Paradigms that combine synthetic (one-word) and periphrastic forms in complementary distribution have loomed large in discussions of morphological blocking (McCloskey and Hale 1983, Poser 1986, Andrews 1990). Such composite paradigms potentially challenge the lexicalist claim that words and sentences are organized by distinct subsystems of grammar. They are of course grist for the mill of Distributed Morphology, a theory which revels in every kind of interpenetration of morphology and syntax. But they have prompted even Paradigm Function Morphologists to introduce syntactic (...) constructions into their morphology. I shall argue, instead, for a lexicalist treatment, which is based on the idea that blocking is a filtering device that applies to the output of the generative system, rather than operating directly on its derivations (Wunderlich 1996). I present this approach to blocking in section 1, and show in section 2 how it deals with the intricate composite verb paradigm of Latin, where the periphrastic perfect passive supplies the missing pieces of an otherwise synthetic inflectional system. This part of Latin verb morphology has recently been treated from the perspective of Distributed Morphology and Paradigm Function Morphology. I compare my solution to these treatments and argue that it is superior in two respects: it predicts the complementarity of the synthetic and periphrastic formations and yet allows their respective morphological and syntactic properties to be captured, and it readily covers some basic data that the other analyses get wrong. (shrink)