Decided: March 14, 1978.
VICTOR D'ARC, PLAINTIFF,
MARY LEA J. D'ARC, DEFENDANT
Imbriani, J.c.c. (temporarily assigned).
[157 NJSuper Page 555]
In a matrimonial action a wife obtained a recording of several telephone conversations allegedly involving her husband. The recording was lawfully made by one of the parties to the telephone conversation, U.S.C.A. , Title 18, sec. 2511(2)(d); N.J.S.A. 2A:156A-4(c); State v. Vizzini , 115 N.J. Super. 97 (App. Div. 1971); U.S. v. Bragan , 499 F.2d 1376 (4 Cir. 1974), who later gave the recording to the wife. Pursuant to court order a voice exemplar was made by the husband. A sound spectrogram, commonly known as a voiceprint, was made, and she now seeks to introduce into evidence the recording and the results of the voiceprint test.
A sound spectrograph is an instrument which produces a spectrogram that can be utilized by a trained examiner to purportedly identify the speaker of an unknown voice. The system was devised in the early 1940s by researchers at Bell Telephone Laboratories, Inc. and operates by an electrical process which produces on a sheet of paper a pictorial representation or graphic display of sounds, which is called a sound spectrogram or voiceprint. Annotation, "Admissibility of Spectrograph Evidence," 49 A.L.R. 3d 915, 917 (1973). Identification of an unknown speaker is made by an examiner
[157 NJSuper Page 556]
who compares the unknown voice with the known voice exemplars of one or more persons. Gorecki, "Comment: Evidentiary Use of the Voice Spectrograph in Criminal Proceedings," 11 Military L. Rev. 167 (Summer 1977). The device does not operate automatically; an individual must operate the machine, adjust numerous controls to obtain readings and subjectively interpret the spectrograms to arrive at an opinion.
The question presented is whether sound spectrograms, or voiceprints, and the opinions of examiners based upon examining and interpreting them, are admissible as valid and accurate scientific evidence in a court of law?
First to review the New Jersey cases. There are three. The precursor is State v. Cary whose proceedings spanned a period of three years. A defendant charged with murder was ordered to provide a voice exemplar so that it could be compared on a spectrograph with an unknown voice on a taped recording. He appealed the order as a violation of his right to privacy protected by the Fourth Amendment. In the first Cary review, 49 N.J. 343, 351 (1967), the court said that "before an intrusion into a person's privacy can be proper within the protections afforded by the Fourth Amendment, the product of the search must have the capacity to be admissible in evidence." The court noted that the only expert offered below was Lawrence G. Kersta, one of the pioneers in the research of sound spectrograms, and in remanding said, "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." Id. at 352.
The trial court then received testimony from four experts (two of whom appeared in the within proceedings), two for each side, and concluded that "at this time there is a lack of general scientific acceptance as to accuracy of the voiceprint technique." 99 N.J. Super. 323, 331 (Law Div. 1968). A second appeal followed, 53 N.J. 256 (1969), and apparently sensing an adverse decision, the State sought
[157 NJSuper Page 557]
an additional opportunity to offer "further expert testimony." Id. at 258. Because of the far reaching implications of such evidence, the request was granted. But when later advised that the State was "not, at this time, able to furnish any new and significant evidence," the finding of the trial court was affirmed, 56 N.J. 16, 18 (1970).
There soon followed State v. Andretta , 61 N.J. 544 (1972). The trial court had concluded "that the State had failed to meet its burden of establishing general acceptance of the voiceprint method." Id. at 546. The Supreme Court observed the increased support for the admissibility of voiceprint evidence since Cary , but noted that the narrow question presented was whether defendant should be compelled to speak for a voiceprint test, not whether such evidence was admissible during the trial. It concluded that defendant should be compelled to submit his voice exemplar. There is no history of what followed on the remand. It is interesting that the court relied upon United States v. Raymond , 337 F. Supp. 641 (D.D.C. 1972), which was appealed and, although the conviction was affirmed, the trial court's admission of voiceprint evidence was held to be erroneous because "[w]hatever its promise may be for the future, voiceprint identification is not now sufficiently accepted by the scientific community as a whole." United States v. Addison , 162 U.S. App. D.C. 199, 498 F.2d 741, 745 (D.C. Cir. 1974).
The most recent case is State v. Perez , 150 N.J. Super. 166 (App. Div. 1977), involving tapes made by electronic surveillance of telephones. A monitoring detective, "although not qualified as an expert in voice identification," id. at 170, was allowed to give his opinion that defendant's voice and that of the taped telephone speaker were the same, in spite of the fact that "[n]o scientific basis for the introduction of the spectrographic test results had been placed in the record." Id. at 170. Thus, Perez appears to be of limited application.
[157 NJSuper Page 558]
At this time, no appellate court in New Jersey has ever concluded that the sound spectrogram, or voiceprint, has attained sufficient scientific acceptance of its accuracy and validity so as to be admissible as evidence during the trial of a case. Cf. State v. Andretta, supra.
Prior to 1971 courts generally refused to admit voiceprint testimony. Thereafter, several courts accepted such testimony for the limited purposes of issuing an arrest and search warrant, State ex rel. Trimble v. Hedman , 291 Minn. 442, 192 N.W. 2d 432 (Sup. Ct. 1971), to corroborate aural identification, Worley v. State , 263 So. 2d 613 (Fla. D. Ct. App. 1972), and to compel a defendant in a criminal case to submit a voice exemplar. State v. Andretta, supra; State v. Olderman , 44 Ohio App. 2d 130, 336 N.E. 2d 442 (App. Ct. 1975). Other courts admitted voiceprint identification evidence as direct and independent proof of the identity of a speaker. Hodo v. Superior Court , 30 Cal. App. 3d 778, 106 Cal. Rptr. 547 (1973); Commonwealth v. Lykus , 327 N.E. 2d 671 (Mass. Sup. Jud. Ct. 1975); United States v. Franks , 511 F.2d 25 (6 Cir. 1975); United States v. Baller , 519 F.2d 463 (4 Cir. 1975). And still other courts rejected such evidence. People v. Kelly , 17 Cal. 3d 24, 130 Cal. Rptr. 144, 549 P. 2d 1240 (Sup. Ct. 1976); United States v. Addison , 162 U.S. App. D.C. 199, 498 F.2d 741 (D.C. Cir. 1974); Commonwealth v. Topa , 471 Pa. 223, 369 A.2d 1277 (Sup. Ct. 1977). See generally, "Admissibility of Spectrograph Evidence," 49 A.L.R. 3d 915 (1973). No clear consensus has emerged throughout the country.
Before a scientific technique may be admitted into evidence, there must be testimony from scientific experts to support and endorse its validity, reliability and veracity. It is not accepted simply because one or more experts urges the court to do so.
The early, and generally widespread, legal test for determining whether a new scientific technique is valid and thus admissible at trial was expounded in Frye v. United States , 54 App. D.C. 46, 293 F. 1013 (D.C. Cir. 1923):
[157 NJSuper Page 559]
Just when a scientific principle or discovery crosses the line between the experimental and demonstrable states is difficult to define. Somewhere in this twilight zone the evidential force of the principle must be recognized, and while courts will go a long way in admitting expert testimony deduced from a well-recognized scientific principle or discovery, the thing from which the deduction is made must be sufficiently established to have gained general acceptance in the particular field in which it belongs." [54 App. D.C. at 47, 293 F. at 1014]
That same court recently went one step further and said that the acceptance of the scientific community must be "as a whole." United States v. Addison, supra , 162 U.S. App. D.C. at 203, 498 F.2d at 745.
While the Frye test is the one most frequently utilized by courts throughout the United States, several states have authorized a less rigorous test, including New Jersey. State v. Andretta, supra , directed the trial court on remand:
To determine whether any identification arrived at through the use of this method is sufficiently reliable to be admissible in light of the proofs 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. [61 N.J. at 551-552]
It would appear that in New Jersey a new scientific technique would be admissible if it meets either the Frye or Andretta test.
The husband offered two experts opposing voiceprint. First, Dr. Louis Gerstman, Professor of Psychology, Speech and Hearing Sciences, at the City University of New York, who testified that from personal communications with colleagues in this field, acceptance of this technique is "low" because they do not believe that the present state of this system is such that it can validly identify an unknown speaker. He said that at least "30 close colleagues" share this view and the only two professional persons with a scientific background who will go into court to support this technique are Professors Tosi and Truby, both of whom appeared in this case in support of the voiceprint technique.
[157 NJSuper Page 560]
The husband also offered Dr. Harry Hollien, Professor of Speech and Linguistics and Director of the Institute for the Advanced Study of Communication Processes at the University of Florida, who particularly questioned as unsubstantiated two assumptions which are the basis of the voiceprint technique. First, that "every individual in the world is unique in their speech," and second, that "they were so unique and so unlike everyone else that they could be in and of themselves identified." Neither assumption, he testified, has ever been scientifically proven. He said that the only major work performed in this field was by Professor Tosi (see Tosi et al., Voice Identification Through Acoustic Spectrography (1971)), which resulted in a 6% error of "false identification" and an additional 12% error of "false elimination," i.e. , the examiner said the speaker's voice was not in a grouping when it actually was. Professor Hollien said that an 18% margin of error was far too high to be acceptable in the scientific community. Moreover, he said that several other scientists who performed more limited tests than Professor Tosi were unable to obtain as low a margin of error as did Tosi. For instance, he discussed a study of Malcolm Hall, a student of Tosi, who had an entertainer, Rich Little, mimic six actors. The actors refused to cooperate, so their voice exemplars had to be taken from film and TV clips. The examiner made false identifications, or could not make an identification, in 25% of the cases.
Professor Hollien testified that after the Tosi report was published he polled 50 of the most respected scientists in this field throughout the world, and asked if any had changed their views after reviewing the Tosi report. He received 41 or 42 responses. Three said they had not read the Tosi report, but all of the others who responded said that Tosi had not changed their opinions that the sound spectrograph technique could not validly identify an unknown speaker. Further, he attended a meeting of the Acoustical Society in 1973 in Miami. His effort to poll the members of a committee dealing with the problem was frustrated
[157 NJSuper Page 561]
by the chairman, but after the meeting adjourned he conducted an "on the spot informal poll." Of some 50 members present, 42 said this technique was not valid; the rest abstained. No one voted to support the validity of "voiceprint" technique. Dr. Truby contested the fairness of this vote, but did not refute that it occurred or that the actual vote was as reported by Professor Hollien. The latter also said that to his knowledge the opinion of the scientific community has not changed since taking those polls. Indeed, Professor Hollien presented a "Status Report of 'Voiceprint' Identification in the United States" to the International Conference on Crime Countermeasures, Science and Engineering held at Oxford, England in July 1977 and concluded that:
To continue to use a procedure that has been so seriously challenged -- and in areas as socially sensitive as are the courts -- is, without question, inconsciencable [ sic ] and unethical. Hence, it is my hope that the "voiceprint" proponents will agree to defer application of their process until they can demonstrate its validity to the scientific community.
The wife offered two witnesses in support of sound spectrography: Professor Oscar Tosi, Director of Speech and Hearing Sciences at Michigan State University, and Dr. Henry Truby, an independent researcher with prior academic background. It was Professor Tosi who did the only major study in this field, which took three years to complete and involved some 35,000 experimental examinations. While Professor Tosi was very impressive and spoke with great passion and certainty of the validity of his findings, his positive conclusions are based simply on his certainty of the validity of his tests and the fact that, in his opinion, other scientists have not proven him wrong. But this is far different from expressed declarations of acceptance by the scientific community of sound spectrography as a valid method of speaker identification. Dr. Truby's opinion appears to be based in large measure open his acceptance of
[157 NJSuper Page 562]
the Tosi report, although he had completed some research of his own. Neither could testify as to any polls supporting acceptance of voiceprint by the scientific community; nor could either point to any significant scientific body that endorsed this technique, except for Truby who named some seven or eight "experts" who he said supported his views. On cross-examination, it appeared that three or four of the "experts" were only technicians, and not professional scientists.
The proofs offered in this case clearly fall far short of the test enunciated in Frye. Not only is there an absence of proof of "general acceptance" by the scientific community, but the contrary appears to be the fact.
Can we nonetheless conclude that Tosi and Truby demonstrated that identification of an unknown speaker by using their technique is sufficiently valid to be admissible under the Andretta test? The response must be negative.
A sound spectrogram is not akin to a fingerprint. The latter have "ridges [that] do not vary. Speech, however, has not been proven to be invariant. Fingerprints cannot be disguised; speech can. Further, fingerprints do not change with age or illness. Whether voices change under these circumstances is uncertain." Note, "Voiceprint Identification", 61 Geo. L.J. 703, 725 (1973). Even Professor Tosi opined that sound spectrography is really an art, not a science, and cannot be compared to fingerprinting. He said it is more similar to a handwriting analysis. Both, he said, are based upon subjective determinations made by an examiner.
What this court finds disconcerting is the paucity of major tests and studies. The first limited tests of Professor Lawrence Kersta were conducted some 20 years ago; regrettably, during the intervening period only Tosi has completed a major study. We must have at least one or two other major studies so that we can compare their results and conclusions with those of the Tosi study. Cary required "something more than the bare opinion of one man." We now have that. But to be assured that we have a scientific technique which is
[157 NJSuper Page 563]
valid and reliable we also need something more than the bare results of one major study.
The Tosi study while voluminous has several shortcomings. He provided the examiner with a "closed set," which means that the examiner was to compare the unknown voice to a limited number of known voices. In a forensic setting there are an unlimited number of voices to be compared to the unknown voice. Obviously, the forensic test is more difficult and the percentage of error will be greater.
Tosi provided his examiners with voices made in a laboratory setting with modern equipment. In a forensic setting there often is considerable ambient noise and, moreover, the recording equipment is not as good. Surely this will result in a higher percentage of error.
Tosi provided his examiners with contemporary exemplars (i.e. , made at or about the same time as the unknown voice). In a forensic setting this is not possible. The greater the time lapse, the more difficult it is to compare voices. No one really knows, but it may be significant that in this case the recordings of the telephone conversations were made in April 1976, while the voice exemplar of the husband was made in February 1978, some 22 months later.
The spectrograph is an instrument that provides a three dimensional function of speech: time, frequency and intensity. But there are other functions of speech which it does not measure, such as pitch and quality of voice. Should we await the development of an instrument which can measure and record all functions of speech?
Further, we do not know if the group of sound spectrograms given to the examiner by Tosi were persons of similar height, weight, age, etc., who may produce similar spectrograms. This would obviously make it more difficult to select the correct comparison. In a jailhouse line-up identification, the object is to obtain persons who are all similar to the accused in height, weight, color, etc.; if the choices were significantly dissimilar, the result may well be meaningless. The
[157 NJSuper Page 564]
same should be done with a voiceprint test. Presumably this was not done by Tosi.
The opponents of the technique also seriously question the qualifications of the examiners. These examiners are technicians who, it is claimed, do not possess the scientific education and background necessary to properly operate and assess the results of the spectrograph. Further, the examiners are trained and qualified by Kersta and Tosi, the proponents of this technique, and have never been tested by anyone else. The opponents say this is not proper or acceptable.
Professor Tosi and his supporters start with the proposition that no two persons make the same sound spectrogram with their voice because everyone in the world is unique in their speech. Yet no one has ever verified this assumption. In addition, no one has ever conducted a thorough test to determine if one can change his sound spectrogram, or voiceprint, by intentionally altering his voice; nor do we know if the voiceprint of a mimic is different from that of the person whose voice he is mimicking.
Finally, only one-third of the Tosi tests were made in what can be termed a forensic setting. It is especially disturbing that if these cases are culled and graded, and these are the types of cases we see in the trial courts, the rate of error for "false identification" rises significantly from 6% to 16%.
Professor Tosi testified that 23 states now accept voiceprint evidence. He did not name the states. No evidence has been submitted to substantiate that claim. This court's independent research did not substantiate that claim. Perhaps the confusion arises because Tosi counted cases where any court in a state, including municipal and justice of the peace courts, allowed such testimony, without a preliminary inquiry into the scientific validity and acceptance of the voiceprint technique as has this court. Perhaps he includes New Jersey because of the Andretta and Perez decisions. If so, his assertions are not entirely accurate and may be
[157 NJSuper Page 565]
nothing more than the normal puffery by the proponent of any new device or instrument.
Suffice it to say that this court perceives many objections and impediments to the admissibility as evidence in a court of law of the results of a sound spectrographic examination. There is particular concern that this evidence, if admissible in this civil case, would obviously also be admissible in a criminal case. While a court can instruct a jury that this evidence is not conclusive and should only be weighed with all of the other evidence, one must recognize that "scientific" evidence of this type has the potentiality to be assumed by many jurors as being conclusive and dispositive. As noted in United States v. Addison, supra , 162 U.S. App. D.C. at 202, 498 F.2d at 744, "scientific proof may in some instances assume a posture of mystic infallibility in the eyes of a jury of laymen." Here we are not dealing simply with a link in the chain of evidence. Here we are dealing with a scientific technique that goes to the very heart of the issue being tried and has the potentiality of usurping, directly or indirectly, the function of the jury.
Admittedly, the burden borne by the proponents of a new scientific technique is a heavy one. But in an era permeated with man's disbelief in man, perhaps rooted in Watergate, and in which mankind has taken for granted the infallibility of science, somewhat rooted in the successes of space exploration, we should require nothing less.