| The present study was motivated by the concern that many current models of phonology (i.e., Structuralist, Generative, and natural Phonology) presented an overly deterministic picture of phonological development (Kiparsky and Menn, 1977). A possible alternative to these models exists among more cognitively-oriented theories (Kiparsky and Menn, 1977; Macken and Ferguson, 1983; Menn, 1976b, 1983a; Schwartz and Leonard, 1982). Cognitively-oriented accounts share many of the same assumptions of the above theories, but differ crucially in the degree of freedom they permit children in constructing phonological systems. Within cognitively-oriented models, rather than following an innately determined path of development, children construct phonologies through general cognitive mechanisms. The theories view children as actively developing hypotheses and imposing structure on the ambient language.;Following Bever (1970), Kiparsky and Menn (1977), Macken and Ferguson (1983), and Menn (1976a), the contention of the paper was that regressions in development constituted the strongest existing evidence for cognitively-oriented theories. Few regressions had been reported in the phonological literature. The present paper studied the phonologies of two children, aged approximately 1;10-2;0 and 2;0-2;2. The research question addressed was: Do the children produce phonological regressions?;Both subjects were found to produce regressions. The Younger child's regressions arose from four rules: Velar Regression, (aI) Rule, Schwa Addition, and Stress Alternation. The older child's regressions arose from three sources: Lexical Incorporation, Semantic Reorganization, and a Morphological Strategy.;The study yielded five general conclusions: (1) The most direct finding was that the children's phonological development was not a succession of steps, each bringing the learner closer to the surface forms of the ambient language. (2) Results were generally in keeping with previous findings as regards type and length of time of regressions. (3) The data generally supported Menn's model of phonological rules, which claims that children's rules function to reduce variation in phonetic output (Menn 1976a, 1976b). (4) The children's regressions appeared difficult to reconcile with phonological models with large deterministic components. (5) The regressions appeared to follow from basic principles of cognitively-oriented models. |
