Authors:

McQueen, J., Norris, D. and Cutler, A.

Reference:

Year of publication:

1998

CBU number:

3791

Abstract:

In four experiments, listeners made decisions to words and nonwords, some of which had been cross-spliced such that they contained mismatching acoustic-phonetic information. In all experiments performance on mismatching items was poorer than on items with no mismatch. There were no differences between words cross-spliced with other words and words cross-spliced with nonwords. In a lexical decision experiment and in only one of three phonetic decision experiments, however, performance was poorer on nonwords cross-spliced with words than on nonwords cross-spliced with other nonwords. These results, together with other recent findings, lead us to reject three models of phonetic decision-making. The Race model (Cutler & Norris, 1979) is unable to account for any lexical involvement in phonetic decisions to nonwords; the standard version of Trace (McClelland & Elman, 1986) is unable to account for the absence of differences between the cross-spliced words; and the third model (Marslen-Wilson & Warren, 1994) is unable to account for the variability of the lexical effect in the cross-spliced nonwords. We outline a new model which can account for these findings.