
Symbols and structures in language acquisition

Carolyn Johnson, Henry Davis, and Marlys Macken

Abstract

Children learning natural languages need to master a symbol
system with both a constituent and a combinatorial structure.
There are at present three main schools of theorizing as to
how this mastery is accomplished: the 'interactive', which
locates language-learning within its context of use; the
'cognitive', which locates language-learning as part of
symbolic development in general; and the 'autonomous', which
emphasizes the independence of the study of grammatical
competence from both language use and general principles of
cognitive development. Additionally, the task of
language-learning is generally investigated under a number of
independent headings: phonological acquisition; lexical
acquisition; morphological and syntactic acquisition; and
pragmatic acquisition.
The phonological (sound) system of a language is rule-based.
Very young human infants can discriminate the majority of the
features which comprise the adult sound system, and can do so
across complex physical dimensions. At a minimum, infants'
abilities provide an initial grid for segmenting, sorting, and
classifying phonetic categories, and for mapping these to the
higher levels of phonological systems. For production, there
is continuity from the babbling period into early language.
The early productive sound systems appear to be based on words
rather than smaller units, words not being initially analysed
into their component phonemes. This analysis begins between
the ages of two and four years, although the child's system
still remains simplified compared to that of the adult. The
process of phonological development follows no invariant
sequence, and can show regressions at the same time as the
system becomes more complex. Some form of 'cognitive' theory
at present provides the best explanatory framework
Verbal communication is preceded by gestural communication,
beginning at about nine months of age. Accounts have been
offered in which gestural communication is claimed to be
necessary for later language development, and continuous with
it; but the evidence on both these points is equivocal.
Meaningful 'words' are used from about one year of age. There
is a spurt in vocabulary development from around eighteen
months. Errors in word use provide the major data source for
theories of semantic development. These data have formed the
basis of a number of theoretical accounts; but none of these
are yet comprehensive in their explanatory powers. Crucially
lacking is an account of how the child learns to relate words
to concepts, and clear criteria for determining when a child's
'words' are truly symbolic.

The acquisition of syntax can be divided into three stages:
presyntactic; syntactic; and postsyntactic. Early
presyntactic word-combinations are of three types: the
combination of words that reflect grammatically relevant
real-world relations; a word plus an intonationally integrated
but meaningless extra syllable; and formulaic or rote-learned
sequences. It is not clear whether there is any continuity
between the combinatorial structures of this stage and those
of the next, syntactic, stage, which lasts from about two to
five years of age. First-language learners make very few
errors in constructing complex grammars - a fact supporting
the claim that there are innate constraints, whether
specifically linguistic or not, on a learner's 'hypothesis
space'. Analysis of errors in learning suggests that
morphological learning may occur via a probabilistic
mechanism, whereas syntactic learning may be based on
different, possibly innate, principles. The final
postsyntactic stage represents the integration of the
newly-emerged syntactic system with the real-world knowledge
the child has accumulated.
Early functions of language are of three types: solicitation
of action, social interaction, and joint attention; expression
of affect; and participation in games. With the advent of
naming, children can also label and request labels. Later
developments, such as threatening, promising, and deceiving,
have not been systematically studied.
The best-supported theories of language development are of the
'autonomous' school. It appears that language-learning skills
are domain-specific, and there are few parallels with
non-linguistic domains, or precursors to the formal systems.
Thus, the autonomous view presents difficulties for both onto-
and phylogenic accounts, in that it is a discontinuity view.
However, the 'interactive' and 'cognitive' schools at present
offer less plausible accounts of the acquisition of linguistic
abilities than 'autonomous' ones, even though they offer more
continuous views of development, and hence might be more
attractive to an evolutionary scenario at first sight
[Eds].