Decided: November 13, 1972.
STATE OF NEW JERSEY, PLAINTIFF-APPELLANT,
THOMAS ANDRETTA AND ARMAND FAUGNO, DEFENDANTS-RESPONDENTS
For reversal -- Chief Justice Weintraub, Justices Jacobs, Proctor, Hall and Mountain, and Judges Conford and Sullivan. For affirmance -- None. The opinion of the Court was delivered by Proctor, J.
[61 NJ Page 545]
This case brings the method of voice identification through the use of the spectrogram, the so-called voiceprint, before us once more. We previously considered this method in State v. Cary, 49 N.J. 343 (1967), on remand 99 N.J. Super. 323 (Law Div. 1968), remanded again 53 N.J. 256 (1969), aff'd 56 N.J. 16 (1970).
On March 31, 1967, the Middlesex County Grand Jury returned indictments against the defendants for threatening to do bodily injury to one Albert Soffer unless the sum of $7,400 should be paid to the defendants, in violation of N.J.S.A. 2A:105-4. Soffer subsequently informed the trial court that he would refuse to testify, even if cited for contempt, because he was fearful for his life and the lives of his family. The State alleges that the police recorded a
[61 NJ Page 546]
telephone conversation made in March, 1967 by Soffer from police headquarters. During that conversation, incriminating statements were allegedly made by the other party or parties to the conversation. The State seeks to compare the defendants' voices with the voice or voices engaged in the conversation through the use of the voiceprint method, and consequently moved for an order compelling defendants to submit to a voiceprint test. After receiving testimony from several witnesses, the trial judge denied the request, holding that the State had failed to meet its burden of establishing general scientific acceptance of the voiceprint method. The judge further found that the length of the five-year time span here involved -- from the recording of the original telephone conversation to the then current date -- precluded use of the method in any case. This Court granted the State's motion for leave to appeal before consideration by the Appellate Division.
Voiceprint analysis is a method of sound identification which utilizes the spectrograph machine. This machine decomposes the sound of the human voice into frequency components which are graphically recorded, thus producing the spectrogram or voiceprint. Voiceprints are then compared by persons trained in the use of the method for possible identification purposes. Although the machine was invented during World War II, it was not until the 1960's that Lawrence Kersta conducted experiments in spectrography at Bell Telephone Laboratories which convinced him that there were unique characteristics of each person's voice which could be used for reliable identification of individuals by spectrograms.
At the time we previously considered this development in State v. Cary, 49 N.J. 343 (1967), the scientific support for the validity of this voice identification method rested solely on the testimony of Kersta. We held that the testimony of one man alone would not be sufficient evidence to establish the reliability of the method's results. We believed it was unreasonable to compel the defendant to submit
[61 NJ Page 547]
to the test until more evidence was produced in support of the reliability of the method for accurate identification. The case was consequently remanded to the trial court for further testimony concerning the method's reliability. We said:
"An actual demonstration which shows the accuracy of the voiceprint process would be desirable. The testimony of experts in addition to Mr. Kersta who believe that voiceprint identification is scientifically sound would be helpful. Although we do not wish to dictate in advance the details of the hearing, we do feel that something more than the bare opinion of one man, however qualified, is required." 49 N.J. at 352.
On remand the State was unable to produce expert testimony to support the opinion of Kersta; the trial court held that testimony based on the voiceprint method of identification would not be admissible at trial, and therefore declined to compel the defendant to submit to the test. 99 N.J. Super. 323 (Law Div. 1968). We remanded the case once more in order to give the State additional opportunity to produce expert testimony, 53 N.J. 256 (1969), but it was unable to do so, and we affirmed, 56 N.J. 16 (1970).
Our reluctance to accept the voiceprint method in Cary on the basis of the testimony of Kersta alone was later supported by scientists who criticized his experiments. See Ladefoged and Vanderslice, "The 'Voiceprint' Mystique," Working Papers in Phonetics (U.C.L.A. 1967); Bolt, and five others, "Speaker Identification by Speech Spectrograms: A Scientists' View of its Reliability for Legal Purposes," 47 Journal of Acoustical Society of America 597 (1970). The criticisms of Kersta were directed to his conclusion that the voiceprint method was sufficiently reliable for legal purposes. Kersta's critics believed that the restricted nature of his experiments produced insufficient results for a conclusion as to the method's reliability, and they suggested a need for further experiments. See, e.g., Bolt, et al., supra at 603.
[61 NJ Page 548]
At the hearing below several prominent scientists testified. Dr. Oscar Tosi, who testified for the State, is a professor of audiology and speech sciences at Michigan State University. Dr. Tosi described the extensive experiments he had conducted from 1968 to 1970 at that university under a grant from the United States Department of Justice. Dr. Tosi's main objective was to meet some of the criticisms of Kersta's conclusions by the use of a broader experimental approach which would be more relevant to a determination of the method's reliability in legal applications. The experiments involved 250 speakers randomly selected from a population of about 25,000 male students. Approximately 35,000 experimental tries at identification through the visual comparison of voiceprints were made. See generally, O. Tosi, et al., Voice Identification Through Acoustic Spectrography (1971). On the basis of this study and Dr. Tosi's additional experience as a consultant to the Michigan State Police Voice Identification Unit, Dr. Tosi testified that he has concluded that the method when used by qualified examiners is highly reliable and could be used for legal purposes. This opinion was premised on the assumptions (a) that the examiners would hear the voices being compared as well as inspect and compare the respective spectrograms; and (b) that the examiners could reject those cases where their examination left them uncertain of the identification.
However, Dr. Tosi conceded that his studies showed a significantly greater degree of error when the compared recordings were made a month apart, and he admitted that further experimentation as to this factor was advisable. He did not think any added time lag over a month would be materially significant.
Much of the testimony below concerned the reaction of other scientists to Dr. Tosi's study. Dr. Peter Ladefoged, a professor of phonetics at U.C.L.A., also testified for the State. Dr. Ladefoged, who had been a severe critic of Kersta, see Ladefoged and Vanderslice, supra, had previously testified against the admissibility of the method in other
[61 NJ Page 549]
courts. See State ex rel. Trimble v. Hedman, 291 Minn. 442, 192 N.W. 2 d 432 (Minn. 1971); State v. Cary, supra, 99 N.J. Super. at 330; People v. King, 266 Cal. App. 2d 437, 72 Cal. Rptr. 478, 489 (Dist. Ct. App. 1968). He has since changed his opinion and testified at the hearing below that he believed that Tosi's study and other subsequent data support the method's reliability for its use in legal proceedings.*fn1 Dr. Ladefoged, as did Dr. Tosi, also stated that the voiceprint method today has scientific acceptance. However, their conclusion was apparently based on the lack of criticism by other scientists of the Tosi study, and not on express declarations of acceptance.
Lieutenant Ernest Nash, head of the Voice Identification Unit of the Michigan State Police, also testified for the State. He has had extensive practical experience in use of the voice print method in criminal investigations, and has testified for the prosecution in cases where he has made positive identifications by the use of voiceprints. See, e.g., United States v. Raymond, 337 F. Supp. 641 (D.D.C. 1972); State ex rel. Trimble v. Hedman, supra. In his procedure, Lieutenant Nash listens to the recorded voices in addition to making a visual comparison of voiceprints, a technique which both Dr. Tosi and Dr. Ladefoged say produces the best results. In Lieutenant Nash's opinion, the method is "extremely reliable."
Dr. Peter Denes and Dr. James Flanagan were called as witnesses by the court. Dr. Denes has been engaged since 1947 in speech research work and is now associated with the Bell Telephone Laboratories. He had been one of the authors of the article cited above which had criticized the conclusions of Kersta. See Bolt, et al., supra. Dr. Denes believed that Dr. Tosi's work was a valid scientific experiment
[61 NJ Page 550]
which advanced the knowledge in the field, but it was only a "small step in the direction of having a general knowledge about how people can do this." Although he doubted that he or his co-authors of the article would change their conclusions concerning the admissibility of the method in legal proceedings, Dr. Denes stated that he could not express a strong opinion since he had not made a careful examination of recent developments. Dr. Flanagan, who is director of the Acoustics Research Department of Bell Laboratories, also considered Dr. Tosi's work to be a carefully done laboratory experiment. Although he said he had not reviewed the Tosi report carefully, he was of the opinion that it was not scientifically acceptable to generalize Dr. Tosi's results to unknown and unspecified field conditions.
Since we last reviewed the voiceprint method in the Cary case two courts have accepted the use of the method. In State ex rel. Trimble v. Hedman, supra, a telephone call was received at police headquarters asking for assistance at a specified address. When a policeman arrived at the location, he was shot dead by an unknown assailant. The call was recorded on tape, and thereafter the police obtained a recording of the voice of the defendant. After the recordings were sent to Lieutenant Nash of the Michigan State Police for voiceprint analysis, arrest and search warrants were issued on the basis of Nash's opinion that the voice of the unknown caller and that of the defendant were the same. The Minnesota Supreme Court, after noting developments in the field including Dr. Tosi's study, held that voiceprint identification would be admissible "at least for the purpose of corroborating opinions as to identification by means of ear alone." 192 N.W. 2 d at 441. Although the Trimble case dealt with whether the evidence produced by voiceprint identification would be sufficient to justify issuance of arrest and search warrants, the court stated that "we are convinced that in the trial of the case spectrograms ought to be admissible for the purpose of corroborating voice identification by aural means if a sufficient foundation is laid
[61 NJ Page 551]
to satisfy the trial judge that the expert whose opinion is sought is qualified to assist the factfinder in coming to the right conclusion." 192 N.W. 2 d at 441.
United States v. Raymond, supra, involved a situation similar to that of the Trimble case, where a policeman was ambushed after responding to a telephone call for assistance. After reviewing developments in the method including the experiments of Dr. Tosi and the work of Lieutenant Nash, the court admitted voiceprints prepared by Lieutenant Nash into evidence. The court noted that its ruling did not "imply that such evidence is mistake-proof or that any voice identification should be admitted," but found that on the detailed record it had before it the particular spectrographic identification in that case was sufficiently reliable to be introduced into evidence. 337 F. Supp. at 645 (emphasis in original).
Certainly the voiceprint method today has much more support for its admissibility as evidence than at the time of Cary. Dr. Tosi's study increases the knowledge we have of this method's reliability, and the admission into evidence of Lieutenant Nash's identifications in Trimble and Raymond demonstrates growing judicial acceptance. However, we need not decide at this time whether results of voiceprint analysis will be routinely admissible at trial. The narrow issue before us is whether the defendants should be compelled to speak for the voiceprint test. The significant scientific experiments and recent judicial acceptance of the voiceprint method since Cary convince us that the support for this method now rests on considerably more than the word of a single man. In light of the developments since Cary, we believe that it is no longer unreasonable to order these defendants to speak for purposes of this test. After this test is completed and the State indicates that it intends to offer the results of the test at trial, the trial judge shall hold an additional pre-trial hearing to determine whether any identification arrived at through the use of this method is sufficiently reliable to be admissible in light of the proofs
[61 NJ Page 552]
which will be adduced as to what the test shows, and such cross-examination of the State's experts and such opposing proofs as defendants may be able to offer.
Our decision in this case to permit the voiceprint test to be conducted while deferring final consideration of its admissibility is justified by both legal and practical considerations. If the experts provided by the State are not able to make a positive identification by the comparison of voiceprints made from the defendants' voices with voiceprints made from the tape recording of the telephone conversations, the application of the State in this case for use of this kind of evidence will of course become moot. On the other hand, if the State's experts purport to be able to make a positive identification, the trial court will have the advantage of the proofs at the hearing directed above concerning the particular conditions under which the test was actually conducted. Thus, as noted above, Dr. Tosi indicated that there was an added potential for error when the recordings compared were as little as a month apart, although he was of the opinion that an even greater time lag would not necessarily make a vital difference. The present time lag will be over 5 years. In the abstract it is difficult to determine whether such an identification will be sufficiently reliable to be admissible. It is not clear that the time lag will produce a risk of error disadvantageous to a defendant; it may be that the lag will operate only to prevent an affirmative identification. We believe, however, that the trial judge will be aided in his task if he has detailed testimony from those persons with experience in the use of this method concerning the potential for error of the specific identification test under examination. In this case, we think the State should at least be given the opportunity to conduct the test and then to sustain its burden of establishing that an identification of the defendants' voice (if any) arrived at is sufficiently reliable to be admissible at trial for the jury's ultimate evaluation.
[61 NJ Page 553]
The order under review is reversed and the matter returned to the Middlesex County Court with directions to require defendants to submit their voices for recording and testing by the State's experts as directed above and for further proceedings in the cause consistent with this opinion.