Zürich Voice Super-Recognizer Test (ZVSRT): Normative Data from Law Enforcement Personnel on Voice Discrimination and Sorting

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

In 2009, Russell et al. identified four “Super-Recognizers” (SRs) with exceptional face processing abilities. The potential existence of voice Super-Recognizers (VSRs) has been discussed both scientifically and related to high-stakes areas like law enforcement, which they could in principle support. Existing methods to formally assess voice identity processing (VIP) are both limited, and oftentimes lack the sensitivity required to reveal superior performance. Moreover, their de facto applied value, e.g. for forensic speaker comparison remains empirically unaddressed. To fill this gap, and drawing on findings from research on (face) SRs, we propose a preliminary working definition for VSRs and report normative data for a novel test battery designed to assess superior VIP: the Zürich Voice Super-Recognizer Test (ZVSRT). Modelling potential VSR deployment scenarios, the ZVSRT involves three difficult perceptual tasks. Its stimuli – derived from a forensic corpus – were selected using a deep learning–based open-source automatic speaker recognition system to ensure the creation of challenging trials befitting its intended purpose. The reported ZVSRT normative data was obtained from a cohort of Swiss law enforcement personnel (N = 204), who completed its constituent components: a voice discrimination, and two voice sorting tasks. These data demonstrate the ZVSRT sensitivity towards capturing the full range of VIP ability – from very low to high performance levels. We discuss future directions, including the utility of the ZVSRT and VSRs’ potential relevance for forensic applications.