Impoverishment is a feature deletion operation that is standardly taken to require a post-syntactic approach to inflectional morphology (like, in particular, Distributed Morphology, as in Halle & Marantz 1993 ). Given that by now a lot of empirical evidence in support of such an operation has been accumulated, impoverishment can be viewed as a good argument in favour of the post-syntactic approach, given a Chomskyan Y-model of grammar in which inflectional morphology is considered an autonomous component of grammar (i.e., not subsumed under syntax proper). Against this background, the main goal of the present contribution is to show that a pre-syntactic approach to impoverishment becomes available if the complete well-formed sets of morpho-syntactic features that are required in any non-syntactic realizational approach to inflectional morphology (cf. Stump 2001 ; Inkelas 2016 ) are not given a declarative characterization (as is normally assumed), but arise incrementally, by iterated addition of features. On this view, pre-syntactic impoverishment is actually premature exponence brought about by late addition of context features, and the interaction between impoverishment and morphological exponence is not one of bleeding (as in post-syntactic approaches) but one of counter-feeding. The new concept of impoverishment is implemented in a pre-syntactic model of inflectional morphology developed in Müller (2020) , which is based on Harmonic Serialism, a derivational version of Optimality Theory (cf. Prince & Smolensky 2004 ). Finally, while pre-syntactic impoverishment is basically a mirror image of post-syntactic impoverishment, it turns out to be more restrictive since it ceteris paribus predicts that there is no impoverishment of inherent features: Features that are not added cannot be added late. I suggest that this prediction is confirmed by evidence from gender features and inflection class features.