Abstract
Pesetsky (1991) proposes that there are two types of C, that and null C, and that the null C is an affix that must move up to the matrix V. This proposal is revived in the minimalist context by Bošković and Lasnik (2003). While assuming that the C-as-an-affix approach is on the right track, I suggest a drastic modification of previous versions of this approach: namely, that (a) there is just one type of C in the lexicon, affixal null C; (b) it can both hop down onto the embedded V and move up to the matrix V; and (c) that/for is inserted at PF as a last resort if affixation is structurally prohibited. This amounts to saying that the English tense and complementizer systems display the same paradigm: both T[+finite] and C[±finite] are affixes, and do and that/for are inserted as a last resort when syntactic affixation is impossible. This approach, especially the C-hopping approach, allows a uniform, principled account for the distribution of that and for, including that-trace effects, ameliorating effects of subject extraction, anti-that-trace effects, For-To Filter effects, and the distribution of that in relative clauses.