Three experiments are described which support the hypothesis that in a lexical decision task, prefixed words are analyzed into their constituent morphemes before lexical access occurs. The results show that nonwords that are stems of prefixed words (e.g., juvenate) take longer to classify than nonwords which are not stems (e.g., pertoire), suggesting that the nonword stem is directly represented in the lexicon. Further, words which can occur both as a free and as a bound morpheme (e.g., vent) take longer to classify when the bound form is more frequent than the free form. Finally, prefixed nonwords took longer to classify when they contained a real stem (e.g., dejuvenate), compared with control items which did not (e.g., depertoire). A general model of word recognition is presented which incorporates the process of morphological decomposition.