Basic-level picture naming is hampered by the presence of a semantically related context word (compared to an unrelated word), whereas picture categorization is facilitated by a semantically related context word. This reversal of the semantic context effect has been explained by assuming that in categorization tasks, basic-level distractor words (e.g., "dog") do not compete with the selection of the correct category label (e.g., "animal"). In this article, we test an alternative account in terms of a congruency effect ("message-congruency"), which arises at the conceptual level when target (e.g., the picture of a cat) and context (e.g., the word "dog") converge on the same to-be-verbalized concept (e.g., "animal"). In four experiments we observed a substantial message-congruency effect in categorization and action naming. Implications for models of spoken-word production are discussed.