“…A novel finding of the current study lies with the pattern of results for “telling apart.” Both for voice identity sorting and face identity sorting, “telling apart” scores usually show very low error rates, regardless of familiarity with the voices/faces. That is, when forming clusters to represent individual identities, unfamiliar and familiar viewers and listeners rarely combine 2 different identities into the same perceived identity cluster (for voices, see J. Johnson et al, 2020 ; Lavan, Burston, & Garrido, 2019a ; Lavan, Burston, Ladwa, et al, 2019b ; Lavan, Collins, & Miah, 2021a ; Lavan, Smith, & McGettigan, 2022 ; Stevenage et al, 2020 ; for faces, see Jenkins et al, 2011 ; J. Johnson et al, 2018 ; Lavan, Collins, & Miah, 2021a , Lavan, Smith, & McGettigan, 2022 ; Redfern & Benton, 2017 ). Where we have previously seen increased “telling apart” errors in voice sorting studies, these have emerged in direct comparisons of different stimulus sets (i.e., highly expressive clips including whispering, shouting, and emotional speech vs. low expressive conversational speech; Lavan, Burston, Ladwa, et al, 2019b ) or task instructions (i.e., when unfamiliar listeners have been instructed to sort the sounds into a two-identity solution; Lavan et al, 2020 ).…”