Hearing voices: forensic speaker identification technology and expert listening in the American courtroom

ABSTRACT

Police wiretaps and taped emergency dispatch calls are just a couple of examples of the kinds of voice recordings that have made their way into criminal and civil proceedings. In some instances, an expert witness may be called upon to identify the person whose voice was captured on tape or digitally recorded. However, this type of forensic analysis – commonly referred to as “speaker identification” – has not been universally accepted by the courts. In this article, I look at several US cases where the efficacy of forensic speaker identification has been brought into question. Using concepts from Science and Technology Studies (STS) and Sociolegal Studies, I examine the attempts made by experts to have their methods of voice identification accepted at trial as valid and reliable techniques, and the decisions made by judges to either admit or exclude this evidence. I demonstrate that the various rulings regarding the admissibility of speaker identification evidence reflect the interplay between law and science and is the direct result of the “boundary-work” undertaken by experts and how judges assess these activities. I argue that forensic speaker identification evidence must be understood and conceptualised as “law-science hybrids” that are co-produced over the course of a trial.

KEYWORDS:

• Voice
• speaker identification
• boundary-work
• experts
• courtroom

Acknowledgments

I am indebted to Karin Bijsterveld and Anna Kviaclova for inviting me to contribute to this Special Issue and for all of their hard work in putting this collection together. I would also like to thank the participants of the Forensic Voices workshop, held at the University of Maastricht, for providing me with constructive feedback on an earlier draft of my work. Lastly, I would like to thank the journal editor and the two anonymous reviewers, whose insightful comments and suggestions have greatly improved this article.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. For example, when calling a customer service line, it is now common to hear a recorded message notifying us that the call may be recorded for training and quality assurance purposes.

2. Much of the material that has found its way into criminal trials comes from audio surveillance recordings where the question of what was said (i.e. analysis of disputed utterances) and/or by whom (i.e. speaker identification) are commonly at issue.

3. Following Edmond, Martire, and San Roque (2011), I have chosen to put the word “experts” in scare quotes to draw attention to the fact that the ability of these witnesses to provide reliable opinions about identity is deeply contested.

4. In the US, the Audio Engineering Society (AES), the International Association for Identification (IAI), and the Scientific Working Group on Digital Evidence (SWGDE) are just a few of the professional organisations that have set professional standards and best practices guidelines for audio forensic experts. The International Association for Forensic Phonetics and Acoustics, highly important in Europe, also has a Code of Practice (see https://www.iafpa.net/about/code-of-practice/). However, none of these groups appear to have specific protocols for speaker or voice identification.

5. As other scholars have pointed out (see, for example, Cole 2002; Jasanoff 1995), challenging the admissibility of forensic evidence requires that lawyers be knowledgeable about how this evidence is collected and analysed, and the various problems and controversies surrounding this practice.

6. To be able to do this, lawyers must know which forensic experts to contact and how to find them. The legal party must also be willing and able to pay for the services of the forensic expert (see, for example, Cole 2002; Jasanoff 1995).

7. The US Supreme Court’s decision in Kumho Tire Co. v. Carmichael (1999) requires judges to apply the admissibility criteria set out in Daubert to both “scientific” and “non-scientific” expertise. Thus, simply claiming that an expert’s opinion is based on experience with no basis in science can no longer be used as a loophole out of the application of the Daubert standard.

8. The spectrograph is a device that produces graphic illustrations of the acoustic signals produced by a sound source called a “spectrogram”.

9. The format of the Tosi study was designed to test varying conditions which could be expected to have an impact on the reliability of the technique in forensic setting. Most notably, unlike Kersta’s experiment, Tosi conducted “open trials” where the known speaker population against which the unknown spectrogram is being compared may or may not contain the author of the known spectrogram. This was thought to be a much closer approximation to real-world conditions as an examiner would never know whether the author of the unknown spectrogram was included in the known speaker population (see Tosi et al. 1972).

10. Although Kersta never utilised aural comparisons in his work (Poza and Begault 2005, 2), it is now incredibly rare for an examiner in this field to rely solely on a visual analysis (i.e. “spectrographic” analysis) of voice spectrograms.

11. While it is important to acknowledge the important role that race and racism played in this case, this discussion is beyond the scope of this paper.

12. “911” is the telephone number used in the United States and Canada to make emergency dispatch calls for police, fire, and ambulance services.

13. At the time of this hearing, Florida was still using the Frye standard. However, during the trial, Florida switched to the Daubert standard which took effect on 1^st July 2013.

14. The defence called on Drs. George Doddington, Peter French, Hirotaka Nakasone, and James Wayman as expert witnesses (State of Florida v. Zimmerman 2012).

15. Dr. French testified that it would be extremely difficult to analyse voices by comparing screaming with a normal voice. During the admissibility hearing, he stated: “I’ve never come across a case in my 13 years where anybody’s tried to compare screaming to a normal voice” (State of Florida v. Zimmerman 2012).

16. This is in addition to allowing jurors to hear the evidence for themselves to determine whether they can identify a match between the recordings in questions.

Additional information

Notes on contributors

Michael Mopas

Michael Mopas is Associate Professor in the Department of Sociology & Anthropology, Carleton University. He specializes in the area of science, technology, and law. Michael’s primary research examines the place of sound in the courtroom. His SSHRC-funded study explored the ways that judges hear and make sense of auditory evidence presented at trial and documented the work of audio forensic experts in rendering them audible. Michael is currently preparing a book manuscript based on this research. Michael has written about rural noise and its regulation and has co-authored a piece, with Amelia Curran, on forensic musicology in the courtroom. Both articles can be found in the Canadian Journal of Law and Society.