United States v. Wright 17 C.M.A. 183, 37 C.M.R. 447, 17 USCMA 183

Opinion

Quinn, Chief Judge:

The accused was brought to trial before a general court-martial at Travis Air Force Base, California, on ten specifications alleging violations of Article 134, Uniform Code of Military Justice, 10 USC § 934. The specifications formed two groups of charges. Four specifications alleged the communication of obscene language by telephone to a female; four other specifications alleged the communication of a threat to the same female. The remaining two specifications dealt with a single telephone call to another woman.

Both females were married to service personnel occupying quarters on the base; and both were volunteer Red Cross workers at the base hospital. The accused was a patient at the hospital, receiving treatment for serious injuries sustained in an automobile accident. At trial, the woman involved in the two-specification group testified she received a single obscene and threatening call. She heard the caller make only a brief statement, before she hung up. Within the hour, she listened to a tape recording of another telephone call. While she was “reasonably certain” the voice of the male speaker was the “same” as the voice of the person who made the obscene call, she acknowledged that at the Article 32 investigation, she had expressed “reservation [s]” as “any reasonable person would.” The court-martial acquitted the accused of the two specifications in this group, but found him guilty of all charges alleging obscene and threatening calls to the other woman. It sentenced him to a bad-conduct discharge, confinement at hard labor for six months, and accessory penalties.

On this appeal, the accused contends he was prejudiced by several evidentiary rulings by the law officer. Two rulings resulted in the admission in evidence of the tape recording of a telephone conversation between the accused and the woman involved in the charges, as to which he was found guilty. A third ruling admitted in evidence expert testimony as to a comparison of “voiceprints” of parts of the tape of this conversation with extracts from a tape recording of a previous obscene telephone call.

The first of the obscene and threatening telephone calls was received on Saturday, January 22, 1966. Other calls were received the next day, and *186the matter was reported to the Air Police. With the assistance of the Pacific Telephone Company, a call received by the victim the following day was traced to a pay station in the basement of the hospital, adjacent to the Red Cross lounge. Lieutenant Robert C. Marcan, Chief of the Law Enforcement Branch of the base police, was notified while the call was still in progress. He went immediately to the site, arriving within four minutes of notification. No one was in the booth, but he observed the accused “walking out of the [Red Cross] lounge.” On February 3, the victim was at the lounge. The accused was also present. Apparently, he was an artist, and at the time was discussing with another woman a possible commission for a sketch of her child. The victim asked the accused to do one of her son, but he refused. In listening to the accused, the victim thought his voice “sounded like” the voice of the person making the obscene calls to her. During the afternoon of February 5, she received another call at home. The caller told her she should not have “called the Air Police, because he was . . . really going to get . . . [her] now.” This call was also traced to the pay booth in the basement of the hospital. On this occasion, Staff Sergeant Caskell Sauls and Lieutenant Marcan went to the hospital. Sauls saw the accused in the Red Cross lounge, and asked him for change of a quarter for use in a coffee machine. The accused said he had no change, but he directed Sauls to a place where it could be obtained. In Sauls’ opinion, the accused’s voice “appeared to be the same voice” he “had heard on the telephone with” the victim in one of the obscene calls. As a result, he asked a hospital corpsman to telephone the accused while he listened in on the conversation. This conversation convinced him the accused’s voice was the “same” as the obscene caller’s. He had the corpsman bring the accused to one of the hospital offices for questioning.

Sauls identified himself and Lieutenant Marcan as Air Police investigators. He informed the accused he was suspected of making obscene telephone calls, and advised him of his rights under Article 31, Uniform Code of Military Justice, 10 USC §831. He also advised him he had a right to counsel. The accused indicated he understood his rights; that he did not desire counsel; and he had no objection to being questioned.

After some questions about the accused’s acquaintanceship with volunteer Red Cross women workers at the hospital and the frequency of his use of the Red Cross lounge and the pay telephone, Sauls asked the accused if he would, “for identification of voice,” talk by telephone with a female who had received obscene calls. The accused agreed. A call was placed to the victim and the accused engaged in conversation with her. In the course of the conversation, the accused was asked to speak more slowly and to make his voice “hoarser” to simulate the speed and pitch of the obscene caller. The conversation was recorded on tape. At trial, defense counsel objected to the admission in evidence of this tape. Testifying in connection with the objection, the accused admitted he “realized” he “had a right to refuse” to talk to the lady, but he did “not mind”; it was his own “choice to make the phone call.” He maintained, and Sauls admitted, that he was not previously informed the conversation would be recorded; but he acknowledged that he “figured” it was taped.

The accused’s initial objection to admissibility of the tape was based on a contention that the warning as to his rights was legally insufficient because he was not directly advised that the telephone conversation would be taped. The objection was expanded on appeal to include an allegation of a denial of due process by the manner in which the Air Police secured identification of his voice by the victim.

So far as the scope of the preliminary advise is concerned, the question is not whether the accused knew the conversation was being recorded, but whether the conversation itself is admissible, Secret use of a record*187ing device by one party in conversation with another, in order “to make an accurate record” of the conversation, does not make testimony as to the content of the conversation inadmissible; -in other words, the content and the circumstances of the conversation determine its admissibility, not the fact that a device is secretly used to make a permanent record of it. Consequently, undisclosed recording of the conversation with the victim did not deny or abridge any substantial right of the accused. Osborn v United States, 385 US 323, 326, 17 L ed 2d 394, 87 S Ct 429 (1966); cf. United States v Greer, 3 USCMA 576, 13 CMR 132. The accuracy of the substantive part of the tape was not disputed. The failure, therefore, to inform the accused, preliminarily, that the test conversation would be recorded did not detract from the legal sufficiency of the warning given him as to his rights during the interrogation.

The due process aspect of the attack on the admissibility of the tape of the test call is constructed upon an analogy between the circumstances of that call and the procedure by which a voice identification was effected in Palmer v Peyton, 359 F2d 199 (CA 4th Cir) (1966). Appellate defense counsel contend the victim’s identification of the accused’s voice as the “same voice as the obscene call voice” was the inevitable result of a statement by Sauls that he had “a suspect” to whom she was to speak. However, neither the Palmer case nor the “psychology of suggestion” (see United States v Wade, 388 US 218, 18 L ed 2d 1149, 87 S Ct 1926 (1967)) is of any help to the accused.

In Palmer, the victim of a rape was not allowed to see the accused, but merely to listen to his voice. Her identification, therefore, was deliberately restricted to only one of many factors that bore upon the issue. In addition, the physical conditions under which the identification was made did not nearly approximate those under which the rapist spoke during the assault. The rapist wore a paper bag over his head; while the bag was in possession of the police it was not used in the identification process. Only the voice of the accused was submitted to the victim for consideration. Finally, at trial, the victim testified only to the station house identification, and made no effort to identify him in open court. In this case, the victim identified the tape of the obscene call of January 24, and stated that the voice was that of the person who had made all the calls to her. She further identified the voice in the recording of the test call of February 5, as that of the person who spoke to her at the request of Sergeant Sauls. Both tapes were played in open court. The witness testified that, to the “best of her knowledge,” the male voice in both tapes was that of the same person. She also declared that she talked to “two people on the phone” for “voice identification.” In her identification of the obscene caller, she referred to similar expressions used in each of the calls. Thus, unlike the situation in Palmer, the witness here was provided with an alternative choice and made an affirmative in-court identification of the obscene caller. Under the circumstances, there is nothing that even hints at the possibility the witness was influenced in her voice identification by the manner in which the test call was made, or by Sauls’ description of the person to whom the victim was to speak as a “suspect.” The test call was. presented as only a test of the general quality of the suspect’s voice. In fact, no effort was made to have him duplicate any of the obscene or threatening words used by the obscene caller, which might perhaps have had a tendency to establish a link between the two. There was no violation of due process by the manner in which the recording of the test call was made, or in the manner in which the victim identified the voice of the accused in the test call with the voice of the obscene caller.

Reference was made earlier to voiceprints of the January 24th tape and of the tape of the test call of February 5th between the accused *188and the victim. Voice identification of a person by human ear is a commonplace experience, and has long been recognized in the courts. United States v Wade, supra; United States v Whisenhant, 17 USCMA 117, 37 CMR 381. What is new in this case is that the voice identification was made by a machine which produced a picture of the voice. Appellate defense counsel contend that the scientific principles underlying the voiceprints are so uncertain in theory and practice as to require their exclusion from evidence as a matter of law. They maintain the record of trial demonstrates that voiceprints suffer from unreliability comparable to that which led to exclusion of the results of lie detector tests, United States v Massey, 5 USCMA 514, 18 CMR 138, and statements by' a witness made under the influence of so-called truth serums. United States v Bourchier, 5 USCMA 15, 17 CMR 15. The board of review considered the issue at length and determined that the voiceprints were admissible, despite some evidence in the record of trial of disagreement in the scientific community as to reliability of the machine which made the prints and the validity of the criteria used to interpret the prints. I reach the same conclusion.

The Manual for Courts-Martial, United States, 1951, indicates a witness may testify as an expert and express an opinion on a state of facts within his specialty if he “is skilled in some art, trade, profession or science or . . . has knowledge and experience in relation to matters which are not generally within the knowledge of men of common education and experience.” Id., paragraph 138e. Judged by that test, the Government’s expert witness, Lawrence G. Kersta, was, as pointed out below, a competent witness as to the process of making and interpreting voiceprints from samples of speech. However, the Manual’s definition of a witness qualified to testify as an expert is not entirely determinative of his competency to testify on a particular subject. In United States v Adkins, 5 USCMA 492, 496, 18 CMR 116, for example, we rejected the idea that years of investigating homosexuals qualifies a law enforcement agent to testify that homosexuals tell the truth, or that they are “ ‘products of broken homes.’ ” We pointed out that “expert opinion should bear some measurable relation to empirical observation.”

Mr. Kersta testified he held a Bachelor of Science Degree in Electronics and had been awarded a Master of Science Degree in Physics by Columbia University. For thirty-nine years he worked in the research department of Bell Laboratories, a research and development division of the Bell Telephone Company. His work was “mainly” in the field of speech and the determination of “individual characteristics of speech.” He was one of five persons at the Laboratory who, in 1943, invented a machine known as the “spectograph.” The spectograph analyzes sounds in various ranges and produces a picture of these sounds by means of a stylus tracing a line pattern. The printed picture of the sound pattern is called a spectogram. Starting in 1961, Kersta began to use the spectograph to experiment with voice identification of humans. In a two-year experiment, one hundred and twenty-three persons were selected from the same “dialectical area” to eliminate obvious voice identification factors. With a group of twelve high school students trained by Mr. Kersta acting as voiceprint interpreters, a single word from voice samples of the one hundred and twenty-three subjects was compared to voiceprints in a file of 14,000 prints obtained from the voices of employees at the Laboratory. The students achieved 96.5 percent success in identifying the speaker. Using a five-word sample, the students achieved 99.65 percent success. The results of this experiment were presented to the Acoustical Society of America and were published in a respected British scientific journal.

In March 1966, Mr. Kersta retired from Bell Laboratories and established his own firm to continue his work on voice identification and other sounds. Before this case, he had prepared between four hundred and five *189hundred voiceprints for use in “40 different cases” of law enforcement, and in various medical and psychiatric applications, ranging from voice analysis to determine disease in the vocal tract to the study of speech sounds as reflective of emotional stress. In the specific area of voice identification, he had obtained verification in “practically all of the cases” and, to his knowledge, had “made no mistakes.”

To establish a basis for voice identification, Mr. Kersta required a minimum of sixteen points of identity in the sound pattern on the spectogram of the questioned voice with the sound pattern of the known voice. He testified he compared the February 5th recording with the tape of the obscene call of January 24th, and found twenty-three points of identity. He also testified that uniqueness of the' voice of one individual from that of others is achieved by differences in the way each person uses the muscles or “articulators,” such as the lips and tongue, within his vocal cavity. These differences are manifested in the different sound patterns made by different voices.

Mr. Kersta’s testimony established that his system of voice identification had, experimentally and in practical application, demonstrated a high degree of accuracy and, further, that he was personally qualified to testify as an expert on comparisons of sound patterns made by human voices. True, two defense expert witnesses expressed reservations as to the complete reliability of Mr. Kersta’s system and procedures. The specifics of their reservations need not detain me. Courts have consistently recognized the admissibility of the testimony of experts in areas where there is neither infallibility of result nor unanimity of opinion as to the existence vel non of a particular condition or fact. For example, the difference of opinion among psychiatrists as to the mental condition of a particular person is very well known. See United States v Henderson, 11 USCMA 556, 29 CMR 372; United States v Carey, 11 USCMA 443, 449, 29 CMR 259. Identifying the author of a questioned document by comparison of the handwriting of the document with other handwritings made by known persons is commonplace in the courts, but it certainly cannot be said that all experts in the field and all techniques of identification are infallible. United States v DeLeo, 5 USCMA 148, 153, 17 CMR 148. In fact, visual examination of a questioned document with handwriting exemplars of the accused may lead the fact finders to an opinion different from that of the expert. United States v Privett, 4 CMR 392. Here, the tape recording of one of the obscene calls and the recording of the test call made by the accused were both before the court-martial. Each was played in open court. Since voice identification by ear is fully acceptable in the courts, the court members could thus determine for themselves the margin of error, if any, in Mr. Kersta’s expert opinion.1 With the board of review, therefore, I am satisfied that the shortcomings of Mr. Kersta’s voice-print system did not render his opinion inadmissible.

Appellant’s final assignment of error charges the law officer with improperly curtailing the testimony. of one of his experts. The witness, Dr. Frank R. Clark, was a senior research psychologist at the Stanford Research Institute, “working in the problem of speech transmission and voice recordings.” Dr. Clark testified he was familiar with various techniques used in voice identification, including Mr. Kersta’s. He acknowledged the Ker-sta system “has been demonstrated to result in far better than chance performance in the identification of talkers,” but, in his opinion, it was still “far from what we’d term highly realistic.” In his testimony he referred to certain research in a Massachusetts *190laboratory which attempted “to duplicate” Kersta’s efforts in a “loose sense.” On objection by trial counsel, Dr. Clark was not allowed to testify to the published results of this study. Appellate defense counsel contend the law officer’s ruling was prejudicial because it would have demonstrated that the Kersta system had not had the “general acceptance of the scientific community.” See United States v Ford, 4 USCMA 611, 16 CMR 185.

Appellate defense counsel concede the Massachusetts study was not admissible as “proof of the matters in the study,” but they insist the results would have been of “immeasurable value” in helping the court members understand the reasons for Dr. Clark’s rejection of Kersta’s claim of near-total accuracy in identification by means of his voiceprint system. See Manual for Courts-Martial, supra, paragraph 138e; 32 CJS, Evidence, § 546(63), page 269; Wigmore, Evidence, 3d ed, § 1700; United States v Williams, 16 USCMA 210, 36 CMR 366. Assuming that the law officer should have allowed defense counsel greater latitude in establishing all the underlying factors that led Dr. Clark to his opinion, still no prejudice appears in the particular rulings challenged by the accused.

Dr. Clark, as noted earlier, testified to his familiarity with the work of others in the field, including the experimentation in the Massachusetts laboratory. Dr. Clark admitted the Massachusetts efforts were only in a “loose sense” similar to those of Kersta’s. Consequently, the results of that study could hardly have contributed materially to Dr. Clark’s opinion as to the fallibility of Kersta’s system. Dr. Clark’s testimony made crystal clear that his criticisms of Kersta .were predicated directly upon his own work in the field of voice identification and the apparent deficiencies of Kersta’s, as revealed in Mr. Kersta’s writings and speeches. Assuming he credited the opinions of the Massachusetts experts (cf. Paschal v United States, 306 F2d 398 (CA 5th Cir) (1962)), the record leaves no doubt that these had only peripheral influence in the development of his own opinion. The record also indicates the court members took the tapes and a recording machine into the closed session deliberations. In his final argument, defense counsel had urged them to listen to certain parts of the tapes. I am reasonably sure, therefore, that the court members considered not only Dr. Clark’s testimony as to the fallibility of Kersta’s system, but determined by ear identification whether his criticisms were justified in this case. In light of these circumstances, the restrictions on Clark’s testimony as to the particulars of the work of others in the field present no fair risk of prejudice to the accused.

The decision of the board of review is affirmed.

Judge Kilday concurs in the result.

FERGUSON, Judge

(dissenting):

I dissent.

Under the theme of the principal opinion, the results of a lie detector test would appear to be clearly admissible. A trained polygraph operator can undoubtedly qualify as an expert in his field. The polygraph machine itself has also “experimentally and in practical application, demonstrated a high degree of accuracy.” See In-bau and Reid, Lie Detection and Criminal Interrogation, 3d ed. Nevertheless, it has been unanimously rejected by the courts as evidence, on the basis that it has not been shown to be generally accepted as reliable in the scientific community. It is by that standard that the product of such devices as the voiceprint machine is to be measured in considering its admissibility and not by the self-evident fact that one of its inventors is an expert in its use and the interpretation Of the results. This is the standard which the principal opinion ignores, but which is the necessary predicate for decision in this case.

The accused stands convicted of eight specifications, charging him with making obscene telephone calls to a *191Mrs. Mclver, in violation of Uniform Code of Military Justice, Article 134, 10 USC § 934. There is no question that the calls were made. The only real issue presented is whether accused is the guilty party.

Airman Wright had been seriously injured in an automobile accident and was a long-time paraplegic patient at the base hospital. In fact, his injuries were so severe that, at the time of his trial, he was awaiting medical discharge from the Air Force. It also appears from the record that he was a talented artist and spent much of his time in the hospital’s Red Cross lounge drawing pictures. The telephone used in making the calls was located in or near this lounge. The victim was a Red Cross aide who had seen the accused there.

Mrs. Mclver testified to receiving the calls in question, as well as listening to the accused speak in a test call made to her and on other occasions. While she expressed the opinion his voice and that of the perpetrator of the oifenses were the same, she expressed some doubts, particularly as to the first call she received. In that instance, the voice “was low, . . . It was hard to understand.” She “can’t say definitely” it was the same individual. In fact, she replied, “I guess not” to the inquiry “whether it was or not or have [you] an opinion?” She likewise conceded that, in the pretrial investigation, she had testified, “how can you say from a voice, I mean it sounded like the same voice, I can’t positively say that is the voice because it’s just hard.” As to whether the voice on any of the calls was that of the accused, she testified that, as she had also stated in the pretrial, she was not certain “beyond a reasonable doubt.” Finally, she admitted she was asked at the investigation if “you cannot say it’s the same person;” to which she replied, “how can anyone, really.”

Lieutenant Mclver, the victim’s husband, listened on an extension to one call and testified that, in his opinion, the voice was the same as that of the accused, with whom he had not been previously acquainted. The two Air Police investigators on the case likewise expressed the view that the accused’s voice and that of the caller on some of the calls were the same.

The foregoing constitutes the prosecution’s evidence regarding identification other than that supplied by Mr. Lawrence G. Kersta, regarding the voiceprints in the case. Kersta does business under the name of Voiceprint Laboratories. Holding a Bachelor of Science degree in Electronics and a degree of Master of Science in Physics, he was employed by Bell Telephone Laboratories until 1966, when he retired and founded his own business. He is one of five inventors of a machine known as a Spectograph, designed in 1943 as a tool for speech research. In 1961, Kersta commenced a project of research into voice identification. He claimed that each human’s speech organs are sufficiently unique that the Spectograph is able to record visual tracing of the differences and thereby identify voices. The visual tracing is known as a “voiceprint,” and he is able to make an identification if two samples share at least sixteen points of similarity.

The basis for Kersta’s conclusion is found in research conducted on one hundred and twenty-three individuals from the same area of the country, speaking the same dialect, and chosen from 14,000 Bell Telephone employees. He claimed approximately ninety-nine percent accuracy in voice identifications made during this sampling. So far as the record reveals, this is the sole basis for the claim that voiceprints are practically the equivalent of fingerprints in reliability.

Kersta’s findings were published in a British periodical entitled “Nature Magazine.” He has been involved in the investigative phases of approximately forty criminal cases for fifteen different police agencies. He has appeared before two grand juries. The voiceprint evidence has been accepted in only on criminal trial — in West-chester County, New York — which ended in a jury disagreement. He agreed that his research and conclusions were totally based on use of “a *192small population group.” He had received praise for his work, and knew of no criticism of it.1

The law officer overruled defense objections to the receipt of Mr. Ker-sta’s opinion, based on voiceprints, that the voice of the caller and that of the accused were the same. The defense based his contention of inadmissibility on the lack of proof of the device’s general acceptance by the scientific community.

For the defense, two expert witnesses appeared. Dr. Frank R. Clark, a senior research psychologist at Stanford Research Institute in the area of speech transmission and voice recording, testified to his familiarity with Mr. Kersta’s work, as well as a number of other techniques in the same area. In his opinion, voiceprint identification was “far from what we’d term highly realistic,” though better than chance performance. Voice print identification was not established to the satisfaction of the scientific community in general. The law officer, however, refused to allow Dr. Clark to testify that, to his knowledge, neither “the professionals in the area, nor the procurement officers [for Government research projects] feel that this is a solved area.” The witness was limited to his “personal knowledge of experiments or results of tests” in giving his opinion on the reliability of voice-print tests. Specifically, he refused to receive testimony regarding experiments conducted by Bolt, Beranek, and Newman, a research laboratory in Cambridge, Massachusetts, with which Dr. Clark stated he was “personally acquainted.”2

Clark concluded his testimony with reference to his own research, which had indicated only sixty percent accuracy by the spectographic method, as compared with ten percent better results through aural voice identification. Other similar devices have shown a rate of error of thirty-two percent.

Dr. Cletus J. Burke testified that his studies had convinced him, as an expert witness, of the statistical unreliability of Mr. Kersta’s claims of the virtual infallibility of the voice-print process. Although he professed familiarity with Kersta’s reputation and work and the degree of its acceptance by the scientific community, he was not permitted to give his view that Kersta “has not really presented convincing evidence on the area that we are concerned with here.” In his personal opinion, the work was still experimental.

I

The first issue before us is whether the voiceprint method of speaker identification is generally accepted by the scientific community as reliable to the point where it should be received as evidence in a criminal trial. That is the principle which must govern us. As was succinctly stated by this Court in United States v Ford, 4 USCMA 611, 16 CMR 185, the inquiry must be whether “the validity of the tests forming the basis of the expert’s opinion . . . are sufficiently well established to merit general acceptance in the particular field.” Id., at page 615. And, as the Chief Judge — contrary to his present position — there noted, this matter is not ended by mere qualification as an expert and expression of an opinion. Ford, supra, at page 613. There must be general acceptance in the scientific community of the basis for that opinion. See, generally, 29 Am Jur 2d, Evi*193dence, §§ 825-831, and Wigmore, Evidence, 3d ed, § 795. See also Annotation, 66 ALR2d 536, 547. Cf. United States v Russell, 15 USCMA 76, 35 CMR 48.

Thus, in People v Morse, 325 Mich 270, 38 NW2d 322 (1949), testimony-concerning a “drunkometer” was held inadmissible, the court declaring, at page 324:

“There is no testimony in the record that there is general acceptance by the medical profession or general scientific recognition of the results of a Harger Drunkometer test as accurately establishing the alcoholic content of a subject’s blood and thus the extent of his intoxication.”

Of another device, in People v Forte, 279 NY 204, 18 NE2d 31 (1938), the court remarked, at page 32:

“. . . The record is devoid of evidence tending to show a general scientific recognition that the path-ometer possesses efficacy. Evidence relating to handwriting, finger printing and ballistics is recognized by experts as possessing such value that reasonable certainty can follow from tests. Until such a fact, if it be a fact, is demonstrated by qualified experts in respect to the ‘lie detector,’ we cannot hold as matter of law that error was committed in refusing to allow defendant to experiment with it.”

Finally, in Frye v United States, 293 Fed 1013 (CA DC Cir) (1923), the Court of Appeals clearly stated the principle applicable here as follows :

“. . . 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 he sufficiently established to have gained general acceptance in the particular field in which it belongs.” [Emphasis supplied.]

Turning to the ease before us, we find the sole evidence on this record to be the claim of one of the developers of the Speetograph that it is virtually infallible in producing voice identification. As to its acceptance by the scientific community, the witness merely remarked that he had heard no criticism of his device and techniques. The evidence which he offered to back up his claims consisted of testing no more than one hundred and twenty-three individuals, all of whom were from the same section of the country and the same social stratum. In short, the prosecution’s evidence is- absolutely devoid of proof of any general acceptance of mechanical and electronic voice identification devices by the scientific community.

Indeed, the defense experts, one in the area of speech and the other in the area of statistics and psychology, testified — as clearly as they could under the strictures imposed by the law officer — that Kersta’s device had not been so accepted; that the statistical basis for his conclusions was erroneous; and that similar experiments, to their personal knowledge, had indicated errors as high as thirty-two percent in voice identification. And; on review, the board of review, alluding to extensive studies by an independent research firm on behalf of military agencies, pointed out that the result of this investigation was the finding that, though the instruments had value, they were not even as accurate as human recognition of voices.

Finally, in State v Cary, 49 NJ 343, 230 Atl 2d 384 (1967), the Supreme Court of New Jersey remanded the case before it for further inquiry into the reliability of voiceprint identification. Though Mr. Kersta there also testified at the trial level as to the infallibility and uniqueness of his machine and procedures, the court expressly noted it felt “that something more than the bare opinion of one man, however qualified, is required. Certainly, the prosecutor must satisfy the trial judge that identification by *194voiceprint technique and equipment has a sufficient scientific basis to produce uniform and reasonably reliable results and will contribute materially to the ascertainment of truth.”

Undoubtedly, Mr. Kersta is an expert in the field of voice identification and his device may well be of value. Indeed, there may be evidence available to support the conclusion that this voiceprint comparison process is generally accepted by scientists as an infallible means of voice identification. But that evidence is not shown in this record; in fact, the contrary testimony of the defense witnesses is all that is made available to us. Hence, I am required to conclude the Government here made no satisfactory showing of the predicate for admissibility of the results of a new and otherwise untried test and that the law officer erred in permitting the v.oiceprints to be admitted and in receiving Mr. Kersta’s opinion based thereon.

It is likewise certain that the error was prejudicial. The trial was virtually concerned with nothing else but the efficacy and infallibility of the voiceprint process. The court’s questions were directed almost completely to its effectiveness and demonstrated the members’ extreme interest in its identification of the accused, to the exclusion of all others. Moreover, the trial counsel made the process the focal point of his argument, devoting approximately sixty percent thereof to its reliability. Finally, the evidence other than the voiceprints is far from compelling, with the victim having expressed doubts at the pretrial investigation and her husband having heard the voice in question for only a brief period. In short, the Government has made it clear from the beginning that its main prop here was the “scientific” evidence on which it relied, and it is certain that its erroneous use presents a fair risk of prejudice to the accused.

II

I am likewise certain that prejudicial error occurred when the law officer restricted the defense expert witnesses, Drs. Clark and Burke, to testimony based on their personal knowledge, as if they were ordinary witnesses. The correct rule has been set forth by a leading authority:

“. . . The data of every science are enormous in scope and variety. No one professional man can know from personal observation more than a minute fraction of the data which he must every day treat as working truths. Hence, a reliance on the reported data of fellow-scientists, learned by perusing their reports in books and journals. The law must and does accept this kind of knowledge from scientific men. On the other hand, a mere layman, who comes to court and alleges a fact which he has learned only by reading a medical or a mathematical book, cannot be heard. But, on the other hand, to reject a professional physician or mathematician because the fact or some facts to which he testifies are known to him only upon the authority of others would be to ignore the accepted methods of professional work and to insist on finical and impossible standards.

“. . . The decisions show in general a liberal attitude in receiving technical testimony based on professional reading.” [Wigmore, supra, § 665b.]

The Manual for Courts-Martial, United States, 1951, itself recognizes this rule, for it provides:

“. . . On direct or cross-examination he [the expert witness] may be required to specify the data upon which his opinion is based. . . .” [Manual, supra, paragraph 138e.]

This provision is binding law. United States v Smith, 13 USCMA 105, 32 CMR 105. In United States v Williams, 16 USCMA 210, 36 CMR 366, we applied it to hold prejudicial error when the law officer denied the accused the opportunity so to test the validity of an expert opinion and attack its weight. We have likewise specifically upheld the right of expert witnesses to base their opinions on matters not in evidence and to ex*195press that foundation in open court. United States v Heilman, 12 USCMA 648, 31 CMR 234; United States v Walker, 12 USCMA 658, 31 CMR 244.

In addition, other authorities fully uphold the right of the expert to refer to scientific publications in giving his views upon a particular subject in issue. Thus, in State v Nicolosi, 228 La 65, 81 So 2d 771 (1955), the court pertinently pointed out that “it would be inconsistent to receive in evidence the opinion of an expert witness but exclude the scientific article on which he partially bases this opinion.” Id., at page 772. See also Reilly v Pinkus, 338 US 269, 94 L ed 63, 70 S Ct 110 (1949), and 20 Am Jur, Evidence, § 797.

Here, the law officer precisely and repeatedly ruled that the two experts would be limited to matters of personal knowledge in their testimony regarding the efficacy of Kersta’s device. Particularly, Dr. Clark was not allowed to refer at all to the study published by Bolt, Beranek, and Newman, which he regarded as the major effort in the area of voice identification and which bore directly on the Kersta method. In short, having received the evidence of voiceprint identification without the slightest evidence of its acceptance as scientifically reliable, the law officer diligently cut off every defense effort to impeach its validity or to attack its allegedly infallible identification. Under the authorities cited, all of which support a basic rule of universal application, these rulings were erroneous.

Again, under the circumstances, there was prejudice to the accused. Kersta’s “scientific” identification of the accused went to the jury virtually unimpeached because of the narrow scope accorded the defense. They were allowed to know nothing of the extensive research and published studies which underpinned Clark’s opinion and that of Dr. Burke. Yet, Kersta was permitted to testify at will concerning the accuracy of his method, the scientific praise it had received, and the fact that he had heard no criticism. In my view, therefore, the defense was never really allowed an opportunity to meet the prosecution’s case. Hence, I would likewise order reversal on this ground.

As these matters serve to dispose of the case, I need not dilate at length on the remaining questions treated by the Chief Judge. As to the test call to Mrs. Mclver, suffice it to say that the record indicates accused was properly warned; that he voluntarily made the call; and that he admitted believing it would be recorded. Knowing these things, he is hardly now in a position to complain that he should have been advised it was to be recorded.

Ill

In sum, then, I cannot agree that there is made out on this record that general scientific acceptance of the voiceprint machine and procedures which justify its reception in evidence. In the words of the New Jersey Supreme Court, “something more than the bare opinion of one man, however qualified, is required.” State v Cary, supra. According to the matters considered by the board of review, the armed services themselves have rejected the technique for voice identification in operational matters. It hardly seems open to us, therefore, to make it the principal basis for a criminal conviction and sentence.

In like manner, I deem it equally prejudicial to have accepted the evidence and then effectively prevented the defense from attacking its accuracy and reliability. So to act is simply to place the accused in the hands of a new and untried device and make its findings conclusive on the fact finders. I, for one, am unwilling to hand their functions over to a machine, and, accordingly, I register my disagreement.

I would reverse the decision of the board of review and order a rehearing.