Chunking mechanisms in human learning.

Pioneering work in the 1940s and 1950s suggested that the concept of 'chunking' might be important in many processes of perception, learning and cognition in humans and animals. We summarize here the major sources of evidence for chunking mechanisms, and consider how such mechanisms have been implemented in computational models of the learning process. We distinguish two forms of chunking: the first deliberate, under strategic control, and goal-oriented; the second automatic, continuous, and linked to perceptual processes. Recent work with discrimination-network computational models of long- and short-term memory (EPAM/CHREST) has produced a diverse range of applications of perceptual chunking. We focus on recent successes in verbal learning, expert memory, language acquisition and learning multiple representations, to illustrate the implementation and use of chunking mechanisms within contemporary models of human learning.

The role of performance limitations in the acquisition of verb-argument structure: an alternative account.

This study investigates the role of performance limitations in children's early acquisition of verb-argument structure. Valian (1991) claims that intransitive frames are easier for children to produce early in development than transitive frames because they do not require a direct object argument. Children who understand this distinction are expected to produce a lower proportion of transitive verb utterances early in development in comparison with later stages of development and to omit direct objects much more frequently with mixed verbs (where direct objects are optional) than with transitive verbs. To test these claims, data from nine children aged between 1;10.7 and 2;0.25 matched with Valian's subjects on MLU were examined. When analysed in terms of abstract syntactic structures Valian's findings are supported. However, a detailed lexical analysis of the data suggests that the children were not selecting argument structure on the basis of syntactic complexity.

Subject-auxiliary inversion errors and wh-question acquisition: 'what children do know'?

The present paper reports an analysis of correct wh-question production and subject-auxiliary inversion errors in one child's early wh-question data (age 2;3.4 to 4;10.23). It is argued that two current movement rule accounts (DeVilliers, 1991; Valian, Lasser & Mandelbaum, 1992) cannot explain the patterning of early wh-questions. However, the data can be explained in terms of the child's knowledge of particular lexically-specific wh-word + auxiliary combinations, and the pattern of inversion and uninversion predicted from the relative frequencies of these combinations in the mother's speech. The results support the claim that correctly inverted wh-questions can be produced without access to a subject-auxiliary inversion rule and are consistent with the constructivist claim that a distributional learning mechanism that learns and reproduces lexically-specific formulae heard in the input can explain much of the early multi-word speech data. The implications of these results for movement rule-based and constructivist theories of grammatical development are discussed.

Subject-verb agreement in Brazilian Portuguese: what low error rates hide.

This study focuses on the acquisition of subject-verb agreement in Brazilian Portuguese. A quantitative analysis of the data produced by a Brazilian child between the ages of 3;02.07 and 3;04.08 is presented. The overall error rate is low. However, a further and more detailed analysis reveals important contrasts both in the frequency of production of different verb inflections (as regards the person/number variables within the verb morphological system) and in the rate of subject-verb agreement errors associated with them. Our findings not only suggest that subject-verb agreement may be acquired piecemeal, but also that the learning of particular verb inflections may itself be a gradual process. Alternatives to the idea of rule-governed production--such as the child's reproducing frozen subject-verb strings previously produced by adults and blending different frozen strings into novel combinations--are discussed as processes which can shed some light on the pattern of both erroneous and correct production shown by this child.

Lexically-based learning and early grammatical development.

Pine & Lieven (1993) suggest that a lexically-based positional analysis can account for the structure of a considerable proportion of children's early multiword corpora. The present study tests this claim on a second, larger sample of eleven children aged between 1;0 and 3;0 from a different social background, and extends the analysis to later in development. Results indicate that the positional analysis can account for a mean of 60% of all the children's multiword utterances and that the great majority of all other utterances are defined as frozen by the analysis. Alternative explanations of the data based on hypothesizing underlying syntactic or semantic relations are investigated through analyses of pronoun case marking and of verbs with prototypical agent-patient roles. Neither supports the view that the children's utterances are being produced on the basis of general underlying rules and categories. The implications of widespread distributional learning in early language development are discussed.

Syntactic categories in the speech of young children: the case of the determiner.

There has been a growing trend in recent years towards the attribution of adult-like syntactic categories to young language-learning children. This is based, at least in part, on studies which claim to have found positive evidence for syntactic phrase structure categories in young children's speech. However, these claims contradict the findings of previous research which suggest that the categories underlying children's early multi-word speech are much more limited in scope. The present study represents an attempt to reconcile the findings of these different lines of research by focusing specifically on Valian's (1986) criteria for attributing the syntactic category of determiner to young children. The aim is, firstly, to replicate Valian's results regarding her determiner criteria on a new sample of seven children between the ages of 1;20 and 2;6; secondly, to investigate the extent to which children show overlap in the context in which they use different determiner types; and, thirdly, to compare this with a controlled measure of the overlap shown by competent adult speakers. The results suggest that Valian's criteria for attributing a syntactic determiner category are too generous and could be passed by children with a relatively small amount of limited scope knowledge. They also provide at least some evidence that a limited scope formula account of children's early determiner use may fit the data better than an adult-like syntactic account.

Reanalysing rote-learned phrases: individual differences in the transition to multi-word speech.

The present study investigates the possibility that the previously documented relationship between referential-expressive and nominal-pronominal styles (Nelson, 1975) may be best explained not so much in terms of 'object-orientation' or 'noun-preference', as in terms of the direction from which different children break into structure, with some children tending to construct patterns by combining two or more items from their single-word vocabularies and others tending to develop patterns by gaining productive control over 'slots' in previously unanalysed phrases. In order to do so it makes use of a methodology for distinguishing between productive and unanalysed multi-word speech proposed in Lieven, Pine & Dresner-Barnes (1992) which is applied to observational and maternal-report data from a longitudinal study of seven children between the ages of 0;11 and 1;8. The results suggest not only that variation in children's early word combinations can indeed be explained in terms of different routes to multi-word speech, but also that, far from being atypical, a strategy involving the breaking down of originally unanalysed phrases may be used by all children, though to varying degrees.

Individual differences in early vocabulary development: redefining the referential-expressive distinction.

The existence of stylistic variation between children in the early stages of language acquisition has been most frequently studied using Nelson's (1973) referential-expressive distinction. While the use of this distinction has generated a great deal of interesting research, there are a number of major problems associated with it. The present study presents a simple scheme, based on formal categories, for coding stylistic variation in the early lexicon. When applied to the first 50 and 100 words of 12 children collected between 0;11 and 2;3, the major dimensions of difference are found to be the relative proportion of common nouns and the relative proportion of frozen phrases. Moreover, the proportion of frozen phrases is also found to be significantly positively related to children's early productivity, suggesting that, rather than being a 'dead-end' in early language development, the acquisition of frozen phrases may provide an alternative route into multiword speech.

How referential are 'referential' children? Relationships between maternal-report and observational measures of vocabulary composition and usage.

This study examines the relationship between maternal-report measures of referential vocabulary and observational measures of referential vocabulary and usage in eight first-born middle-class children at 50 and 100 words. The results indicate that although maternal-report measures of vocabulary composition can be reasonably reliable, provided some attempt is made to restrict variation in the number of vocabulary items upon which they are based, such measures tend to exaggerate the relative importance of common nouns for the child in two ways; firstly, in the sense that they reflect differential maternal sensitivity to such word-types in comparison with other less-'referential' items; and, secondly, in the sense that they overestimate the extent to which such word-types actually occur in the child's spontaneous speech.

Referential style at thirteen months: why age-defined cross-sectional measures are inappropriate for the study of strategy differences in early language development.

This study examines the relationship between cross-sectional measures of referential style taken at 1; 1 and measures based on the first 50 words in 12 first-born children. Since no significant relationship is found it is argued that age-defined cross-sectional measures are inappropriate for the study of strategy differences in early language development because they confound such differences with variation due to differences in development level.