On the representation and processing of compound words: Automatic access to constituent morphemes does not occur

Three lexical decision experiments using a variant of the semantic priming technique tested the hypothesis that compound words are morphologically decomposed during recognition. If a compound constituent is accessed during processing, an associative prime will facilitate that access and hence recognition of the whole word. Contrary to the predictions derived from the automatic decomposition hypothesis, Experiment 1 revealed no priming effects for semantically opaque compounds (buttercup) and pseudo-compounds (boycott), primed either on their initial or final constituent. The data from Experiment 2 suggested that both constituents in semantically transparent compounds are accessed (teaspoon). Experiment 3 was a replication experiment, confirming that final constituents of opaque compounds are not accessed, whereas those of transparent compounds are. The overall pattern of data refutes the notion of automatic morphological decomposition proposed by Taft and Forster (1976). However, a revised decomposition procedure would be compatible with the results. Morphemes might only be accessed if no other lexical representations match the orthographic description of the parsed stimulus part. In this account, only semantically transparent compounds lack an independent lexical representation.