Lexical activation (and other factors) can mediate compensation for coarticulation

A key dispute in theories of spoken word recognition is whether activation of a lexical representation can affect the perception of sublexical components, such as phonemes. Elman and McClelland (1988) provided evidence for such top–down processing by showing that a prelexical process (compensation for coarticulation) could be affected by lexical activation. However, Pitt and McQueen (1998) reported that the observed compensation effects were in fact due to the transitional probability of certain phonemic sequences, rather than lexical activation. Part I of the current study shows that perceptual grouping must be considered in assessing compensation effects, reopening the question of whether lexical activation can drive compensation. Part II shows that when perceptual grouping is taken into account, lexically mediated compensation can indeed be observed, confirming the interactive nature of speech processing. Collectively, the results make it clear that compensation for coarticulation is affected by a number of factors, making it difficult to isolate lexical influences on the phenomenon.