Principle B Effects in English —Conventionalized Division of Labor—

For many years, the most prominent approach to Principle B effects in English has been the two-layered approach initiated by Reinhart (1983). Recently, however, serious doubt has been shed on the empirical claims that support this approach. This paper proposes an alternative account. Principle B effects in English are argued to be manifestations of a conventionalized division of labor: marked pronouns are used to indicate that a reflexive interpretation is intended, while unmarked pronouns are used to indicate that such an interpretation is not intended. The account is spelled out in detail, and compared with related proposals in the literature. 1 Background: the Two-Layered Approach Much attention in linguistics is devoted to certain puzzling restrictions on the interpretation of anaphoric expressions. One such a restriction is that English pronouns cannot be anaphorically related to their coarguments. For example, (1) can not be taken to mean that John voted for John, and (2) cannot be taken to mean that every girl hated herself. (1) John voted for him. (2) Every girl hated her. The most influential approach to this phenomenon is the one initiated by Reinhart (1983). This is a two-layered approach: it assumes that there is a syntactic constraint which prohibits binding of pronouns by coarguments, and that there is an additional mechanism which prohibits coreference whenever this yields an interpretation that is indistinguishable from the one that would be obtained by coargument binding. The syntactic constraint on coargument binding is usually referred to as Principle B, while the additional mechanism is usually referred to There are several versions of this generalization. The one assumed here is adapted from (Büring, 2005a, p.55): a pronoun cannot be codetermined with one of its coarguments. Two expressions A and B are codetermined iff (i) A binds B, (ii) A and B corefer, or (iii) there is a third expression C which is codetermined with both A and B (cf. Heim, 1998; Roelofsen, 2008). Two expressions are coarguments iff they have the same caseor θ-role assigner.