Re-Examining the Effect of Top-Down Linguistic Information on Speaker-Voice Discrimination

The current study replicated and extended the results from a study conducted by Narayan, Mak, and Bialystok (2017) that found effects of top-down linguistic information on a speaker discrimination task by examining four conditions: rhymes (day-bay), compounds (day-dream), reverse compounds (dream-day), and unrelated words (day-bee). The original study found that participants were more likely to judge two words to be spoken by the same speaker if the words cohered lexically (created lexical compounds such as day-dream) or were phonologically related (rhymes, such as day-bay), but their study contained two limitations: (a) Same- and different-speaker trials were analyzed separately, which obscures effects of response bias, and (b) cross-gender pairs were used in the different-speaker trials, potentially inflating performance. The current study addresses these limitations by including only within-gender trials and by examining sensitivity and bias using signal detection theory. Our results not only provide support of the original study but also provide clear evidence that listeners are biased to judge two words as being produced by the same person when they share either phonological information (rhymes) or lexical-semantic coherence (compounds). Thus, the current study provides an important modified replication of previous research.