"Is there a Device for
The aim of this paper is to question
some of the assumptions underlying recent psycholinguistic thought about
first-language acquisition, as exemplified in David McNeill's paper
presented at a conference in the United States in 1966 on "Language development
in children". We also intend to examine some other ideas on language
structure with a view to presenting some alternative proposals to those
of McNeill. We shall begin with a brief summary of McNeill's position.
[For footnotes please click on corresponding underlined links]
Summary of McNeill's paper
Acknowledging the now generally
held view that children's language reflects the development of their own
grammatical system rather than a poor imitation of adult grammar (cf. also
Smith 1971: 51-55), McNeill categorises the components of infants' first
two-word utterances as members of a pivot and open class respectively,
and cites sample utterance-components recorded in three studies (Braine,
Brown, Ervin) to illustrate his proposed classification. According
to another study (Brown & Bellugi) cited by McNeill, the 'pivot' class
(which, unlike the 'open' class, could not alone comprise an utterance)
is subsequently differentiated -- first to distinguish articles and demonstratives,
then adjectives and possessive pronouns.
In the absence of any other satisfactory
explanation, McNeill takes the pivot/open class distinction as the child's
attempt at a "generic classification of adult grammatical categories" (1966:
31), and relates it to the notion of 'semi-grammaticality' described by
Chomsky (see McNeill 1966: 32-35) in terms of category hierarchy.
Proceeding to the analysis of three-word utterances, he formulates two
hierarchical rules, which, while they "do not generate well-formed sentences
according to adult grammar" (1966: 44), at least generate the major constituents
of such sentences ('NP' and 'PredP').
McNeill rejects the notion that
these rules, categories and hierarchy could be derived through imitation
of or inference from adult speech, or even on the basis of distributional
evidence in adult speech (in which the hierarchy is completely unmarked),
and feels obliged to work from the hypothesis that the 'basic grammatical
relations' manifest in the child's initial structures "are part of the
innate linguistic capacity" (1966: 45).2
He further represents them as "substantive universals", part of the 'Language
Acquisition Device' proposed by Chomsky and Katz as the mechanism innate
in every human being for developing linguistic competence (see, for example,
Chomsky 1965: 52-58).
McNeill: discussion & criticism
It is worth noting that these 'basic
grammatical relations', allocated by Chomskyan transformational theory
to the level of 'deep structure' (or semantic interpretation -- cf. Chomsky
1965: 136), are still derived from surface structure and continue
to owe their existence as relations to the external structure of human
language. They are linguistic, and not psychological concepts.
(What could be the meaning of main verb, for example, outside the
context of actual language?) Yet the implication of McNeill's theory
is that children begin constructing utterances from these innate 'basic
grammatical relations', with meaning somehow being plugged into
them along the way.
Frank Smith's recently published
study on the reading process (1971) includes some useful insights on the
nature of language development in children. The following paragaph
gives an interesting contrast to the order of learning implied by McNeill
(Smith 1971: 52): "Contrary to popular ... belief, the child is not learning
words and then finding meanings for them. Instead he is acquiring
or inventing words ... to meet his own particular requirements and represent
meanings which he needs to express."3
Would not this relative order of learning extend to all aspects of surface
structure, including the 'basic grammatical relations'?
Charles Fillmore apparently
thinks so. In his discussion and rejection of "earlier approaches
to the study of [grammatical] case" he criticises Redden's analysis of
Walapai on the grounds that the meaning or 'functions' of cases "are not
taken as primary terms in the description" but merely fitted into the framework
of already identified surface-structure forms (Fillmore 1968: 9).
He further finds "reasons for questioning the deep-structure validity of
the traditional division between subject and predicate, a division which
is assumed by some to underlie the basic form of all sentences in all languages"
Another attack on the use of surface
structure as a starting-point for probing the language-thought relationship
has come from psychologist Norman Segalowitz, who argues as follows
The error which I think linguists have
made is this. They have focussed on the wrong property of language
to base their theory on. So far, the main concern of linguistics
has been to show how sentences are related to each other through their
internal structures. The structures which emerged in these theories
are always subordinated to the task of relating one type of surface structure
to another. The psychologist on the other hand is interested not
so much how surface structures are related ... but in the very nature of
the internal structure itself.
It is our view that any serious
consideration of the acquisition of language in the young child, inasmuch
as it parallels the development of the child's thinking processes, cannot
afford to neglect the nature of this 'internal structure' -- the thought
component in the language-thought relationship. And yet the answer
cannot lie entirely in the realm of psychology either, for the language
component is equally important. What is needed is a theory of linguistic
development that will take fuller account of the link between the two.
Some other ideas on language
We believe that the basis of such
a theory is to be found in Fillmore's own approach to transformational
grammar, as set forth in his 1968 article "The case for case". While
it is true that his study deals largely with what has been termed "case
grammar" and is applied primarily to adult language, the fact that his
whole treatment of grammatical theory (and
is rooted in his concept of a fundamental relationship between language
and thought -- an "underlying syntactic-semantic relationship" as he puts
it (21) -- makes his model a most interesting source of enlightenment on
the subject of child-language acquisition.
On the language side, Fillmore
sees basic sentence structure as consisting "of a verb and one or more
noun phrases, each associated with the verb in a particular case relationship";
the sum of these constitutes the proposition of the sentence as
distinct from the modality (features affecting the sentence as a
whole -- negation, tense etc.). These 'case categories', however,
in contrast to the Walapai study mentioned above, are not treated as the
primary units of analysis but are related to a 'conceptual framework' of
relationship categories which are presumably non-linguistic in origin and
form part of one's innate cognitive structure.6
As examples of these categories he lists Agentive, Instrumental, Dative,
Factitive, Locative and Objective -- the latter "not to be confused
with the notion of direct object, nor with the name of the surface case
synonymous with accusative" (1968: 25).
Fillmore then proceeds to classify
sentences according to which of these conceptual categories they express,
illustrating his classification with a 'case frame': e.g., the verb run
could take the frame [___A], give [___ O + D + A]
etc. (initial letters refer to the case categories given above).
Another point of departure from Chomskyan theory is that adjectives are
treated as verbs, e.g., sad [___ D].
While all case categories (including
the Agentive) are given following the verb, Fillmore has stated that no
linear ordering of deep structure is implied. In surface structure,
however, one of the categories is 'subjectivalised' and moved to the beginning
of the sentence. For example, the sentence John opened the door
as its underlying proposition the structure V + O + A, with
becoming the subject in surface structure. The proposition underlying
door opened, on the other hand, is V + O, and O becomes
This supports Fillmore's basic assumption that the notion of 'subject'
and other components of syntactic relations are aspects of surface structure,
not deep structure (see p. 17).
Fillmore: discussion & criticism
It is interesting to note that Segalowitz
also sees in Fillmore's grammar the possibility of a closer union between
linguistics and psychology in regard to the understanding of the language-thought
relationship. His criticism of Fillmore's approach is that, radical
though it is in using "explicitly psychological aspects of language", it
doesn't go far enough -- it is "still subordinated to the task of relating
surface structures to each other (Segalowitz 1970: 15).
Indeed, Fillmore devotes comparatively
little attention to the non-linguistic analysis of his case categories,
concentrating rather on their relationship to surface structure.
Even the definitions of these categories are all predicated on something
called the verb, which for some reason is left undefined in terms
of cognitive structure. The definitions, moreover, leave something
to be desired in terms of their relationship to the overall cognitive structure.
One category that appears to be missing entirely is that expressed in the
initial component of 'copula' sentences (John is a teacher, That is
my brother) -- a category that might possibly be termed, for want of
a better expression, the Referential. Another discrepancy
is that determiners are introduced into the deep-structure phrase markers
without any statement as to their origin or semantic correspondence.
All of which means in effect that Fillmore's theory as stated is
not fully satisfactory as a model for language acquisition.
Segalowitz (ibid.) focuses
on the concept of action-object relationship inherent in Fillmore's
grammar as a "specific property of language" that could be significant
to the relation of language to thought. In the matter of action-object
relationships, an interesting comparison may be found in a discussion of
primordial development of thought and language in the human species by
the Russian psychologist-anthropologist A. A. Leont'ev (1963).
In Leont'ev's view, the very first step in the development of thought beyond
the level of the pre-human ape was the "separation of action from object"
[otdelenie dejstvija ot ob"ekta] (1963: 50). We might also
speak of the separation of the object from the situation.
Another major step in this development described by Leont'ev was the subsequent
separation of a generalised concept of an object from its individual manifestation
(1963: 119-21). We believe that these two primary mental processes
Leont'ev proposes may contribute toward an alternative theory to McNeill's
regarding language acquisition in the young child.
Toward an alternative theory
of language acquisition
Let us look for a moment at McNeill's
analysis of the two-word utterances which he divides into pivot-
and open-class components (1966: 22). Braine's study is especially
interesting, since his list, unlike the other two, reportedly represents
the very first two-word combinations ever produced by his subjects.
We may first of all note that the
second component of each utterance (ascribed by Braine and McNeill to the
'open' class) is a noun denoting some object,8
while first-component 'pivot'-class words, none of them nouns, comprise
adjectives, verbs and other expressions relating to the following noun:
pretty, more, my, see, allgone, byebye, nightnight, hi. Remembering
Fillmore's treatment of adjectival expressions as verbs, we might classify
these initial constructions as V + O -- i.e., a verb followed by
a noun in the Objective case category.
The only difficulty in this manner
of approach is that we are still talking about language-bound concepts.
Fillmore's Objective category, it will be recalled, was predicated on the
'verb', and just what this 'verb' might mean to an adult, let alone a child,
is as yet unexplained. We still need something to tie in Fillmore's
categories more closely with the child's basic cognitive structure.
Let us look at Braine's list in
a different way. We have said that all the second-component words
denote objects, whereas none of the first-component words do. The
latter, however, all may be considered to have a strong semantic correspondence
with situations relating to the objects denoted by the second component
-- situations which, like the objects, form part of the child's non-linguistic
cognitive experience -- e.g., disappearance of an object (allgone, byebye),
its physical features (pretty, big), a desire for an object (more)
etc. Seen in this light, do not these first combinations point to
what might be a prior or simultaneous first step in the development of
the child's thinking processes, a first step not unlike the one proposed
by Leont'ev for the original evolution of human thought? Do not utterances
like allgone shoe, big boat, more milk etc. express the first mental
separation of an object from the situation in which the object appears?9
Therefore, instead of introducing language-bound, psychologically meaningless
classifications such as 'pivot' and 'open' classes -- or even psychologically
vague ones such as 'verb' and an 'Objective' predicated on the verb, could
we not rather formulate the child's first grammatical rule in terms of
the cognitive dichotomy of object and situation, possibly:
> (Sit) + Obj ?10
Looking at the list compiled from
Brown's study (McNeill 1966: 22), however, we find a couple of items which
do not appear to have any semantic correspondence at all, let alone a correspondence
to a 'situation' -- namely, the articles a and the.
How are these to be accounted for in terms of our postulated rule?
The answer appears to lie in the
fact that the children in Brown's study were somewhat older (in terms of
linguistic development) than those examined by Braine; it is reported (1966:
20), in fact, that they were already producing three- and four-word utterances
even before the study began. If Brown & Fraser's study (1964:
63) may serve as an analogy, these children were probably producing utterances
such as See a boot, There the man, in which the above-formulated
rule could very well have given see, there, as situations
and a boot, the man, as objects. The latter expression
could have been obtained alone simply through the non-inclusion of the
optional situation component (as provided for in the rule); they almost
certainly represent a very different grammatical structure. Such
is the danger of treating all two-word utterances alike at a time when
their utterers have progressed beyond the two-word stage and must of necessity
have expanded their hierarchy of rules.
How then might their hierarchy have
been expanded to account for constructions with a or the
as first component? They appear to be very similar to the determiner-plus-noun
constructions of adult grammar, which Fillmore introduces as a sub-classification
of his case categories, but even he is unable to provide any psychological
correspondence for them; they simply appear in his deep-structure phrase-markers
Again we turn to Leont'ev's analysis
of thought-development in the human species. It will be recalled
that a second major development was cited -- that of separation of a generalised
concept of an object from its individual manifestation. Now, could
not the articles, both definite and indefinite, be taken to mark (at least
at this stage of the child's linguistic development) recognition of an
individual manifestation of a particular object in contrast to the generalised
concept of it? This distinction might give rise to the following
rule: Obj > (Ind) + Obj -- where Ind would signify individualisation).
The fact that Ervin's list includes
adjectives and verbs as second components of two-word utterances suggests
(a) that her subjects were even more linguistically advanced than those
studied by Brown, and (b) that more information about their longer utterances
is needed before an attempt is made at analysis on the basis of the rules
The two grammatical rules we have
now formulated on the basis of presumed mental processes have, in our view,
definite implications for the subsequent development of language acquisition
all the way into adult speech. We can already foresee specific expansions
of the rule hierarchy to account for the three-word utterances analysed
by McNeill (1966: 40ff), as well as for those studied by Brown & Fraser
(1964: 62ff). We believe we are also capable at this point of explaining
the initial negations and questions dicussed by McNeill (54ff).
As to a model for still later linguistic
development, we are not absolutely convinced as to the greater validity
of Fillmore's grammar over Chomsky's, although at this point it has the
advantage of being an attempted closer approximation to the non-linguistic
cognitive structure in which human language
We certainly do not see the involvement, however, of anything like a 'Language
Acquisition Device' containing innate 'basic grammatical relations'.
Rather, we see the above-formulated rules as pointing the way toward a
more mentally-oriented theory of grammatical structure which might supply
the missing psychological relationship for either Chomsky's or Fillmore's
grammar, or suggest the development of an entirely new grammatical system.
For example, we might explore the
relationship of the concept of situation suggested by Leont'ev's
study to Fillmore's 'verb' (or even some of his case categories), or alternatively
to Chomsky's notion of 'predicate'. Does Leont'ev's 'object' develop
into Fillmore's 'Objective' or any of his other categories, or is it the
forerunner of Chomsky's 'NP'? Could it be that the process of individualisation
mentioned above underlies the whole concept of determiner in linguistic
structure? The exact nature of these relationships is beyond the
scope of the present paper but is definitely recommended for future explanation.
The major thrust of our present
paper, as we see it, has been to examine what we might call the whole 'surface-structure
approach' to language acquisition, as seen in the McNeill presentation
discussed above. Specifically, we have questioned these two basic
assumptions: (1) the validity of the 'pivot-open' classification of children's
first two-word utterances; (2) the existence of a 'Language Acquisition
Device' separate from the rest of one's cognitive structure and the correlative
hypothesis that 'basic grammatical relations' are to be treated as innate
Two points have been made in regard
to the first assumption: (a) the classification has been erroneously
applied to utterances of children who have already progressed beyond two-word
utterances and beyond a single grammatical rule; (b) the classification
is meaningless as a description of the real 'deep structure' -- the mental
processes -- underlying these initial two-word utterances. Instead,
an alternative formulation has been proposed which attempts to take account
of this internal structure.
In regard to the second assumption,
it has been shown that no separate device is necessary. Language
acquisiton can be handled by the same processes that govern the non-linguistic
aspects of cognitive structure, and the same rules can, at least in the
initial stages of acquisition, account for the development of both thought
© John Woodsworth, 1972.