In semantic studies of argument structure alternations as well as in psycholinguistic studies on syntactic priming, lexical alternation biases are typically measured at the level of the verb lexeme. This study explores the hypothesis that the proper locus of subcategorization probabilities is the verb sense. It investigates the effects of lexical polysemy on the subcategorization probabilities of Dutch dative alternating verbs as reflected in frequency data from natural language corpora and from a priming experiment. A sense-based distinctive collexeme analysis on the corpus data indicates that distinct senses of the same verb may indeed display different alternation biases. The response patterns in our priming experiment suggest that language users keep track of verb subcategorization preferences at different levels of schematization.