Abstrakt
Auditory identification of speakers by unfamiliar lay listeners is notoriously problematic. The construction of a voice parade is therefore, not only for this reason, to be considered only in strictly defined conditions.
When designing a voice parade, it is always recommended that the victim or witness describe the voice of the perpetrator in as much detail as possible. Forensic phoneticians who employ, at least to some extent, auditory analysis during their voice comparison typically rely on an analytical protocol.
A structured, analytical procedure is recommended. Many experts have focused on the auditory analysis of voice quality, with Laver's Voice Profile Analysis still the standard, despite recent attempts to adapt it for forensic purposes.
The aim of this study is to examine whether an adapted analytical protocol could facilitate voice description of the unfamiliar speaker for naïve listeners. We compiled a protocol inspired mostly by Hollien's protocol (2001) and by the SVPA and "translated" it into layman's terms, so that naïve listeners could make informed assessments about the speech and voice characteristics.
To evoke denasalization, for instance, the listener were asked Did the speaker sound as if he had a cold? Attempts at enhancing earwitnesses' description of heard voices have been, to our knowledge, relatively scarce. Öhman et al. (2013) compared the effects of the Cognitive Interview and a Swedish Security Service protocol with a baseline, instruction-free procedure on speaker identification but found no improvement in identification; the protocol was, understandably but unfortunately, not made available. Recordings of four male speakers were selected from the database of Common Czech so that their speech included at least one remarkable characteristic (e.g., a low or high pitch of the voice, a non-neutral phonatory modification, speech rate, a speech defect).
A oneminute reading task was chosen to ensure that the content was identical. The recording of each speaker was played to three listeners; all 12 listeners were students of philological majors like Anglophone studies, but had no experience with describing voices.
The listeners' first task was to listen to the speaker and, after a two-minute pause, to describe his voice in as much detail as possible without any cues from the experimenter (first author). When the listener has finished, the second task consisted in the experimenter guiding them through the adapted protocol.
She described the individual speech and voice characteristics and, for some of them (e.g., breathy voice quality, palatalization), she used illustrations of these qualities previously recorded by the third author instead of imitating them herself to avoid any possible bias. The analysis focused on how consistently the same speaker's voice is described by the three listeners, which speech or voice characteristics are picked up on most readily and, especially, to what extent the guided description with the protocol qualitatively differs from the baseline, uninstructed description.
An improvement in the guided condition, if revealed by our analyses, may be of significance for future voice parade constructions. A preliminary analysis of the results confirms our expectation that the naïve, uninstructed descriptions would be very simple; they usually focus only on two or three characteristics, both phonetic, such as articulation features, and social-indexical, for example gender or age.
Some listeners also described their overall impression, such as "the speaker sounded nervous". However, the listeners were able to use the protocol to make considerably more detailed descriptions, and most of them managed to notice the most salient characteristics of the speakers' voices.