Talker Identification Learning and New Jersey Dialect Discrimination

Previous studies have demonstrated that individuals can be trained to identify a speaker by voice. Other research has found profound regional differences in dialect across the US, with individuals being able to generally discriminate between them at above chance levels. However, little research has investigated the ability to discriminate between two similar dialects or the effect of using bisyllabic words in talker training. The present study aims to explore the patterns of talker learning and dialect discrimination which arise from training talkers on the two predominate dialects spoken in New Jersey. To investigate such factors, the current research trained 24 listeners (12 female) over several days to learn to identify speakers grouped across dialect conditions by voice. Participants were trained on a list of 80 bisyllabic words. Before and after training, listeners were tested on ability to discriminate between Northern and Southern New Jersey dialects. Results displayed a significant overall increase in ability to identify talkers over the course of training, showing a pattern of learning largely consistent with previous findings. However, no correlation between talker identification and dialect discrimination ability was found. Dialect discrimination tasks displayed no difference in score before and after training. Rates were significantly below chance levels of 50%. No evidence of a listener response bias was found, allowing the possibility that listeners were consistently discriminating between dialects but using incorrect labeling. This study fills a gap in training research in regards to bisyllabic words and provides a foundation for future research investigating discrimination between neighboring dialects