Towards a screening test for earwitnesses

Abstract

The present study explores the feasibility of a screening test for earwitnesses. The underlying assumption is that lay listeners differ in their voice-processing capabilities and might consequently not be equally suited for a standardised voice parade. One hundred British participants took part in an online AX discrimination task with the aim of obtaining a representative sample of the population. For the stimuli, two 10s-long recordings were taken from 48 speakers of the DyViS corpus. Participants differed markedly in recognition accuracy (mean = 75%, range = 50–93.8%). Two potential ‘super recognisers’ were identified as well as four participants at the opposite end of the spectrum. The test serves to establish a baseline for more complex investigations of witness-dependent estimator variables.