| Summary: | In Experiment 1, participants read a long list of nonwords and made ratings of pronunciation variability. Experiment 2 compared speeded naming of nonwords presented alone or mixed with words. In both experiments, at least 85% of nonwords based on inconsistent word bodies were pronounced according to spelling–sound rules, but regular pronunciations were assigned to only 30–40% of nonwords with no regular word-body neighbors. Contradicting the prediction of parallel-distributed-processing models of nonword naming, the probability of an irregular pronunciation was a function of the proportion of irregular word neighbors, not the frequency of these neighbors. The dual-route cascade model overestimated the proportion of regular nonword pronunciations but successfully predicted the particular nonwords most likely to be pronounced irregularly. These results highlight the issues that must be addressed in future refinements of models of nonword naming. Reprinted by permission of the publisher.
|
|---|