descriptions of distinctive and non-distinctive voices over time

8
ORIGINAL PAPERS Descriptions of distinctive and non-distinctive voices over time A DANIEL YARMEY Department of Psychology, University of Guelph, Guelph, Ontario, Canada NlG 2W1 The quality of an eyewitness' description of a perpetrator's face is known to be reliably less complete and less detailed, after an interval of 24 hours or more, than descriptions given within an hour of the observation period. In contrast, the quality of earwitness' descriptions of a perpetrator's voice has not been extensively examined. This study tested 295 under- graduate students' rated descriptions of selected voice charac- teristics for either a distinctive voice or a non-distinctive voice heard for 36 seconds and retained for a maximum of one week. Results showed that with one exception, rate of speaking, descriptions of distinctive voices were reliable over a one-week retention period. In contrast, descriptions of non- distinctive voices were more adversely affected by delays in testing. It was concluded that witness descriptions of non- distinctive voices should be given, if possible, within 24 hours of an incident. Key Words: Voice descriptions; Speaker identification; Ear- witness memory. Journal of the Forensic Science Society 1991; 31: 421 -428 Received 17 October 1990; accepted 5 June 1991 Introduction Under ideal eyewitness circumstances, observational conditions are favour- able and allow "optimal" processing of information. Similarly, under ideal interrogation of witness conditions, police record highly detailed, complete and accurate eyewitness reports. In order to construct "fair" lineups and photo-spreads, a necessary condition is that they are constructed from the descriptions a witness first gives to police when memory is strongest and least likely to be influenced by post-event interferences [I, 21. According to the US Supreme Court, the more accurate the initial description of a suspect by a witness, the greater the likelihood of identification accuracy [3]. Observational conditions and witnesses' ability to recall accurately and completely are seldom ideal. Apart from the fact that a number of situational and witness factors may contribute to eyewitness mis- identification, the imprecision of language in which a witness describes what he or she saw and heard may introduce distortions and errors in identifica- tion [4-61. Witnesses' attempts to describe their memory of a suspect's face, JFSS 1991; 31(4): 421 -428 42 1

Upload: a-daniel-yarmey

Post on 02-Jul-2016

216 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Descriptions of distinctive and non-distinctive voices over time

ORIGINAL PAPERS

Descriptions of distinctive and non-distinctive voices over time A DANIEL YARMEY

Department of Psychology, University of Guelph, Guelph, Ontario, Canada N l G 2W1

The quality of an eyewitness' description of a perpetrator's face is known to be reliably less complete and less detailed, after an interval of 24 hours or more, than descriptions given within an hour of the observation period. In contrast, the quality of earwitness' descriptions of a perpetrator's voice has not been extensively examined. This study tested 295 under- graduate students' rated descriptions of selected voice charac- teristics for either a distinctive voice or a non-distinctive voice heard for 36 seconds and retained for a maximum of one week. Results showed that with one exception, rate of speaking, descriptions of distinctive voices were reliable over a one-week retention period. In contrast, descriptions of non- distinctive voices were more adversely affected by delays in testing. It was concluded that witness descriptions of non- distinctive voices should be given, if possible, within 24 hours of an incident. Key Words: Voice descriptions; Speaker identification; Ear- witness memory. Journal of the Forensic Science Society 1991; 31: 421 -428 Received 17 October 1990; accepted 5 June 1991

Introduction Under ideal eyewitness circumstances, observational conditions are favour- able and allow "optimal" processing of information. Similarly, under ideal interrogation of witness conditions, police record highly detailed, complete and accurate eyewitness reports. In order to construct "fair" lineups and photo-spreads, a necessary condition is that they are constructed from the descriptions a witness first gives to police when memory is strongest and least likely to be influenced by post-event interferences [I, 21. According to the US Supreme Court, the more accurate the initial description of a suspect by a witness, the greater the likelihood of identification accuracy [3].

Observational conditions and witnesses' ability to recall accurately and completely are seldom ideal. Apart from the fact that a number of situational and witness factors may contribute to eyewitness mis- identification, the imprecision of language in which a witness describes what he or she saw and heard may introduce distortions and errors in identifica- tion [4-61. Witnesses' attempts to describe their memory of a suspect's face,

JFSS 1991; 31(4): 421 -428 42 1

Page 2: Descriptions of distinctive and non-distinctive voices over time

for example, may fail to capture the richness of detail of their mental representations. Ellis, Shepherd and Davies [7] found that subjects' verbal descriptions of two faces were significantly shorter, less detailed, and less likely to lead to an accurate identification after a 24-hour interval or more, than those descriptions made one hour after seeing the faces. Clearly, accuracy and completeness of recall memory for faces declines over time.

Although verbal descriptions of suspects' faces are an important part of police investigations there are some crimes, such as obscene phone calls, ransom demands, hooded rape, or crimes committed in darkness, in which witnesses' memory of the perpetrator's voice is the only evidence available to guide the police investigation [8,9]. Analyses of subjects' descriptions of perpetrators' voices in several experimental studies (eg, [lo, 111 indicate that suspects' voices are very difficult to describe. Subjects typically describe them in very brief and incomplete ways, using general characteristics such as soft, hoarse, gentle, mellow, pleasant, calm, etc. Similarly, victims of sexual assault in hooded rape cases or crimes committed in darkness (eg, People v Robin Scott Dasenbrock [12]) describe the voice of their attacker in very non-specific, general terms. Nevertheless, voice identification testimony is treated as direct evidence of identity and has been accepted by courts of law for several decades [13-151.

One purpose of the present study was to test whether the reliability of victim-subjects' descriptions of suspect voices from a simulated abduction scenario changes as a function of delay. Rated descriptions of voice characteristics, in contrast to free descriptions, were used because ratings can be efficiently administered and are quantitatively representative of specific voice characteristics, such as rate of speech, tremor, pauses, etc [16,17]. The standard or baseline condition in this study was subjects' ratings of voice quality information completed when the suspect voice was perceptually available. An issue of interest, therefore, was to see which, if any, rated characteristics constructed from memory significantly change over time relative to ratings given in the perception condition.

A second purpose of this study was to investigate the influence of distinctiveness of suspects' voices on verbal descriptions over time. The facilitating role of distinctiveness of facial encoding on memory has been thoroughly investigated [18-201, but distinctiveness has received little attention by researchers interested in voice memory [9,21,22]. It is probable that voices varying in distinctiveness also differentially influence memory. Voices that are highly distinctive may be intrinsically easier to remember, or may attract heightened attention which facilitates both the encoding and retrieval operations involved in verbal descriptions. It was predicted that verbal descriptions of non-distinctive voices, in contrast to distinctive voices, would be more adversely affected by time. However,

422 JFSS 1991; 31 (4): 421-428

Page 3: Descriptions of distinctive and non-distinctive voices over time

there are no empirical or theoretical reasons to make differential predictions for the reliability of specific ratings over time for the ten selected voice characteristics.

Experimental

Subjects A total of 295 introductory psychology students volunteered for this study and received course credit for their participation. All subjects were Caucasian and none reported hearing disabilities.

Materials Separate groups of subjects served as judges to rate 40 tape-recorded male voices for distinctiveness. Voices were defined as either distinctive or non-distinctive based on their extremity of rated characteristics. Voices were rated on a ten-point scale with end points of "not at all distinctive" and "very distinctive". Four suspect voices were selected from these ratings. Two voices consistently rated 7 or above were randomly selected as "distinctive" and two voices rated consistently between 3 and 6 were randomly selected as "non-distinctive" from the judges' mean ratings. All voices were played on a Sony BM-12 tape recorder at a fixed volume level.

Procedure Subjects were randomly assigned to one of four delay conditions: a perception condition; an immediate memory condition; a 24-hour delay condition; and a one-week delay condition. Subjects were tested in groups varying in size from two to five with one of the four suspect voices. Upon entering the laboratory, subjects were seated and given the following instructions:

You are the daughter or son of a wealthy businessman and you have just been kidnapped. You did not see your assailant as he approached you from behind and he was wearing a mask. You were blindfolded and gagged and taken to an unfamiliar building where you were bound to a chair. Once you assess the situation, you realize that you will only be able to help the police if you can recognize your kidnapper's voice. Therefore, you are to listen very closely to the following tape-recorded monologue.

Except for subjects in the perception conditions, subjects were not forewarned about the testing procedures for recall of voice descriptions. All participants in the three memory conditions heard the tape recording for a total of 36 seconds. After listening to the tape, subjects in the memory conditions were either tested immediately for their ratings of voice characteristics, or were asked to return to the laboratory 24 hours later, or one week later to complete the study. Subjects in the perception condition

JFSS 1991; 31 (4): 421 -428 423

Page 4: Descriptions of distinctive and non-distinctive voices over time

were allowed to listen to the abductor's voice as they simultaneously completed their ratings.

Ten voice characteristics, selected from the "Voice Profile Check List" developed by Handkins and Cross [16], were rated on 10-point scales, as shown in Table 1. The Voice Profile Check List was constructed from written descriptions given by 50 listeners for five highly distinctive voices. Composite descriptions were then used by 20 new listeners to identify, within a series, the voice described. Ratings of these voices were averaged to produce linearly scaled descriptors of perceived voice quality [13].

TABLE 1. Analysis of rated characteristics which showed significant differences for two distinctive and two non-distinctive voices

Voice characteristic

Scale Distinctive Non-distinctive 1-10 Ff1 , 94) P Ff1 , 18-51 p

Rate of speech Rate variation

Pitch Expressive style

Age Enunciation Inflection Tremor Pauses Nasality

very slow-very fast smooth flowing, steady- halting jerky, irregular monotone-highly varied none-very expressive 18-40 years very poor-very good none, flat-very much very steady-very shaky very short-very long none-very nasal

Results The judges' ratings for voice distinctiveness were confirmed. Distinctive voices differed significantly from non-distinctive voices on each of the ten characteristics at the 0.01 level of confidence. A separate analysis of the rated characteristics (Table 1) showed that, as expected, the two randomly chosen distinctive voices differed significantly from each other on several characteristics. The two non-distinctive voices differed significantly from each other on fewer characteristics. There were no significant differences in inflection or tremor between the two distinctive voices and no significant differences in rate variation, pitch, expressive style, enunciation, inflection or pauses between the two non-distinctive voices.

Differences in voice characteristics over time Subject's mean ratings for the ten voice characteristics as a function of the time of test, for the distinctive voices and for the non-distinctive voices, are presented in Table 2.

Among the distinctive voices, significant differences for delay of test were

424 JFSS 1991; 31(4): 421-428

Page 5: Descriptions of distinctive and non-distinctive voices over time

found for only one voice characteristic, rate of speaking (F(3,94) = 3.75, p < 0.01). Subjects in the perception condition rated the distinctive voices as moderately fast in rate of speech, which differed significantly from the one-week delay group and the immediate memory group who recalled the voices as slower in rate of speech ( p < 0.05, Newman-Keuls).

TABLE 2. Mean ratings of voice characteristics over delay conditions for distinctive and non-distinctive voices

Voice Delay conditions characteristics Perception Immediate 24-hour 1 week

Distinctive Rate Rate variation Pitch Expressive style Age Enunciation Inflection Tremor Pauses Nasality

Non-distinctive Rate Rate variation Pitch Expressive style Age Enunciation Inflection Tremor Pauses Nasality

Analysis of variance on ratings for the non-distinctive voices revealed reliable differences over time for five voice characteristics.

Rate. Non-distinctive voices were recalled over time as being slower in rate of speech (F(3,185) = 7.02, p < 0.001). No significant differences were found between the perception group and the immediate memory group but both conditions differed reliably from the one-week group. The 24-hour delay group also differed significantly from the one-week delay group.

Since rate also differed for the distinctive voices, analysis of variance was conducted combining ratings for the distinctive and non-distinctive voices. Distinctive voices were rated significantly slower in rate of speech than non-distinctive voices (F(1,287) = 18.60, p 0.01). The main effect for time

JFSS 1991; 31(4): 421-428 425

Page 6: Descriptions of distinctive and non-distinctive voices over time

also was significant (F(3,287) = 6.83, p < 0.01). The one-week delay groups recalled the voices reliably slower than all other groups. The interaction effect between distinctiveness and time was not significant (F(3,287) = 1-98, p > 0.05).

Rate variation. A significant effect for time in the recall of rate variation (F(3, 185) = 6-02, p < 0.001) was qualified by a reliable interaction effect with the two different voices (F(3, 185) = 3-44, p < 0.05). In contrast to the perception group, who perceived non-distinctive Voice 1 as relatively smooth flowing and steady, all of the memory groups recalled Voice 1 as more halting, jerky, and irregular over time. A reliable increase in irregularity of speech was also found between the one-week delayed group and the 24-hour group. For non-distinctive Voice 2, the perception group did not differ from the one-week delayed group but did differ reliably over time from the immediate memory group and the 24-hour delayed group.

Tremor. Ratings of voice tremor increased over time (F(3, 185) = 6.18, p < 0.001). Whereas the perception group perceived the voices as relatively steady, the 24-hour and the one-week delayed group recalled reliably greater shakiness in the voices. The one-week group also differed sig- nificantly from the 24-hour group.

Pauses. Subjects recalled longer pauses in speech over time (F(3, 185) =

11-83, p <0-001). Significant differences in recall of pauses were found between the one-week delayed group and all other groups.

Nasality. Non-distinctive voices were recalled with greater nasality over time ( F ( 3 , 185) = 8.60, p <0.0001). The perception group differed significantly from all other groups and the one-week delay group also differed sig- nificantly from the other two memory conditions.

Discussion The results of this study support the hypothesis that verbal descriptions of non-distinctive voices in contrast to distinctive voices are more adversely affected by the length of the retention interval. In contrast to ratings made in an immediate perception condition, descriptions of non-distinctive voices made over a one-week interval showed a reliable decrement on five of the ten voice quality characteristics. Specifically, rated descriptions of non- distinctive voices changed reliably over time for rate of speech, rate variation, tremor, pauses, and nasality. Furthermore, all of these five voice characteristics showed a reliable change in memory between the 24-hour test condition and the one-week test condition which suggests that wit- nesses' descriptions should be given, if possible, within 24 hours of an incident. Nasality proved to be a very poor voice descriptor since the perception condition differed reliably from all of the memory conditions, including immediate memory. Ratings of pitch, expressive style, age,

426 JFSS 1991; 31(4): 421-428

Page 7: Descriptions of distinctive and non-distinctive voices over time

enunciation, and inflection were similar to the ratings made by the perception group and did not change significantly over time. These results suggest that these latter voice characteristics may be considered reliable descriptors when non-distinctive voices are recalled over a one-week interval.

For distinctive voices the only rated characteristic to show significant changes over one week was rate of speaking. This finding suggests that within a one-week period most descriptions of distinctive voices will be reliable. However, this study showed, as expected, that distinctive voices vary considerably in voice quality from each other and from non-distinctive voices. Similarly, non-distinctive voices differ considerably from each other in voice quality characteristics. .These results have implications for the construction of voice lineups by law enforcement officials. When choosing foils for voice lineups, police must be aware that not all witnesses' descriptions of voice characteristics of the perpetrator upon which the lineup should be based will be as reliable as others. Furthermore, police must be sensitive to both the variability among voice characteristics across individuals who may have either distinctive or non-distinctive voices, and to the variability among police officers who must listen to the suspect's voice and select matching foils. (Brigham and Ready [23] discuss potential bias in lineup construction). Certainly, foils should fit the same general description as the suspect, as given by witnesses. But, in addition, less confidence can be given to statements about rate, rate variation, tremor, pauses, and nasality, at least for suspects having non-distinctive voices.

Finally, it must be noted that any generalizations made from this ex- perimental study are restricted to non-stressful events that are approxim- ately a half minute in duration, and to events in which the witness is set to try to remember the suspect's voice.

Acknowledgments This research was supported by a grant from the Social Sciences and Humanities Research Council of Canada. The author gratefully acknowledges the assistance of Eva Matthys.

References 1. Wells GL. Eyewitness Identification: A System Handbook. Toronto: Car-

swell, 1988. 2. Yarmey AD. Understanding Police and Police Work: Psychosocial

Issues. New York: New York University Press, 1990. 3. Neil v Biggers, 409 US 188 (1972). 4. Ellis HD. Practical Aspects of Face Memory. In: Wells GL and Loftus EF,

eds. Eyewitness Testimony: Psychological Perspectives. New York: Cam- bridge University Press, 1984: 12-37.

5. Pigott M and Brigham JC. Relationship between accuracy of prior description and facial recognition. Journal of Applied Psychology 1985; 70: 547-555.

JFSS 1991; 31(4): 421-428 427

Page 8: Descriptions of distinctive and non-distinctive voices over time

6. Wells GL. Verbal descriptions of faces from memory: are they diagnostic of identification accuracy? Journal of Applied Psychology 1985; 70: 619-626.

7. Ellis HD, Shepherd JW and Davies GM. The deterioration of verbal descriptions of faces over different delay intervals. Journal of Police Science and Administration 1980; 8: 101-106.

8. Bull R and Clifford BR. Earwitness voice recognition accuracy. In: Wells GL and Loftus EF, eds. Eyewitness Testimony: Psychological Perspectives. New York: Cambridge University Press, 1984: 92-123.

9. Kiinzel HJ. Phonetische Untersuchungen zur Spechererkennung durch lin- guistisch naive Personen. Stuttgart: Franz Steiner Verlag, 1990.

10. Yarmey AD. Verbal, visual and voice identification of a rape suspect under different levels of illumination. Journal of Applied Psychology 1986; 71: 363-370.

11. Yarmey AD and Matthys E. Voice identification of an abductor. Applied Cognitive Psychology 1991; in press.

12. People v Robin Scott Dasenbrock, Orange County, California, Superior Court Case No C-66882, 1990.

13. Deffenbacher KA, Cross JF, Handkins RE, Chance JE, Goldstein AG, Hammersley R and Read JD. Relevance of voice identification research to criteria for evaluating reliability of an identification. The Journal of Psychology 1989; 123: 109-119.

14. Hollien H. The acoustics of crime. New York: Plenum, 1990. 15. Tosi 0 . Voice Identification: Theory and Legal Applications. Baltimore:

University Park Press, 1979. 16. Handkins RE and Cross JF. Can a voice line-up be too fair? Presented at

the meeting of the Midwestern Psychological Association, Chicago, IL, May 1985.

17. Voiers WD. Perceptual bases of speaker identity. Journal of the Acoustical Society of America 1964; 36: 1065-1073.

18. Cohen ME and Carr WJ. Facial recognition and the von Restorff effect. Bulletin of the Psychonomic Society 1975; 6: 383-384.

19. Going M and Read JD. Effects of uniqueness, sex of subject and sex of photograph on facial recognition. Perceptual and Motor Skills 1974; 39: 109-110.

20. Light LL, Kayra-Stuart F and Hollander S. Recognition memory for typical and unusual faces. Journal of Experimental Psychology: Human Learning and Memory 1979; 5: 212-228.

21. Clifford BR, Rathborn H and Bull R. The effects of delay on voice recognition. Law and Human Behavior 1981; 5: 201-208.

22. Thompson CP. Voice identification: speaker identifiability and a correction of the record regarding sex effects. Human Learning 1985; 4: 19-27.

23. Brigham JC and Ready DJ. Own-race bias in lineup construction. Law and Human Behavior 1985; 9: 415-424.

JFSS 1991; 31(4): 421-428