Grammatical effects in picture–word interference experiments have been argued to reflect grammatical constraints during lexicalisation. Alternative views hold that those effects derive from the coincidence of semantic and grammatical differences between candidates. We present three experiments conducted in Spanish. Semantic relatedness between target and distracters (related or unrelated), as well as grammatical class (nouns or verbs) and semantic domain (objects or actions) of the distracters were manipulated in infinitive or inflected action naming tasks. Whereas related action-words, but not object-nouns, produced longer reaction times irrespective of their grammatical class in the infinitive condition, only related verbs slowed latencies in the inflected condition. Our results suggest that speech production relies on the exclusion of candidate responses that do not fulfil task-pertinent criteria like membership in the appropriate semantic domain or grammatical class. Taken together, these findings are explained by a response-exclusion account of speech output. This and alternative hypotheses are discussed.