Earwitness descriptions and speaker identification

Abstract

Some 160 men and women selected from public locations agreed to participate in a voice identification experiment. Participants were instructed to listen carefully to the tape-recorded voice of a perpetrator committing a simulated armed robbery of a business establishment. Two minutes later they were asked to describe the voice characteristics of the perpetrator, to recall exactly what he said, and then attempt to identify the speaker from a six-person perpetrator-present or perpetrator-absent voice line-up. Half of the participants in each line-up heard a sample of identical phrases and the other half heard phrases non-identical to those used in the robbery. Accuracy of speaker identification was significantly better than chance; however, there were no significant differences in performance on either line-up as a function of the type of voice sample employed. The confidence-accuracy of identification correlation proved to be non-significant. No significant correlations were found between accuracy of speaker identification and completeness of voice descriptions, or speaker identification and percentage accuracy of recall of actual words used by the perpetrator, or speaker identification and percentage accuracy of recall of idea units contained in the perpetrator’s monologue. It was concluded that voice lineups should be constructed of non-identical phrases rather than the identical phrases reportedly used by the perpetrator.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 74 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.