Long-term auditory memory: Speaker identification.

Abstract

While working on a cover task, 120 female college students overheard a taped female voice from an adjoining room answer a phone and talk in an angry tone for approximately 11 sec. On a test for speaker identification, subjects listened to tape recordings of the target speaker and four female speakers used as foils, all repeating the original message. The target speaker was presented talking in the original hostile tone or talking in a more normal conversational tone. Two of the distractor voices were hostile in tone and two were conversational. Results showed that participants prepared for the recognition test were superior to uninformed subjects in identifying the target. No significant differences were found between subjects tested immediately and those tested after a delay of 24 hours. However, the change in tone of the target's voice reduced identification to chance, even for subjects tested within minutes of first hearing the speaker. In addition, a small but significant point-biserial correlation, .26, between subjects' certainty of response and accuracy of recognition was found. It is concluded that the criminal justice system should treat long-term speaker identification with caution.The pre-Socratic philosopher Heraclitus noted that the "Eyes and ears are bad witnesses for men, since their souls lack understanding" (Donner, Eble, & Helbling, 1967, p. 65). More recently psychologists and criminal justice investigators have documented an extensive body of research indicating that eyewitness identification of faces is extremely fallible (Buckhout, 1974;Goldstein, 1977; Yarmey, 1979b). In contrast, few studies have been conducted on speaker identification (see Bull, 1978), yet testimony of voice recognition often is trusted by both law enforcement agencies and the courts (Tosi, 1979). Our article adds empirical support to the doubts raised by legal critics about this practice, by examining through laboratory simulation three factors inherent in the legal problem of speaker This research is based on a thesis submitted by the first author to the Faculty of Graduate Studies, University of Guelph, in partial fulfillment of the requirements for the master's degree.

Hearing Faces and Seeing Voices: The Integration and Interaction of Face and Voice Processing

“…The results of this study provide the strongest indication yet of a decline in performance when emotional tone changed, irrespective of whether the emotional tone was presented at study (EN) or at test (NE). Consequently, these data confirmed the findings of Saslove and Yarmey (1980) and Read and Craik (1995) and suggested that emotional voices did not produce the levels of performance associated with 'strong' voices, especially when emotional tone changed from one instance to the next.…”

“…However, surprisingly, the results tend to run counter to expectation. For example, Saslove and Yarmey (1980) conducted a voice line-up task. At the initial study phase, participants overheard an angry speaker for 11 seconds.…”

Multiple Independent Identification Decisions: A Method of Calibrating Eyewitness Identifications.

“…M e r 2 weeks, accuracy 4 1 dropped to 48%, then to 47% afier 4 weeks and 45% afier 8 weeks. Saslove and Yarmey (1980) looked at shorter delay intervals. The listeners were initially asked to Iisten to a voice sample that lasted 10 seconds and then were asked to atternpt to identify that voice tiom a total of 4 voices, played one at a time.…”

Analysis of human scream and its impact on text-independent speaker verification

“…The forensic implications of speaker identification under extreme emotional conditions have long been a motivation for research (Saslove and Yarmey, 1980). A significant number of these studies date before many of the recent advancements in speaker recognition.…”