Bobaljiks (2002) Copy Theory: Pronunciation of Chains
Typically, in a nontrivial chain, the highest or the lowest copy is pronounced, as in the Hindi example in (1) and the Punjabi example in (2).
(1)
(2)
In Bobaljik’s (2002) approach, the syntax module of grammar generates a single representation that is interpreted by LF and PF. LF and PF are treated equally in the sense that they can both privilege a certain copy in a given chain. In other words, LF can choose to interpret the higher copy or the lower copy, and PF can choose to pronounce either copy as well. As a result, the distinction between overt and covert movement operations is just the matter of which copy is pronounced. Bobaljik (2002, 199) suggests that there is a four-way typology of movement operations, demonstrated in table 1. Pattern (a), in which the higher copy is interpreted and pronounced, is the structure of ‘‘overt’’ movement in syntax such as the questions and relative constructions in (3)–(4). Pattern (b) is the structure of ‘‘covert’’ movement in which the higher copy is interpreted, but the lower copy is pronounced. This kind of movement is illustrated in (2). Pattern (c) is an instance of the reconstruction effect (e.g., Barss 1986; Chomsky 1995), in which the higher copy is pronounced, while the lower copy renders the interpretation. This pattern is demonstrated in (3), an example given by Chomsky (1995, 74) regarding the binding condition.
Table 1.
(3)
Here the antecedents for himself and each other must be Bill and we, respectively, even though both John and they are possible antecedent candidates. In order to capture the correct binding condition, both anaphors have to be interpreted at the lowest position, t. Finally, Bobaljik argues that object shift in Germanic languages conforms to the pattern in (d), in which both pronunciation and interpretation target the lower copy.
A number of authors (see, e.g., Bobaljik 1995; Brody 1995; Pesetsky 1998) have questioned why PF always targets the highest copy, which Chomsky (1993) assumes without argument. They have convincingly shown that in certain cases PF can choose either the highest or the lowest copy of a nontrivial chain, a result that captures ‘‘overt’’ and ‘‘covert’’ movement in syntax. Two questions arise at this point: (1) Instead of the highest or the lowest copy being pronounced, is it possible for PF to choose to pronounce an intermediate copy of a nontrivial chain that consists of more than two copies? (2) Can PF allow more than one copy of a nontrivial chain.
