David Patterson, Asst. U.S. Atty., (Robert B. Fiske, Jr., U.S. Atty., Robert J. Jossen, Asst. U.S. Atty., Southern District of New York), New York City, for plaintiff-appellee.

Richard G. Rosenbaum, New York City, for defendant-appellant Isiah Williams.

Phylis Skloot Bamberger, New York City (The Legal Aid Society, Federal Defender Services Unit, New York City, of counsel), for defendant-appellant Michael Manning.

Before Van GRAAFEILAND, Circuit Judge, MARKEY, Chief Judge, U.S. Court of Customs and Patent Appeals,* and DOOLING, District Judge, U.S. District Court for the Eastern District of New York.**

MARKEY, Chief Judge, U. S. Court of Customs and Patent Appeals:

Isiah Williams and Michael Manning were convicted of violating federal narcotics laws, 21 U.S.C. § 846 (1972) and 21 U.S.C. § 841 (1972). The admission of spectrographic voice-identification evidence is challenged for the first time in this circuit. We affirm.

On November 12, 1976, an undercover police officer attempted unsuccessfully to purchase heroin from a man introduced to him as "Biggie," an event witnessed by three surveillance officers, including Detective Copeland, who followed "Biggie" into a bar four days later. Copeland asked "Biggie" for identification, pretending he was investigating a complaint that a man named "Biggie" was taking numbers. "Biggie" admitted that his name was Isiah Williams, and that he was called "Biggie," but denied taking numbers. Explaining his unemployment, Williams showed Copeland a support truss he was wearing. Surveillance officers also observed these events.

On November 30, 1976, Officer Lopez arranged to purchase heroin from "Biggie." The two drove to a building. "Biggie" entered the building. He emerged with another man and both joined Lopez in the car. "Biggie" introduced the other man as "Red." Lopez handed "Biggie" money and "Red" gave the heroin to Lopez. "Red" was later identified as Manning. Except for the transactions in the car, these events were observed by the same surveillance officers.

On December 10, 1976, Lopez purchased additional heroin from "Biggie," in a similar transaction observed by the same surveillance officers.

On December 13, 1976, "Biggie" telephoned Lopez concerning another heroin sale. Lopez taped the call. The proposed sale never materialized. Two days later, Lopez called "Biggie" to arrange another date, and taped that conversation. Two more meetings between Lopez and "Biggie" were observed by the same surveillance team.

At trial, Lopez and the surveillance officers unequivocally identified Manning as "Red." Lopez was unable to make an in-court identification of Williams as "Biggie" upon viewing Williams and his arrest photo, a difficulty apparently caused by Williams' having drastically altered his appearance. Lopez did identify a 1972 photo of Williams as "Biggie." All three surveillance officers identified Williams as "Biggie." At the time of his arrest, Williams recognized Copeland from the "numbers complaint ruse," and was found to be wearing a truss.

After his arrest, Williams gave telephone voice exemplars which were taped. After a pre-trial hearing, voice analysis evidence was ruled admissible.1 At the trial spectrographic voice identification expert Frank Lundgren made aural and spectrographic comparisons of Williams' voice in the exemplars and of "Biggie's" voice in the taped conversations with Lopez.

Referring to appellant Manning in his rebuttal summation, the prosecutor told the jury:

The only important thing is that you have the testimony of Agent Lopez, who saw this man clearly on two occasions, and who is able to recognize him, there being no substantial change in appearance. We know, if we know nothing else in this case, that Agent Lopez is a careful and completely honest, scrupulous man, and would not make such an identification if he were not absolutely sure.

I suggest you bear that in mind above all else when considering Mr. Manning.

The sole issue in Williams' appeal is whether it was error to admit a spectrographic voice analysis as identification evidence.2

The sole issue in Manning's appeal is whether the quoted portion of the prosecutor's rebuttal summation constituted vouching for a government witness.

Sound consists essentially of pressure waves of varying frequencies and amplitudes. The pressure waves associated with speech are initiated when air is exhaled past the vocal cords. The resulting vibration of the vocal cords produces the pressure waves.

Frequency and intensity are important speech characteristics involved in voice analysis. Frequency determines the pitch of the sound. It is delineated by the time interval between successive vocal cord vibrations and the speed at which air molecules are vibrated thereby. Intensity is loudness, and is a function of the number of air molecules vibrating at a given frequency. All speech is composed of several frequencies produced simultaneously; a fundamental frequency and several overtones having frequencies which are even multiples of the fundamental.

An individual's speech is created by a complex physiological and mechanical operation. The waves generated by the vocal cords are modified by vocal cavities (throat, nose, and cavities formed in the mouth by positioning the tongue), and by articulators (lips, teeth, tongue, palate and jaw muscles). The vocal cavities act as resonators which cause sound energy to be reinforced in specific sound spectrum areas, dependent upon the size, shape and interrelationship of the cavities. The articulators cooperate in a controlled dynamic interplay in the production of intelligible speech. The manner in which each of us manipulates his articulators when speaking has been developed by a process of imitation and trial and error.

Voice analysis thus rests on the non-likelihood that two individuals would have identical vocal cavities and identical dynamic patterns of articulator manipulation, and on the inability of an individual to change or disguise the particular voice characteristics created by his unique combination of cavities and articulator manipulative patterns. Spectrographic voice analysis involves the reflection of voice characteristics in a "spectrogram" produced by a "spectrograph."4

The spectrograph is an electromagnetic instrument which analyzes sound and disperses it into an array of its time, frequency and intensity components. The array is graphically displayed in a spectrogram.5

The spectrograph operator is supplied with two magnetic tapes one with a known, the other with an unknown, voice. He listens for similar words and phrases on both tapes. The preferred cue words are: THE, TO, AND, ME, ON, IS, YOU, I, A, and IT. Spectrograms are then made of the portions of the tapes on which the selected words and phrases occur.

In producing spectrograms, a tape is placed in the spectrograph. The spectrograph electronically scans the tape and generates electronic signals representative of the components of the sound. The signals are fed to a variable filter, which adjusts the position of a stylus. The stylus burns thin parallel lines on current-sensitive paper wrapped around a rotating drum. The stylus traces a horizontal line, representing a single frequency, the darkness of the line-trace varying as it progresses. At the end of each line, the stylus returns to trace out another line, representing a slightly higher frequency, and so on, producing a bar spectrogram.

The spectrograms of the same words and phrases are then compared visually, to determine whether they were made by the same speaker. The bar spectrogram indicates time along the horizontal axis, frequency along the vertical axis, and intensity by varying shades of darkness in the pattern. The unique speech characteristics of the individual whose voice is being analyzed produce unique spectrogram patterns of vocal energy at the various frequency levels. Though it is not necessary that two spectrograms be identical, there must be exhibited a sufficient number of similar spectrogram patterns, called "matches," to warrant a conclusion that they were produced by the same person.

Though the weight of authority6 supports the admissibility of spectrographic voice identification evidence, we rest our decision on an independent evaluation.

There is no clearly defined, universal, litmus test for the general admissibility of all "scientific" evidence.

An oft-cited test, first laid down in Frye v. United States, 54 App.D.C. 46, 293 F. 1013 (1923), requires "general acceptance in the particular field to which it belongs." The "Frye" test is usually construed as necessitating a survey and categorization of the subjective views of a number of scientists, assuring thereby a reserve of experts available to testify. Difficulty in applying the "Frye" test has led a number of courts to its implicit modification. See United States v. Baller, supra note 6.

We deal here with the admissibility or non-admissibility of a particular type of scientific evidence, not with the truth or falsity of an alleged scientific "fact" or "truth."7 Hence the established considerations applicable to the admissibility of evidence come into play, and the probativeness, materiality, and reliability of the evidence, on the one side, and any tendency to mislead, prejudice, or confuse the jury on the other, must be the focal points of inquiry.

In the present case, probativeness and materiality present little difficulty. If the "unknown" voice be established as having been employed in the commission of, or in relation to, the crime alleged, I. e., a proper foundation has been laid, evidence that the voice is that of the accused is both probative and material. Our concern is therefore directed to (a) reliability, and (b) tendency to mislead.

(a) Reliability

A determination of reliability cannot rest solely on a process of "counting (scientific) noses." In the present case, Williams presents a list of 10 scientists classified as favoring use of spectrographic analysis in the courtroom and 17 scientists classified as opposed. Williams admits, however, that there are differing shades of opinion within each classification, and that many names could be added to each. Selection of the "relevant scientific community," appears to influence the result. Acceptance of the technique appeared strong among scientists who had worked with spectrograms and weak among those who had not. Further, some of the scientists now classified as favoring had previously been opposed. Lastly, unanimity of opinion in the scientific community, on virtually any scientific question, is extremely rare. Only slightly less rare is a strong majority. Doubtless, a technique unable to garner Any support, or only minuscule support, within the scientific community would be found unreliable by a court. In testing for admissibility of a particular type of scientific evidence, whatever the scientific "voting" pattern may be, the courts cannot in any event surrender to scientists the responsibility for determining the reliability of that evidence.

Nor need it be found that spectrographic evidence is infallible. The sole question is whether spectrographic analysis has reached a level of reliability sufficient to warrant its use in the courtroom.

One indicator of evidential reliability is the potential rate of error. In the most comprehensive study of spectrography made to date, involving experiments at Michigan State University with voices recorded directly and over the telephone, with and without background noise, with words spoken in context and in isolation, and with recordings made contemporaneously and at different times, Dr. Tosi arrived at a false identification rate of 6.3%, a rate reduced to 2.4% When doubtful comparisons were eliminated.8

Another reliability indicia is the existence and maintenance of standards. The International Association of Voice Identification, an organization concerned with training and certification of spectrograph examiners and with procedures, requires that ten matches be found before a positive identification can be made.

A third reliability factor can be the care and concern with which a scientific technique has been employed, and whether it appears to lend itself to abuse. In spectrogram analysis only five possible conclusions are permitted. The two voices are: (1) the same; (2) different; (3) probably the same; (4) probably different; and (5) no conclusion is possible. Lundgren, the expert who testified in this case, has rendered a positive identification in only 8% Of the 200 cases in which he has been involved. He found the voices different in 6%, and reported an inability to reach any conclusion in 76% Of the cases.

A further indication of the reliability of spectrographic analysis is its analogous relationship with other types of scientific techniques, and their results, routinely admitted into evidence. Like handwriting exemplars and gun barrel striations, spectrograms are variable, but contain sufficient points of similarity or dissimilarity to enable a trained expert to reach a conclusion.9

Lastly, a convincing element in determining reliability is the presence of "failsafe" characteristics. A poor quality in the original tapes, poor recording conditions, deterioration of the tapes over time, and changes in mental or physical state of the speaker, can affect spectrogram comparisons, but such factors are more likely to result in different rather than in similar spectrograms. The result would then more likely be that no conclusion is possible, or an erroneous conclusion that two voices are different when they are the same, rather than vice-versa. Thus, inaccuracies in the spectrograms, or failure to accurately reflect the voice of the accused, is more likely to redound to his benefit than to his detriment. Moreover, the listed factors affecting spectrogram accuracy are subject to exposition at the trial, or in a preliminary hearing. If the attack on accuracy goes only to the weight to be given the evidence, admission to jury consideration would appear proper. If the attack establishes such fundamental inaccuracy as to render the particular spectrographic evidence of no assistance to the jury, the potential to mislead may be so great as to warrant an exclusion order.

(b) Tendency to Mislead

Williams says there is a tendency of spectrographic voice analysis evidence to mislead or confuse the jury because lay jurors are awed by an "aura of mystic infallibility" surrounding "scientific techniques," "experts," and the "fancy devices" employed. The objection is raisable, of course, whenever an expert witness in any field of science gives conclusions based on his expertise, and on scientific test results produced with technological equipment. It has less force, however, in respect of spectrographic voice analysis evidence than it might in some other fields. As above discussed, the critical step in spectrographic voice analysis is the simple step of visual pattern-matching, a step easily comprehended and evaluated by a jury.

Moreover, in the consideration of spectrographic voice analysis evidence, it is not expected that the jury will participate only as passive listeners. The objective components supplied by spectrography are subject to direct evaluation by the jury. The tapes can be played, and, if need be, replayed for the jury.10 The spectrograms themselves can be examined and compared by the jury. In addition, the normal safeguards are fully available. The expert's qualifications, the reliability of his equipment, the reliability of the technique itself, and in essence the accuracy and reliability of the spectrograms, are all subject to challenge and attack through cross-examination and the testimony of opposing experts. Finally, the jury can be instructed that the expert's opinion is solely for their assistance, and subject to their complete rejection if they consider it unreliable.

It bears reiteration that admissibility alone is under consideration. The jury remains at liberty to reject voice analysis evidence for any of a number of reasons, including a view that the spectrographic voice analysis technique itself is either unreliable or misleading. We hold only that spectrographic voice analysis evidence is not so inherently unreliable or misleading as to require its exclusion from the jury's consideration in every case.11

The record in this case demonstrates that virtually all of the safeguards designed to assure reliability, and to prevent a misleading of the jury, were employed.12 The exemplar tape of Williams' voice was recorded over a telephone line, duplicating as closely as possible the conditions under which the questioned tape, I. e., the tape of the unknown voice, was made. In making the exemplar tape, Williams read from a transcript of the questioned tape. The opinion of spectrograph expert Lundgren was based on both an aural comparison of the voices on the two tapes and on a visual comparison of the spectrograms. The tapes were played for the jury. The spectrograms were shown to the jury. Lundgren's opinion was based on thirteen matches between spectrograms of corresponding words and phrases on the questioned and exemplar tapes, three more than the established standard for identity. A "consistency" was found among the remaining spectrograms. The expert took all the time necessary to make a complete analysis. He was subject to cross-examination concerning his qualifications, his methods, and his equipment. The expert was knowledgeable, his equipment was accurate and reliable, and the spectrograms were proper subjects for comparison. The jury was instructed that the expert's opinion was for their assistance only and could be rejected if found unreliable.13

Williams' attack here on Lundgren's qualifications, and on the quality of his equipment, are without merit. All of the experts testifying at the evidentiary hearing, including those presented by Williams, attested to Lundgren's qualifications. Lundgren is certified as an examiner by the International Association of Voice Identification and has extensive practical experience. No evidence supports the assertion that the spectrograph equipment was defective at the time Lundgren made his analysis, Williams' assertion being based entirely on speculative inferences drawn from past difficulties with the machine. Nor is there evidence to support a finding that the tapes were in any manner defective. On the contrary, all of the experts attested at the hearing to the high quality of the tapes.

Manning's argument that the prosecutor vouched for the credibility of the witness hinges on the interpretation to be given the word "we" appearing in the prosecutor's rebuttal summation (quoted above). Impermissibly taken out of context, the challenged portion of the prosecutor's remarks would be at worst ambiguous. There is no proof or indication that the term "we" was used, or intended, or taken by anyone, to mean any person or group other than the total accumulation of people gathered in the courtroom, or that it was used to suggest personal knowledge of the prosecutor outside the record. On the contrary, the word "we" appears repeatedly, without objection, at other points in the prosecutor's two summations, as clearly referring to all persons present. The reference to Lopez's honesty and carefulness was a permissible effort to highlight for the jury the significance of Lopez's inability to make an in-court identification of Williams. Moreover, the trial judge, who was in the best position to observe the effect on the jury, expressly instructed the jury that it could totally reject the prosecutor's arguments and that such arguments were not evidence.

Spectrographic voice analysis evidence is admissible. The prosecutor's rebuttal summation did not constitute vouching for the credibility of a witness. Accordingly, we find no error below, and the judgments of conviction of Williams and Manning are affirmed.