Language Variation and Change, 2 (1990), 1-18. Printed in the U.S.A. © 1990 Cambridge University Press 0954-3945/90 $5.00 + .00 The development of a morphological class GREGORY R. GUY Stanford University SALLY BOYD Goteborgs Universitet ABSTRACT English has a class of "semiweak" verbs, which in the past tense have root-vowel ablaut as well as reflexes of the apical stop suffix, for example, kept, told. This study traces the development of this class in a sample of speakers aged 4-65. The evidence is derived from the variable rates of occurrence of the final -t,d in these words in the speech of individuals of different ages. The rate of -t,d absence in semiweak verbs systematically declines with increasing age. We iden- tify three ontogenetic stages in the development of the class. In children's speech, these segments rarely appear, suggesting they are underlyingly absent. In young adults they appear but undergo the variable -t,d deletion process of English at the same high rate as noninflectional -t,d in words like west, old, implying that such speakers do not treat them as affixes. Finally, some adult speakers show a lowered deletion rate, suggesting that they accord the final stops separate morphemic status. The age distribution of this pattern implies that speakers only arrive at this analysis in adult life, after the age when ac- quisition is often assumed to be complete. One problem that must be confronted in studying child language develop- ment is that of describing a system that is rapidly changing but that retains structure and systematicity in the midst of the change. This is similar to the problem that has arisen in studies of language change in society. In the so- ciolinguistic arena, two basic concepts that have facilitated successful ap- proaches to the study of changing structure are orderly heterogeneity (Weinreich, Labov, & Herzog, 1968), which highlights the systematic nature of linguistic differences between individuals or social groups, and the vari- able rule (Cedergren & Sankoff, 1973; Labov, 1969), which makes it possi- ble to model quantitatively the alterations that a changing system passes Earlier versions of this article were presented (as Boyd & Guy, with a somewhat different title) at the Linguistic Society of America meeting in 1979 and the Australian Linguistic Society meet- ing in 1980. Those versions have already been cited in print, in articles such as Sankoff and Labov (1979), Romaine (1986), and Labov (1989). The preparation of the present version has been hin- dered by the fact that the authors have been living in different hemispheres since 1979. We grate- fully acknowledge the constructive criticism of Bill Labov, Tony Kroch, Charles Ferguson, Eve Clark, and the anonymous reviewers of this journal. Address correspondence and offprint requests to: Gregory R. Guy, Department of Linguis- tics, Stanford University, Stanford, CA 94305-2150. Email: guy@csli.stanford.edu. 1