Relational Frame Theory and Early Language Learning: Evidence from Outside the Experimental Analysis of Behaviour Rick Dale Department of Psychology, Cornell University, [email protected]

Abstract Verbal behaviour, however, involves other levels of In early language (1;3 to 2;0) 50% of verbs were constrained organization just as amenable to theoretical discussion in BA as to one construction type; verbs with similar “semantics” This poster outlines research by Michael Tomasello and his elementary naming behaviour. One such level is syntax, a exhibited disparate usage; even verbal inflection (-ing, -ed, colleagues (see Tomasello, 2000, for a review) concerning the perennial problem (Lashley, 1951) around which much of etc.) appeared as constrained to particular verbs. nature of early language learning. It is argued that the data modern cognitive science has revolved.2 should not merely be seen as consistent with RFT, but as 92% of all utterances can be accounted for by 25 initial important evidence that verbal behaviour emerges in accordance A possible strategy3 is to show emerging verbal behaviour in constructed patterns. with RFT principles. accordance with RFT. In what follows, I argue that data from discontinuity theorists (Tomasello, 2000) reveal that syntactic Reinventing the Wheel properties of early language are predicted by RFT principles. Cognitive Approaches to Language Learning Continuity theorists of the Chomskian sort rely on the syllogism, crudely put: Children are so neat and I don’t know why, and if I RFT prediction: Continuity theorists (e.g., Pinker, 1984; Chomsky, 1986) argue don’t know why, then it must be because of innate internal that children use innate, full-fledged grammatical categories Words and their syntactic (autoclitic) frames are initially tied machinery. The inference is intended to explain language during language acquisition. The acquisition process merely to exemplars, and gradually decontextualize into complex acquisition. 4 involves aligning this innate, linguistic machinery to the relational frame networks akin to adult syntax. appropriate properties of the child’s language. Tomasello’s results challenge the initial premise of this argument. Children are not magically endowed with adult Continuity prediction: Challenging Chomskians grammatical categories. Instead, he explains his results using a Verb-Island hypothesis: Early syntactic competence develops Early language looks very much like adult language. Tomasello (2000) reviews data showing that learning revolves around verbs and their nominal arguments. around concrete exemplars, not abstract grammatical categories. There exist many theories that involve discontinuity between Most studies involve simply introducing a new word to children child language acquisition and adult language (e.g., Tomasello, However, RFT introduces theoretical tools that provide both in some grammatical context, and observing what children do breadth of understanding, and comparative specificity in 2000; Seidenberg & MacDonald, 1999). Central to this with that word in the same or different grammatical contexts (in approach is the importance of the primary linguistic data to the prediction. Just as naming emerges, early syntax is constrained production or comprehension). This general technique yields by exemplars, and gradually becomes a sophisticated repertoire child. In general, each theory purports that language acquisition interesting results: proceeds according to some general function of the child’s of syntactic relations. language experience. Young children (1;6 to 3;0) do not use novel verbs in new The properties central to Tomasello’s approach (item-specific ways; this skill emerges later (3;0 and later). Discontinuity prediction: learning, imitation, etc.) have been recognized in BA since Skinner’s maligned text. A post-Skinnerian approach through 1 Novel transitive constructions are highly constrained to the Early language can look very different from adult language. RFT may adequately enhance fundamental principles to go model exemplar in young children (2;0 to 3;0). beyond existing cognitive theory. These extreme views lie at opposite ends of a continuum of theories, with many a mixture of continuous and discontinuous Younger children (2;0) form sentences based on an exemplar References features (e.g., Newport, 2002; Fisher, 2002). The focus of their model, despite its being unusual; older children (4;0) correct Akhtar, N. (1999). Acquiring basic word order: evidence for data-driven learning of syntactic structure. debate, the role of experience, is taken for granted in with the usual word order (figure 1). Journal of Child Language, 26, 339-356. Chomsky, N. (1959). Review of Skinner’s Verbal Behavior. Language, 35, 26-58. behavioural approaches. Chomsky, N. (1986). Knowledge of Language. Berlin: Praeger. Figure 1: Mean proportion of order by model condition (from Akhtar, Fisher, C. (2002). The role of abstract syntactic knowledge in language acquisition: a reply to 1.2 1999) Tomasello (2000). Cognition, 82, 259-278.

Relational Frame Theory 1 Hayes S.C., Barnes-Holmes, D. & Roche, B. (2001). Relational Frame Theory: A Post-Skinnerian Account of Human Language and Cognition. New York: Kluwer Academic/Plenum Publishers.

Behaviour analysis (BA) has been experiencing a revived 0.8 SVO Horne, P.J. & Lowe, C.F. (1996). On the origins of naming and other symbolic behavior. Journal of the interest in the nature of verbal behaviour. The debate in BA has SOV Experimental Analysis of Behavior, 65, 185-241. 0.6 VSO centred around the phenomenon of naming, an unfortunate fact Lashley, K. (1951). The problem of serial order in behavior. In L.A. Jeffries (ed.), Cerebral Mechanisms in Behavior: The Hixon Symposium, pp. 112-136. New York: Wiley. 0.4 given good arguments against its status as central to verbal Lipkens, G., Hayes, S.C. & Hayes, S.C. (1993). Longitudinal study of derived stimulus relations in an Mean proportion matches behaviour (e.g., Wittgenstein, 1958; Quine, 1960). RFT (Hayes 0.2 infant. Journal of Experimental Child Psychology, 56, 201-239. Newport, E. (2002). Statistical Learning in Human Adults, Infants, and Nonhuman Primates: et al., 2001) and alternative accounts of naming (e.g., Horne & 0 2 3 4 Constraints on Learning Adjacent and Nonadjacent Regularities. Paper presented at the Cornell Lowe, 1996) have gone to great lengths to explain basic Age (years) University Symposium on Statistical Learning Across Cognition, April, 2002. equivalence performance, such as naming. Pinker, S. (1984). Language Learning and Language Development. Cambridge, MA: Harvard University Press. 2. Consider Chomsky, 1959, and both its real and imagined effects on the study of behaviour Quine, W.V.O. (1960). Word and Object. Cambridge, MA: MIT Press. 1. Discontinuity is still consistent with the continuity prediction, should it bear out. However, only 3. Lipkens et al. (1993) contributed to this debate by studying a preverbal child, and demonstrating Seidenberg, M. & MacDonald, M. (1999). A probabilistic constraints approach to language acquisition discontinuity could account for its prediction, however much of a difference between early and adult behaviour in accordance with RFT principles. and processing. Cognitive Science. 23, 569-588. language could be discovered with future research. Tomasello, M. (2000). Do young children have adult syntactic competence? Cognition, 74, 209-253. 4. RFT might predict words and their sequential relations becoming increasingly complex as relational responding emerges through multiple-exemplar training. Wittgenstein, W. (1958). The Blue and Brown Books. New York: Harper & Row.