noting the inevitable divergence of scientific opinion, and according greater weight to experts who had empirical experience with voiceprint analysis than to theoretical scientistsSummary of this case from United States v. Downing
January 7, 1975.
March 27, 1975.
Present: TAURO, C.J., REARDON, QUIRICO, BRAUCHER, HENNESSEY, KAPLAN, WILKINS, JJ.
Evidence, Spectrograph test; Opinion: expert; Telephone conversation. Witness, Expert. Identification.
In deciding whether to admit in evidence expert opinions deduced from scientific principle, the standard set forth in Frye v. United States, 293 F. 1013 (D.C. Cir. 1923), and subsequent Massachusetts cases respecting general acceptability is satisfied if the principle is generally accepted by those who would be expected to be familiar with its use. [196-203] At the trial of indictments, there was no error in admitting the opinion of an expert for the Commonwealth that the voice of the unknown speaker of "tapped" recorded telephone calls, whispered for the most part, concerning a missing youth and the payment of ransom, was the voice of the defendant where the opinion was based in part on a visual comparison of spectrograms, "voiceprints" produced by a spectrographic machine, depicting by graphic patterns the recorded telephone calls and recorded exemplars of the defendant's voice, and in part by listening to the recorded voices; admission of the opinion was justified by evidence as to reliability and general acceptance of spectrography introduced at an extensive voir dire hearing, by judicial opinions from other jurisdictions, and by relevant scientific writings.  QUIRICO, J., concurring [205-206]. KAPLAN, J., in a separate opinion, concluded that the voice identification was not admissible, on the ground that the standard of Frye v. United States, 293 F. 1013, 1014 (D.C. Cir. 1923), had not been met [206-213].
THREE INDICTMENTS found and returned in the Superior Court, two on February 6, 1973, and the third June 8, 1973.
The cases were tried before McGuire, J.
The cases were submitted on briefs.
Ronald D. Harper for the defendant.
Philip A. Rollins, District Attorney, for the Commonwealth.
In this case of first impression in this Commonwealth, the single issue argued by the defendant concerns the judge's allowing in evidence, from an expert witness introduced by the Commonwealth, opinions as to voice identification based in part on visual analysis of spectrograms, commonly referred to as "voiceprints." We conclude that the evidence was properly admitted.
The defendant was convicted, after a jury trial, on three indictments charging him with the kidnapping and first degree murder of Paul Cavalieri, and extortion from the victim's father. The appeal is under G.L.c. 278, §§ 33A-33G.
The extortion indictment specifically charged that the defendant, without lawful authority, did forcibly confine and imprison one Paul Cavalieri, with intent to cause him to be held to service against his will, with intent to extort money or other valuable things thereby.
The evidence in the case was as follows. Paul Cavalieri, thirteen years old, left his house in the early evening of November 2, 1972. He did not return home that night, or ever thereafter. Two days later the boy's mother received a telephone call in which a voice said in a whisper, "you will receive instructions." Although Mrs. Cavalieri had known the defendant for seventeen years, she did not recognize the voice.
With permission of the Cavalieris, a "tap" was placed on their telephone for the recording of incoming calls, and a so called "trap" was also placed on the telephone to assist in the tracing of such calls.
A few days after that, a ransom note was received demanding $50,000. In the ensuing several days, the victim's father received several telephone calls all concerned with his son and the payment of ransom. He did not recognize the caller's voice. Following instructions in the note and in the telephone calls, the father left the ransom money at two different locations; the money was not picked up. As a result of the third telephone call, he left the money at a new and different location. At 12:10 A.M. on November 10, 1972, a white male was observed by police picking up the money. Sometime later, and not far away, other officers observed a man carrying a white bag come out of the woods and get into a red station wagon. By tracing the vehicle, they learned the identity of the driver, one Tardiff. From him, they learned of the identity of the defendant as a person who, on a pretext, had persuaded Tardiff to leave the defendant on a highway near the place where the ransom had been left, and to pick up the defendant later, at a prearranged place.
The defendant was arrested. The police warned him of his constitutional rights in terms that are not now contested, as to adequacy, by the defendant. The defendant admitted to the police that he had picked up the package, but he contended that he had done it for pay of $500 from a man in the drug traffic business. He denied having made telephone calls to the Cavalieris.
The defendant told of leaving the money where he had been instructed to by the man who had hired him. The police recovered the money later (not at the place told to them by the defendant), but rather as a result of information received from a man named Delaney, in whom the defendant had confided the location of the money. A package of black bags, identified as the same kind as the bag in which the ransom had been recovered by the police, was found in the defendant's vehicle. The defendant's personal effects and his automobile were stained by the dye that had been put on the ransom money.
On April 12, 1973, the body of Paul Cavalieri was found in a wooded area in North Attleborough. Death had been caused by gunshot wounds. In addition to the evidence outlined above, it was shown at trial that a .38 caliber Colt revolver had been found by police in the defendant's apartment. An expert testified that the bullets found in the victim's body could have been fired from that gun, as shown by rifling impressions on the bullets, but due to oxidation on the bullets there were not sufficient microscopic marks for identification. Ammunition for the gun was also discovered in the defendant's apartment and this ammunition, like the bullets found in the victim's body, was of a rare type. In addition, five staples found in the ransom note were examined and determined to be consistent in manufacture and size with a stapler and staples owned by the defendant. About the time of the victim's disappearance, the defendant was admittedly near the place where the victim was last seen alive. The defendant's car was seen several times near the place where the "drop" of ransom money was first made, at the time when the money, in response to telephoned instructions, had been left at that location.
Eight witnesses listened at the trial to one of the recorded telephone calls. Seven of these witnesses had heard the defendant speak on the telephone on prior occasions. Six of them identified the voice as that of the defendant. The eighth witness had not heard prior telephone talks of the defendant; he also identified the voice as that of the defendant. The witnesses were particularly definite in identifying certain words which were said, not in a whisper, but clearly. Only one of the eight witnesses, called as to this issue, could make no voice identification.
All of the "tapped" recorded telephone calls to the Cavalieris had been whispered, at least for the most part, by the caller. Several exemplars of the defendant's voice reading the words of the previously recorded telephone calls, including exemplars in the defendant's whispering voice, were given to the police. The telephone tapes and the exemplars were sent to Lt. Ernest W. Nash of the Michigan State police.
We turn now to the issue raised in these appeals: the admission in evidence of the voice identification opinions based on voice spectrograms. In an extensive voir dire hearing as to the admissibility of the evidence, Dr. Oscar Tosi, described as "a professor at Michigan State University and director of the Speech and Hearing Research Laboratory," and Lt. Ernest W. Nash of the Michigan State police, described as "of the Voice Identification Unit, which is a unit of the Scientific Criminal Laboratory, stationed at East Lansing, Michigan," both introduced by the Commonwealth, testified at length in support of the admissibility of the evidence. Of the two only Lt. Nash subsequently testified before the jury, after the judge had ruled that his expert opinions were admissible. Dr. Louis J. Gerstman, introduced by the defendant, testified in general that voice spectrogram analysis is unreliable and that, at this time, voice spectrograms should not be used as a means of identification in a forensic situation. No issue is raised as to the qualifications of Lt. Nash, or, indeed, any of the three expert witnesses. The sole issue relates to whether the expert evidence of Lt. Nash offered here has been shown to be sufficiently reliable to be admissible in evidence.
In Commonwealth v. Fatalo, 346 Mass. 266, 269 (1963), and Commonwealth v. A Juvenile, 365 Mass. 421, 425-426 (1974), cases concerned with the admissibility of polygraphic evidence, we adopted the standard of admissibility as to expert opinions based on scientific discoveries as first stated in the case of Frye v. United States, 293 F. 1013, 1014 (D.C. Cir. 1923): "Just when a scientific principle or discovery crosses the line between the experimental and demonstrable stages 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" (emphasis added).
In the Fatalo case, supra, at 269, we said: "Judicial acceptance of a scientific theory or instrument can occur only when it follows a general acceptance by the community of scientists involved. When supported by substantial authority establishing scientific reliability, this court has not hesitated to accept the benefits of science." See also Commonwealth v. Stappen, 336 Mass. 174, 176-177 (1957), and Commonwealth v. Devlin, 365 Mass. 149, 153, fn. 3 (1974).
In accordance with the standards articulated in the Fatalo case, we, as a preliminary matter, consider the nature of the scientific machine involved. Like some other scientific devices, a spectrographic machine signifies its results by producing graphic patterns on a paper, patterns which are then the subject of analysis by the examiner. The spectrogram portrays three main parameters of speech: time (horizontal axis), frequency (vertical axis), and relative amplitude or intensity (degree of shading in the different time/frequency regions). Identification is attempted by both aural and visual means. The operator compares voices by listening to recordings of known and unknown voices, and by visually analyzing high-speed sound spectrograms of each voice. If he finds sufficient points of similarity, he will indicate that two voice exemplars were made by the same person.
The examiner thus draws his conclusions not only from a visual but also from an aural comparison. It can be said that, at least in these aspects and considering the relative importance of subjective reasoning by the examiner, the process bears a closer resemblance to polygraph testing than it does to fingerprint or handwriting analysis. In some important respects, however, the claims made in support of the polygraph are fundamentally different from those made for the spectrograph, and in those distinctions lies part of our reasoning in treating the two differently. Most important is the breadth of the inference urged from the reading of the respective machines. Relying in part on voice characteristics demonstrated and measured by the spectrograph, the examiner there seeks to do no more than compare voices. In contrast, from the measurements reflected by the polygraph, the examiner then extrapolates to arrive at a judgment of something not directly measured by the machine, that is, the credibility of the person examined. In so doing, polygraphic evidence, with its purported ability to discern truth in testimony, may constitute, in any case, a force which intrudes far into the jury's most important functions of determining credibility of witnesses and finding facts. We are aware that scientific proof may in some instances assume "a posture of mystic infallibility in the eyes of a jury of laymen." United States v. Addison, 498 F.2d 741, 744 (D.C. Cir. 1974). For the reasons we have cited, that consideration does not lead us to exclude the voice identification opinions here nor to impose so restrictive a standard of admissibility as we applied to polygraphic evidence in Commonwealth v. A Juvenile, 365 Mass. 421 (1974).
In reaching our conclusion that the evidence was properly admitted, we have examined relevant decisions from other jurisdictions; relevant writings in scientific journals and other sources; and the evidence produced at the trial as to reliability and general acceptance of the principle. We start with the concepts that neither infallibility nor unanimous acceptance of the principle need be proved to justify its admission in evidence.
As to the judicial acceptance of spectrography, it is undoubtedly more useful to consider the holdings of various courts since the year 1971 than prior thereto. The conclusions from two years of extensive laboratory tests supervised by Dr. Tosi were first made available in 1971. No comparable study was made before or has been made since this one.
Of the appellate courts which have considered the admissibility issue since these results were published, courts in three States (Minnesota, Florida, and California) have ruled that such voice identification evidence is admissible: State ex rel. Trimble v. Hedman, 291 Minn. 442 (1971) (voiceprints admissible to establish probable cause and for corroborative purposes); Worley v. State, 263 So.2d 613 (Ct. App. Fla. 1972) (voiceprints properly admitted to corroborate voice identification); Alea v. State, 265 So.2d 96 (Dist. Ct. App. Fla. 1972); Hodo v. Superior Court, 30 Cal.App.3d 778 (1973). In addition, the Supreme Court of New Jersey indicated that growing acceptance of the technique might prompt it to reconsider its prior rejection of the evidence. State v. Andretta, 61 N.J. 544 (1972). Two appellate courts which have ruled since the Tosi study have decided that the evidence is inadmissible. United States v. Addison, 498 F.2d 741 (D.C. Cir. 1974). People v. Law, 40 Cal.App.3d 69 (1974). The Law case goes no further than to say that such evidence is not admissible where the voice to be identified was disguised or mimicked.
Prior to the Tosi study, appellate courts in New Jersey and California had ruled the evidence to be inadmissible. State v. Cary, 56 N.J. 16 (1970). People v. King, 266 Cal.App.2d 437 (1968). Compare the Andretta case, supra, the most recent from the New Jersey Supreme Court and the still more recent Hodo case, supra, from California. In 1972, the Florida Fourth District Court of Appeal, in deciding the Worley case, supra, commented on the Cary and King cases, supra, as follows: "Both of these cases are over three years old. Since then, impressive scientific data has been amassed as to the voiceprint's reliability. We hypothesize that, based on the changes in Dr. Tosi's testimony alone, [the] Cary and King [cases] would be decided differently today. For these reasons, we decline to follow these cases." 263 So.2d at 614. This statement may prove to be prophetic in view of the subsequent Andretta (New Jersey) and Hodo (California) cases.
Literature of the field does not reflect unanimous approval of the technique on the part of those who, from their writings, claim to be qualified to speak. Favorable comments in strong support have, of course, been written in the reports of the studies directed by Dr. Tosi. While some scientists have written in approval, others appear from their writings to have serious reservations as to the issue of admissibility.
The results of Dr. Tosi's study were published in The Journal of the Acoustical Society of America. See Tosi, Dyer, Lashbrook, Pedrey, Nichol, and Nash, Experiment on Voice Identification, 51 J. Acoustical Soc. Am. 2030 (1972). Other articles in support of the Tosi study and vouching for the reliability of spectrographic analysis include Kersta, Voiceprint Identification, 196 Nature 1253 (1962). Cf. Stevens, Williams, Carbonell, and Woods, Speaker Authentication and Identification: A Comparison of Spectrographic and Auditory Presentations of Speech Material, 44 J. Acoustical Soc. Am. 1596 (1968).
Originally critical of spectrographic technique, Dr. Peter Ladefoged, Professor of Phonetics at the University of California at Los Angeles, appeared at certain pre-1971 trials to oppose the admissibility of voiceprints. See Ladefoged and Vanderslice, The "Voiceprint Mystique," reprinted from Working Papers in Phonetics (Dept. of Linguistics, U.C.L.A., November, 1969). However, based on the results of the Tosi study, Dr. Ladefoged has now expressed general support for the voiceprint method. The most recent testimony and written comments of Dr. Ladefoged are referred to in United States v. Addison, 498 F.2d at 744-745, fn. 9 (D.C. Cir. 1974), and People v. Law, 40 Cal.App.3d 69, 80-81, fn. 12 (1974).
In the District Court opinion in the Addison case ( United States v. Raymond, 337 F. Supp. 641 [D.D.C. 1972]), the trial judge made the following significant findings as to Dr. Ladefoged's judgments in this field: "Dr. Tosi's study has substantially changed the opinion expressed by the scientific community as to the reliability of voice spectrograms as a means of identifying an unknown voice. A striking example of this can be seen in the case of Dr. Peter Ladefoged, Professor of Phonetics at U.C.L.A. Dr. Ladefoged was co-author of a leading article which criticized the Kersta study and conclusions, Ladefoged and Vanderslice, The `Voiceprint' Mystique, supra, and even testified as an expert against the admission of spectrograms into evidence in the Trimble case, supra. After examining the Tosi study, however, Dr. Ladefoged stated he now believes that spectrograms have been established as a reliable method of voice identification, and testified in favor of the admission of spectrograms in the case at bar" (emphasis added). Id. at 644-645.
Literature critical of spectrographic analysis includes Bolt, Cooper, David, Denes, Stevens, and Pickett, Speaker Identification by Speech Spectrograms: A Scientists' View of its Reliability for Legal Purposes, 47 J. Acoustical Soc. Am. 597 (1970); Bolt and others, Speaker Identification by Speech Spectrograms: Some Further Observations, 54 J. Acoustical Soc. Am. 531 (1973). But see Black, Lashbrook, Nash, Oyer, Pedrey, Tosi, and Truby, Reply to "Speaker Identification by Speech Spectrograms: Some Further Observations," 54 J. Acoustical Soc. Am. 535 (1973). See also Hazen, Effects of Differing Phonetic Contexts on Spectrographic Speaker Identification, 54 J. Acoustical Soc. Am. 650 (1973) (i.e., disguised voices).
See, generally, Jones, Danger — Voiceprints Ahead, 11 Am. Crim. L. Rev. 549 (1973); Note, Voiceprint Identification, 61 Georgetown L.J. 703 (1972-1973); Comment, Evidence: Admissibility of Spectrographic Voice Identification, 56 Minn. L. Rev. 1235 (1972). Compare Thomas, Voiceprint — Myth or Miracle (The eyes Have It), 3 U. of San Fernando Valley L. Rev. 15 (1974) (opposed to use), with Boren, Voiceprint — Staging A Comeback, 3 U. of San Fernando Valley L. Rev. 1 (1974) (in favor of use).
The voir dire hearing as to admissibility conducted by the judge in this case was lengthy and comprehensive. Evidence reflected not only the qualifications and opinions of Dr. Tosi and Lt. Nash, but also appraised the qualifications (or lack of same) of those who have written in opposition to the use of the spectrograph in voice identification. At the hearing, a substantial showing of reliability was made from the results of the Tosi laboratory experiment which culminated in 1971. In that study a total of approximately 35,000 trials of identification were conducted. The 250 speakers, all males, and all English speaking, were selected at random from a homogeneous population of students at Michigan State University. Visual examinations of spectrograms were conducted by twenty-nine trained examiners. Results, as published, indicated an error of six per cent false identifications and approximately twelve per cent false eliminations. Other correlated and detailed opinions as recorded by the examiners supported a conclusion that, if the examiners had been able to express no opinion when in doubt, errors of identification would be about two per cent.
From the testimony of Dr. Tosi and Lt. Nash the trial judge could well conclude that forensic experience of the witnesses, since the major laboratory study, had reduced the chance of error in their findings to a negligible amount. First, it was shown that the "real life" examiner listens to the voices in addition to visually inspecting the spectrograms, and it is the combination of both aural and visual comparison which enhances reliability. Second, the forensic examiner has experience and training in audiology and speech science beyond that of a usual laboratory worker. Third, the professional examiner, unlike the laboratory worker, can take all the time he wants to make his decision, and he can examine and reexamine all the words he believes to be necessary, if these words are available for study. Fourth, the professional examiner, unlike the laboratory examiner, can select any one of several alternative types of decisions and, most important, he is free to choose to come to no decision in any given case. Lieutenant Nash testified that he had examined forensically 3,300 voices in the hope of identifying them; that he had positively eliminated slightly more than 1,900 of them, and had positively identified only slightly more than 500 of them. He thus came to no decision in almost a third of the cases.
Identification of telephone voices by witnesses familiar with the voice of the identified person has long been permitted by the law of the Commonwealth. Commonwealth v. Bonomi, 335 Mass. 327, 343 (1957). Chartrand v. Registrar of Motor Vehicles, 345 Mass. 321, 325 (1963). Expert testimony as to voice identification may also be admitted. Thus, in this case Dr. Gerstman, an expert introduced by the defendant, testified that he compared the two voices aurally using three sorts of tests, not including a spectrogram. He further testified that he reached the firm conclusion that the defendant was not the telephone speaker. In this regard it could be said that experts bring, even to aural examinations, knowledge and experience in voice identification superior to those of lay persons, and to this extent no great departure from traditional evidentiary principles is made when a voiceprint expert's opinion as to aural analysis, apart from his spectrographic examination, is admitted. This, of course, depends on the degree to which reliance is placed on the independent aural comparison. Thus, in one sense the issue before us in this case would be stated as whether the opinion of an expert as to voice identification may be received in evidence if that expert relied in part on a spectrogram as well as aural comparison to reach his conclusion.
All of these conclusions have bearing on the evaluations of the criticism leveled by those writers who oppose use of spectrography in judicial identifications. For the most part they contend that the percentage of error will increase when conditions depart from the laboratory and are considered in the real world. To the extent that a convincing case is made, as in this case, that the professional examiner will achieve more reliable results than the laboratory examiner, then to that extent the opposition can be discounted.
The requirement, as in the Frye and Fatalo cases of general acceptance in the scientific community assures that those most qualified to assess the general validity of a scientific method will have the determinative voice. See United States v. Addison, 498 F.2d 741, 743-744 (D.C. Cir. 1974). Additionally, the application of the rule protects the parties by assuring a reserve of experts who may be needed to testify.
It can be argued that both of these goals are endangered when only a limited number of persons claim special knowledge in the field, as appears to be the case with spectrography. Limited in number though the experts may be, the requirement of the Frye rule of general acceptability is satisfied, in our opinion, if the principle is generally accepted by those who would be expected to be familiar with its use. In admitting evidence of a procedure called the Nalline test it was said: "No experts were called by the defendants and the expert testimony as above summarized stands uncontradicted in the record with this exception: Each of the People's experts did admit on cross-examination that the medical profession generally is unfamiliar with the use of Nalline and therefore it cannot be truthfully said that the Nalline test has met with general acceptance by the medical profession as a whole, general acceptance being at present limited to those few in a specialized field who deal with the narcotic problem. Should this fact render the testimony inadmissible? We believe not. All of the medical testimony points to the reliability of the test. It has been generally accepted by those who would be expected to be familiar with its use. In this age of specialization more should not be required" (emphasis added). People v. Williams, 164 Cal.App.2d Supp. 858, 861-862 (1958).
It has been suggested that the requirement of general acceptance, as in the Frye and Fatalo cases, should be modified or abandoned. See McCormick, Evidence § 203, p. 491 (2d ed. 1972), where it is said, "`General scientific acceptance' is a proper condition for taking judicial notice of scientific facts, but not a criterion for the admissibility of scientific evidence. Any relevant conclusions which are supported by a qualified expert witness should be received unless there are other reasons for exclusion." The suggestions by this author and others (see the concurring opinion of Mager, J., in the Worley case, supra, 263 So.2d 613, 615 ) urge that the opinions of a qualified expert should be received and that the considerations similar to those expressed in the Frye and Fatalo cases should be for the fact finder as to weight and value of the opinions.
The Commonwealth, in essence, suggests that the Frye and Fatalo rule may have already been modified by our holdings in Commonwealth v. Devlin, 365 Mass. 149 (1974), and Commonwealth v. Gilbert, 366 Mass. 18 (1974). We make no comment at this time as to whether the Devlin or the Gilbert case has application in any circumstances other than the precise scientific principles involved in these cases.
There is no need for modification of the general principle of the Frye or Fatalo cases in order to uphold the judge's ruling in this case. Examination of (1) the evidence as to admissibility presented before the judge, (2) judicial opinions from other jurisdictions, and (3) relevant scientific writings provides convincing proof to justify admission of the evidence. The considerable reliability proved by the Tosi experiment, the greatly added reliability induced by the application of further skills by the experienced examiner working under forensic conditions, and the totality of the evidence received at the voir dire hearing which tended to minimize the importance and weight of adverse or skeptical writings all serve to support a conclusion of general acceptability as required by the rule of the Fatalo and Frye cases.
While we do not agree with the conclusion expressed in Justice Kaplan's separate opinion that "turbulence and discord" sound the measure of the scientific community's acceptance of spectrographic analysis, we agree that there certainly is not uniform and total acceptance of the method and the separate opinion accurately expresses the doubt of several experts in the field. Yet the Fatalo and Frye standard does not require unanimity of view, only general acceptance; a degree of scientific divergence of view is inevitable. In this case we are disposed to give greater weight to those experts who have had direct and empirical experience in the field of spectrography. Among such scientists who have had empirical experience and are supportive of the method are, as previously noted, Dr. Peter Ladefoged, Professor of Phonetics at the University of California at Los Angeles; John W. Black, Professor of Speech Sciences at Ohio State University; Dr. Henry Truby, formerly of the University of Miami, Florida; Drs. Herbert J. Oyer and Charles Pedrey of Michigan State University, who participated in the Tosi study; and Dr. Max D. Steer, Professor of Speech Sciences at Purdue University, who acted as consultant to the Tosi study.
We are further buttressed in our decision by the fact that the "judgment" of six experts quoted at some length in the separate opinion as critical of the conclusions claimed in the Tosi study consists in substance of a theoretical review of the Tosi report and is in the form merely of a four page letter in 54 J. Acoustical Soc. Am. 531 (1973). See fn. 3, supra. Thus, we find the evidence presented in support of the reliability of voiceprints, particularly as expressed in Dr. Tosi's study, sufficiently persuasive to outweigh the criticism expressed by certain other scientists in the field of acoustics.
We hold that there was no error in the admission of the contested opinions of Lt. Nash. Nothing that we have said in this opinion is intended to narrow the discretionary function of the trial judge in appraising the qualifications of an expert witness, or in excluding expert opinions for other reasons within the judge's traditional area of discretion. We add that the admission of expert testimony as to spectrographic analysis should be subject to the closest of judicial scrutiny, particularly in any case where there is an absence of evidence of voice identification other than that of the voiceprint or where, but for the voiceprint, there would be insufficient evidence to warrant any inference of the defendant's guilt. And, of course, as is traditional, once the voiceprint is admitted in evidence the jury may give it such weight as they deem proper.
We have reviewed the entire record in accordance with our duties under G.L.c. 278, § 33E, and we observe nothing which in justice indicates that we should modify the results reached by the jury.
In Commonwealth v. A Juvenile, 365 Mass. 421, 441 (1974), I expressed the view, in a dissenting opinion, that the art of polygraph testing had not yet advanced to the point that the admission of polygraphic evidence should, under the limitations imposed by a majority of the court, be left to the discretion of the trial judge. This view was predicated in large part on the reason that, because of "an almost impenetrable aura of scientific infallibility" surrounding the polygraph machine, there was "a grave risk that the jurors will regard . . . [polygraphic] opinion testimony as resolving the ultimate question of the defendant's guilt or innocence, not considering polygraphic evidence as they do other expert testimony and, contrary to the court's instructions, not in fact weighing it with all the other evidence presented," 365 Mass. at 447 (Quirico, J., dissenting). I also joined in the dissenting opinion of Justice Kaplan in that case.
From what is said in the present case, in the opinion of the court as well as in the separate opinion of Justice Kaplan, it is clear that spectrographic analysis shares with polygraphic analysis an incomplete, though significant, acceptance in the scientific community. As the majority opinion points out, however, "the claims made in support of the polygraph are fundamentally different from those made for the spectrograph," in some important respects. These differences exist both in regard to the breadth of the inference and extrapolation connecting the machine readings to the expert's final testimony and in regard to the depth of the evidence's intrusion into the jury's historic functions of determining the credibility of witnesses and finding facts. In short, I do not believe spectrographic evidence is fraught with the dangers inherent in polygraphic evidence. Accordingly, while not entirely free from doubt, I am persuaded that the testimony of properly qualified expert witnesses as to the results of spectrographic analysis is, in the careful discretion of the trial judge, properly admissible in evidence.
KAPLAN, J. (separate opinion).
Although the court's opinion makes a notably fair and well reasoned argument to the contrary, I believe it was error to admit in evidence the voice identification based in part on visual analysis of speech spectrograms. My conclusion can rest simply on the proposition that the standard of Frye v. United States, 293 F. 1013, 1014 (D.C. Cir. 1923), has not been met: the spectrogram method used by the Tosi group and employed in the present case cannot, in my view, be said to be derived from a scientific principle "sufficiently established to have gained general acceptance in the particular field in which it belongs."
Some chronology is useful here. Before Dr. Tosi's work, speech spectrography had not been sufficiently validated to qualify its results for forensic use. The Technical Committee on Speech Communication of the Acoustical Society of America had requested six authorities in this country in the field of acoustics to study the question. In 1970 the experts — R. Bolt, F. Cooper, E. David, P. Denes, J. Pickett, and K. Stevens — gave their answer in an article published in the Journal of the Society. It was negative. They concluded that "the available results are inadequate to establish the reliability of voice identification by spectrograms. We believe this conclusion is shared by most scientists who are knowledgeable about speech; hence, many of them are deeply concerned about the use of spectrographic evidence in the courts."
Speaker Identification by Speech Spectrograms: A Scientists' View of Its Reliability for Legal Purposes, 47 J. Acoustical Soc. Am. 597, 603 (1970). This criticism, like all subsequent ones, challenged not the technique of recording speech spectrograms but rather the validity of the conclusions drawn by the examiner of spectrograms upon viewing and comparing them.
Dr. Tosi carried out his experiments after this unfavorable judgment about the state of the art. He and his coworkers (H. Oyer, W. Lashbrook, C. Pedrey, J. Nicol, and E. Nash) published their results in 1972. Thereafter a number of courts admitted in evidence identifications based on spectrograms appraised by the Tosi method. Dr. Tosi himself was a principal witness in these cases. See State ex rel. Trimble v. Hedman, 291 Minn. 442 (1971); Worley v. State, 263 So.2d 613 (Dist. Ct. App. Fla. 1972) (use for corroboration); Alea v. State, 265 So.2d 96 (Dist. Ct. App. Fla. 1972) (following Worley); Hodo v. Superior Court, 30 Cal.App.3d 778 (1973). Contra, People v. Chapter, No. 4516, Super. Ct. Marin Cty., Cal., July 23, 1973. It can fairly be said, however, that when the cases were decided the scientific community had not had sufficient time to study Dr. Tosi's work and reach conclusions as to its possible advance over the previous work in the field. See People v. Law, 40 Cal.App.3d 69, 81-82 (1974). The decisions thus reflected less a consensus in the relevant scientific community that the Tosi method was acceptable, than an absence of study on which an informed opinion could be based one way or the other.
Experiment on Voice Identification, 51 J. Acoustical Soc. Am. 2030 (1972).
Although the Trimble and Worley cases were decided before the publication of Dr. Tosi's results, Dr. Tosi testified about his completed experiment in both cases, and both courts relied on that testimony in accepting the spectrographic comparisons.
By the time of the trial in the present case, the scientific community had begun to react, and it appears to me that the reaction has been prevailingly negative. Thus the authors of the 1970 paper, reviewing Dr. Tosi's work, declared they were troubled by several difficulties and summed up their judgment as follows:
They mentioned substantial increase in error rate in comparisons of voice samples taken at different times, as against comparisons of contemporaneous samples; substantial increase in error rate when words taken from sentences were used for comparison, as against words spoken in isolation; failure to consider the problems of mimicking or disguising of voices, changes of voice levels, and changes of emotional state of the speaker; increase of error rate as number of samples were increased from which match to unknown voice was to be made.
"The Tosi study has improved our understanding of some of the problems of voice identification from spectrograms by indicating the influence of several important variables on the accuracy of identification. In uncovering factors that tend to increase identification errors, however, the study has not given us a definitive answer to the question: `How reliably can a person be identified by examining the spectrographic patterns of his speech sounds?' Under certain laboratory conditions and for some selected sample of the population, the probability of making an error in identification can be stated. But for the less-than-ideal conditions encountered in forensic situations, the indications are that the probability of error will increase substantially. Further studies are needed, with particular attention to the examiner's decision criteria, the selection of speaker population, the time lapse between voice samples, background-noise conditions, and the psychological condition of the speaker.
"As scientists rather than lawyers, we offer no judgment as to whether or to what extent speech spectrograms should be used for identification in the courts. We wish only to point out that present methods for such use lack an adequate scientific basis for estimating reliability in many practical situations and that laboratory evaluations of these methods show increasing errors as the conditions for evaluation move toward real-life situations. We hope that our explanations of some of the factors that affect speaker identification will provide the legal profession with helpful information on which to base its own judgments concerning the admissibility of the spectrographic method."
Speaker Identification by Speech Spectrograms: Some Further Observations, 54 J. Acoustical Soc. Am. 531, 533-534 (1973).
Dr. Tosi and his group, seeking to reply to the criticism that forensic comparisons would tend to be less accurate than those made under the more aseptic experimental conditions, pointed to some factors which in their opinion worked in quite the opposite direction. The dispute is not resolved.
See Reply to "Speaker Identification by Speech Spectrograms: Some Further Observations," 54 J. Acoustical Soc. Am. 535 (1973).
In the forensic situation Lt. Nash was allowed to make aural as well as visual comparisons of the voices, to allow himself a longer period of time to make a decision than was allowed the examiners in the Tosi experiments, and to conclude, if he so felt, that he could not reach a decision. These facts, together with the fact that Lt. Nash had had much more training and experience in spectrographic identification than the examiners used in the Tosi study, have led Dr. Tosi to assert, in reply to his critics, that forensic comparisons by Nash or a person of Nash's training are more, not less, accurate than the experimental comparisons. But except for evidence that the error rate declines as an examiner is allowed to express no opinion when in doubt, Tosi's assertion does not appear to be experimentally supported. Moreover, to the extent that Nash depends on aural comparisons for his conclusions, the weight that should be accorded his accompanying, more "scientific" spectrographic analysis, which could have large influence with a jury, becomes questionable.
Dr. Tosi stated at voir dire in the present case that if confronted with a need to compare a whispered telephone conversation in a noisy environment with a later recording in a quiet laboratory — the actual situation in our case — he would "rely very much on my ears."
More recent scientific writing from outside the Tosi group — writing not available to the judge below at voir dire — is unfavorable to the Tosi method, although at least one authority, Dr. P. Ladefoged, has become better disposed to it after being at first negative.
C. LaRiviere, Speaker Identification from Turbulent Portions of Fricatives, 29 Phonetica 246 (1974); H. Hollier, Peculiar Case of "Voiceprints," 56 J. Acoustical Soc. Am. 210 (1974).
Recent law review commentary is also generally opposed to the use of speech spectrograms. Compare note, Voiceprint Identification, 61 Georgetown L.J. 703 (1972-1973); Jones, Danger — Voiceprints Ahead, 11 Am. Crim. L. Rev. 549 (1973); and Thomas, Voiceprints — Myth or Miracle (The Eyes Have It), 3 U. of San Fernando Valley L. Rev. 15 (1974) (opposed to use), with Boren, Voiceprint — Staging a Comeback, 3 U. of San Fernando Valley L. Rev. 1 (1974) (in favor of use).
Judge McGowan in United States v. Addison, 498 F.2d 741 (D.C. Cir. 1974) (discussed below), took Dr. Ladefoged's position as expressed at the trial of that case to be one of "abatement of skepticism," not complete acceptance. This estimate of Dr. Ladefoged's testimony has been doubted (see n. 3 of the court's opinion in the present case), but Dr. Ladefoged's indorsement of Tosi still seems to me incomplete if I am to judge from the court's remarks in People v. Law, 40 Cal.App.3d 69, 84 (1974), where Dr. Ladefoged again testified. See n. 11 below.
To sum up, opinion is divided on the Tosi method; the journal material shows turbulence and discord rather than that "general acceptance" which the Frye case lays down as a precondition of admissibility. Nor can it be plausibly said that those with adverse views are either unqualified to have opinions worthy of respect or are strangers to the relevant scientific "field."
True, the expert panel that criticised the Tosi experiments had not dealt, or dealt so much, with the specific material in the laboratory as the Tosi group. But it is common in scientific fields for experts in a general discipline to be called on to review a particular experiment or technique in the light of their broad experience.
A decision of the United States Court of Appeals for the District of Columbia Circuit rendered on June 6, 1974, reviews the current situation and rejects the results of speech spectrogram analysis as evidence. United States v. Addison, 498 F.2d 741. Judge McGowan, speaking for the court, held that the Frye standard had not been met; he wrote interestingly in explanation and defense of that standard. The case of People v. Law, 40 Cal.App.3d 69, decided on June 25, 1974, held that spectrogram analysis was unacceptable at least as applied to a voice mimicked or disguised. The reliability of such analysis seems also diminished when a voice is whispered, as in the present case. The only recent decision admitting speech spectrogram evidence is United States v. Sample, 378 F. Supp. 44 (E.D. Pa. June 10, 1974), which allowed it at a probation revocation hearing on the reasoning that the prosecution could succeed there on a "preponderance of the evidence" without proof beyond a reasonable doubt; but it appears that no evidence was presented by the defense to counter that of Lt. Nash, Dr. Tosi's coworker.
See at pp. 743-744, also discussing Rule 703 of the new Federal Rules of Evidence. For clarity's sake I should say that Judge McGowan's remarks about Dr. Ladefoged's position, see n. 9 above, are not crucial to his opinion in the Addison case.
Lt. Nash's testimony in the present case indicates he relied heavily on the whispered exemplar in making the identification. Dr. Tosi conceded at voir dire that further studies were needed on disguised voices, including whispered voices. The papers critical of Tosi have made the same point.
It should be observed also that Nash in the present case did not compare the unknown voice with numerous samples but only with the Lykus samples furnished by the police. Although Nash undoubtedly strove for impartiality, his conclusion is subject to discount for possible bias (even if unintended) due to any assumption or realization on his part that the police thought they had the guilty man.
I have taken account of the specific expert testimony offered below, but, especially when considered together with the journal writings, it does not, as I see it, make a case that satisfies the Frye requirement. I take less comfort than my brethren from the fact that the results of speech spectrogram analysis are not so close to the ultimate probanda of a criminal trial as the results of polygraph analysis typically are, whence it is suggested that in the former case the function of the jury is not so seriously invaded. I should add that since the opinion of the court, while referring to Professor McCormick's proposition, does not embrace it, I do not enter upon the reservations that I would have about admissibility in the present case if the McCormick thesis were to be accepted and applied here.
I have indicated elsewhere my belief that the law might preferably proceed by a "commission" procedure to handle questions of validating new methods of scientific measurement or demonstration intended for use in a court room. See Commonwealth v. A Juvenile, 365 Mass. 421, 452, 452-453 (1974) (Kaplan, J., dissenting). The present case seems to me a nice illustration of the point. Lastly it may be noted that, failing a "commission" procedure, it was open to the trial judge, and might have been of help to him in the quandary in which he found himself, to call independent experts, of whom a few with very high qualifications could have been located within a short radius of the court house.
As I am outvoted on the question of admissibility, I need not attempt to answer the difficult question whether the erroneous (as I judge it) reception of the evidence would require reversal of the conviction in light of the rest of the proof.