Abstract This article argues that post-syntactic feature insertion is insufficiently motivated at present and should therefore not yet be accepted as a morphological process within the theory of Distributed Morphology. Post-syntactic feature insertion has been proposed for Nimboran verbal morphology (Noyer 1998), Kiowa agreement prefixes (Harbour 2003a), Leísta Spanish doubled clitics (Nevins 2007), and Romanian gender agreement (Farkas 1990). I show that these datasets can be analysed without invoking post-syntactic feature insertion, and discuss what a dataset would have to look like to motivate it. Such a dataset would have a number of parallel circumstances, each with a true elsewhere exponent and a specific environment in which another exponent behaved like an elsewhere one, exponing two feature bundles that are the polar inverse of each other. A theory of morphology that does not allow for post-syntactic feature insertion predicts such data to be impossible.
Hannah Jane Middleton (Sat,) studied this question.