======================================
NOTE: This paper was never published, but it has occasionally been cited 
in published work. For that reason, I'm making it available here, though 
it has some flaws (mostly of presentation and framing) that I never got
around to addressing to my own satisfaction.
======================================
Recent studies within Distributed Morphology (DM; Halle and Marantz 
1993; 1994) have argued that contextual allomorphy is 
morpho-syntactically local: allomorphic dependencies can only hold 
between morphemes within the same cyclic domain, where 'cyclic' is 
defined in terms of functional syntactic heads like n and C (e.g. 
Marantz 2007; Embick 2010). In this paper I argue that reduplication 
poses a serious problem for this view of allomorphic locality. Two 
widely attested patterns of reduplication --- category-changing 
reduplication and aspectual reduplication --- are shown to be non-local 
under the proposals set out in Embick (2010), given that they involve 
dependencies across two cyclic heads. Further, an amended theory 
allowing for such interactions makes incorrect predictions about the 
range of possible allomorphic dependencies. Theories of locality relying 
on cyclic syntactic heads are thus either too restrictive, or not 
restrictive enough. I conclude that phonological adjacency, rather than 
morpho-syntactic locality, is the relevant constraint on allomorphic 
locality.