ABSTRACT
One of the longest running debates in psycholinguistics centres around the question of whether processes such as syntax semantics, word recognition, and phoneme identification interact or not. Connectionism has given this old debate a new lease of life. Once it was easy to criticise interactive theories on the grounds that they never presented a clear specification of how interaction between processes might actually take place. Connectionism provides a clear answer to such criticism. A connectionist model such as TRACE (McClelland & Elman, 1986) gives us an explicit computational account of the mechanism whereby phoneme and word recognition interact. No longer is there any room for debate over what that theory can explain, or exactly what its predictions should be.
