Based on previous work by Bošković (2001, 2002, 2004a,b) and Nunes (1999, 2004), this chapter discusses a considerable amount of evidence involving A-movement, A'-movement, head movement, and remnant movement that points to the conclusion that "traces" (i.e. copies structurally lower in the syntactic representation) may be phonetically realized. In addition, the issues regarding phonetic realization of copies are shown to be determined by conditions of the phonological component and not of syntax (movement) per se. As a result, the chapter is able to explain a variety of complex phenomena that cannot be captured by trace theory. The chapter starts by reviewing several pieces of evidence that show that the phonetic realization of copies is similar to the LF interpretive procedure in the sense that it allows activation of lower copies, as well as instances of "scattered deletion", where different pieces of different chain links are realized. It is argued that convergence requirements related to linearization and morphological fusion interact with economy computations regarding applications of deletion, yielding a complex crosslinguistic pattern whereby chains in the general case have only their highest link phonetically realized, but they may also trigger pronunciation of a lower link or even pronunciation of multiple links if convergence so demands. * Previous versions of this paper were presented at the Mayfest 2002: The Minimalist Fact (University of Maryland, 5/16-18/02) and at the Workshop "The Copy Theory of Movement on the PF Side" (Universiteit Utrecht, 12/14-15/04). We are thankful to these audiences and an anonymous reviewer for their comments. Special thanks to Marcel den Dikken for his detailed comments and suggestions. Thanks also to Carol Petersen for her editorial help. During the writing of this paper, the second author received support from CNPq (grants 300897/1996-0 and 200422/04-9). On the LF side we thus have some choice in deciding where deletion should take place in nontrivial chains. By contrast, it is standardly assumed that no such choice is available in PF, the 1 From now on, copies will be annotated with superscripts. 2 Chomsky (1993) actually assumes that LF deletion in cases such as (2a) needs to leave a variable (without internal structure) behind, at least for θ-theoretic reasons. This however might become unnecessary under Bošković