Reduplication and Syllable Transfer in Sanskrit and Elsewhere

ing away from the effect of the phonological rules mentioned, the template of Nicobarese reduplication is a syllable PuC. A standard copyand-association analysis involving a prespecified CVC template will have to somehow insure that the coda segment in a form like cim ends up in coda position. Since prelinked segments, if used at all, must be interpreted conditionally, we are faced with the choice of derivations shown below: (79) c i m c i m c i m c i m [X [X XI] [X [X X]] [X [X XI] [X [X X]I I I I I u ? u output: rnum-cim output: *Ac-cim In the right-hand derivation association proceeds left-to-right. Since the initial c of cim cannot link up to the first C of the prefix, it is expected that it will continue to search for an eligible C position: but when the empty C is found, we get *Puc-cim (surface *Pc-cim) instead of Pum-cim. The correct result is achieved in the left-hand derivation, where association proceeds right-to-left. The problem here is that we have derived the right form at the cost of reversing the expected direction of association: left-to-