Using criterio voice familiarity to augment the accuracy of speaker identification in voice lineups

Voice familiarity is a principal factor underlying the apparent superiority of human-based vs machine-based speaker identification. Our study evaluates the effects of voice familiarity on speaker identification in voice lineups by using a Familiarity Index that considers (1) recency (the time of last spoken contact), (2) duration of spoken contact, and (3) frequency of spoken contact. Three separate voice-lineups were designed each containing ten male voices with one target voice that was more or less familiar to individual listeners (13 per lineup, n = 39 listeners in all). The stimuli consisted in several verbal expressions varying in length, all of which reflected a similar dialect and the voices presented a similar speaking fundamental frequency to within one semitone. The main results showed high rates of correct target voice identification across lineups (>99%) when listeners were presented with voices that were highly familiar in terms of all three indices of recency, duration of contact, and frequency of contact. Secondary results showed that the length of the verbal stimuli had little impact on identification rates beyond a four-syllable string.