It is widely believed that reduplication is constrained in certain formally explicit ways. A classic constraint, expressed in the Prosodic Morphology Hypothesis, maintains that reduplicants must conform to invariant affixational templates which are themselves defined “in terms of the authentic units of prosody.” Recent work has shown that in some languages reduplication is atemplatic. All known examples of atemplatic reduplication involve a copying domain that targets part of a prosodic unit. Reduplicants derived in this manner can thus be characterized as subtemplatic. By contrast, in at least three Formosan languages reduplicants are formed by joining a syllable to a preceding coda to form a non-prosodic morphological constituent, and hence constitute examples of what can properly be called ‘supertemplatic’ reduplication. Together these observations make it clear that although there is a general tendency for the form of reduplication to be shaped by considerations of prosody, this tendency can be violated by underextension or overextension of a prosodic domain. Other Austronesian languages violate theoretical expections on the form of reduplication with regard to notions of markedness or alignment.