Abstract
This article explores the possibility that the distinction between interpretable (valued) and uninterpretable (unvalued) features has grammatical manifestations beyond its role in feature checking. I argue that both selection and lexical insertion are sensitive to this distinction; thus, a head may determine not only which features its complement must bear but also whether they should be interpretableor not. Empirical consequences are explored in Hebrew, where infinitival complements to negative verbs (‘refrain’, ‘prevent’) display a number of surprising syntax-semanticscorrelations.Those are tracedto the operation of negative features in the Comp position. The analysis also provides insight into the recalcitrant prevent DP from V-ing construction in English.