Abstract: The morpheme repeated prime paradigm was used to probe the mental representation of compound words. The targets were 24 semantic transparent words, 24 opaque words, and 48 nonwords (each consisted of two characters). Five types of primes were used. namely : (1) the targets themselves, (2) the first morphemes of the targets, (3) the second morphemes of the targets, (4) nonwords made of the second and first morphemes of the target words, and (5) morpheme-related nonwords. After a prime showing on the monitor for 100 msec, a target replaced it and lasted for 400 msec. Subjects were instructed to decide whether a target was a word or not as correctly and rapidly as possible. The primes for transparent words were significantly larger than those for opaque ones in the conditions of (2), (3), and (5). The result demonstrated the role of semantic transparency in the mental representation of compound words. It implied that the connection between a transparent word and its morpheme was different from that between an opaque word and its morpheme: the former might be excitatory and the latter inhibitory.

Key words: compound words, morpheme, mental representation, semantic transparency, repeated prime