constraints have played a role in determining the forms
that grammars take today. Where did the tiger get his stripes? Why do
 148

grammars have relative clause markers? A great deal of work in

functionalist linguistics is of this historical sort, in particular studies of
"grammaticization" (e.g. Givon, 1979; Bybee, 1985). Although this work is
extremely interesting in its own right, claims at the historical level have
no necessary implications for current language use by adults, language
acquisition by children, or the proper characterization of grammatical
knowledge. Like the large-scale forces that operate to
create mountains and rivers across geological time, the forces that
operate across many individuals to bring about historical language change
may not be detectable (or even operative) in every individual case.

Level 2 is a synchronic variant of Level 1, focussing on the causal

relationship between form and function in real-time language use by adult
speakers of the language. Much of our own work with adults is of this
sort: we manipulate competing and converging sets of grammatical forms
as "causes" to see what interpretations our subjects derive; conversely,
we manipulate competing and converging meanings in picture and film
description, to see what expressive devices our subjects produce to meet
these demands. However, even if we could show a perfect
cause-and-effect relation in adults, we could not immediately conclude
that children are able to perceive or exploit these relations.

Level 3 presupposes but goes beyond Level 2, focussing on

the causal role of cognitive and communicative functions in language
acquisition by children. The cause-and-effect work of Level 2 must be
repeated at every stage of language acquisition, to determine empirically
if and when children are sensitive to the form-function correlations
available in the adult model. Furthermore (as we noted earlier), we need a
well-articulated theory of the learning process, one that can adequately
describe, predict and explain the stages that children go through on their
way to adult performance.

Finally, Level 4 is reserved for the claim that facts from Levels 1 3
play a direct role in the characterization of adult linguistic competence. A
variety of competence models of this sort have been proposed within the
functionalist tradition, ranging from Eastern European functionalism (i.e.
theso-called Prague School Derso, 1972; Driven and Fried, 1987; Firbas,
 149

1964; Firth, 1951), British functionalism (e.g. Halliday, 1966), the

American school of generative semantics (e.g. Fillmore, 1968; Chafe,
1971), to more recent proposals that include cognitive grammar
(Langacker, 1987; Lakoff, 1987), construction grammar (Fillmore, 1987),
role and reference grammar (Foley and Van Valin, 1984), and several
other approaches that either retain the simple term "functionalism" or
elect to avoid labels altogether (e.g. Dik, 1980; Kuno, 1986; Givon, 1979.
For the sake of simplicity, we will refer to these otherwise rather
disparate linguistic theories with the single term functional grammar.
Although functional grammars are not designed to account for real time
processing, they are most compatible with highly interactive models of
performance, i.e. with models like ours. For obvious reasons, "modular"
theories of performance are instead more compatible with "modular"
theories of competence, that is, w:in linguistic theories that emphasize
the autonomy of various components and subcomponents of the grammar
(c.f. Berwick and Weinberg, 1984; Bresnan, 1982; Pinker, 1984). It is quite
possible that there will ultimately be a convergence between
some Level 4 version of functional grammar, and the performance model
that we have developed to account for data at Levels 1 to 3. But it is also
possible, at least in principle, that there may be a rapprochement between
a functionalist model of performance and the various rules and
representations that have been proposed within the
many-times-revised-and-extended school of generative grammar.

In short, we are not anti-linguistic, nor is our work directly

relevant to any particular rlass of competence models. We are consumers
of linguistic theory, and we have our own bets about which linguistic
theory or class of theories will ultimately prevail. But we are much too
preoccupied with problems of a different sort to enter into the linguistic
fray. This is an exciting new era in languag acquisition research, and
time is too precious to be wasted on battles that are best waged
elsewhere.

14
 150

References

Bates, E. (1976). Language and context. New York: Academic Press.

Bates, E. (1979). The emergence of symbols. New York: Academic Press.

Bates, E. & Devescovi, A. (in press). A functionalist approach to sentence

production. In B. MacWhinney & E. Bates (Eds.), The cross-linguistic
study of sentence processing. New York: Cambridge University Press.

Bates, E. & MacWhinney, B. (1979). A functionalist approach to the

acquisition of grammar. In E. Ochs and B. Schieffelin (Eds.),
Developmental pragmatics. New York: Academic Press.

Bates, E. & MacWhinney, B. (1982). A functionalist approach to grammar.

In E. Wanner & L. Gleitman (Eds.), Language acquisition: the state of the
art. New York: Cambridge University Press.

Bates, E. & MacWhinney, B. (1987). Competition, variation and language

learning. In B. MacWhinney (Ed.), Mechanisms of language acquisition.
Hillsdale, N.J.: Lawrence Erlbaum.

Bates, E. & MacWhinney, B. (in press). Functionalism and the Competition

Model. In B. MacWhinney & E. Bates (Eds.), The cross-linguistic study of
sentence processing. New York: Cambridge University Press.

Berwick, R. & Weinberg, A. (1984). The grammatical basis of linguistic

performance. Cambridge, Mass.: MIT Press.

Bresnan, J., (Ed.). (1982). The mental representation of grammatical

relations. Cambridge, Mass.: MIT Press.

Bybee, J. (1985). Morphology: a study of the relation between meaning and

form. Amsterdam: John Benjamins.

Chafe, W. (1971). Meaning and the structure of language. Chicago:

University of Chicago Press.

15
 151

Derso, L. (1972). Bevetes e mondattani tipologiaba. (Introduction to

linguistic typology). Budapest: TIT Kozpontja.

Dik, S. (1980). Studies in functional grammar. New York: Academic Press.

Driven, R. & Fried, V. (1987). Functionalism in linguistics. Amsterdam: J.

Benjamins, 1987.

Elman, J. (1988). Finding structure in time. CRL Technical Reports. La

Jolla: Center for Research on Language, University of California, San
Diego.

Fillmore, C. (1968). The case for case. In E. Bach & R. Harms (Eds.),
Universals in linguistic theory. New York: Holt, Rinehart & Winston.

Firbas, J. (1964). On defining the theme in functional sentence. Travaux

linguistique de Prague, 1:267 - 280.

Firth, J. (1951). Modes of meaning. London: The English Association.

Foley, W. & Van Valin, R. (1984) Functional syntax and universal grammar.
New York: Cambridge University Press.

Givon, T. (1979) On understanding grammar. New York: Academic Press.

Halliday, M. (1966). Notes on transitivity and theme in English: Part 1.

Journal of Linguistics. 2:37 71.

Karmiloff- Smith, A. (1979). A functional approach to child language. New

York: Cambridge University Press.

Kuno, S. (1986). Functional syntax. Chicago: University of Chicago Press.

Lakoff, G. (1987). Women, fire and dangerous things. Chicago: University

of Chicago Press.

Langacker, R. (1987). Foundations of cognitive grammar. Stanford, Ca.:

Stanford University Press.
 152

MacWhinney, B. (1987). The competition model. In B. MacWhinney (Ed.),

Mechanisms of language acquisition. Hillsdale, N.J.: Lawrence Erlbaum
Associates.

MacWhinney, B. & Bates, E. (Eds.), The cross-linguistic study of sentence

processing. New York: Cambridge University Press.

Maratsos, M. (1982). The child's construction of linguistic categories. In E.

Wanner & L. Gleitman (Eds.), Language acquisition: the state of the art.
New York: Cambridge. University Press.

McDonald, J. (1986). The development of sentence comprehension

strategies in English and Dutch. Journal of Experimental Child
Psychology, 41: 317 - 335.

McDonald, J. & MacWhinney, B. (in press). Maximum likelihood models for

sentence processing. in B. MacWhinney & E. Bates (Eds.), The
cross-linguistic study of sentence processing. New York: Cambridge
University Press.

Peirce, C. S. (1932). Language and context: Collected papers. Cambridge,

Mass.: Harvard University Press.

Pinker, S. (1984). Language learnability and language development.

Cambridge, Mass.: Harvard University Press.

Rumelhart, D., McClelland, J. & the PDP Research Group. (1986). Parallel
distributed processing: explorations in the microstructure of cognition.
Cambridge, Mass.: MIT/Bradford Books.

Slobin, D. I. (1982). Universal and particular in the acquisition of language.

In E. Wanner & L. Gleitman (Eds.), Language acquisition: the state of the
art. New York: Cambridge University Press.

Taraban, R., McDonald, J. & MacWhinney, B. (in press). Category learning in

a connectionist model: learning to decline the German definite article.
In Sixteenth annual UWM symposium on linguistic categorization.

17