We present a series of three analyses of young children's linguistic input to determine the distributional information it could plausibly offer to the process of grammatical category learning. Each analysis was conducted on four separate corpora from the CHILDES database (MacWhinney, 2000) of speech directed to children under 2;5. We showthat, in accord with other findings, a distributional analysis which categorizeswords based on their co‐occurrence patterns with surroundingwords successfully categorizes the majority of nouns and verbs. In Analyses 2 and 3, (...) we attempt to make our analyses more closely relevant to natural language acquisition by adopting more realistic assumptions about howyoung children represent their input. In Analysis 2, we limit the distributional context by imposing phrase structure boundaries, and find that categorization improves even beyond that obtained from less limited contexts. In Analysis 3, we reduce the representation of input elements which young children might not fully process and we find that categorization is not adversely affected: Although noun categorization is worse than in Analyses 1 and 2, it is still good; and verb categorization actually improves. Overall, successful categorization of nouns and verbs is maintained across all analyses. These results provide promising support for theories of grammatical category formation involving distributional analysis, as long as these analyses are combined with appropriate assumptions about the child learner's computational biases and capabilities. (shrink)
We recognize today's deep neural network (DNN) models of language behaviors as engineering achievements. However, what we know intuitively and scientifically about language shows that what DNNs are and how they are trained on bare texts, makes them poor models of mind and brain for language organization, as it interacts with infant biology, maturation, experience, unique principles, and natural law.
We found a direct relationship between variation in informants' grammaticality intuitions about pronoun coreference and variation in the same informants' use of a clause segmentation strategy during sentence perception. It has been proproposed that ‘c‐command’, a structural principle defined in terms of constituent dominance relations, constrains within‐sentence coreference between pronouns and noun antecedents. The relative height of the pronoun and the noun in the phrase structure hierarchy determines whether the c‐command constraint blocks coreference: Coreference is allowed only when the complement (...) structure containing the noun is attached higher than the pronoun. We collected informants' judgments on pronoun‐noun coreference in which the noun antecedent was contained in a complement structure dominated by either the Sentence‐node (S‐node) (higher than the pronoun) or the Verb‐phrase‐node (VP‐node) (not higher than the pronoun). We also assessed each informant's perceptual clause‐closure tendency using an auditory word‐monitor paradigm. Informants who strongly segmented clauses in the perceptual task did not differentiate between an S‐ and VP‐attachment of sentence complements, as revealed in their coreference judgments, but rather appeared to attach all sentence complements to the S‐node. Informants with relatively weak perceptual segmentation differentiated their coreference judgments according to the node attachment of the complement structure. These results indicate that the linguistic universal controlling within‐sentence coreference applies to the perceptually available structure for a sequence, not to its pure linguistic structure. Hence, linguistic intuitions result from the interaction of three independent faculties: language‐specific knowledge, perceptual processes, and linguistic universals. (shrink)
These essays by some of the most prominent figures in linguistics, artificial intelligence, and psychology explore the problems involved in creating a general cognitive science that will treat language, thought, and behavior in an integrated fashion. They address the fundamental questions of the relations between linguistic structures and cognitive processes, between cognitive processes and language behavior, and between language behavior and linguistic structure. Contents: Introduction, Thomas G. Bever (Columbia University), John M. Carroll and Lance A. Miller (IBM Thomas J. Watson (...) Research Center). "Philosophy and "Linguistics: An Outline of Platonist Grammar, Jerrold J. Katz (CUNY). Sense and Reference in a Psychologically Based Semantics, Ray Jackendoff (Brandeis University). Some Thoughts on the Boundaries and Components of Linguistics, Charles J. Fillmore (University of California, Berkeley). Psychology: Approaches to the Study of the Psychology of Language, Walter Kintsch (University of Colorado). Toward An Abstract Performance Grammar, Charles E. Osgood (University of Illinois). Upgrading a Mind, David Premack (University of Pennsylvania). Computational "Models: "Memory, Meaning, and Syntax, Roger Schank (Yale University) and Lawrence Birnbaum (Yale University). Some Inadequate Theories of Human Language Processing, Mitchell P. Marcus (AT&T Bell Laboratories). (shrink)
Christiansen & Chater ignore the many linguistic universals that cannot be reduced to processing or cognitive constraints, some of which we present. Their claim that grammar is merely acquired language processing skill cannot account for such universals. Their claim that all other universal properties are historically and culturally based is a nonsequitur about language evolution, lacking data.