respectively, "*(it) seemed the dog to bark", where each verb takes its own selected subject, and na:tural, with the same sound structure of na:ture, plus al. The question in each case is what causes the observed distortions which, as it turns out, are themselves regularities of the language ("raising", and "trisyllabic shortening", respectively). The answer in each case could in principle be either: (i) Representational constraints on the larger units (phrases or words) that prevent achievement of the strict compositionality; or: (ii) Transformational operations that take the result of strict compositionality as their input and produce the observed structures as their output. In the 60's, (ii) was uniformly taken to be the right answer, both in syntax and phonology. But at that time, the mere discovery of formally statable regularities was a considerable achievement in itself, and that success understandably obscured the important empirical issue of choosing between (i) and (ii) that

Introduction One of the important goals of Chomsky's (1993) "minimalist" program is that of simplifying the theory of syntax by reducing all Case assignment to Spec(ifier)-head agreement, in the manner illustrated in (1a-c). (1) a. Nominative subject: AGR P AGR ' s s 6# # ########### # # # #### 5 5 5 Spec AGR V s He 3SG talks b. Nominative object: AGR P AGR ' s s 6# # ########## # # # #### # # #### # # ### 5 5 5 # 5 Spec AGR V NP PP s There 3PL are people at the door # # #######<####################### LF movement c. Accusative object: AGR P AGR P AGR ' s o o 6# # ## # # # ########### # # # #### # ### 5 5 5 5 5 Spec Spec AGR V NP o John reads a book # # ######<################### LF movement In (1a), the "functional" head AGR enters into a