Understanding how input matters: verb learning andthe footprint of universal grammarJeffrey Lidza,*, Henry Gleitmanb, Lila GleitmanbaDepartment of Linguistics, Northwestern University, 2016 Sheridan Road, Evanston, IL 60208-4090, USAbDepartment of Psychology and Institute for Research in Cognitive Science, University of Pennsylvania,MYLINEPhiladelphia, PA 19104, USAReceived 18 April 2001; received in revised form 11 January 2002; accepted 17 September 2002AbstractStudies under the heading “syntactic bootstrapping” have demonstrated that syntax guides youngchildren’s interpretations during verb learning. We evaluate two hypotheses concerning the origins ofsyntactic bootstrapping effects. The “universalist” view, holding that syntactic bootstrapping falls outfrom universal properties of the syntax–semantics mapping, is shown to be superior to the “emergen-tist” view, which holds that argument structure patterns emerge from a process of categorization andgeneralization over the input. These theories diverge in their predictions about a language in whichsyntactic structure is not the most reliable cue to a certain meaning. In Kannada, causative morphologyis a better predictor of causative meaning than transitivity is. Hence, the emergentist view predicts thatKannada-speaking children will associate causative morphology (in favor of transitive syntax) withcausative meaning. The universalist theory, however, predicts the opposite pattern. Using an act-outtask, we found that 3-year-old native speakers of Kannada associate argument number and notmorphological form with causativity, supporting the universalist approach.q 2003 Elsevier Science B.V. All rights reserved.Keywords: Word learning; Causatives; Kannada; Input; Syntactic bootstrapping; Universal grammar1. IntroductionHow do children acquire the vocabu lary of their language? Two conclusions about thechild’s acquisition of a first lexicon are obvious from the outset of inquiry. First, the wordsare learned as a tight function of the input, for children manifestly learn French words fromFrench input and Igbo words from Igbo input. Second, the learner plays an active role,weakening and distorting any “simple” description of the input–output relation. It is hope-J. Lidz et al. / Cognition 87 (2003) 151–178 151Cognition 87 (2003) 151–178www.elsevier.com/locate/cognit0010-0277/03/$ - see front matter q 2003 Elsevier Science B.V. All rights reserved.doi:10.1016/S0010 -0277( 02 )0023 0 -5* Corresponding author.E-mail address: [email protected] (J. Lidz).less, for example, to suppose that learning is responsive (solely) to input frequency,because the first word in the English vocabulary is not the. Some “theory of the child”is necessary, then, to bring the facts about the input – how adults speak to children – andthe facts about the output – how children speak – into some sort of responsible alignment.This alignment is what linguistic theory is designed to be about (Chomsky, 1965). Under-standing how environment and nature in this sense come together to explain the course andoutcome of language learning is a crucial question. Indeed, it is the only question worthasking in this domain given that children’s exposure to a language is finite and limited, andyet they come to say anything they choose (or at least anything they can get away with). Inthe present paper, we examine the effects of input and output in understanding a particu-larly revealing subcomponent of lexical learning: the child’s acquisition of the verbvocabulary. Two features of verb learning, which we now discuss in turn, make this aparticularly appealing testbed for comparing learned (input-responsive) and unlearned(learner-driven) aspects of language acquisition.2. Structural correlates of verb learningAs is now well attested, the verbs of the exposure language are acqui red in lockstep withacquisition of those features of the clause-level grammar having to do with the relationbetween a verb’s semantic argument structure and its syntactic structure. Children whounderstand the English verb swim appreciate that it prototypically involves only a singleparticipant and surfaces as an intransitive; it takes two participants for an act of killing,however, and so this verb is transitive; as for giving, because it involves three entities (thegiver, the givee, and the given), it usually surfaces as a ditransiti ve, e.g.Max gave a hamsandwich to Pat.This coincidence ofstructural and semantic learning poses a classic chicken–egg problem,with some authors arguing that syntax (at this level) provides a fundamental guide to youngchildren’s interpretations of the verbs they are learning (Bloom, 1999; Fisher, 1996; Fisher,Hall, Rakowitz, & Gleitman, 1994; Gleitman, 1990; Landau & Gleitman, 1985; Naigles,1990) andothersaverring that argumentstructure patterns emerge from generalizationsmadeafter significant item-based learning (Goldberg, 1999; To masello, 2000).So stated, these positions need not be and have not usually been interpreted as in directconflict. Instead, many authors have emphasized com plementary (or “trade-off”) relationsbetween acquisition of the verb meanings and their associated syntactic properties (see,e.g. Bowerman, 1982; Gillette, Gleitman, Gleitman, & Lederer, 1999; Pinker, 1984,1989). But examined more closely, two currently viable hypotheses concerning the acqui-sition of this form–meaning interface reflect starkly differing viewpoints about the natureof children and thus about their language learning. On the “universalist” view, the correla-tion between lexical and syntactic development falls out from universal properties of themapping between syntax and lexical semantics (Gleitman, 1990; Lidz, 1998a; Lidz, Gleit-man, & Gleitman, 2001; cf. Pinker, 1989). According to this view, many argument struc-ture patterns do not have to be learned independently of the syntax of the language. Thealternative, “emergentist”, view holds that the relations between argument structure andsyntactic structure are learned through a process of categorization and generalization overJ. Lidz et al. / Cognition 87 (2003) 151–178152the input (Goldberg, 1999; Tomasello, 2000). According to the stronges t version of thislatter view, the synt ax –semantics patterns could have been anything at all but, whateverthey turn out to be in any single language, the child will pick up the patterns via inductivedistributional
View Full Document
Unlocking...