Success in spoken word processing relies not only on accurate word recognition but also the veracity with which words are maintained in memory. However, research on word retention is still scarce, especially in tonal languages and phonologically impaired populations. To address these gaps, the present study administered an auditory order recall task to native Cantonese speakers with and without amusia. Stimuli intrinsic (segmental similarity, suprasegmental similarity, and lexicality) and extrinsic (cognitive load) factors were manipulated. As expected, word recall was significantly impaired by increased cognitive load, and, similar to the retention of non-tonal languages, rime similarity exerted opposite effects on Cantonese real and pseudo word recollection. However, no reversal was observed in the suprasegmental dimension: Lexical tone similarity was persistently detrimental to recall accuracy. Tonal similarity Effects were also more robust against increased cognitive load than segmental similarity effects, implying differential weighting of suprasegmental and segmental cues in spoken word retention. These results are consistent with the view that distinct processing mechanisms exist for segments and tones. Besides, this study found it useful to combine accuracy and response time analyses. The potential of the combined analysis to capture the nature of phonological deficits is discussed in relation to amusia.

Index Terms: spoken word retention, phonological similarity, cognitive load, phonological deficit, amusia