Proceedings of the Annual Meeting of the Cognitive Science Society

The lexical representation of words with multiple pronunciation variants has been widely debated: while single storage accounts propose that all variants of a word are represented by a single, canonical, representation, multiple storage accounts include representations for different pronunciation variants. Previous work has provided evidence for the representation of noncanonical variants, consistent with multiple storage models; however, this work has focused on highly frequent noncanonical variants, which some single storage models propose can also be lexically represented. To test predictions of multiple storage models more rigorously, we examined the representation of low-frequency noncanonical variants of the Persian uvular stop [É¢]. Results from three experiments support the multiple storage model, in that they provided evidence for lexical storage of low-frequency variants. Implications of these findings are discussed.