Orne, M. T., Thackray, R. I., & Paskewitz, D. A. On the detection of deception A model for the study of the physiological effects of psychological stimuli. In N. S. Greenfield & R. A. Sternbach (Eds.), Handbook of psychophysiology. New York: Holt, Rinehart & Winston, 1972. Pp.743-785.
It is widely believed that lying is accompanied by specific perceptible physiological or behavioral alterations. Thus a mother tells her child, "Look at me," in order to assure his telling the truth, because she believes that lying is associated with averting the gaze. Other idiosyncratic behavioral manifestations may be recognized by those close to an individual when he purposefully decides to lie. These may include signs of tension, agitation, respiratory changes, a catch in one's voice, blanching, blushing, etc. In addition to physiological changes readily visible to an observer, more subtle ways of determining physiological responsivity may be used. In ancient China an individual was required to speak with his mouth full of rice in order to test the truth of his statement (Boring, 1942). This rather primitive "lie detector" might have worked, in some instances at least,
743
744 APPLICATIONS OF TECHNIQUES AND FINDINGS
because emotion could interfere with salivation, thus differentially increasing the liar's difficulty in enunciating under those circumstances. (See Larson, 1932, for an excellent review of the history of the detection of deception.)
Shortly after the turn of the century, modifications of Galton's (1879) word association technique were used to detect lying by investigating associations and differential reaction times to stimuli associated with material about which an individual hopes to deceive (Jung, 1906; Wertheimer & Klein, 1904). Somewhat later, Luria (1932) showed that psychomotor coordination can be impaired while lying, using a technique which required subjects to hold one hand steady while depressing a plunger with the other. These techniques, while of theoretical interest, have had very limited practical application. It remained for Keeler (1930) to develop and combine the work of several previous workers in a simple polygraph to measure physiological responsivity as a technique for the detection of deception. His device measured respiration, "relative blood pressure," and the galvanic skin response (GSR).
PRINCIPLES UNDERLYING THE DETECTION OF DECEPTION
Despite the widespread use of the polygraph technique by police, governmental agencies, and industry to detect deception, considerable misunderstanding persists about its mode of operation. Contrary to popular superstition, no specific physiological response has ever been identified that is unique to lying. On the contrary, the kinds of behavioral and physiological responses associated with deception are also characteristic of arousal, anxiety, stress, etc. The detection of deception depends upon a comparison of the subject's responses to two or more stimuli matched in their presumed ability to arouse the subject. Some of these stimuli are known not to be associated with deception, whereas one or more of the others may be associated with deception. If the individual's responses to these basically similar stimuli are considerably different in the case of those stimuli about which deception is suspected, a diagnosis of lying is made. The parameter on which this comparison is made can range from reaction time (as in Jung's experiments); to psychomotor coordination (as in Luria's); and to physiological parameters, as in a more typical "lie detection" procedure. Regardless of the parameter being measured, the model remains a comparison of matched stimuli, one set of which may involve deception. The procedure rests on a number of assumptions, which include an adequate matching of stimuli, as well as differential responsivity on one or more of the parameters being measured.
The Usefulness of the Paradigm
It is striking that a phenomenon such as the detection of deception has received so little systematic investigation. Perhaps this is partly explained by the psychologist's irritation with laymen who characteristically refer to all polygraphs as "lie detectors." Yet the phenomenon is probably a fundamental paradigm for psychophysiology. Consider the card test, a demonstration used both in certain field applications and in laboratory studies. A subject is asked to select and remember one of six cards. As the experimenter names each of the cards, he is required to answer "No" each time, thereby "lying" about the card he actually did select. The physiological response (typically GSR) to each of the stimuli is compared with the other responses and, under appropriate conditions, the greatest GSR will have been evoked by the card selected. Such a simple procedure remains one of the best illustrations of psychosomatic relationships. One assumes that prior to the selection of any one card, each of them would have evoked an essentially similar response, whereas afterward a greater physiological response is elicited by the card that has been selected. The process of selecting the card, then, differentially alters the individual's physiological responsivity to it. This procedure presents a rather simple way of studying the differential effect of past experience on the physiological responsivity of the individual. It should be emphasized that this simple procedure will not automatically guarantee that the greatest physiological response is associated with a selected card. On the contrary, as will be discussed later, situations varying the consequences of deception, the kind of response required of the subject, his antecedent
745 On the Detection of Deception
experience, the kind of parameters being tested, etc., will dramatically affect the incidence with which a selected card will evoke a differentially greater physiological response.
It is possible, therefore, to view the apparently simple detection of deception situation as a meaningful way of investigating the contingencies under which an apparently matched stimulus can, by specifiable previous experience, become capable of arousing differential physiological responsivity. Whereas in most psychophysiological research one attempts to infer the fact that a stimulus has been meaningful from the physiological response of the individual, this paradigm allows an investigator to specify the meaningful stimulus (i.e., the selected card) and determine the particular kind of psychological situation under which it will lead to a differentially augmented physiological response. As will be shown later, psychological rather than physiological factors tend to determine the presence of differentially greater responsivity. This paradigm then becomes a technique for systematically investigating the contingencies under which previously neutral stimuli become capable of eliciting alterations in physiological responses.
Sources of Information Concerning the Detection of Deception
There are essentially two bodies of literature dealing with the detection of deception: a small number of studies by psychologists published in professional journals and a relatively large number of manuals, journals, and reports written by "professional lie detection experts." The concerns of these two groups have, as a rule, been dissimilar. The practitioner involved in "lie detection" tries to structure a situation in which he can successfully make a diagnosis of deception. His concern typically centers about reducing the number of cases in which he cannot make a clear-cut decision in his own mind on whether an individual is guilty or innocent. Because of a variety of reasons, not the least of which is the technical difficulty of establishing "truth" in an absolute sense, few reports deal with attempts at validating findings. No attempts are made to vary procedures in a systematic fashion, in order to establish the relative contribution of different aspects of the procedures to the diagnosis of deception. The diagnosis of guilt or innocence is not typically made on the basis of a single measure or test, but represents a global decision, including all of the information known about the case.
In direct contrast to reports from the field are laboratory studies that are characteristically concerned with varying procedures and studying the effect on detection, using explicitly defined criteria as dependent variables. Not only are the motivations of subjects and the context in which these studies are carried out different, but neither the procedures nor the dependent variables are readily comparable. The problems raised by using, in one context, the data derived from another will become more evident in later sections. It seems most appropriate to discuss the field situation separately, since it is more closely akin to a clinical art developed by practitioners who are faced with the task of making the diagnosis of deception. In order to evaluate the field situation, it is necessary to appreciate the context in which the detection of deception is carried out, the procedures that are utilized, and the kinds of criteria that are employed.
LIE DETECTION IN THE FIELD
Typically, "lie detection" procedures are carried out by individuals who are trained interrogators but who may have only limited formal training in psychology, psychophysiology, and related fields. The technology has developed empirically, largely through the influence of Keeler (1930), his associates and successors, Inbau and Reid (1953), and Backster (1963). A body of knowledge, phrased in interrogation terminology, has been developed to explain and teach the rationale and the application of this technique. Usually, the interrogators have taken one or more courses from individuals who have had considerable experience with the technique. Currently, various attempts are being made to specify standards of training for interrogators who are to use this procedure.
Polygraphic interrogation techniques have been used in three major contexts: (a) the interrogation of suspects during criminal investigation procedures; (b) the screening of individuals for security
746 APPLICATIONS OF TECHNIQUES AND FINDINGS
purposes in the government; and (c) the screening of individuals who are or may be placed in positions to convert goods or defraud their employers on a more or less routine basis, in order to identify individuals guilty of crimes, who might not otherwise be detected.
Interrogation Procedure
For a detailed discussion of various procedures the reader is referred to Reid and Inbau (1966). It seems useful, however, to summarize certain aspects of the procedure. We shall describe the situation characteristic of lie detection in the context of criminal investigation. Typically, the polygraph operator sees the suspect in a relatively quiet, plainly furnished, comfortable office. He is acquainted with the suspect's dossier and has obtained as much factual information as he can. He is introduced to the suspect and is usually alone in the room with him, although the interview may be recorded and observed through a one-way screen.
The pretest interview Once the interrogator meets the suspect, he explains that he will shortly conduct a polygraph examination and that the "lie detector" is a scientific instrument which will record his physiological reactions and indicate when he is lying. He tends to state flatly that this can be done with certainty, implying that deception is pointless because the record will reveal this deception. A concerted effort is made to structure the interview as an opportunity for the suspect to prove himself innocent by telling his side of the story. The machine is usually visible in the background but is not directly mentioned at this point. The whole tone of the initial conversation is designed to maximize the suspect's belief in the infallibility of the "lie detector," a belief that, to some extent at least, tends to be shared by the interrogator. In the context of the impending polygraph examination, the interrogator goes over the material in the suspect's history. Since he is trained in his task, he will typically sympathize with the suspect, provide face-saving rationalizations for any crimes that might have been committed, and indicate subtly that he can understand how someone might murder but that lying is the lowest form of degeneracy. Such a procedure may sound farfetched, yet in the hands of an expert it is remarkably convincing. Against the background of the forthcoming examination, he goes over all of the crucial points and at various stages casually indicates that the suspect might as well tell the truth now, since it will certainly come out later, and that by confessing he will earn the interrogator's respect.
Unlike other interrogation procedures, inconsistencies in the suspect's answers and possibly deceptive replies are not challenged by the interrogator during the pretest interview, except perhaps subtly during the course of formulating questions. It is important that the suspect not feel that the interrogator is trying to obtain admission of guilt, but merely eliciting information essential for the conduct of an adequate examination. This behavior is quite different from the usual behavior of an interrogator, where inconsistencies are used as leverage to force the suspect to reveal the truth. Since interrogators are trained to bear down when they suspect deception, and the pretest interview specifically prohibits such behavior, many polygraph practitioners are genuinely convinced that interrogation is not the real purpose of this interview. The examiner's belief that this is the case helps maximize the effectiveness of the interview, which often, in fact, elicits unexpected admissions of guilt prior to the actual test. It is, of course, a matter of semantics whether one wishes to consider these cases as examples of successful lie detection. While the polygraph record does not lead to the detection of deception in these instances, the existence of the polygraph creates a context without which a confession might not have been obtained.
The interrogator will then carefully discuss with the suspect the questions he will ask. There are, of course, different procedures used by different individuals in this regard, but the most widely used procedures, "peak of tension," "zone of comparison," and their various adaptations, all employ a similar pretest strategy. Approximately 12 questions are usually formulated, including a number of known, neutral control items such as: "Is your name John Doe?" "Are you now in the United States?" "Is today_______________ ?" Also included will be certain questions designed to elicit a nonspecific emo-
747 ON THE DETECTION OF DECEPTION
tional response, questions such as: "Have you ever stolen anything?" Finally, two or three crucial questions relating to the specific purpose of the interrogation will be included.
The interrogator will go over these questions in detail with the suspect, defining for him exactly what each aspect of the question should mean, to be certain that it will not be necessary for him to lie unless he is guilty. The pretest interview and the discussion of the questions to be asked during the test provide an unusual opportunity to elicit information. The interrogator usually approaches the suspect in a very calm, matter-of-fact fashion. The questions he asks in the process of clarifying the subsequent "lie detection" test are for the ostensible purpose of protecting the individual from giving a false guilty response.
The formulation of questions The formulation of appropriate questions is considered to be a crucial part of the procedure. It is in the formulation of questions that an attempt is made to take into account the individual's overall state of tension. An innocent individual may have cause to be highly aroused and very frightened in the test situation. The interrogator will attempt to include several emotion-arousing questions, not related to the specific crime, in order to have an estimate of the individual's responsivity to embarrassing questions. In an examination concerning a particular murder, for example, he might ask the individual about some other murder. This second murder might well be a fiction made up by the interrogator but to the innocent suspect it will be as real as is the first murder. Asking about a fictitious murder permits the interrogator to compare the responses to the two questions about the murders, one in which the suspect is known to be innocent and the other in which his involvement is questioned. It is, of course, essential that the interrogator ask questions in a manner which convinces the person that he is really suspected of each crime.
In an ideal case, some crucial questions may be asked about material that is known only to the guilty person and the interrogator. If, for example, the murdered person has been stabbed with a letter opener, and this fact has not been released to the public, questions may be formulated asking whether the victim was shot, strangled, or stabbed. The innocent individual's responses should be similar for all these questions, whereas the guilty individual can be expected to respond differently to the relevant item.
It is difficult to provide a real picture of the pretest interview that is such a crucial part of "lie detection." This interview carried out prior to the test, with the machine discreetly in the background and yet very much present as the ultimate scientific arbiter of truth, is useful not only to elicit information but also allows a quasi-alliance to be formed between the interrogator and the suspect vis-a-vis the machine. The very process of formulating questions and clarifying them with the suspect is immensely effective in maximizing the conviction that the machine will, in fact, detect lying unless both the interrogator and the suspect make absolutely certain that no lie is involved. In this context, a lie is carefully defined. As an example, an individual might be accused of stealing several hundred dollars from his employer. During the pretest interview, the interrogator asks him about a previous job in a supermarket. While formulating the questions, the interrogator asks, "Did you ever steal from the market?" If the suspect says, "No," the interrogator asks, "Did you ever, for example, run out of cigarettes and take a pack off the shelf?" If the suspect admits, after a few minutes, that he probably has, although he cannot remember specific instances, the interrogator then asks whether perhaps he might have helped himself to some cookies from a broken package. Again the suspect might admit that this could have happened. The interrogator then explains that this is not what was meant by stealing, but that stealing meant only the actual taking of money or merchandise of significant value, say in excess of $5, for his own use or to sell. He would not, then, mean an occasional cookie, cigarette, or piece of chewing gum. Thus he says, "When I ask you, 'Did you ever steal from the market?', you now know that I will be referring to taking merchandise valued in excess of $5 or to taking a similar or greater amount of money." Note that the interrogator apparently takes great care to prevent the suspect from giving responses not elicited by the issue in question. In so doing he implies that it is vital that there be no confusion about the
748 APPLICATIONS OF TECHNIQUES AND FINDINGS
meaning of the questions, lest the polygraph pick up lying, and, incidentally, that he, himself, has some understanding of trivial transgressions. He provides ample opportunity for the suspect to mention that he might have stolen, just so that there will be no confusion, since this stealing was quite irrelevant to the crime under investigation. The situation is so designed that a truthful interchange seems obviously essential for the suspect's own good.
Only after the questions that are to be asked have been carefully worked out does the actual examination begin. The pretest discussion may take an hour or more. It is emphasized that the suspect should be put at ease as much as possible prior to the examination; he should be properly rested and not medicated. The pretest interview is intended to reassure the suspect about the infallibility of the polygraph, in the hope that this will reduce the anxiety of the innocent and maximize the responsivity of the guilty.
The test procedure The typical commercial polygraph measures respiration using a pneumograph, a form of occlusion plethysmography of the arm often erroneously called "relative blood pressure" because it uses a pneumatic system with a blood pressure cuff,1 inflated to a point midway between systolic and diastolic pressure, and GSR with simple, stainless steel electrodes without paste. Because the inflated blood pressure cuff soon becomes uncomfortable, the actual test is limited to about 10 or 12 questions taking only about 2 or 3 minutes.
After the polygraph has been attached, the interrogator may demonstrate its use by asking the suspect to select one of a number of cards and requesting that he answer "No" to all questions about these cards, forcing him to lie. He will then run through a number of questions concerning which card the suspect selected. At the conclusion of this demonstration, he will invariably be able to tell the suspect which card he selected. Certainty is possible in this instance because the interrogator ascertains the card in advance, the purpose of this procedure being to underline the infallibility of the polygraph. While a card test helps in these measures, it is far less important -- or effective -- than the pretest interview. If it is felt that the suspect already has the right attitude, the card test may be omitted. 2
During the actual test the suspect is seated so that he cannot see the record on the machine or the interrogator. Each of the questions the suspect is to answer has been discussed in detail in advance and is formulated to require a simple "Yes" or "No" answer. The interrogator waits long enough between each question to allow the three measures to return to a fairly stable baseline. After the interrogator has gone through the 12 questions once, the blood pressure cuff is deflated, and the suspect is given a rest period which, however, may be used as an integral part of the procedure. If, for example, a response to some item suggests guilt, the interrogator may well ask the suspect about it, in an attempt to account for this physiological response. Here again the interrogator takes the role of trying to help the suspect vis-a-vis the machine. In the case discussed earlier, for example, the suspect might have shown a reaction to the question, "Did you ever steal from the supermarket?" The interrogator might say, "Look, what really happened? You didn't tell me. You have a response there." The suspect may then recall that on one particular day he did, in fact, take $10 because he was short of money. He might say that he had always meant to return the money and hadn't thought of it as theft but he didn't get around to it. The interrogator might then say, "Fine, now let's be clear that the next time I ask you about this we will not think of stealing as meaning this $10 which you borrowed and intended to return." The test is then repeated, after clarifying what is meant by stealing, and the physiological response previously associated with the question may now no longer be elicited. Any apparent unexplained responses in the test are utilized in this manner to elicit more information. Such, of course, is also the case for the material concerning the crime. In most cases, the suspect is tested at least twice on the questions. Between the first and second tests, however, the interrogator may attempt to have the suspect reconcile and
1 This test may instead be given following an initial examination on the relevant material.
2 See p. 764 for a further discussion of this point.
749 ON THE DETECTION OF DECEPTION
explain the reason for any undue physiological response that might indicate knowledge of the crime.
Scoring The scoring of the physiological parameters is not specified in quantitative terms. Although examples of various types of deceptive responses are given in sources such as Reid and Inbau (1966), it is difficult to say that they indicate much more than an undue response. A change in respiration, blood pressure, or GSR beyond what would seem normal is considered important, but no specific physiological change is uniquely associated with lying. Instead, deception is diagnosed by comparing the subject's response on relevant items with his response on items that should be equally arousing or anxiety-provoking to an individual who is not lying.
If some particular item of information is available that would be known to the suspect but not to an innocent individual, the "peak of tension" technique may be used. In this procedure, a series of questions is asked in an order previously established with the suspect; the crucial item is typically placed third or fourth. The operator looks for a gradual increase in the amount of tension elicited by each item as the crucial one is approached and a sudden relaxation once it is passed. Items of information that involve numbers are particularly suited to this procedure. If, for example, $212 were taken in a theft and this amount is not generally known, the suspect might be asked, "Was the amount taken $56, $102, $160, $212, $418?" The suspect who knows the correct amount should not only show his maximal response to $212 but also demonstrate a gradual increase in tension that peaks at $212, with a sudden relaxation as the crucial item is passed.
Screening procedures When the polygraph is used in screening employees or prospective personnel, the procedures are much the same as in criminal investigations but there are certain differences. The problem facing the interrogator in a criminal case is to develop appropriate control questions to enable him to interpret the responses to questions concerning the crime. In screening, the interrogator does not have a set of ready-made questions about a specific event. The interview for pre-employment (or "routine" screening) is designed to discover whether or not the particular person being tested is guilty of certain transgressions, involvements, or tendencies that would render him unsuitable for the position. The polygraph, then, not only serves as a powerful lever in the pretest situation but serves as an "independent" measure of the veridicality of statements concerning the questions asked. The kinds of issues that are explored in the screening procedure tend to include material of a private, personal, and confidential nature, which under most circumstances would not legitimately be viewed as relevant to an individual's normal employment. The use of the polygraph in screening has been subject to considerable criticism (House of Representatives, 1964). The issues, however, relate less to its utility in such a context than to the kinds of questions explored. It should be realized that the questions asked, rather than the attempted validation of the answers, are objectionable to individuals. The issue of what aspects of an individual's life may be searched, and under what conditions, either by an employer or by the government, is an ethical and moral problem, separate from the problem of the effectiveness of the instrument and, therefore, beyond the scope of this presentation.
Evaluation of the Technique
When one considers the use of the polygraph in a real life situation, it becomes clear why it is difficult to evaluate. The interrogation procedure is an art rather than a science -- much depends upon the manner in which the individual interrogator formulates his questions, the tone of his voice, and the subtle cues he may provide in the interrogation, etc. The experienced interrogator must intuitively deal with problems inherent in such issues as individual differences in reactivity, response specificity, and the overall state of the person being questioned. Interrogators who have worked with "lie detection" will readily admit that the psychological edge the machine provides during the pretest interview is, from their point of view, of major importance. Some go on to say that, even if the machine were out of order, it would still be a major asset. Scientific evaluation of the validity of the "lie
750 APPLICATIONS OF TECHNIQUES AND FINDINGS
detection" procedure is very difficult, since the test itself is only a small part of the total procedure, and much of the interrogator's attachment to the polygraph undoubtedly is related to the psychological effect it will have on the suspect.
In the past, the literature on "lie detection" written by the professional expert has tended to emphasize the scientific aspect of this technique. The polygraph is presented to the suspect as a scientific way of recording his physiological responses, which indeed it is. It is also implied -- and at times even stated -- that the polygraph is a scientific way to record lying. No adequate evaluation of the validity of the polygraph in this regard is yet available (Orlansky, 1962). Perhaps there has been an unfortunate disinclination toward evaluation by practitioners in this area, yet it must also be acknowledged that to do so is extremely difficult. How does one appropriately determine percentages of correct detection? Assume that one individual admitted a theft and polygraph examinations were carried out among 100 suspects, with all individuals apparently innocent; does this justify the claim of 99% accuracy that has been made in one instance? If, indeed, the interrogator considered the contingencies in each instance to be 50-50 instead of 100 to 1, such a claim might not be quite as absurd as it seems.
In evaluating "lie detection," it would perhaps be most appropriate to ask not how well the procedure works under ideal circumstances but rather to consider the whole technique as a unit -- the interrogator, the test, and his interpretation of the test -- and ask how effective this combination is in determining the truth. In the absence of any satisfactory way of establishing what truth really is, such an evaluation cannot be carried out. To validate against a jury's decision of guilt is a dubious procedure at best since criminal trials are adversary proceedings carried out in a highly formalized manner. The courts are, of necessity, more concerned about making certain that only legally obtained evidence is used according to correct procedure than in the guilt or innocence of any one particular individual. The purpose of a jury is to provide a socially useful criterion of fact in the absence of more reliable ways of determining truth. A panel of experts evaluating all of the facts in a case, including those inadmissible in court, might be more accurate. A criterion measure of this sort was employed by Bersh (1969), who found better than 90% agreement between polygraph examiners' judgments of deception, based on both examinations and knowledge of details of the case, and unanimous judgments of guilt made by a panel of four judge Advocate General attorneys. This panel independently reviewed case files from which all polygraph material was deleted. Bersh felt that such a panel provided the best practical criterion against which to compare examiner judgments. Although an excellent study, errors of classification are possible, even with such a panel. Other ways have been proposed to validate the procedure. These include subsequent confessions, the uncovering of evidence that would prove innocence after the test, and uncovering circumstantial proof of crimes by such a procedure. Each of these presents serious problems. In real life, polygraph examinations given during a period when some doubt as to guilt exists are seldom those in which such doubt is subsequently resolved.
An evaluation of the total procedure in such a retrospective fashion is greatly complicated by the fact that the procedure has an effect on the disposition of the case; thus in some instances the fact that a suspect appears to be telling the truth during a polygraph test is the basis upon which further investigation is deemed unnecessary. Obviously, a high correlation between apparent innocence on a polygraph examination and innocence in real life is virtually guaranteed. The same problem exists in the determination of guilt or lying. As has been pointed out, the interrogator has available to him as much information as possible and this information will include the investigative findings and convictions of his colleagues about the guilt of a suspect (Orlansky, 1962). These beliefs and convictions could certainly bias the interpretation of the records.
If the bias affected only the interpretation of the polygraph record, it would be possible to circumvent this bias by blind analysis of the records or a reevaluation by judges after the fact. The problem is far more complex. The procedure is by no means standardized, and the importance of the pretest interrogation in making the test itself work is uni-
751 ON THE DETECTION OF DECEPTION
versally acknowledged. The importance of formulating proper questions is also stressed in all of the literature written by "lie detection" experts. Since the technique depends upon the comparison of physiological responses to critical questions with those to control questions, the choice of control questions is of crucial importance. The manner in which questions are asked can certainly influence the physiological response. The interrogator wishing to obtain a record with physiological evidence of deception is likely to be able to do so. In view of the work by Rosenthal (1966) on experimenter bias, it seems extremely likely that the interrogator could unwittingly communicate some aspects of his conviction to the subject. In fact, suspects whom the interrogator believes to be innocent might well be taking quite a different test from that taken by those about whom the interrogator has strong convictions regarding their guilt. If there were some way of knowing absolute truth and if one could evaluate a large series of decisions for their accuracy, this issue would be less serious. In any retrospective attempt at evaluation, however, where one is using primarily the polygraph record itself, it cannot be ignored. The extent to which experimenter bias enters into a decision can be investigated. It is possible to ask polygraph operators to administer examinations to individuals who are definitely innocent, and by providing accurate, but slanted, investigative material convince the interrogator that the individual is probably guilty. A group of individuals tested in this fashion would yield evidence about the extent to which the polygraph examiner might unwittingly bias the test in actual practice. It would not be possible to obtain evidence about the effect of the interrogator's conviction of innocence on the guilty subject's record, but this is perhaps of less concern than false-positive tests.
Experimenter bias is not the only factor that may cause innocent individuals to be classified as guilty. An excellent example of the problem is cited by Dearman and Smith (1963), in which an individual felt guilty about legitimate financial dealings and was, as a result, suspected of stealing from his employer. Research directed at the problem of false-positive classifications seems of great importance.
The reliability of the procedure may also be evaluated. The record of physiological responses that is obtained can be examined by different individuals and judged as to whether deception is or is not present. Such studies, although not presently available, would seem feasible. A high degree of concordance in judgments, however, would in no way deal with the problems of validity outlined. It seems likely that different practitioners develop different feelings of what might be indicative of deception, partially based on an awareness of the kind of interrogation they carried out. The importance of interrogator reliability about the purely physiological evidence of lying could thus be overestimated.
No fully satisfactory way is available at this time for evaluating the overall effectiveness of the technique, and it is probable that no such answer will be forthcoming in the near future from real life situations. Perhaps most sensibly, one must compare the total technique with alternative ways of ascertaining whether an individual is telling the truth. One could, for instance, compare the effectiveness of clinical interviews with that of polygraph interviews, although no comparative data are yet available. In the clinical setting, the professional makes certain that there is little if any advantage to the patient in lying. The interrogator questioning a suspect, however, tends to feel that the polygraph yields a considerable increment to his effectiveness. Just what proportion of this increment is based on the physiological record, and what proportion on the psychological leverage provided by the technique itself, remains an open question.
Ethical issues, as well as practical ones, are raised by the use of polygraphs in a real life situation (Sternbach, Gustafson, & Colier, 1962). Whether such a technique ought to be used is beyond the scope of a scientific discussion. The matter of whether or not the technique works is, however, subject to scientific inquiry. If, as the preponderance of evidence seems to indicate, the technique is effective -- at least more effective than alternative methods of ascertaining truth -- it still remains to be determined how accurately and under what conditions. It should be noted, however, that were further research to establish a technique which is 100% effective in determining the truthfulness of
752 APPLICATIONS OF TECHNIQUES AND FINDINGS
a statement, the ethical difficulties would not be diminished. Thus wire-tapping, when it yields incriminating evidence, is 100% effective, but by no means does this accuracy resolve the ethical issues raised by its use.
LABORATORY RESEARCH
The mechanisms involved in the detection of deception are subject to laboratory research, although major differences exist between laboratory situations and those in the field, and inferences from one to the other will need to be drawn cautiously (Orne, 1969). Yet, to the extent that we are investigating mechanisms, they may be meaningfully explored under controlled conditions. When the polygraph is used for criminal interrogation, security screening, or in pre-employment interviews, the consequences of the test are significant and real. These consequences are in sharp contrast to those in most laboratory research, where the consequences of detection tend to be minimal. Even if they are maximized by experimental manipulations, they tend to be from a universe of discourse different from that of the field situation. These differences in consequence are by no means the only differences, however. Researchers have rarely used situations that are comparable in any way to those found in the field. The tendency has been to use the kind of situation characterized by the card test described earlier. The subjects are asked to select one of several cards and then requested to lie about the identity of the one selected. The GSR to the selected card is typically compared with that evoked by cards not selected. Occasionally, investigators such as Lykken (1959), Kubis (1962), and Davidson (1968) have used a pseudocrime which is enacted by the subject, or in which he might participate. Here, again, a fixed number of stimuli are given to the subject, stimuli that include some highly relevant to the crime he observed and others less relevant to it. In the laboratory experiment, therefore, the formulation of questions characteristic of the field situation and a discussion of the reasons for a suspect's undue response are replaced by the use of closely matched stimuli. The psychological leverage characteristic of the field situation is typically not exploited in the laboratory. With some exceptions (Ben Shakhar, Lieblich, & Kugelmass, 1970; Kugelmass & Lieblich, 1966), the subjects tend to be college students, a group that differs markedly from the clients of the "lie detection" experts. Finally, the relationship between the experimenter and the subject is radically different from that of the interrogator and the suspect.
The Effect of Motivation
In a field situation the suspect's concern about the test and its consequences for him may be taken for granted. The consequences of being detected have been emphasized repeatedly as significant determinants of detection (Burtt, 1921; Chappell, 1929; Marston, 1917). Because this view is widely held, a number of investigators have offered rewards to subjects who could successfully deceive during a polygraph examination (Davidson, 1968; Kubis, 1962; Lykken, 1960), or have administered electric shock during the experiment, presumably to increase the affective response to the relevant question (Lykken, 1959). The effectiveness of such manipulations is not known, however, since these investigators have not compared the detection rates obtained under these circumstances to those which have been obtained without such motivating techniques.
In a study designed to explore the role of motivation as a determinant of detection, Gustafson and Orne (1963) compared two groups of normal, paid college volunteers in a card experiment, using GSR as the response variable. Both groups were asked to select one card and were then played a tape recording that included a reference to the critical card, as well as to control cards. Before selecting the card, one of the two groups listened to a tape that contained the following information: (a) the experiment was designed to see how well the subject could keep information from the experimenter; (b) that this was extremely difficult and only persons with superior intelligence and great emotional control could succeed; (c) they were to try as hard as they could to beat the experimenter and the equipment; and (d) if they were successful, they would be paid an extra dollar. The experiment was carried out in a very impersonal fashion. Detection rates were found to be no greater than chance in the control group but significantly greater than
753 ON THE DETECTION OF DECEPTION
chance in the motivated group. It would appear, then, that detection is in large part a function of the subject's attempt to deceive and that incentives to do so result in an increased ease of detection.
On the basis of these findings it could be hypothesized that motivated subjects who are told they had not been successful in their attempt to deceive, and who are thereby even more highly motivated to escape detection, are more easy to detect. This logic would seem to be the implicit assumption underlying the use of the card test in a field situation, where the interrogator goes to considerable pains to convince the subject of the virtual infallibility of the polygraph in the hope of maximizing the differential responsivity.
The wide use of the card test in field situations makes the results obtained by Ellson, Davis, Saltzman, and Burke (1952) puzzling. In a study on the consequences of previous detection, they report that subjects detected on an earlier trial and informed of this fact are more difficult to detect on a subsequent trial. Gustafson and Orne (1965a) attempted to resolve this apparent paradox. They observed that there is a widely held conviction among college students that lie detection techniques are effective with all normal subjects, and that only habitual liars or psychopaths fail to be detected. In a laboratory situation, therefore, subjects might be motivated to be detected. Once they had obtained proof that detection was possible, their level of motivation would decrease, and they would then become more difficult to detect. If they were given proof that they were not detected, however, their concern about possibly having psychopathic tendencies would heighten the level of motivation and facilitate detection. In a real life context, any such problem would be overshadowed by concern for the consequences of detection, so that the results might be reversed.
To test these concepts, Gustafson and Orne (1965a) designed a study using GSR as an indicator of deception. Two groups of subjects were treated identically, except for different pre-experimental instructions designed to create two different mental sets toward the experiment. One group was told: "This is a study of lie detection. As you know, a lie detector is a scientific instrument. However, psychopaths and habitual liars are very difficult to detect. In the study we are assessing the adequacy of the technique. We want you to try to fool the lie detector. It is your job to fool the machine. Good luck!" The other group was presented a tape with motivating instructions similar to those employed in the earlier study, omitting only the monetary reward. The tape said: "This is a study in lie detection. Only very stable and emotionally mature individuals are able to fool the lie detector. We want you to try your best to fool the machine.
754 APPLICATIONS OF TECHNIQUES AND FINDINGS
Good luck!" All subjects were shown cards with numbers and were asked to select one, apparently without the experimenter's knowledge. A tape recording was then played on which all the possible numbers were repeated in the form of questions. The subject was instructed to reply "No" to each number, so that one of his responses would necessarily be a lie. Table 19.1 illustrates that on this first trial the two groups were not significantly different.
After the first trial, regardless of the subject's actual physiological responses, the experimenter entered the room and told half the subjects what their card had been -- thus indicating that they had been detected. To the other half, the experimenter named the wrong card, so that they believed they had gone undetected. A second trial was then given to each subject in order to determine the effect of feedback on subsequent detection. Table 19.2 summarizes the findings.
The two groups showed exactly opposite reactions to the same information. For those subjects who had been striving to be detected because "psychopaths are difficult to detect," the information that they had been detected eliminated the differential responsivity the next time that they were required to lie. The identical information given to those subjects who hoped to go undetected because "only mature and stable individuals are able to fool the lie detector," however, greatly augmented the response the next time that they were required to lie, making their detection much easier. Exactly the opposite result occurred when the experimenter purposively picked an incorrect card, communicating to the subject that he had effectively deceived the machine.
It should be emphasized that the physiological criterion for detection was the GSR, objectively scored, and that the experimental treatment was identical for both groups once the pre-experimental instructions had been given.
These findings illustrate the problems of inferring from laboratory findings to the field situation. Clearly, the identical experimental treatment and stimulus conditions can yield exactly opposite results depending upon the subject's attitude and set as he enters the experiment (Orne, 1962). In the above study, the subject's set was manipulated, but in many studies it is not investigated and can be inferred only with difficulty. In the field situation the individual's beliefs, attitudes, and sets are of great importance, but the suspect's motivations tend to be to prove himself innocent. A similar motivation must be assumed for laboratory studies from which we expect to draw inferences to a real life context. Recent work by Thackray and Orne (1967) underscores the need for care in generalizing about motivation. In the context of one of several studies altering the consequences of detection, the
755 ON THE DETECTION OF DECEPTION
subjects were administered "need to deceive" instructions and then were divided into several subgroups receiving differential feedback of information regarding the polygraph's effectiveness. The subjects receiving information that they were detectable were easier to detect in a subsequent interrogation than those subjects who believed that they were not detected. This difference was true, however, only for the later part of the interrogation. These studies used a situation in which an attempt was made to create a high degree of subject involvement. It is possible that, when a high level of arousal exists, the effect of differential feedback of information may be masked, becoming evident only after the subjects habituate sufficiently to the interrogation procedure. In summary, it appears that not only degree of subject motivation but the object of the motivation as well are important determinants of deception. The task that the subject perceives to be his in an experiment and the subject's perception of the total situation can exercise a profound influence on the outcome of an experimental treatment.
The Experimental Paradigm in the Detection of Deception
As has been indicated earlier, no specific physiological responses are pathognomic of lying. If one asked a suspect, "Did you murder Mr. X?" and observed a physiological response to the question, one could not infer deception, since any suspect could be expected to respond to such a question. The problem is to find additional questions which, when asked of innocent persons, would lead to responses equivalent to those obtained from questions related to the crime under investigation. The development of questions that will have a differential impact only on the individual possessing what Lykken (1959) calls "guilty knowledge" constitutes the major effort on the part of the interrogator employing this technique. His task is made considerably more difficult, as Lee (1953) points out, by the fact that the suspect has usually been questioned extensively prior to the test regarding the details of the crime and, further, that relevant details are often known to both innocent and guilty alike through news media. The situation in the laboratory, however, is more easily controlled. If a subject is asked to choose a card, the alternative choices that he did not make provide ready-made questions. It is easy to lose sight of the major differences between such a laboratory paradigm and the field situation.
When the subjects are asked to take a card, and the names of these cards are then presented to the individual, all the subjects are known to be guilty and the experimenter wishes to know the specific nature of the information about which they are guilty. This situation has been termed the "guilty information paradigm" (Gustafson & Orne, 1964). On the other hand, the usual field situation is most typically one in which the examiner and the suspect know the nature of the critical questions and the intent of the procedure is to determine whether or not the individual is lying about these questions. This situation has been called the "guilty person paradigm" (Gustafson & Orne, 1964); the task of the experimenter is to differentiate the guilty person(s) from those who are innocent. An analogous distinction is made by Ben Shakhar, Lieblich, and Kugelmass (1970) in a discussion of the altered criteria of detection necessary in applying a signal detection model to the detection of deception situation.
The task that a subject in the laboratory faces is very different for one paradigm as opposed to the other. In one instance he sees his task as concealing specific information that he is known to have; in the other instance he sees his task as one of convincing the interrogator that he does not have any special information at all. The paradigm employed will determine to a large extent the kind of strategy subjects use to avoid detection. In a study of possible countermeasures to avoid detection, Kubis (1962) indicates that it is easier for subjects to produce willfully augmented autonomic responses to irrelevant questions than to suppress their responses to the critical questions. The typical laboratory study using the guilty information paradigm invites the use of this countermeasure. Under the guilty person paradigm, however, such attempts would lead an operator to suspect deception. Moreover, when a polygraph examination is carried out by an interrogator in a real life situation, the effectiveness of this strategy is even more severely limited because the suspect is asked about the
756 APPLICATIONS OF TECHNIQUES AND FINDINGS
reason for any undue physiological response and the situation constrains him to explain such a response. Having done so, the question may be rephrased in such a manner as to reduce the possibility of responses occurring for such a reason. It would appear that in the hands of a good interrogator a suspect's attempt purposively to augment his responses to neutral stimuli has distinct limitations as a countermeasure.
It is not necessary for the subject and the experimenter to view the experiment as following the same paradigm. In the laboratory one may create a situation that, when seen from the subject's point of view, has many of the elements of the guilty person paradigm, while retaining the structure of the guilty information paradigm in the manner of stimulus presentation and selection of control questions. Such a mixed paradigm is applicable to both of the commonly used laboratory tasks, the card or number test and the mock crime. In the usual card or number test situation (Alpert, Kurtzberg, & Friedhoff, 1963; Block, 1957; Block et al., 1952; Burtt, 1921; Geldreich, 1941; Kubis, 1962; Kugelmass, 1967; Landis & Wiley, 1926; Langfeld, 1921; Obermann,1939; Van Buskirk & Marcuse, 1954; Violante & Ross, 1964), all possible cards (including the one that the subject has selected) are presented as stimuli, a procedure which automatically requires that the subject lie concerning the card he actually selected. This situation is clearly one in which the subject and the experimenter both know that the subject has some special knowledge and that the experimenter is attempting to identify the nature of that knowledge. Under these circumstances, if the subject voluntarily produces an augmented physiological response, he successfully interferes with detection. The identical experimental procedure may be transformed into the guilty person paradigm, from the subject's point of view, by explaining that there are blank cards as well as numbered cards in the deck from which he is to draw a card. His task will be to convince the interrogator that he has selected a blank card. If he shows an undue response to any of the cards, regardless of whether he selected it or not, he will convince the interrogator of his guilt. This is much the same situation as is met in real life, where the suspect's task is to convince the interrogator that he is innocent of any crime, rather than that he has committed another crime from that of which he is accused. The subject is thus constrained from responding to neutral stimuli as a countermeasure, in that he sees the task of the experimenter not as that of determining which card was chosen but rather whether or not the card chosen contained any significant information.
Gustafson and Orne (1964) carried out a study comparing the successful detection of information under two conditions, one in which the subject saw the experiment as attempting to detect guilty information and the other in which he saw the experiment as attempting to detect whether he had information or not. Subjects operating under the guilty person paradigm were significantly easier to detect, overall, than were subjects operating under the guilty information paradigm. In addition, the way in which the subject perceived his task interacted with the manner in which the stimuli (cards) were presented in the interrogation. Clearly, it is important to consider the subject's perception of his task as it relates to the type of information which he believes the experimenter seeks to uncover.
Next to the rather simple card selection procedure, the most frequently used experimental model is one in which the subject observes or participates in a mock crime and is subsequently interrogated concerning this crime (Baesen, Chung, & Yang, 1949; Berrien, 1942; Berrien & Huntington, 1943; Burtt, 1936; Chappell, 1929; Davidson, 1968; Kubis, 1962; Landis & Wiley, 1926; Lykken, 1959; Marston, 1917; Obermann, 1939; Runkel, 1936). This experimental procedure can also be presented to the subject as two different paradigms. In the guilty information paradigm the subject can be told that several crimes are possible and that the interrogator will try to find out which crime he committed. In the guilty person paradigm, the subject can be told that the interrogator will try to discover whether or not he has committed the crime. Most of the mock crime situations involve the use of the guilty person paradigm. Some, however (Baesen, Chung, & Yang, 1949; Burtt, 1936; Davidson, 1968; Kubis, 1962; Lykken, 1959), provide situations where several crimes or several roles in a crime are involved. In such situations it is possible for the subject to see
757 ON THE DETECTION OF DECEPTION
his task as either to appear innocent or to avoid revealing his role in the crime. In the Baesen, Chung, and Yang (1949) study, pairs of subjects committed the mock crime; one of the subjects actually carried it out and the other observed. The subjects were told that the purpose of the experiment was to find out which of them had carried out the crime and which had observed. No subject was innocent of the crime. In contrast, Lykken (1959) had some of his subjects commit two crimes, some only one, and some committed no crime. The subjects were told that they were to be questioned about the two crimes, but no report was given concerning their perceived task in the situation. Burtt (1936) required his subjects to look into one of two boxes containing several miscellaneous articles. The investigators then attempted to identify the box into which the subject had looked, clearly a guilty information paradigm. Davidson (1968) ran four groups of subjects, one actually committing a crime, one attempting the crime but failing, one planning the crime but not attempting it, and one group of naive subjects. The subjects were told that they could keep the proceeds of the crime if they could successfully avoid detection. Kubis (1962) created three roles: thief, lookout, and innocent suspect. The subjects who were assigned to the thief and lookout roles were told to appear completely innocent when questioned.
Apart from the obvious necessity of being aware of the paradigm under which these studies were run, it is also important to examine the degree of stimulus control present in each situation. As Lykken (1959) indicates, the subject in a mock crime situation cannot always be relied upon to notice all of the details about which he is questioned. Furthermore, certain aspects of the crime may have a greater impact on the subject than others. Orne and Thackray (1967) have developed an experimental situation that allows for a high degree of stimulus control and yet is flexible in terms of the paradigm. The subjects are asked to play the role of a courier with secret information in the form of code words. The words are thoroughly learned by the subjects and may be assigned without the knowledge of the interrogator. Furthermore, the code words may be words with little meaning to the subject outside of the experimental context. The subjects can be told either to appear innocent of being a courier or to avoid revealing the code words. Orne and Thackray (1967) found that such a model led to high subject motivation and involvement and felt that a realistic interrogation resulted when the subjects played the courier role. The results of experiments using this procedure may be easily considered either in terms of identifying the guilty person or the guilty information.
The lack of uniformity between the various investigations in regard to the paradigm used has made cross comparisons of the results difficult. It is important to consider the paradigm under which the research is conducted, in light of the demonstrated effects of subject motivation and expectancies on detection. It would seem that to render the laboratory situation more like that encountered in the field, the subject should view his task as following the guilty person paradigm. For purposes of statistical treatment of results, however, one can seek to identify both information and guilty subjects.
The Nature of the Stimulus Material
Since the detection of deception depends upon the comparison of significant stimuli with neutral control stimuli, the nature of the critical stimuli used will affect the probability of a physiological response. On one end of a continuum is the selection of a card or number about which the individual is expected to lie. The stimulus has only the affective value given it by the experimental situation. On the other end of the continuum, the individual's actual name or similar pertinent facts about him, his family, or other aspects of his identity possess considerable affective value apart from the laboratory. Using personal material (mother's name, father's name, name of high school) Lykken (1960) was able to obtain 100% accuracy in matching subjects to arrays of stimuli. This high degree of success was probably a function not only of the nature of the stimuli but also the level of motivation and the limited number of subjects with whom the subject was compared.
Mock crimes have been used to provide stimuli that have affective value because of their association with the totality of an involving context. Baesen, Chung, and Yang (1949), Berrien and Hunt-
758 APPLICATIONS OF TECHNIQUES AND FINDINGS
ington (1943), Davidson (1968), Ellson et al. (1952), Kubis (1962), Lykken (1959), and Marston (1921) reported rates of detection in studies employing mock crimes ranging from 94% (Lykken, 1959; Marston, 1921) to 50% (Landis & Wiley, 1926). The median rate of detection from these studies is 80%. If we compare this rate with the rates in most studies that have used neutral stimuli such as numbers, letters, or geometric designs (Alpert, Kurtzberg, & Friedhoff, 1963; Block, 1957; Block et al., 1952; Burtt, 1921; Geldreich, 1941; Landis & Wiley, 1926; MacNitt, 1942; Ruckmick, 1938; Van Buskirk & Marcuse, 1954), detection rates range from 40% (Landis & Wiley, 1926) to 83% (Ruckmick, 1938) with a median rate of 73%. While this might appear to be slightly lower, comparisons of this kind have little meaning. In a mock crime situation the subject may be told that a $26 sum was stolen, thereby causing the number 26 to become a meaningful stimulus because of its association with the context. This manipulation is not much different from having a subject select a card, thereby making the stimulus meaningful. Furthermore, as has already been pointed out, differences between studies in experimental paradigms, methods of presentation, physiological response used, etc., make valid comparisons impossible.
A systematic attempt to compare contextual and personal stimuli in the same situation, matching for involvement, was carried out by Thackray and Orne (1968b) using the courier situation. The subjects were told that they would play the role of couriers and would overlearn three critical code words. Having done so, they were told that they would be interrogated by someone who, because of a leak in the courier system, had the first names, last names, and dates of birth of six agents and their respective code words. The interrogator, however, would not know the subject's actual identity nor would he know whether the subject was one of the couriers or one of the subjects whose names he did not have and who were not familiar with any code words. They were told that their task would be to appear innocent; i.e., to appear not to be one of the individuals about whom they would be interrogated, nor were they to recognize any of the code words. Each subject was introduced to the interrogator as subject Number 26. Further, the interrogator was in fact blind as to the identity of the subjects. Using this situation it was possible to compare the rate of detection to first name, last name, and date of birth with the rate to three random words that the subject had overlearned. The mean detection rank for the personal material (1.44) was significantly better (p < .025) than for the code words (1.62).
Lieblich (1966, 1969) systematically varied the relevance and frequency of usage of stimuli in an information detection situation. He found striking and consistently ordered differences in detection between conditions using the subject's own name and conditions using either an irrelevant name or one that is rare. Detection was greatest when the subject's own name was compared to alternative names known not to be related to the subject. Lower rates of detection were found when the subject's name was compared with names from within his family circle, and still lower rates when an irrelevant or rare name given to the subject was compared with names from within his family circle.
These findings support the common-sense assumption that subjects should be more responsive to personally relevant stimuli reinforced by the context than to previously neutral stimuli made relevant only within the experimental context. Even before any experimental manipulation, the subject's own name or date of birth should evoke a physiological response. This component is probably additive to that component engendered by the experimental situation.
In the field setting, one deals with stimuli that tend to be of great concern to the individual, and which have usually been associated in the individual's mind with their involving context many times. Such stimuli are not briefly associated with the context, as in an experiment, nor are they typically as enduring a part of an individual as is his own name. There are no systematic studies relating the ease with which differential physiological responses may be elicited under varying degrees of association between a stimulus and the context within which it became meaningful. It would be of great interest to know the effects of time on such an association and the effects of repeated interrogation. Probably some of the increase in detection rate with the courier situation used by Thackray and
759 ON THE DETECTION OF DECEPTION
Orne (1967), when compared to the usual card selection situations (84% versus 73%) results from the procedure of requiring the subjects to overlearn the stimulus words. Even the spontaneous rehearsal of such connections may ease the task of detection. Such rehearsal may form the basis for the suggestion made by Reid and Inbau (1966) that, after each question session, the examiner leave the room for a few minutes to allow "a lying subject ... [to] .. . develop greater concern with respect to whatever lies he may be contemplating . . . [p. 28]." In the laboratory context the results of such incubation are probably much more closely approximated by the subject's own name or personal material than by a card, number, or other similar stimuli.
For the most part, laboratory studies have not begun to concern themselves with the nature of the stimuli used in the detection of deception and their effect on the rates of detection. Several studies Jones & Wechsler, 1928; Smith, 1922; Syz, 1926) have explored differential responsivity in other contexts, but laboratory studies within the detection of deception framework are rare. It would seem that, in light of the central role played by the stimulus material in detection, such studies are in order.
Methods of Stimulus Presentation
Most laboratory research has utilized what has been called the "relevant-irrelevant" method of stimulus presentation (Inbau & Reid, 1953; Lee, 1953), where critical and neutral stimuli are presented to the subject either in a completely random order or in some other sequence unknown to the subject. Such a procedure has many obvious design advantages, including the facilitation of counterbalancing and the equating of stimuli.
In field situations, on the other hand, the tendency has been to use questions that the suspect knows in advance; both the nature of the questions and the order in which they are presented are familiar to him. Such techniques arise from the use of the pretest interview and are partly the consequence of having to develop appropriate control questions. They also permit the interrogator to discuss the meaning of the questions in such a manner as to minimize misunderstandings.
Each of the procedures has certain advantages. In the case of the relevant-irrelevant method, it is difficult for the individual to prepare himself in advance, since he does not know what question to expect. The element of surprise plays a role in all of his answers. On the other hand, the overall level of responsivity remains very high, since the subject does not know exactly what to expect next. In a situation of high motivation and concern, the response to the relevant question may be augmented, but even the irrelevant questions will, because of the manner in which they are presented, evoke considerable physiological response.
A procedure in which a suspect knows what to expect, especially when the questions are discussed in great detail, has the advantage that little or no response should be evoked by the neutral control questions. Any undue response, even to neutral questions, can therefore become a meaningful indication of possible deception. The overall level of responsivity can be greatly reduced and, while this would also affect the responsivity to the critical items, the remaining response to these items would tend to be far more conspicuous in relation to the control questions once the surprise element is eliminated. Such a procedure, on the other hand, permits the suspect to prepare himself for the critical items and, to the extent that he is able to suppress a response, might make detection more difficult. In the most characteristic form of interrogation of this kind, the peak-of-tension technique, the interrogator not only compares the relative response to the critical item but takes advantage of the suspect's tendency to relax once it has passed, looking for (as the name implies) a peaking of tension at the time of the critical item.
Marcuse and Bitterman (1946), discussing the peak-of-tension technique, feel that experimenter biases (voice inflections, subtle mannerisms, etc.) may unduly influence the course of an examination. The same sources of error are, however, inherent in the relevant-irrelevant method as well. In the laboratory these can be controlled by making certain that the person conducting the examination is unaware of the identity of the critical items; in the field no such ready solution is available.
Any comparison between stimulus methods of presentation must be carried out under comparable conditions. Gustafson and Orne (1964) compared
760 APPLICATIONS OF TECHNIQUES AND FINDINGS
the two procedures by asking subjects to select a numbered card. Numbers were then presented either in a random fashion, following the relevant-irrelevant procedure, or in ascending and descending order, following the peak-of-tension procedure. Each of these methods of stimulus presentation was studied under both the guilty person and the guilty information paradigms. When the guilty information paradigm was used, the relevant-irrelevant method was significantly more effective than the peak-of-tension technique; when the guilty person paradigm was employed, however, both techniques became more effective and any advantage of the relevant-irrelevant method was obviated. The peak-of-tension method appeared particularly vulnerable to the creation of false-positive responses in the laboratory context. Such responses, however, would not be a very effective stratagem in the real life situation and, when the laboratory design restrains subjects from their use, the findings appear more in line with the reported experience of professional interrogators that the two techniques are similar (Reid & Inbau, 1966).
The Effect of Habituation
As will be discussed later, the probability of detection is raised by increasing the number of trials available comparing different significant stimuli to neutral controls. One simple way in which such an increase may be augmented is to repeat the same stimuli. To the extent that one is dealing with random responses, a repetition will tend to clarify which stimuli evoke a differentially greater response. Habituation, however, may diminish the differential responsivity. The crucial question, then, concerns the relative rates of habituation between significant and control stimuli. If the critical stimuli habituate more rapidly than do control stimuli, little will be gained by extensive repetition of the same stimuli.
A study reported by Ellson et al. (1952) compared 10 subjects who were administered two series of five trials, each consisting of the names of six months, one of which the subject had previously selected. The average number of successful detections was 80% for the first series and 70% for the second series. Although somewhat inconclusive due to the small N, this finding suggests greater habituation to the critical stimuli and is similar to a finding of Gustafson and Orne who, in an unpublished study, repeated a simple card test five times, motivating the subjects as described earlier. It was observed that combining the first and second repetition yielded a somewhat increased accuracy of detection over the first alone but that adding the third repetition did not augment detection. The addition, in turn, of the fourth and fifth repetitions actually decreased accuracy, indicating that in that situation at least, detection was not facilitated by habituation.
The relationship of habituation to detection, however, is by no means simple. As mentioned earlier, Thackray and Orne (1967) observed that a differential treatment effect became clear only at the end of a long series of stimuli. This study was characterized by a great deal of responsivity to neutral stimuli, which seemed to attenuate more rapidly under those circumstances than the responsivity to the significant stimuli. The effect of habituation on differential responsivity may be a function of the involvement of the individual. Peterson and Jung (1907) report the results of three successive trials on a word association test, during which the skin potential response (SPR) was monitored. During the first trial a number of stimuli elicited responses, one of which was to a stimulus with which the subject was emotionally involved. On the second trial the response to this stimulus was the largest and only a few other stimuli evoked significant responses. On the third trial, the same stimulus led to an only slightly attenuated response, while the other stimuli now produced very little response. Jones and Wechsler (1928), again using the word association technique, found differential habituation even within one trial. When a list of mixed emotional and neutral words was given, the responses to the neutral words became less for words at the end of the list when compared to the same words when placed at the front of the list, while emotion-provoking words showed no such position effect. These results, different from those of the Ellson et al. (1952) and the unpublished Gustafson and Orne studies, both obtained with card selection tests, raise the possibility that successive trials may serve to accentuate the signal-to-noise ratio when more involving stimuli are utilized.
761 ON THE DETECTION OF DECEPTION
If it is possible to habituate a subject to the neutral stimuli while minimizing or preventing habituation to the critical stimulus, then detection should be dramatically improved. Geldreich (1941) demonstrated this process. The subjects in one group were asked to choose a card from a set of five and then were administered a single series of questions concerning which card they had drawn. Using the GSR, he correctly identified the chosen card for 74% of the subjects. The other group was also required to select a card, following which they were administered from 20 to 50 cards, none of which were in the original set of five, until virtually no GSR was elicited. Then the original set of five cards was presented. Under these conditions, the subjects differentially habituated to the noncritical stimuli and essentially failed to habituate to the critical stimulus. As one would predict, the GSRs to the noncritical stimuli in the final set were substantially less than GSRs to comparable noncritical stimuli in the group without habituation. The average response to the critical stimulus, however, was essentially the same in both groups. As a result, the correct card was identified for 100% of the subjects in the second group.
The studies mentioned so far in this section have used the GSR as a measure of responsivity. The detection of deception in the field, however, typically employs the heart rate, "relative blood pressure," and respiration changes in addition to, or in place of, the GSR. Differential habituation is likely to vary with the parameter being examined. In an unpublished study, Solomon et al. (1958), focusing on the conditioning of the heart rate in humans to highly dramatic stimuli, observed that the heart rate and respiratory changes were difficult to condition and extinguished very rapidly, whereas differential GSRs persisted over a long period of time. J. P. Seward and G. H. Seward (1934) supported these observations in a study tracing the course of habituation to shock stimuli. They found that, within the same session, body movement and respiration responses to the shock diminished more quickly than did the GSR.
It seems likely that differential responsivity on parameters such as the heart rate or blood pressure would demand a very high level of arousal in the subject, perhaps sufficiently great to interfere with the easy recognition of a differential GSR responsivity. A lower level of arousal, on the other hand, might facilitate the recognition of differential GSR responsivity, while even lower levels of arousal again reduce differential responsivity. Darrow (1936) felt that adrenin (natural secretion of which adrenalin is an analog) inhibited the GSR and that blood pressure might therefore be a better measure at high levels of activation. Systematic work on the differential responsivity of measures with differing levels of activation is lacking.
The Role of Verbal Response
Although a few studies have not required a verbal response (Gustafson & Orne, 1963; Lieblich, 1966; Lykken, 1959), most studies require subjects to respond "Yes" or "No" to the questions. The procedure is usually arranged such that the subject will be forced to lie concerning the critical item. The extent to which verbal lying influences detection rates has been investigated in several studies (Ellson et al., 1952; Gustafson & Orne, 1965b; Kugelmass, Lieblich, & Bergman, 1967; Lieblich, 1966), all employing the GSR as the physiological measure. The conclusions to be reached from examining these studies are, at best, equivocal. Part of the confusion arises out of the diversity of procedures among the studies. Some studies have compared saying "Yes" to each question with saying "No" to each question (Ellson et al., 1952; Kugelmass et al., 1967). Others have compared saying "No" to remaining mute (Gustafson & Orne, 1965b; Lieblich, 1966). Gustafson and Orne (1965b) have also compared these two latter conditions with a condition in which subjects were required to respond with the first word that came to mind in a free association procedure.
The results are as varied as the procedures. Ellson et al. (1952) found in their pilot study that four of their eight subjects were detected when required to say "No" to each question, while only two of the eight (chance) were detected when replying in the affirmative. No assessment of significance was possible because of the small N and the use of the subjects as their own controls. Kugelmass, Lieblich, and Bergman (1967) replicated these conditions and compared trials in which individuals were required to say "No" to all questions with
762 APPLICATIONS OF TECHNIQUES AND FINDINGS
trials in which they were required to say "Yes" to all questions, counterbalancing for order. Under both conditions, the subjects were detected at a level well above chance, and no significant difference was found between conditions.
Gustafson and Orne (1965b), in a study comparing three groups of separate individuals, in which all subjects were motivated to deceive the experimenter, found that both those subjects who responded with "No" and those remaining silent were detected at a level significantly above chance, with the "No" group exhibiting the higher detection rate. The subjects in the free association group, however, were not detected better than chance. Lieblich (1966) ran two conditions out of six, in which his subjects were requested to remain mute. Although the material used in these two conditions was rather neutral, detection in the mute conditions was reduced from that in similar conditions in which the subjects were required to lie.
Clearly, the differentially augmented physiological response that forms the basis for the detection of deception is not contingent upon the subject's actual lying. It can occur in the absence of any verbal response and may also occur when subjects tell the truth to critical stimuli and lie to irrelevant ones (Kugelmass et al., 1967; Reid & Inbau, 1966). It is tempting to explain these discrepancies by indicating that it is the deceptive intent, rather than an overt lie, that is responsible for the increased physiological response. There appear, however, to be at least two other components in the detection of deception situation that are often confounded with the effect of lying or deception. The first of these components is that the subject is sensitized to an essentially neutral stimulus by the mere fact of drawing attention to it as the selected stimulus. This effect is in contrast to that seen when a stimulus has inherent arousal value, which has been discussed earlier. The sensitization effect is a situation-specific arousal inherent in the process of selecting any given stimulus as critical in the set of stimuli. The selected material becomes the figure in a figure-ground relationship among stimuli, with the remaining stimuli constituting the ground.
Only when the stimuli are all intrinsically neutral, and both lying and deceptive intent are eliminated, does the effect of sensitization become self-evident. If a subject is asked to select a numbered card from among several, to place it in full view of the experimenter, and then to respond honestly when asked by a tape recording about several numbers, the mere selection of a particular number will result in some increase in the GSR to the chosen number. In such an experiment, the subject responds "No" to each number except the one he had selected, and it could be argued that the augmented physiological response is a function of the different verbal response rather than of the sensitization to a particular number. Similar findings are obtained, however, if the same experimental procedure is followed, the chosen number exposed in full view, and the subject then asked to say "No" to each number in a context where his "lie" could not conceivably have deceptive intent. This type of sensitization to a selected number is present in all card test situations, and is generally confounded with the consequences of attempting to deceive. It is an effect which becomes obvious only in the kind of situation described above, where the stimuli are truly neutral and matched, the situation simple, and the time interval between card selection and the presentation of the stimuli is kept very brief to prevent the effect from being swamped by other factors.
The same kind of an experimental procedure can be easily altered to demonstrate the effect of trying to deceive. Merely instructing the subject not to show the card to the experimenter and to try his best to conceal its identity in his physiological responses -- adding perhaps that this is difficult but some intelligent subjects are able to do so -- leads to a significantly greater physiological response to the selected card. The sensitization effects, the effects of the intrinsic arousal value of the stimuli, and the motivation to deceive, as well as the consequences of detection, all appear to be independent and additive in the degree they augment the differential physiological response.
The second component is the amount of attention the subject pays to the entire set of stimuli. While this effect is partly determined by the motivational factors discussed earlier, instructing the subject to respond "No" (or "Yes") identically to each stimulus allows a strategy of ignoring the
763 ON THE DETECTION OF DECEPTION
saliency of the selected stimulus. (This is why in a field context questions are phrased to require both yes and no answers in order to assert innocence.)
The effect of whether or not the subjects pay attention to the stimuli was noted in the Gustafson and Orne study (1965b). Postexperimental inquiries with subjects revealed that, when the individual had the task of providing free associations, it was possible for the subject to decide in advance on a number of associates which he then used as responses in a mechanical fashion; this strategy actually helped the subjects to ignore the content of the stimuli. Subjects reported that they found it easier to ignore the stimuli when carrying out this task than when they were required either to say "No" to every stimulus or to attend without giving any verbal response.
It may also be possible, however, for subjects who are required to say "No" to each stimulus to do so in a completely mechanical fashion, responding with "No" to the sound rather than to the actual word. Instances have been observed, for example, where subjects answered "No" when instructed that the test was completed, or when they heard some extraneous noise. No such response is observed with subjects who are differentially verbalizing, since it is essential for these subjects to listen to the stimuli as they are presented in order to respond correctly.
The additional factors that seem to play a major role in laboratory situations appear less relevant in the field situation because of the profound consequences of the interrogation. Furthermore, questions in the field situation are frequently arranged so that both "Yes" and "No" answers may be required, depending on the way in which the question is asked. Such a procedure forces attention to the questions asked, so as not to answer incorrectly. Even during procedures in which all answers may be the same, the suspect is so deeply involved in the situation that he is likely to attend to every cue and stimulus present in order to avoid making a mistake. In the laboratory the role of lying and the response required of the subject is, as yet, an unsettled question. Studies must separate the factors of involvement and attention from those concerning the response. It appears that, in the laboratory, either a "Yes" or a "No" response facilitates detection over an unrelated response or no response. Detection is possible, however, at a level significantly better than chance without any verbal response, provided that the subject is motivated to deceive and the stimuli have sufficient impact. The clarification of the specific effect of the verbal response on the likelihood of detection remains a task for future research.
Response Variables
To evaluate the state of the subject with respect to the stimuli presented him, one must have some measure of his emotional response. By far the most frequently used measures are those several correlates of autonomic nervous system activity which are subject to rapid change and a rapid return to baseline. Some investigators have used motor activity of the hand, eye movements, or verbal responses, measures usually considered to be more nearly under voluntary control. Such experiments rely on the psychological effect of the stimuli to produce abnormalities in these more usual modes of response. In view of the reliance placed on autonomic measures in the field, these will be discussed first.
Measures of autonomic activity While studies have demonstrated differences between individual patterns of autonomic activity (Lacey & Lacey, 1958; Schnore, 1959; Wenger, Clemens, Coleman, Cullen, & Engel, 1961) and differences between various emotional states and the patterns produced (Ax, 1953; Sternbach, 1960), there is no evidence at the present time that the autonomic changes which accompany deception differ qualitatively from those produced in other emotional states involving sympathetic nervous system activation. In general investigators feel that deception responses most closely resemble those characteristic of emotional excitement (Chappell, 1929) or fear (Marston, 1917), although even simple, auditory stimuli have been shown to produce autonomic patterns which are essentially the same as those occurring during deception (Davis, Buchwald, & Frankmann, 1955). A great deal of the research that has been done on the psychophysiological correlates of emotion is relevant to the detection of deception problem, but
764 APPLICATIONS OF TECHNIQUES AND FINDINGS
a discussion of most of this work is beyond the scope of this section. (See Arnold, 1960; Woodworth & Schlosberg, 1954.)
Most studies in the detection of deception use the detection rate as the validating criterion for judging the usefulness of the measures employed. As the preceding sections of this chapter tried to make clear, however, both subject and task variables are significant determinants of detection rates. Since few studies have concerned themselves exclusively with response variables per se, the effectiveness of any given physiological or behavioral measure of deception is generally confounded with these other experimental manipulations. Only when several response variables are included in a single study is it possible to draw meaningful conclusions concerning their relative effectiveness. To the extent possible, emphasis will be given to those studies which provide a comparison of more than one response variable within a single design. 3
Cardiovascular measures Commercial "lie detectors" have used a measure often referred to as "relative blood pressure," obtained by inflating an arm or wrist cuff to a point approximately halfway between systolic and diastolic pressure. Davis (1961) has pointed out that this technique can be both painful and dangerous, unless frequently interrupted, and is not a true measure of either systolic or diastolic pressure. Perhaps for this reason, attempts are currently underway among field experts to change from the term "relative-bloodpressure" tracing to the term "cardio" tracing. This change reflects the growing awareness that this tracing does not really reflect changes in blood pressure but volumetric changes in the arm and heart rate. It is important that cardio measures and true measures of systolic pressure not be confused, since the two are quite different. Most laboratory studies and several early field studies have used systolic blood pressure rather than the cardio tracings, taking measurements intermittently during the course of questioning (Chappell, 1929; Landis & Gullette, 1925; Landis & Wiley, 1926; Langfeld, 1921; Marston, 1917). A frequent response is a rise in pressure that is greater during deception than when the subjects are telling the truth. As expected, because of individual differences in reactivity and differences in procedure, the magnitude of the mean difference varies across the studies. Using systolic blood pressure alone, Chappell (1929) obtained 87% correct discrimination between subjects telling the truth or lying concerning details of a mock crime. Marston (1921), also using a mock crime, obtained 94% accuracy in differentiating between liars and those choosing to tell the truth. Recent investigators have not reported results as outstanding as these. Ellson et al. (1952) found systolic blood pressure measures about equal to respiration measures, and far inferior to the GSR. Thackray and Orne (1968a) found no evidence of discrimination with a measure of systolic blood pressure.
A reason for the differences may stem from the temporal relationships between the stimuli and measures of systolic blood pressure. The time necessary for a maximal rise in systolic pressure following a stimulus has not been well established, although Rushmer (1961) notes that rises in blood pressure are rapidly modulated through the baroreceptor reflex. The ability of a procedure, such as systolic pressure monitoring, to detect deception may well depend largely upon the time at which the pressure is measured. Both Ellson et al. (1952) and Thackray and Orne (1967) used devices to measure systolic pressure automatically at relatively fixed intervals. When the lack of success with the measure in these two studies is contrasted with the success reported by Chappell (1929) and Marston (1917), it seems possible that the manual determination of blood pressure is effective because it time-locks the measure in a fixed relationship to the stimuli, whereas an automatic system introduces a variable time interval between the stimulus and the point of measurement.
In the past few years a number of measures have been investigated for the purpose of augmenting or replacing blood pressure as a measure of cardiovascular changes during deception. Perhaps the most common technique is digital plethysmography. Brown (1967) has covered this measure in detail and states that "the plethysmogram is an extremely sensitive measure of internally and ex-
765 ON THE DETECTION OF DECEPTION
ternally induced changes, surpassing even the electrical skin phenomena in this respect [p. 55]." The experience of investigators in the area of the laboratory detection of deception does not, however, show the measure to be as useful as GSR.
Although the heart rate is easily derived from either the electrocardiogram (EKG) or from records of finger or pulse volume and is a commonly used measure in psychophysiology, it has received little attention in laboratory studies of deception. Changes in the heart rate are difficult to see unless special circuitry is employed to convert the interpulse time into a measure of rate.
None of these cardiovascular measures appears to discriminate well between guilty and innocent individuals in the laboratory setting, as indicated in the few studies comparing these measures with others, notably with the GSR. Kubis (1962) reported pulse volume to be slightly superior to respiration in supplying cues to examiners attempting to differentiate guilt from innocence. Violante and Ross (1964), on the other hand, found respiration to be superior to pulse volume and found little or no discrimination with finger volume. In a study comparing several measures of autonomic activity during deception, Thackray and Orne (1968a) observed that finger volume discriminated between innocent and guilty subjects at a better than chance level, while pulse volume did not. In all of these studies the GSR was the best discriminator. While Kugelmass and Lieblich (1966) and Kugelmass, Lieblich, Ben-Ishai, Opatowski, and Kaplan (1968) found the heart rate to discriminate no better than chance, Ellson et al. (1952) indicated that heart rate was slightly superior to pulse volume but inferior to the GSR and systolic pressure. Even pulse volume was found to discriminate at a level better than chance. They indicated that the heart rate response consisted of an initial acceleration, followed by deceleration, with the magnitude of the deceleration becoming greater during lying than during truth. Maximum discrimination with the plethysmographic technique occurred when they measured the rate at which subjects recovered from increased peripheral resistance during the 12-22 sec poststimulus period, with non-lie responses recovering more rapidly than lie responses. This measure was employed after discovering that equal vasoconstriction responses were evident to both critical and noncritical stimuli. They suggested that a maximal response may have occurred in both cases. In the light of these data, Brown's (1967) statement that the plethysmogram is highly sensitive to change may indicate that the measure is too sensitive to yield a differential response in the detection of deception. The lack of agreement across studies concerning the usefulness of these measures suggests that greater effort should be directed toward an understanding of cardiovascular changes during deception.
Psychological changes in the subject are reflected both in changes in the cardiac output and in the amount of peripheral resistance. Blood pressure itself is the end product of the complex interplay of heart rate, stroke volume, and vasomotor influences. An increase in the heart rate alone, unaccompanied by other circulatory changes, affects diastolic blood pressure but has relatively little effect on systolic pressure. On the other hand, an increase in the volume of blood pumped per beat will affect primarily the systolic blood pressure, consequently increasing pulse pressure. An increase in peripheral resistance results in an increase in the mean blood pressure, but the more intense effect is on diastolic pressure, due to the longer period of the diastole relative to the systole (Best & Taylor, 1950). The primary cause of increased cardiac output may be an increase in stroke volume, rather than in the heart rate (Rushmer, 1961). The "cardio" measure, used with success in the field situation, is largely a plethysmographic measure. The record reflects complex interactions between the blood pressure and volumetric changes in the arm (Woodworth & Schlosberg, 1954). Reid and Inbau (1966) indicate that the usual criterion for deception consists of a rise in the baseline of the tracing; a decrease in pulse amplitude may or may not be present. Changes in the baseline level of the tracing are likely to reflect an increase in blood flow, which in turn may be accompanied by various different kinds of blood pressure changes. During emotionally mediated activation of the sympathetic nervous system, vasoconstriction takes place in the skin and in the splanchnic viscera, while vasodilation occurs in the skeletal muscles (Brod, 1964). At least part of the increase in cuff pressure observed during
766 APPLICATIONS OF TECHNIQUES AND FINDINGS
deception in the field situation is probably a result of the increase in volume produced by the increased blood flow to the muscles that make up the bulk of the upper arm tissue and greatly outweigh any reduction in blood flow to the skin of the upper arm (Brod, Hejl, & Ulrych, 1963). Ansley (1959) gave some support to this contention. He compared records taken with the standard arm cuff to those taken with a wrist cuff; in most cases the wrist cuff records showed markedly reduced responses, although the majority of records could be readily scored for deception. Placing the cuff over the musculature of the upper arm rather than the forearm may increase responsivity, since about 85% of the tissue in this area is skeletal muscle (Barron & Ruch, 1960). If some measure of the relative concentrations of blood in the viscera and skeletal musculature were available, it might serve as a sensitive measure of deception. The measurement of visceral vasoconstriction, however, would seem difficult at best.
A few other measures of cardiovascular activity have been suggested for use in the detection of deception, among which are pulse velocity (Dana & Barnett, 1957; Ellson et al., 1952; J. G. L. Williams & B. Williams, 1965) and blood flow. Reid and Inbau (1966) suggest the use of blood flow on the basis of preliminary investigations using an infrared sensor placed on the neck over the carotid artery. Their evidence indicates that the measure is differentially sensitive to stimuli during a card test and suggests that a Doppler shift measurement of blood flow in the arm might prove useful as well.
Respiratory measures Measures of respiration were among the first to be used in the detection of deception (Trovillo, 1939). With the exception of the early work of Benussi (1914), virtually no studies of deception have used respiration measures as a sole physiological index, perhaps because of the difficulty involved in an attempt to quantify the respiratory pattern. Although the practice among field examiners is to regard any marked change in respiratory patterns following critical questions as indicative of deception (Joseph, 1957; Reid & Inbau, 1966; Trovillo, 1942), laboratory studies have usually employed either Benussi's inspiration-expiration (I/E) ratio (Burtt, 1921; Landis & Gullette, 1925; Landis & Wiley, 1926) or more simple measures of average respiration amplitude and breathing cycle time (Ellson et al., 1952). There seems to be an increase in the ratio of the time of inspiration to the time of expiration (I/E ratio) following deception. This increase is closely related to the increase in breathing cycle time found by Ellson et al. following deception. These latter investigators also found breathing amplitude to be reduced during deception, when compared to nondeception responses. Some of the difficulties in quantifying respiration may be the result of difficulties in obtaining the I/E ratio. Ellson et al. (1952) reported that they attempted to determine the ratio from several records and found that no clear delineation existed between periods of inspiration, expiration, and rest. The lack of a distinctive point of change, they feel, can easily lead to large random variations in the ratio.
Apart from the measurement of respiratory parameters by means of a sensor stretched around the thoracic or abdominal region, very little work has been done on monitoring respiration within the area of deception detection. Reid and Inbau (1966) report the use of Doppler techniques to obtain the usual respiration tracing without sensor contact. Ellson et al. (1952) employed a thermocouple placed so as to measure the temperature of the subject's breath as he inhaled and exhaled. They found that baseline changes in breath temperature occurred but they were not significantly related to the stimuli. Measures of tidal volume or gaseous saturations in the breath, which require the subject to breathe through a mouthpiece, would seem to be impractical in the detection of deception situation. Impedance pneumography, however, may have potential as an alternative technique (Stein & Luparello, 1967).
Galvanic skin response The galvanic skin response has long been recognized to be one of the most sensitive indicators of autonomic activity. 4 (See Edelberg, 1967, and Chapter 9 of this book for techniques and theory.) Its use in the detection of deception has had an interesting and controversial history. Many persons with field experience claim
767 ON THE DETECTION OF DECEPTION
the GSR to be useful in the laboratory but of much less value than respiration and blood pressure in criminal investigations (Larson, 1932; Lee, 1953; Marston, 1938; Reid & Inbau, 1966). They appear to reflect the belief that the GSR is too responsive to any stimulus, making unequivocal decisions difficult in real life situations. Others, however, have supported the use of the GSR in the field. Kugelmass et al. (1968) found both relative blood pressure and the GSR about equal in detecting card choices with criminal suspects. Summers (1939) strongly supported the use of the GSR in both laboratory and field applications, and reported detection rates of over 98% with this measure alone. Likewise, MacNitt (1942) also supported the GSR and reported detection rates of equal magnitudes in both laboratory and field situations.
Interestingly enough, most of the laboratory studies that have compared GSR with one or more additional variables agree that the GSR is superior to other variables in the detection of deception. Ellson et al. (1952), for example, found it far superior to all of the other variables they investigated, as did Thackray and Orne (1968a). Likewise, Kubis (1962) reported GSR to be considerably better than either pulse volume or respiration, while Violante and Ross (1964), using measures similar to those of Kubis, also found GSR to discriminate better than other measures. Kugelmass and Lieblich (1966) obtained significant discrimination with GSR but not with pulse rate.
All of these studies tried to create highly motivating or involving situations for the subjects, in an attempt to approximate the real life situation. Ellson et al. (1952) and Kubis (1962) employed mock crimes, while Violante and Ross (1964) used aversive stimuli (noise) in a conditioning paradigm. Kugelmass and Lieblich (1966) specifically tried to create a highly stressful situation by employing policemen as subjects and informing them, as one of the experimental conditions, that their future occupational advancement might depend upon their success in controlling emotions in the polygraph test. Despite the variety of conditions and stresses, all found the GSR to be highly effective.
It is, of course, possible that the highly emotional circumstances of actual examining conditions in a criminal case were not even approached in these studies, and thus they offer no real argument against the contentions of field authorities. The problems of simulating real life situations are too well known to require elaboration. A lack of involvement on the part of the subjects in all of the studies, however, particularly the Kugelmass and Lieblich (1966) study, is difficult to believe.
It has been argued that the equipment employed for measuring GSR in the field is inadequate, accounting for the lack of success with this measure (Ellson et al., 1952; Higley, 1958). In an unpublished study conducted by Gustafson, GSR detection rates based upon measures obtained from a Stoelting Deceptograph (an instrument typical of those used in the field) were compared to those obtained using a Beckman Dynograph with a constant current bridge and zinc-zinc sulphate electrodes. No difference in detection rates was found. While the equipment employed by field examiners to measure GSR may leave much to be desired in terms of current psychophysiological practice, it is doubtful that its inadequacies are the sole explanation for the convictions held by field personnel regarding the effectiveness of the GSR in guilt detection.
The relative effectiveness of GSR in both laboratory and field situations remains to be evaluated. No simple explanation such as equipment insufficiencies or level of arousal can fully account for the discrepancies. It is clear, however, that in laboratory settings, particularly with low subject involvement, GSR has been the most effective discriminator. It is also true that GSR is difficult to interpret with extremely anxious subjects; and under these circumstances respiration and the cardio tracings remain easily interpreted by inspection. The tendency to discount the GSR in the field situation is probably also related to the fact that some individuals are very unresponsive on this variable. Whether this lack of GSR responsivity is situational or a matter of autonomic specificity has not been investigated in this context. Despite all this, some individuals with extensive field experience are convinced that GSR is the most useful index of deception. In the absence of definitive data, the possibility that reports concerning the effectiveness of the GSR as a discriminator in field situations are related in large part to operator preference and response specificity cannot be excluded.
768 APPLICATIONS OF TECHNIQUES AND FINDINGS
Oxygen saturation A measure of the oxygen saturation in the blood can easily be obtained by means of a photoelectric sensor attached to the pinna of the ear or placed on the neck (Dana, 1958; Dana & Barnett, 1957). This measure has been shown to reflect emotional changes (Lovett Doust & Schneider, 1955), although changes in oxygen saturation take place at a rate somewhat slower than the rate for other measures (Dana & Barnett, 1957). Thackray and Orne (1968a) compared discrimination with the oxygen saturation measure to discrimination with several other measures; they found the discrimination significantly better than chance for experimental material (code words) but not for personally relevant material. The relatively slow rate of change in this measure, however, probably precludes its use in field situations.
Pupillary response The response of the pupil to changes in autonomic activity in emotional states has been known for some time (Bender, 1933; Lowenstein & Friedman, 1942) and has more recently become the object of considerable research (Hess, 1965; Hess, Seltzer, & Shlien, 1965; Woodmansee, 1966). Berrien and Huntington (1943) compared pupillary responses to a measure of relative blood pressure in a deception situation involving a mock crime. The most typical response consisted of a slow, negatively accelerated dilation, lasting 1-5 sec, followed by a rapid constriction. Although this response also occurred spontaneously following control questions, it was most frequent when deception was involved. A second pupillary indicator of deception consisted of an increase in pupillary fluctuation in guilty subjects when the critical questions were introduced. When scored independently, both pupillary response and blood pressure yielded correct discrimination in over 70% of the cases and, since misidentifications of subjects on one measure were not necessarily the same as those on the other measure, the authors note that the two measures serve to complement each other, in combination yielding about 80% detection. Simultaneous significant changes to critical stimuli in both measures occurred in only 48% of the responses.
The most severe drawback in the use of the pupillary response is the problem of measurement. Photographic recording of the response has long been common (Bender, 1933), sometimes employing infrared lighting and film to avoid illumination differences (Lindsley, 1951); recently, devices have been developed using photocells to scan the eye (Lowenstein,1956). Such techniques, in addition to the possible use of closed circuit television, should facilitate further research with this interesting and potentially significant measure.
Electroencephalogram A number of investigators have reported a blocking or reduction in the alpha activity of the electroencephalogram (EEG) with stimulation of an emotion-producing sort (Lindsley, 1951; Thiesen, 1943; Williams, 1939). The only known study employing the EEG in the detection of deception was conducted by Obermann (1939), with two separate experiments reported, one involving neutral stimuli (numbered cards) and the other involving knowledge of the details of a fictitious crime. Monopolar recordings from a left occipital placement were obtained with the eyes closed, and judges were asked to rank the records on the basis of their likelihood of being indicative of deception, using as a criterion any disturbance in the record. Although the statistical treatment of the data makes them difficult to interpret in terms of the discrimination achieved, the detection of deception with EEG criteria seems possible. Interestingly enough, more objective measures of amplitude or percentage of alpha activity did not prove as efficient as did the subjective judgment.
Miscellaneous autonomic measures The use of several other measures of autonomic nervous system activity has been suggested, among them salivary secretion, pilomotor response, and gastrointestinal motility. No evidence is available, however, concerning the use of these measures in the detection of deception. It would seem that the understanding and refinement of some of the more usual measures would be of greater value than would the proliferation of a number of esoteric measures, unless a significant improvement in detection can be demonstrated (Orlansky, 1962).
Behavioral measures of deception Although the emphasis in current interrogation seems to focus on physiological measurements that are recorded during the examination period, field operators have long recognized the important information to be
769 ON THE DETECTION OF DECEPTION
gained from observations of a suspect's actions, manner, and vocal responses (Reid & Arther, 1953; Reid & Inbau, 1966). While aspects of the subject's behavior no doubt add to the interrogator's ability to classify him correctly (Kubis, 1950), a number of studies have specifically investigated overt behavioral measures in the detection of deception situation. A clear separation does not exist between these measures and those covered previously. Respiration, for instance, is under voluntary control, as well as subject to autonomic influences, and several of the measures in the present section are probably influenced, in part, by autonomic activity. Nevertheless, the following behavioral measures have been employed to detect the presence of deceptive responses.
Reaction time The use of associative reaction time to indicate deception was one of the earliest methods employed (Jung, 1907; Jung, 1910; Wertheimer, 1905; Wertheimer & Klein, 1904; Yerkes & Berry, 1909). A review of this early literature has been compiled by Crosland (1929). If a subject is asked to associate freely to a list of words, some critical to a crime and others neutral, it is expected that he will either give himself away through revealing or bizarre responses or, in suppressing such reponses, will give long reaction times and signs of emotion. As Woodworth and Schlosberg (1954) point out, however, it is vital that normative data from known innocent subjects be available, since critical words are typically less common than are the neutral words and will result in long reaction times, due to a lack of associations.
Marston (1920) studied reaction time in a laboratory situation, in which the subjects were instructed to either obey or disobey requests concerning arithmetic operations to be performed on sets of numbers. The time required to answer each item was measured, and it was predicted that response times would be longer when the subject disobeyed the experimenter and tried to deceive him. He found three groups of subjects: those whose responses were longer as predicted, those whose reaction times were inconsistent, and those who gave consistently shorter reaction times. He felt that both those subjects with shorter reaction times and those with longer response times were displaying a typical deception response. Goldstein (1923), however, felt that those subjects with shorter reaction times were simply not involved in the task of deceiving the experimenter. The whole argument raised considerable controversy (English, 1926; Goldstein, 1923; Marston, 1925; Rich, 1926) but, as a general conclusion, whenever considerable affect is generated with deception, reaction times will probably increase.
The use of associative reaction time as an indicator of deception has been both supported and questioned by investigators. In a statistical combination with other indicators, Crosland (1929) found reaction time to be useful in determining the guilt of a person in an actual dormitory theft. Winter (1936) likewise found increased reaction times to critical words in a person guilty of an actual dormitory theft, although several innocent suspects also exhibited increases. Respiratory and blood pressure measures were also used in this instance; blood pressure was found to be the most useful, while respiration records were the least useful. In a study comparing deception responses to a mock crime, Runkel (1936) measured both the reaction time and motor response disturbances and found that for every subject the mean reaction time was greater to critical stimuli than it was to noncritical stimuli. Larson (1922) feels that the reaction time measure is not as satisfactory as is a respiration tracing and gives evidence that reactions occur in the tracings which are not reflected in longer reaction times.
In addition to measuring the reaction time to a word association task, the reaction time in answering questions has been measured. Ellson et al. (1952) instructed subjects to remove a sum of money from a box and then questioned them concerning the amount taken. The subjects responded by pressing a lever to the right for an affirmative answer and to the left for a negative answer. In addition to measuring reaction times, muscle potentials were obtained from the right forearm, since the authors felt that deception would result in motor conflict which would be reflected both in the reaction time and in the electromyogram (EMG). Measures of the reaction time significantly differentiated between truthful and deceptive responses. A measure of EMG amplitude, weighted and combined across five time periods during the stimulus presentation and the response, enabled
770 APPLICATIONS OF TECHNIQUES AND FINDINGS
the investigators to discriminate deceptive affirmative responses 90-95% of the time and deceptive negative responses 65-70% of the time.
Interference with voluntary motor behavior Considerable interest was generated in the early 1920s concerning the use of semivoluntary muscular indicators of deception. This interest was primarily engendered by the considerable success in criminal interrogation reported by the Russian psychologist Luria (1932) in recording the voluntary and involuntary movements of both hands in response to word associations. The rationale underlying this technique is that during heightened emotion there is a breakdown in motor control. When, as in Luria's case, the subject is required to press on a rubber bulb with his right hand as he responds verbally and to keep his left hand on another bulb, variations in response patterns and tremors or disturbances of the left hand may be recorded pneumatically. Significant responses usually consisted of tremors in both the right and left hands and rough or partial responses on the part of the right hand. Often a motor response would begin with the termination of the stimulus and fail to be completed, being followed later by a complete motor response and a verbal response. Luria felt that these premature responses indicated that certain words were being suppressed on the part of the subjects. Luria's techniques stimulated considerable research, most of which has been concerned with the nature of disturbance in abnormal subjects. Yates (1961) has summarized much of this literature, coming to the conclusion that increasing degrees of mental disorder are related to increasing levels of both left- and right-hand motor disturbance.
Few studies have specifically applied the technique to the detection of deception. Burtt (1936) found the most diagnostic measure of deception to be the extent to which the motor response in the right hand preceded the verbal response and was able to discriminate correctly in 69% of the cases. The nature of the responses, if any, elicited by the other hand was not reported. Runkel (1936) also found responses in the left hand of little discriminative value, being infrequent. He attributes this lack of distinctive left-hand responses to the relatively weak emotion generated by the experimental procedure. He found distinctive right-hand motor responses to occur to critical stimuli an average of 50% of the time across subjects, the responses consisting of a wavelike tremor preceding the motor response. As Morgan and Ojemann (1942) point out, the determination of significant responses depends to a large extent upon the skill of the investigator.
Voice quality Quite a number of studies have explored the relationship between aspects of vocal expression and the emotional state of the person speaking (Kramer, 1963). Only two known studies have explored the use of vocal measures of deception, one employing subjective judgments of deception, the other an objective amplitude analysis. Using the subjective judgments of persons hearing other persons lie or tell the truth over a public address system, Fay and Middleton (1941) found that the two types of statements could be differentiated correctly about 55% of the time (chance was 50%). Alpert, Kurtzberg, and Friedhoff (1963) employed a deception situation with neutral stimuli and measured the amplitude of the subject's verbal responses with a voltmeter. Both a full frequency band (100-6000 Hz) and a filtered low frequency band (100-250 Hz) were measured. No differences were found in the full frequency band during deception, but amplitude increases in the low frequency band to the critical stimuli were significantly different from those to noncritical stimuli. Such an easily obtained and potentially useful measure should not be ignored.
Eye movements Eye movements and eye blinks have been noted as possible indicators of emotion (Lindsley, 1951; Orlansky, 1962). While no evidence exists on the use of eye blinks in the detection of deception, recordings of eye movements have been made and found useful in the differentiation of guilty and innocent subjects. Eye movements may be recorded by placing appropriate electrodes near the eye and amplifying the resulting signals, or by shining a light onto the cornea and projecting the reflected light through lenses onto photographic film. Using this corneal reflection method, Berrien (1942) explored eye movement variations in response to questions concerning a mock crime situation, but found no differences between the amount of eye movement present
771 ON THE DETECTION OF DECEPTION
during critical questions and during neutral questions. During the prequestion period, however, guilty subjects seemed to decrease in ocular tremor, while innocent subjects did not. When judges were asked to classify as guilty those subjects whose records showed this decrease and as innocent those subjects who exhibited no change or an increase, the judges were, on the average, about 70% correct in discriminating guilty from innocent subjects.
Differences in steadiness during the question period did not separate the subjects as well as did this criterion. Furthermore, the criterion successfully discriminated between guilty and innocent subjects when both groups were told the details of the mock crime before the crime was carried out, and between two subjects who were not tested until the day after performance of the mock crime.
Ellson et al. (1952) report a somewhat different approach to the determination of guilty responses. The subjects were instructed to remove a coin located in one quadrant of a box, and their eye movements were recorded during questioning, again using the corneal reflection method. Some subjects were instructed to tell the truth during questioning, while others were told to lie concerning the quadrant from which they had taken the coin. Eye movements to the guilty quadrant were greater in number than to the other quadrants for all subjects. The difference, however, was greater for lying subjects than for truthful subjects. Critical questions caused more eye movements and eye movements of greater magnitude toward the guilty quadrant for subjects who were told that they could keep the coin if successful in deceiving the experimenter than for subjects not told this. During neutral questions the subjects who had taken the coin made fewer movements toward the guilty quadrant than did a group of subjects who were not presented a coin in the box. It may be that in cases in which guilty subjects know details of crime not known to innocent subjects, an analysis of eye movements in response to the simultaneous presentation of several similar stimuli, one of which is critical, might provide evidence of guilty knowledge.
Muscular tension Reid and Inbau (1966) report that the measurement of the pressure with which the suspect bears down on the arms of the chair can often serve as an indicator of deception. These authors have developed pneumatic and Doppler sensors for detecting these movements, as well as movements of the thighs. They report that suspects will, at times, tense their muscles in an attempt to raise blood pressure and thereby confuse the polygraph operator. They consider such deliberate actions indicative of guilt. As noted earlier, Ellson et al. (1952) were able to derive a measure of muscular activity from EMG recordings that allowed some discrimination between truthful and deceptive responses.
Nonverbal behavior Based on data derived from observations of psychiatric patients, Ekman and Friesen (1969) have studied those aspects of a subject's behavior that communicate deception to the observer. In studying strictly nonverbal behavior, they find that persons who observed the hands and feet were more likely to detect what they refer to as deception cues and leakage of genuine underlying affect than those persons who observe the face alone. Furthermore, they reported that facial expression, when studied in slow motion, reveals another kind of deception clue. Observed at normal speed, the face will reveal one affect, while the very rapid "micro" expressions that become clearly visible only in slow motion may show the very opposite expression.
It would seem that these studies address themselves to the kinds of cues which allow an observer to become suspicious about the presence of deception. Certainly, this approach is of great interest, but as yet little is known about the power of these techniques. Further, such an approach must still be compared with other techniques that have been used to study the presence of deception.
Quantitative Evaluation of Responses
The current practice of field examiners is to rely on subjective impressions of whether or not deception has occurred. The examiner studies the record and notes any significant changes or differences in the patterns of responses between the critical and control stimuli. The tendency to consider as potentially significant almost any change in the response pattern (Joseph, 1957) leads to problems in interpretation. To the experienced examiner working on-line, small changes may have
772 APPLICATIONS OF TECHNIQUES AND FINDINGS
greater significance than do larger, more observable changes. Such may not be the case for the inexperienced examiner or someone examining the record post facto. In addition, examiners differ in the emphasis they place on the various measures, some relying primarily on respiration, some on "cardio," and some on the GSR (see Hathaway & Hanscom, 1958).
Laboratory studies, on the other hand, have tended to rely heavily on objective and quantifiable information such as the magnitude of the GSR (Block et al., 1952; Geldreich, 1941), GSR ranks (Lykken, 1959,1960; Thackray & Orne, 1967), the I/E ratio (Burtt, 1921; Landis & Wiley, 1926), and systolic blood pressure (Chappell, 1929; Marston, 1917). Kugelmass et al. (1968) have attempted to score the "cardio" channel by connecting the "notches" of the tracings and measuring deflections from the baseline. Subjective impressions of several judges have even been quantified (Kubis, 1962). In order to quantify the responses, however, experimenters are forced to choose those measures that are obvious and amenable to quantification. With respiration, for example, laboratory studies generally rely on relatively crude measures of amplitude, frequency, or time relationships. Many possibly significant changes such as suppression, blockage, irregularities, or other subtle pattern changes are lost because they seem to defy ready quantification. Furthermore, invariant criteria are often applied to the responses of all subjects. In discarding much of the information available to the field examiner, however, a gain in reliability may be offset by the possible loss in validity. Kubis (1950) reports higher detection rates for independent experts who analyzed records resulting from an experimental situation than for persons using purely objective criteria, and the highest detection rates of all for operators interpreting the records on-line. Studies are needed exploring the possibility of quantifying the less regular aspects of changes in physiological records. While it is possible that a computer analysis might reveal measures contained in the variables, the computer must be programmed to derive these measures. The human investigator, therefore, is responsible for initially specifying the relationships and can rely on the computer only for information as to their usefulness in the differentiation between guilty and innocent subjects.
Another use of the computer is as an aid in discovering combinations of measures that will allow increased differentiation. Kugelmass and Lieblich (1968) applied a discriminant function analysis to data derived from the three channels of a field polygraph but found poorer detection with this combination than with a simpler sum of ranks. They felt that the low intercorrelations they obtained between channels in many subjects was a contributing factor in this poorer detection. Kubis (1962) applied individual discriminant analyses to examiner's ratings of the significance of GSR, respiration, and pulse volume responses. For the individual subjects, higher rates of detection were achieved with these discriminants. Because of the subjective criteria employed in the ratings, however, there was little equivalence in the discriminant weights among those who rated the same records. Kubis (1962) concludes, "In view of the fundamental problems which are yet to be solved in the objective measurement of the physiological patterns ... the concern for obtaining optimal discriminant functions is a small matter indeed [p. 60]."
Definition of the Detection Rate
As mentioned earlier, the analysis of data gathered in a detection of deception experiment depends upon the paradigm under which the study has been conducted. The definition of the detection rate differs for the two paradigms. When all of the subjects in a study are known to possess guilty information (the guilty information paradigm), it is the information rather than the individuals that the investigator seeks to detect. Studies of this sort usually compare the relative rates of detection resulting from experimental treatments such as different levels of motivation, the subject's perception of his task, or orders of presentation. In such cases, it is quite satisfactory and appropriate to define the detection rate as the percentage of times the significant stimulus evokes the greatest response when compared to otherwise matched
773 ON THE DETECTION OF DECEPTION
control stimuli. 5 The fact that by chance alone this percentage could be, for instance, 20% does not diminish the importance of findings in which the experimental manipulation significantly affects detection. When all of the subjects are known to have guilty information, a greater response to a noncritical stimulus is a failure to detect the information sought and should affect the detection rate. Large responses to critical items result in the detection of the information and are thus successes.
An entirely different state of affairs exists when innocent subjects are introduced into an experiment. Such an experiment, following the guilty person paradigm, seeks to detect both guilty and innocent subjects, i.e., to discriminate between the two groups. To define the detection rate in such an experiment in the same way as for an experiment involving only guilty subjects, or even to redefine it as the percentage of individuals who exhibit their greatest response to the critical stimulus, is to ignore the chance factors operating in the case of the innocent subjects. Given a situation employing five stimuli, one of which is critical, and where some subjects may be innocent, all of the guilty subjects could exhibit their largest response to the critical stimulus. Detection of the guilty group, therefore, would be 100%. By chance alone, however, 20% of the innocent subjects would also show their largest response to the critical stimulus and would be misclassified as guilty. Such a situation results from overlapping response patterns in the two groups. In selecting the response pattern that allows maximal detection of the guilty subjects, in this case the greatest response to the critical stimulus, innocent subjects are unavoidably misclassified.
The Use of Multiple Indices of Guilt
There are a variety of strategies applicable in the laboratory situation to maximize the discrimination between guilty and innocent subjects. Each of these strategies involves changing the situation from the simple example just presented. Each also demands an understanding of the parameters that enter into making the decision of guilt or innocence. Perhaps the simplest way to increase discrimination is to increase the number of neutral stimuli (Lieblich, Kugelmass, & Ben Shakhar, 1970). With the same 100% detection of the guilty subjects, if the number of stimuli is increased to 10, one of which is critical, the percentage of innocent subjects misclassified will drop to 10%. It is just as possible to increase the number of critical stimuli, but, in such a case, the detection of guilt would require either a greater response to all critical stimuli when compared with responses to the neutral stimuli, or at least some proportion of the responses must be greater. The probability of classifying an innocent subject as guilty would depend upon the criterion used for determining guilt; chance factors would involve joint probabilities of occurrence (assuming that each stimulus is independent). Suppose that 8 neutral and 2 critical stimuli were presented and that those subjects who responded with GSRs greater to the 2 critical stimuli than to the others were classified as guilty. Then by chance, 10% of the innocent subjects would respond most to 1 of the 2 critical stimuli and 10% of 10%, or 1%, of the innocent subjects would respond more to both critical stimuli than to the neutral stimuli and would be misclassified as guilty. As the number of critical stimuli is increased, the likelihood of an innocent subject's reacting significantly to all of the stimuli would decrease rapidly. In reality, however, it would be unlikely for all of the guilty subjects to show their greatest responses to all of the critical stimuli. To maximize discrimination, therefore, it is often desirable to widen somewhat the definition of a guilty subject to include response patterns in which only a specified number of stimuli yield greater responses. In this way the number of falsenegative subjects (guilty subjects classified as innocent) is reduced, while the number of false-positive subjects (innocent subjects classified as guilty) is increased. Similar arguments could be advanced for the case of additional trials. If several trials are given, each with the same stimuli, con-
774 APPLICATIONS OF TECHNIQUES AND FINDINGS
tinued high responses to critical stimuli would be unlikely to occur by chance, assuming minimal adaptation to the stimuli.
Separate distributions of guilty and innocent subjects can be established in the laboratory situation and each response pattern can be expressed in terms of the probabilities of falling into the guilty and innocent distributions. Each subject, then, could be classified as falling into that group for which the probability was highest (Hathaway & Hanscom, 1958). It is also possible to take into account factors such as the relative importance of avoiding false-positives and false-negatives. The elimination of false-positives would be more important in criminal investigations than in screening for a sensitive security position (Ben Shakhar et al., 1970). Much of the difference between the two situations arises from the fact that in criminal investigations only a few individuals, often only one, properly belong in the guilty group, whereas in screening, a large number of individuals may be in this category. The mechanism at work here is analogous to the concept of the selection ratio found in personnel selection (Cronbach & Gleser, 1965). Little work has been done in the laboratory with regard to the selection ratio problem. One would expect that, if the laboratory interrogator knew that only one subject in a group could be correctly classified as guilty, his approach would differ from his approach in the usual situation in which a number of subjects may be assigned to this category.
The process of combining probabilities to increase discrimination may be used not only across stimuli within one individual but also across individuals presented the same stimulus, provided that a response is available which can be readily averaged. The idea is analogous to that used in studying evoked cortical potentials with intact human subjects except that instead of time-locking multiple stimuli within the same individual, one observes simultaneously the response of multiple individuals to the same stimulus. If these individuals share the same guilty information, the averaged response to the critical stimuli should be maximized. Responses to the control stimuli, which may have idiosyncratic meaning to one subject or another, will tend to be minimized by averaging the responses. As a result, the signal-to-noise ratio will be markedly augmented.
A study by Orne and Thackray (1967) illustrates the use of this procedure. In this experiment the subjects were interrogated in groups of six subjects each, some groups being innocent and some guilty. Both individual and group GSRs, averaged across the individuals, were taken. The separation between the guilty and innocent subjects was shown to be very significantly improved through averaging. One would expect that with larger groups discrimination would improve still more, provided that all subjects are known to be either guilty or innocent. When the groups may contain both guilty and innocent subjects, one should be able to discriminate those groups containing guilty subjects, i.e., subjects with shared information, even if only a small proportion of the group consists of these individuals. The innocent subjects would merely add to the overall level of noise. While individual responsivity is difficult to predict, the responsivity of groups might well approach a mathematical function.
The Effect of Awareness
All of the research discussed so far is based on the assumption that the subject knows that someone is measuring physiological parameters to detect whether or not he is telling the truth. We have indicated, moreover, that motivation to deceive appears to be a major factor in making detection possible. One may legitimately wonder, therefore, how effectively we could detect deception if we had access to the physiological responses of an individual without his awareness that he was being recorded. This question, in view of present-day technology, cannot be considered totally unrealistic. It would be possible to sense at a distance parameters such as facial temperature, respiratory rate, and pupillary size. Appropriate concealed sensors could also measure the degree of fidgeting, some form of the cardioballistogram, and heart rate. Reid and Inbau (1966) report the results of research into techniques for recording respiration and body movements without the use of attached sensors. They report that results with the Doppler technique, whereby sound energy reflected from the subject may be analyzed to reveal relative
775 ON THE DETECTION OF DECEPTION
movement, corresponded closely to respiration tracings obtained in the usual manner, and that movements of the limbs could be accurately recorded. These measurements can be obtained without the knowledge of the subject if the sensors are concealed. Even ignoring the ethical considerations, one might wonder how helpful such equipment would be in the detection of deception. Some preliminary indication concerning this problem was yielded in the course of a study by Thackray and Orne (1967). Prior to a peak-of-tension series in a group experiment, the stimuli were presented to the subjects, ostensibly so that they would know the stimulus order in the subsequent test. During this presentation the subjects were led to believe that the equipment, located in an adjacent room, was turned off. Immediately after the presentation, the subjects were interrogated about these same words; i.e., they were requested to answer "No" to each item. The difference in detection rates between the preinterrogation presentation and the actual interrogation, while in favor of the actual interrogation situation, only approached significance (p < .10). Any interpretation of this finding must be tempered by the realization that the procedure confounded a number of variables and no systematic effort was made to establish whether or not the subjects believed the experimenter when they were told that they were not being recorded.
In the context of another study an attempt was made to get at the latter issue more directly. Thackray and Orne (1968b) arranged the courier situation so that as the subject was entering the interrogation room a technician was turning off the polygraph. A number of electrodes were then attached to the subject, and it was explained to him that it would be necessary for some of the sensors, particularly the oximeter, to stabilize over a period of time before the test could take place. Prior to turning on the polygraph and actually taking the test, the subject and the interrogator discussed the questions in a way similar to that of a real life detection situation, ostensibly in order to avoid confusion and surprise. During this time the subject's GSR responses were telemetered to another polygraph four rooms away. Once the discussion was completed, the actual examination took place. A subsequent attempt was made to determine whether the subjects believed that a recording was made during the pretrial test period. Only one subject reported such a belief. Nevertheless, it was not possible to demonstrate a significant difference between detection rates under these two conditions. Novelty and habituation worked against the hypothesis, there being no easy way to make plausible a reversal in the order of presentation. Furthermore, in talking with the subjects it became clear that, while they tended to believe that they were not being recorded during the initial presentation, they were very uncomfortable and felt strange with all of the sensors attached. It is difficult, therefore, on the basis of these studies, to answer the question of how subjects would behave if they were not in a polygraph situation. The findings do suggest that not merely the conscious belief that one is being monitored but the situation in its totality may be sufficient to evoke differential responsitivity.
Studies are needed that allow monitoring of physiological responses, particularly the GSR, without the subject's feeling that he is being monitored, preferably with little or no electrode attachment. We know remarkably little about the effect of being monitored. One suspects that the extensive habituation characteristic of physiological recording is partly a result of habituation of the subjects to the psychological situation of being recorded (Woodworth & Schlosberg, 1954). This response is usually confounded with the experimental treatment itself. With the development of small, portable, and easily carried telemetering equipment, it should be possible to minimize the effects of being monitored and approximate the recording of physiological responses in an unaware individual. There has been some recording of physiological parameters over many sessions during the course of psychotherapy (Coleman, Greenblatt, & Solomon, 1956; Malmo, Boag, & Smith, 1957). Watson and Kanter (1956) report an experiment in which both the patient and the therapist were monitored throughout the treatment. While the therapist reported the monitoring as having a considerable effect on him, at least initially, the patient was able to make therapeutic progress despite this situation. The importance of the material to him was probably such that it made less salient any
776 APPLICATIONS OF TECHNIQUES AND FINDINGS
concern he might have had about the physiological monitoring. His major concern with the experimental situation centered around the presence of human observers, not the physiological measures. Again the heart rate, respiration, and EEG of astronauts have been monitored routinely throughout their flights, and without their reporting concern about the attached electrodes (Adey, Kado, & Walter, 1967; NASA Manned Spacecraft Center, 1962).
The issue of monitoring without awareness, however, is of importance in order to clarify some of the mechanisms involved in the detection of deception. The field situation is designed to take advantage of any psychological effects on the physiological response resulting from the suspect's awareness of being recorded on a machine. Many of the laboratory data mentioned here support the view that detection is possible only because of the individual's attempt to prevent detection. It is not clear whether this effort on the part of the subject is directed toward preventing detection through the physiological measures, or if he attempts to escape detection in the more general sense of controlling his tone of voice, manner, and overall behavior. Differences in physiological responsivity to highly affect-laden stimuli might emerge even more clearly if the individual were not aware that he was being monitored. In a like manner, conditioned GSR responses or anticipatory GSRs associated with the threat of shock should occur regardless of the subject's awareness of being monitored. On the other hand, there is no clear reason why a selected card or number should evoke a differential physiological response simply because an individual is required to "lie" and deny its selection.
The effectiveness of the lie detection procedure probably depends both on the subject's response to the content about which he is lying, as well as his awareness that his physiological responses are actually being recorded. Furthermore, there may be an interaction between the response to the arousal quality of the stimulus and the awareness of trying to deceive. The field interrogation takes place under conditions in which the stimuli have high intensive meaning, and the subject is made maximally aware of his attempt to deceive and the consequences of being detected, probably the situation that most favors detection. Studies by Gustafson and Orne (1963) and Thackray and Orne (1967, 1968b) have compared this situation with others in which involvement or awareness was manipulated, but no studies have specifically investigated the situation in which the subject is not concerned with the evasion of detection and is unaware of being monitored. Such a study could clarify the relative importance of both variables in the detection of deception.
Means of Minimizing Detection
The problem of minimizing detection is the obverse of procedures whereby detection is made more likely. Implicit in considering factors of subjects and situations that enhance detection is a consideration of factors that may also serve to minimize detection. These factors have often been grouped together under the heading of "countermeasures." Inasmuch as successful detection involves responses to relevant stimuli that differ in some way from responses to nonrelevant stimuli, countermeasure techniques attempt to reduce this difference. There are two principal means to reduce the difference in responses, increasing the level of response to irrelevant items or reducing the level of response to the critical items. Either method would bring about a reduction in the signal-to-noise ratio. The first method implies either an increase in the overall activation level, until additional responsiveness to critical items is small, or the production of false-positive responses to irrelevant items. The second method implies a reduction in responsivity on either a general or a specific level, brought about by voluntary suppression or by techniques involving drugs, hypnosis, or conditioning.
The voluntary production of false-positive responses The experimental literature on countermeasures has focused on the production of falsepositive responses to neutral stimuli (Kubis, 1962; Lykken, 1960). Kubis (1962) trained subjects in three types of countermeasures, two of which involved the production of false-positive responses. The subjects were told either to generate exciting mental imagery to the neutral stimuli or to clench their toe muscles in response to the same stimuli. Using
777 ON THE DETECTION OF DECEPTION
GSR, respiration, and pulse volume as response measures, both imagery and muscle tension resulted in a significant reduction in detection rates. Kubis (1962) concluded that "the most important result underlying the successful use of countermeasures is the inability of a lie detection operator to distinguish the physiological reactions in deception from those in simulated deception. In other words, the tensing of the muscles and the bringing to mind of exciting thoughts produce a physiological pattern that cannot be distinguished readily from that found in lying [p. 104]." More (1966), however, claims to have replicated the Kubis (1962) study and reported that the same countermeasures led to a much smaller reduction in the detection rate than found by Kubis. More argues that his results "indicate that a carefully trained examiner can interpret a polygram even though countermeasures are being taken by the subject [p. 75]."
Block et al. (1952) used a conditioning paradigm to produce false-positive GSR responses to neutral stimuli. During a card test the subjects were shocked for "No" responses to cards other than the one they had chosen, presumably on the assumption that shocking subjects for telling the truth would lead to heightened autonomic activity when telling the truth in subsequent trials without shock. Responses to neutral stimuli did not increase relative to critical stimuli, as was expected, but there were fewer measurable responses to all stimuli and an increase in the magnitude of the GSR during deception.
A primary disadvantage in using false-positive responses as a countermeasure stems from the nature of the field examination. As Reid and Inbau (1966) point out, unusual responses, particularly those to irrelevant stimuli, should be discussed with the suspect in an attempt to uncover plausible explanations for these responses. Unexplainable responses, movement, or a lack of attention to the interrogator may be considered prima facie evidence of guilt. These authors endorse the use of arm and thigh movement detectors to uncover evidence of muscle tension (Reid, 1945) and caution operators to be constantly aware of the possibility of conscious acts by the suspect to distort the record. An individual with a good knowledge of the examination procedure may produce responses at appropriate times and thus deceive the operator. Such an effort would require a knowledge of the order of the questions presented, as pointed out by Gustafson and Orne (1964). These investigators also note that detecting guilty information, such as the card choice in the Kubis (1962) study, allows the subject to mislead the experimenter by producing false responses, but the detection of a guilty person does not lend itself to such a strategy. The least suspicious pattern of false responding would seem to be the production of responses to all stimuli, relevant and irrelevant alike.
Experimental work dealing with increased general activation as a countermeasure has not been reported, although it is known that highly anxious subjects are often difficult to detect (Reid & Inbau, 1966). It would seem that increasing the activation of a subject should lead to a lessened difference between critical and noncritical responses because the subject is already at a level where further stimulation would have little effect. Obviously, however, a person would have to be capable of sustaining a high level of activation throughout an entire examination period. Reid and Inbau (1966) suggest that highly anxious or overly reactive suspects should be questioned about a fictitious crime, as well as about the crime in question. A lack of reactivity to questions concerning the fictitious crime would indicate deception concerning the actual crime.
The suppression of autonomic responses The entire basis for the detection of deception rests on the premise that the autonomic nervous system responds in an involuntary fashion; i.e., the conscious control of this system is difficult. There is, however, an increasing body of literature concerning the control of autonomic activity, by means of drugs, hypnosis, and particularly conditioning techniques. (See, for instance, Razran, 1961, for a review of Russian literature on conditioning.) Little research, however, has been directly concerned with response suppression in detection of deception situations.
The voluntary suppression of responses Kubis (1962) has published the only study directly involved with the voluntary suppression of responses during the detection of deception. As part of the
778 APPLICATIONS OF TECHNIQUES AND FINDINGS
previously mentioned study on countermeasures, his subjects attempted to assume a detached, "yoga" sort of attitude. He found these subjects no more difficult to detect under this condition than when given the usual instructions. Numerous authors have speculated on the inability of subjects to suppress affect at will (see Gustafson & Orne, 1964; Lykken, 1959).
The most important factor in this seeming inability to control autonomic responses is probably that the subject receives no feedback from the response he gives (Brener & Hothersall, 1966). Several studies indicate that providing feedback to subjects of heart rate information allows them to reduce heart rate variability (Hnatiow & Lang, 1965) and to increase or decrease the heart rate (Brener & Hothersall, 1966). Such feedback may take place naturally with the GSR (Reiser & Block, 1965). Kamiya (1969) has demonstrated that alpha activity in the EEG may be controlled by subjects who are informed of its appearance. The process by which subjects may learn to control autonomic activity is dependent upon providing appropriate feedback to the individual about changes and/or their direction in the parameter under investigation. Much of this work has been conceptualized as instrumental conditioning since in human studies the presentation of the appropriate feedback stimulus can serve as a reinforcer, and little effort has been made to separate reinforcement from informational aspects of the feedback stimulus. Again, partial reinforcement schedules are difficult to distinguish from intermittent feedback. It remains to be established, however, whether changes in the amount of feedback parallel those observed with varying the ratio of reinforcement contingencies. Nonetheless, it seems likely that learning to control autonomic functions in man is best understood as the consequence of feedback in individuals motivated to achieve such control rather than being due to the effects of discrete simple reinforcers.
The conditioned suppression of responses A great deal of research has been done concerning the conditioning of autonomic responses, primarily the GSR (Crider, Shapiro, & Tursky, 1966; Grings & Carlin, 1966; Kimmel & Hill, 1960; Stewart, Stern, Winokur, & Fredman, 1961) and heart rate (Brener, 1966; Notterman, Schoenfeld, & Bersh, 1952), although work has been done with other responses as well (Razran, 1961). These efforts have involved both classical and instrumental paradigms. The majority of these investigators have sought to establish responses to previously neutral stimuli, rather than extinguish responses already present. No one appears to have used the classical paradigm to suppress an autonomic response.
Several operant conditioning studies have examined the suppression of a response, particularly the GSR. Shapiro and Crider (1966) report that they were able to demonstrate increases in GSR activity with reinforcement (money) contingent upon the production of GSRs and also decreases in GSR activity with the reinforcement contingent upon inhibition of GSRs. An inspection of their data, however, shows the increase in activity to be much more striking than the decrease in activity. When reinforcement was contingent upon a lack of response, activity decreased greatly from the level during reinforcement for responding. The level during reinforcement for lack of responding was not, however, much less than it was during periods of nonreinforcement. Senter and Hummel (1965) performed an experiment in which one group of subjects received an electric shock each time a spontaneous GSR occurred. A control group received similar shocks at times when their experimental partners had received them. During the subsequent test period the experimental subjects showed a significant reduction in the number of spontaneous GSRs emitted, when compared to the number emitted by their yoked control subjects, whose number of GSRs actually increased. They also found that abrupt changes in the GSR consistently occurred between the spontaneous GSR and the shock in the experimental subjects and conclude that the spontaneous emission of a GSR "bears a sensory correlate which can act as a stimulus element in operant conditioning [p. 4]." Using the heart rate as a response, Engel and Hansen (1966) demonstrated that some subjects, particularly those who were not aware of the contingency between the feedback and their heart rate, were able to lower their rate significantly below that of their yoked control subjects.
The ability of operant conditioning techniques to suppress a response pattern to critical emotional
779 ON THE DETECTION OF DECEPTION
stimuli has not been demonstrated. Even if a person is conditioned to suppress responses to GSR, heart rate, respiration, and other variables separately, he could experience difficulty in the simultaneous control of all of these measures while maintaining normal external appearance to the interrogator (Orlansky, 1962). It may be possible to condition a lack of concern for the critical issues of the examination. There is some evidence that a lack of concern leads to low rates of detection, both in laboratory settings (Gustafson & Orne, 1963) and in field settings involving abnormal personalities (Floch, 1950; Kubis, 1957). Outside of the one laboratory study just mentioned, though, little research has been directed toward this topic. Studies of subjects who are difficult to detect or who show low reactivity to critical stimuli indicate them to be less theoretical and abstract in their thinking than sensitive subjects (Iwahara, Miseki, Shiokawa, & Yoshida, 1960). They are also characterized as cool, independent, opportunistic, practical, and realistic by their responses to the Gough adjective checklist (Block, 1957). Ethnic background also seems to be a determiner of differential reactivity (Kugelmass, 1967; Kugelmass & Lieblich, 1966).
The use of drugs The use of drugs as aids to interrogation has been adequately reviewed elsewhere (Gottschalk, 1961) and will not be considered. This literature has dealt with the use of sedatives, stimulants, tranquilizers, or hallucinogenic drugs as possible aids in maximizing the effects of interrogation. The use of drugs to minimize detection, on the other hand, especially when the polygraph is employed, has not received much attention in print. Davis (1961) considers alcohol, barbiturates, and perhaps tranquilizers to be of possible use in minimizing detection, but the lack of empirical investigation renders this view speculative. On the basis of known evidence relating the effects of various depressants and tranquilizers to autonomic and central nervous system activity and behavior (Trouton & Eysenck, 1961), it could be predicted that the use of such drugs would impair detection. One of the more troublesome aspects of the use of drugs as countermeasures, however, is the possibility that the interrogator will discover that drugs have been taken. Unless a plausible reason can be found for the taking of a drug, this act may be considered prima facie evidence of the intent to deceive. Reid and Inbau (1966) state that doses of tranquilizers sufficient to depress responsivity usually produce obvious behavior abnormalities and suggest that the examination be postponed until the interrogator can be sure that no drugs have been taken which would alter the effectiveness of the test.
The use of hypnosis Hypnosis has been considered as a technique for either maximizing or minimizing detection (Orlansky, 1962), although evidence suggests that the use of posthypnotic amnesia to suppress autonomic responses is generally ineffective. Germann (1961), for example, induced posthypnotic amnesia in five subjects, screened for deep trance capability. Clinical evaluation of the GSR, blood pressure, and respiration responses to the critical stimuli showed that amnesia did not prevent an increased autonomic response to these items. Bitterman and Marcuse (1945), in a series of hypnotic inductions with one subject, were likewise unable to demonstrate any reduction in autonomic response resulting from posthypnotic amnesia. However, no data are available about other possible uses of the hypnotic phenomenon to prevent detection. Hypnosis might, for example, be used as a means of decreasing an individual's anxiety, either by direct suggestion or along the lines of systematic desensitization, making the subject more sensitive about a pseudocrime that was never committed by the use of appropriate suggestions. Alternatively, it could be used to induce a state of high anxiety that would make differential responsivity difficult to recognize, or to vary the effect of the meaning of the stimulus, rather than its recognition. Studies of this kind, carried out with appropriate waking controls, might help clarify the nature of hypnosis, as well as help to shed light on some of the mechanisms involved in causing the differential responsivity upon which the detection of deception is based.
CONCLUSIONS
In this review independent variables have been categorized for the purpose of organization. It must be recognized, however, that such categorization distorts some of the more complex effects these
780 APPLICATIONS OF TECHNIQUES AND FINDINGS
variables may have. For example, there is almost no increase in responsivity when subjects are asked to lie about a card they have selected, unless they are motivated to deceive; yet in another study in which the stimulus material has great intrinsic meaning, differential responsivity occurs even when the subject is telling the truth. Such findings are not contradictory, although one might well be tempted to draw such a conclusion; rather, they indicate that several factors play a role in the detection of deception. Either the attempt to deceive or the intrinsic significance of the stimulus may be responsible for the physiological response, and an experimental study may emphasize one or the other factor by the nature of its design.
Most studies to date have only begun to examine the effect of independent variables singly, and certainly such research is essential. Yet a full understanding of the phenomena involved in the detection of deception will need to take into account some of the significant interactions between these variables. Many of the apparent inconsistencies found between studies may ultimately be resolved when the effects of these interactions are clarified. We have tried to discuss the factors that seemed likely to play a role in this process, but it has become clear that even the apparently simple laboratory situation is highly complex and many aspects of the field situation remain to be subjected to meaningful analysis.
The detection of deception has not, however, been an area of much systematic research. Perhaps because of its applied implications, the field has been avoided by psychologists and physiologists alike. Psychophysiologists, in the search for professional identity, have seemed least interested in the phenomenon that the public most closely identifies with the polygraph. As a result of the increased status of psychophysiology, the intrinsic interest of the problem may receive the attention it deserves. There is little doubt that both psychological and physiological variables play large roles in determining the ability of a person using polygraphic techniques to judge whether or not another individual is guilty of deception. The conditions prevailing in the detection of deception situation seem to provide an almost unparalleled opportunity for psychologists to study those contingencies under which a psychological stimulus will or will not evoke an augmented physiological response. Clarification of these problems may be of interest to individuals professionally engaged in "lie detection." This in no way, however, diminishes the scientific importance of the phenomenon. It provides a paradigm for studying the effect of cognitive, motivational, attentional, and learning factors upon the physiological responsivity to verbal stimuli. It would seem appropriate that, instead of viewing the practical application as a detriment, we recognize its potential advantage in providing a means of testing the ecological validity of our findings outside the laboratory context. Because the detection of deception model allows systematic quantitative analysis, it provides a meaningful way of learning about the variables involved and the interactions, at least as they relate to one set of psychophysiological contingencies. An understanding of these relationships is likely to facilitate the use of physiological techniques to clarify other, and perhaps ultimately more important, aspects of psychological functioning.
REFERENCES
ADEY, W. R., KADO, R. T., & WALTER. D. O. Computer analysis of EEG data from Gemini flight GT-7. Aerospace Medicine, 1967, 38, 345-359.
ALPERT, M., KURTZBERG, R. L., & FRIEDHOFF, A. J. Transient voice changes associated with emotional stimuli. Archives of General Psychiatry, 1963, 8, 362-365.
ANSLEY, N. A comparison of arm-cuff and wrist-cuff blood pressure patterns in polygraph charts. Journal of Criminal Law, Criminology, and Police Science, 1959, 50, 192-194.
ARNOLD, M. B. Emotion and personality. New York: Columbia University Press, 1960.
AX, A. F. The physiological differentiation between fear and anger in humans. Psychosomatic Medicine, 1953,15, 433-442.
BACKSTER, C. Polygraph professionalization through technique standardization. Law & Order, 1963, 11(4), 63-64.
BAESEN, H. V., CHUNG, C. M., & YANG, C. Y. A lie-detector
781 ON THE DETECTION OF DECEPTION
experiment. Journal of Criminal Law and Criminology, 1949, 39, 532-537.
BARRON, D. H., & RUCH, T. C. Circulation through special regions. In T. C. RUCH & J. F. FULTON (Eds.), Medical physiology and biophysics. (18th ed.) (Ch. 33.) Philadelphia: Saunders, 1960.
BENDER, W. R. G. The effect of pain and emotional stimuli and alcohol upon pupillary reflex activity. Psychological Monographs, 1933, 44, No. 2(No. 198).
BEN SHAKHAR, G., LIEBLICH, I., & KUGELMASS, S. Guilty knowledge technique: Application of signal detection measures. Journal of Applied Psychology, 1970, 54, 409-413.
BENUSSI, V. Die Atmungssymptome der Luge. Archiv fur die gesamte Psychologie, 1914, 31, 244-273.
BERRIEN, F. K. Ocular stability in deception. Journal of Applied Psychology, 1942, 26, 55-63.
BERRIEN, F. K., & HUNTINGTON, G. H. An exploratory study of pupillary responses during deception. Journal of Experimental Psychology, 1943, 32, 443-449.
BERSH, P. J. A validation of polygraph examiner judgments. Journal of Applied Psychology, 1969, 53, 399-403.
BEST, C. H., & TAYLOR, N. B. The physiological basis of medical practice. (5th ed.) Baltimore: Williams & Wilkins, 1950.
BITTERMAN, M. E., & MARCUSE, F. L. Autonomic response in posthypnotic amnesia. Journal of Experimental Psychology, 1945, 35, 248-252.
BLOCK, J. D. A study of affective responsiveness in a lie-detection situation. Journal of Abnormal and Social Psychology, 1957, 55, 11-15.
BLOCK, J. D., ROUKE, F. L., SALPETER, M. M., TOBACH, E., KUBIS, J. F., & WELCH, L. An attempt at reversal of the truth-lie relationship as measured by the psychogalvanic response. Journal of Psychology, 1952, 34, 55-66.
BORING, E. G. Sensation and perception in the history of experimental psychology. New York: Appleton-CenturyCrofts, 1942.
BRENER, J. Heart rate as an avoidance response. Psychological Record, 1966, 16, 329-336.
BRENER, J., & HOTHERSALL, D. Heart rate control under conditions of augmented sensory feedback. Psychophysiology, 1966, 3, 23-28.
BROD, J. Circulation in muscle during acute pressor responses to emotional stress and during chronic sustained elevation of blood pressure. American Heart Journal, 1964, 68, 424-426.
BROD, J., HEJL, Z., & ULRYCH, M. Metabolic changes in the forearm muscle and skin during emotional muscular vasodilation. Clinical Science, 1963, 25, 1-10.
BROWN, C. C. The techniques of plethysmography. In C. C. BROWN (Ed.), Methods in psychophysiology. (Ch. 2.) Baltimore: Williams & Wilkins, 1967.
BURTT, H. E. The inspiration-expiration ratio during truth and falsehood. Journal of Experimental Psychology, 1921, 4, 1-23.
BURTT, H. E. Motor concomitants of the association reaction. Journal of Experimental Psychology, 1936, 19, 51-63.
CHAPPELL, M. N. Blood pressure changes in deception. Archives of Psychology, New York, 1929, 17, 1-39.
COLEMAN, R., GREENBLATT, M., & SOLOMON, H. C. Physiological evidence of rapport during psychotherapeutic interviews. Diseases of the Nervous System, 1956, 17, 71-77.
CRIDER, A., SHAPIRO, D., & TURSKY, B. Reinforcement of spontaneous electrodermal activity. Journal of Comparative and Physiological Psychology, 1966, 61, 20-27.
CRONBACH, L. J., & GLESER, G. C. Psychological tests and personnel decisions. (2nd ed.) Urbana: University of Illinois Press, 1965.
CROSLAND, H. R. The psychological methods of word-association and reaction-time as tests of deception. Oregon University Publications, Psychology Series, 1929, 1(No. 1).
DANA, H. J. It is time to improve the polygraph: A progress report on polygraph research and development. In V. A. LEONARD (Ed.), Academy lectures on lie detection. Vol. 2. (Ch. 8.) Springfield, Ill.: Charles C Thomas, 1958.
DANA, H. J., & BARNETT, C. C. The emotional stress meter. In V. A. LEONARD (Ed.), Academy lectures on lie detection. (Ch. 7.) Springfield, Ill.: Charles C Thomas, 1957.
DARROW, C. W. The galvanic skin reflex (sweating) and blood-pressure as preparatory and facilitative functions. Psychological Bulletin, 1936, 33, 73-94.
DAVIDSON, P. O. Validity of the guilty-knowledge technique: The effects of motivation. Journal of Applied Psychology, 1968, 52, 62-65.
DAVIS, R. C. Physiological responses as a means of evaluating information. In A. D. BIDERMAN & H. ZIMMER (Eds.), The manipulation of human behavior. (Ch. 4.) New York: Wiley, 1961.
DAVIS, R. C., BUCHWALD, A. M., & FRANKMANN, R. W. Autonomic and muscular responses, and their relation to simple stimuli. Psychological Monographs, 1955, 69, No. 20 (No. 405).
DEARMAN, H. B., & SMITH, B. M. Unconscious motivation and the polygraph test. American Journal of Psychiatry, 1963, 119, 1017-1020.
EDELBERG, R. Electrical properties of the skin. In C. C. BROWN (Ed.), Methods in psychophysiology. (Ch. 1.) Baltimore: Williams & Wilkins, 1967.
EKMAN, P., & FRIESEN, W. V. Nonverbal leakage and clues to deception. Psychiatry, 1969, 63, 88-106.
ELLSON, D. G., DAVIS, R. C., SALTZMAN, I. J., & BURKE, C. J. A report of research on detection of deception. Tech. Rep. prepared for Office of Naval Research, Contract N6onr-18011, Indiana University, 1952.
ENGEL, B. T., & HANSEN, S. P. Operant conditioning of heart rate slowing. Psychophysiology, 1966, 3, 176-187.
ENGLISH, H. B. Reaction-time symptoms of deception. American Journal of Psychology, 1926, 37, 428-429.
FAY, P. J., & MIDDLETON, W. C. The ability to judge truthtelling, or lying, from the voice as transmitted over a
782 APPLICATIONS OF TECHNIQUES AND FINDINGS
public address system. Journal of General Psychology, 1941, 24, 211-215.
FLOCH, M. Limitations of the lie detector. journal of Criminal Law and Criminology, 1950, 40, 651-653.
GALTON, F. Psychometric experiments. Brain, 1879, 2, 149-162.
GELDREICH, E. W. Studies of the galvanic skin response as a deception indicator. Transactions of the Kansas Academy of Science, 1941, 44, 346-351.
GERMANN, A. C. Hypnosis as related to the scientific detection of deception by polygraph examination: A pilot study. International Journal of Clinical and Experimental Hypnosis, 1961, 9, 309-311.
GOLDSTEIN, E. R. Reaction times and the consciousness of deception. American Journal of Psychology, 1923, 34, 562-581.
GOTTSCHALK, L. A. The use of drugs in interrogation. In A. D. BIDERMAN & H. ZIMMER (Eds.), The manipulation of human behavior. (Ch. 3.) New York: Wiley, 1961.
GRINGS, W. W., & CARLIN, S. Instrumental modification of autonomic activity. Psychological Record, 1966, 16, 153-159.
GUSTAFSON, L. A., & ORNE, M. T. Effects of heightened motivation on the detection of deception. Journal of Applied Psychology, 1963, 47, 408-411.
GUSTAFSON, L. A., & ORNE, M. T. The effects of task and method of stimulus presentation on the detection of deception. Journal of Applied Psychology, 1964, 48, 383-387.
GUSTAFSON, L. A., & ORNE, M. T. Effects of perceived role and role success on the detection of deception. Journal of Applied Psychology, 1965, 49, 412-417. (a)
GUSTAFSON, L. A., & ORNE, M. T. The effects of verbal responses on the laboratory detection of deception. Psychophysiology, 1965, 2, 10-13. (b)
HATHAWAY, S. R., & HANSCOM, C. B. The statistical evaluation of polygraph records. In V. A. LEONARD (Ed.), Academy lectures on lie detection. Vol. 2. (Ch. 11.) Springfield, III.: Charles C Thomas, 1958.
HESS, E. H. Attitude and pupil size. Scientific American, 1965, 212(4), 46-54.
HESS, E. H., SELTZER, A. L., & SHLIEN, J. M. Pupil response of hetero- and homosexual males to pictures of men and women: A pilot study. Journal of Abnormal Psychology, 1965, 70, 165-168,
HIGLEY, B. R. Interrogation with instrumentation. In V. A. LEONARD (Ed.), Academy lectures on lie detection. Vol. 2. (Ch. 10.) Springfield, III.: Charles C Thomas, 1958.
HNATIOW, M., & LANG, P. J. Learned stabilization of cardiac rate. Psychophysiology, 1965, 1, 330-336.
HOUSE OF REPRESENTATIVES, EIGHTY-EIGHTH CONGRESS, Subcommittee of the Committee on Government Operations. Use of polygraphs as "lie detectors" by the federal government. Washington, D.C.: U.S. Government Printing Office, 1964.
INBAU, F. E., & REID, J. E. Lie detection and criminal interrogation. (3rd ed.) Baltimore: Williams & Wilkins, 1953.
IWAHARA, S., MISEKI, H., SHIOKAWA, N., & YOSHIDA, R. GSR to forced lying and personality traits. Psychological Abstracts, 1960, 6, 269. (Abstract)
JONES, H. E., & WECHSLER, D. Galvanometric technique in studies of association. American Journal of Psychology, 1928, 40, 607-612.
JOSEPH, C. N. Analysis of compensatory responses and irregularities in polygraph chart interpretation. In V. A. LEONARD (Ed.), Academy lectures on lie detection. (Ch. 9.) Springfield, Ill.: Charles C Thomas, 1957.
JUNG, C. G. Die psychologische Diagnose des Tatbestandes. Juristisch-psychiatrische Grenzfragen, 1906, 4(2), 1-61.
JUNG, C. G. On psychophysical relations of the associative experiment. Journal of Abnormal Psychology, 1907, 1, 247-255.
JUNG, C. G. The association method. American Journal of Psychology, 1910, 21, 219-269.
KAMIYA, J. Operant control of the EEG alpha rhythm and some of its reported effects on consciousness. In C. TART (Ed.), Altered states of consciousness: A book of readings. (Ch. 35.) New York: Wiley, 1969.
KEELER, L. A method for detecting deception. American Journal of Police Science, 1930, 1, 38-51.
KIMMEL, H., & HILL, F. Operant conditioning of the GSR. Psychological Reports, 1960, 7, 555-562.
KRAMER, E. Judgment of personal characteristics and emotions from nonverbal properties of speech. Psychological Bulletin, 1963, 60, 408-420.
KUBIS, J. F. Experimental and statistical factors in the diagnosis of consciously suppressed affective experience. Journal of Clinical Psychology, 1950, 6, 12-16.
KUBIS, J. F. Instrumental, chemical, and psychological aids in the interrogation of witnesses. Journal of Social Issues, 1957, 13(2), 40-49.
KUBIS, J. F. Studies in lie detection: Computer feasibility considerations. Tech. Rep. 62-205, prepared for Air Force Systems Command, Contract No. AF 30 (602)2270, Project No. 5534, Fordham University, 1962.
KUGELMASS, S. Reactions to stress. Air Force Office of Scientific Research Rep. 67-0530, filed with Defense Documentation Center AD-647467, January 1967.
KUGELMASS, S., & LIEBLICH, I. Effects of realistic stress and procedural interference in experimental lie detection. Journal of Applied Psychology, 1966, 50, 211-216.
KUGELMASS, S., & LIEBLICH, I. An analysis of mechanisms underlying psychophysiological detection. Paper presented at the 16th International Congress of Applied Psychology, Amsterdam, The Netherlands, August 18-22, 1968.
KUGELMASS, S., LIEBLICH, I., BEN-ISHAI, A., OPATOWSKI, A., & KAPLAN, M. Experimental evaluation of galvanic skin response and blood pressure change indices during
783 ON THE DETECTION OF DECEPTION
criminal investigation. Journal of Criminal Law, Criminology and Police Science, 1968, 59, 632-635.
KUGELMASS, S., LIEBLICH, I., & BERGMAN, Z. The role of "lying" in psychophysiological detection. Psychophysiology, 1967, 3, 312-315.
LACEY, J. I., & LACEY, B. C. Verification and extension of the principle of autonomic response-stereotypy. American Journal of Psychology, 1958, 71, 50-73.
LANDIS, C., & GULLETTE, R. Studies of emotional reactions: III. Systolic blood pressure and inspiration-expiration ratios. Journal of Comparative Psychology, 1925, 5, 221-253.
LANDIS, C., & WILEY, L. E. Changes of blood pressure and respiration during deception. Journal of Comparative Psychology, 1926, 6, 1-19.
LANGFELD, H. S. Psychophysical symptoms of deception. Journal of Abnormal Psychology, 1921, 15, 319-328.
LARSON, J. A. The cardio-pneumo-psychogram and its use in the study of the emotions, with practical application. Journal of Experimental Psychology, 1922, 5, 323-328.
LARSON, J. A. Lying and its detection: A study of deception and deception tests. Chicago: University of Chicago Press, 1932.
LEE, C. D. The instrumental detection of deception. Springfield, III.: Charles C Thomas, 1953.
LIEBLICH, I. Identification of variables related to the efficiency of the detection of information within the organism, using a physiological index. Unpublished doctoral dissertation, Hebrew University of Jerusalem, 1966.
LIEBLICH, I. Manipulation of contrast between differential GSR responses through the use of ordered tasks of information detection. Psychophysiology, 1969, 6, 70-77.
LIEBLICH, I., KUGELMASS, S., & BEN SHAKHAR, G. Efficiency of GSR detection of information as a function of stimulus set size. Psychophysiology, 1970, 6, 601-608.
LINDSLEY, D. B. Emotion. In S. S. STEVENS (Ed.), Handbook of experimental psychology. (Ch. 14.) New York: Wiley, 1951.
LOVETT DOUST, J. W., & SCHNEIDER, R. A. Studies on the physiology of awareness: An oximetrically monitored controlled stress test. Canadian Journal of Psychology, 1955, 9, 67-78.
LOWENSTEIN, O. Pupillography: Methods and diagnostic system. Archives of Ophthalmology, 1956, 55, 565-571.
LOWENSTEIN, O., & FRIEDMAN, E. D. Pupillographic studies: I. Present state of pupillography; its method and diagnostic significance. Archives of Ophthalmology, 1942, 27, 969-993.
LURIA, A. R. The nature of human conflicts: Or emotion, conflict and will. Translated by W. H. Gantt. New York: Liveright, 1932.
LYKKEN, D. T. The GSR in the detection of guilt. Journal of Applied Psychology, 1959, 43, 385-388.
LYKKEN, D. T. The validity of the guilty knowledge technique: The effects of faking. Journal of Applied Psychology, 1960, 44, 258-262.
MacNITT, R. D. In defense of the electrodermal response and cardiac amplitude as measures of deception. Journal of Criminal Law and Criminology, 1942, 33, 266-275.
MALMO, R. B., BOAG, T. J., & SMITH, A. A. Physiological study of personal interaction. Psychosomatic Medicine, 1957, 19, 105-119.
MARCUSE, F. L., & BITTERMAN, M. E. Minimal cues in the peak of tension procedure of determining guilt. American Journal of Psychology, 1946, 59, 144-146.
MARSTON, W. M. Systolic blood pressure symptoms of deception. Journal of Experimental Psychology, 1917, 2, 117-163.
MARSTON, W. M. Reaction-time symptoms of deception. Journal of Experimental Psychology, 1920, 3, 72-87.
MARSTON, W. M. Psychological possibilities in the deception tests. Journal of Criminal Law and Criminology, 1921, 11, 551-570.
MARSTON, W. M. Negative type reaction-time symptoms of deception. Psychological Review, 1925, 32, 241-247.
MARSTON, W. M. The lie detector test. New York: R. K. Smith, 1938.
MORE, H. W. Polygraph research and the university. Law & Order, 1966, 14(3), 73-78.
MORGAN, M. I., & OJEMANN, R. H. A study of the Luria method. Journal of Applied Psychology, 1942,26,168-179.
NASA MANNED SPACECRAFT CENTER. Results of the first U.S. orbital space flight, February 20, 1962. Washington, D.C.: U.S. Government Printing Office, 1962.
NOTTERMAN, J. M., SCHOENFELD, W. N., & BERSH, P. J. Conditioned heart rate response in human beings during experimental anxiety. Journal of Comparative and Physiological Psychology, 1952, 45, 1-8.
OBERMANN, C. E. The effect on the Berger rhythm of mild affective states. Journal of Abnormal and Social Psychology, 1939, 34, 84-95.
ORLANSKY, J. An assessment of lie detection capabilities. Tech. Rep. 62-16, prepared for Institute for Defense Analysis, Research and Engineering Support Division, AD 603860. Springfield, Va.: Clearing House of Scientific and Technical Information, 1962.
ORNE, M. T. On the social psychology of the psychological experiment: With particular reference to demand characteristics and their implications. American Psychologist, 1962, 17, 776-783.
ORNE, M. T. Demand characteristics and the concept of design controls. In R. ROSENTHAL & R. ROSNOW (Eds.), Artifact in behavioral research. (Ch. 5.) New York: Academic, 1969.
ORNE, M. T., & THACKRAY, R. I. Group GSR technique in the detection of deception. Perceptual and Motor Skills, 1967, 25, 809-816.
PETERSON, F., & JUNG, C. G. Psycho-physical investigations with the galvanometer and pneumograph in normal and insane individuals. Brain, 1907, 30, 153-218.
784 APPLICATIONS OF TECHNIQUES AND FINDINGS
RAZRAN, G. The observable unconscious and the inferable conscious in current Soviet psychophysiology: Interoceptive conditioning, semantic conditioning, and the orienting reflex. Psychological Review, 1961, 68, 81-147.
REID, J. E. Simulated blood pressure responses in lie-detector tests and a method for their detection. Journal of Criminal Law and Criminology, 1945, 36, 201-214.
REID, J. E., & ARTHER, R. O. Behavior symptoms of liedetector subjects. Journal of Criminal Law and Criminology, 1953, 44, 104-108.
REID, J. E., & INBAU, F. E. Truth and deception: The polygraph ("lie detector") technique. Baltimore: Williams & Wilkins, 1966.
REISER, M. F., & BLOCK, J. D. Discrimination and recognition of weak stimuli: III. Further experiments on interaction of cognitive and autonomic-feedback mechanisms. Psychosomatic Medicine, 1965, 27, 274-285.
RICH, G. J. Dr. Marston on deception types. American Journal of Psychology, 1926, 37, 307-309.
ROSENTHAL, R. Experimenter effects in behavioral research. New York: Appleton-Century-Crofts, 1966.
RUCKMICK, C. A. The truth about the lie detector. Journal of Applied Psychology, 1938, 22, 50-58.
RUNKEL, J. E. Luria's motor method and word association in the study of deception. Journal of General Psychology, 1936, 15, 23-37.
RUSHMER, R. F. Cardiovascular dynamics. Philadelphia: Saunders, 1961.
SCHNORE, M. M. Individual patterns of physiological activity as a function of task differences and degree of arousal. Journal of Experimental Psychology, 1959, 58, 117-128.
SENTER, R. J., & HUMMEL, W. F. Suppression of an autonomic response through operant conditioning. Psychological Record, 1965, 15, 1-5.
SEWARD, J. P., & SEWARD, G. H. The effect of repetition on reaction to electric shock: With special reference to the menstrual cycle. Archives of Psychology, New York, 1934, 25 (No. 168).
SHAPIRO, D,, & CRIDER, A. Operant electrodermal conditioning: Some effects of multiple schedules of reinforcement. Tech. Rep. 14, prepared for Office of Naval Research, Group Psychology Branch, AD 641699. Springfield, Va.: Clearing House of Scientific and Technical Information, 1966.
SMITH, W. The measurement of emotion. London: Paul, 1922.
SOLOMON, R. L., BLACK, A. H., WATSON, P. D., HUTTENLOCHER, J., TURNER, L., & WESTCOTT, M. R. Some autonomic correlates of human traumatic avoidance learning. U.S. Public Health Service Progress Rep. USPHS M-1246, February 1958.
STEIN, M., & LUPARELLO, T. J. The measurement of respiration. In C. C. BROWN (Ed.), Methods in psychophysiology. (Ch. 3.) Baltimore: Williams & Wilkins, 1967.
STERNBACH, R. A. A comparative analysis of autonomic responses in startle. Psychosomatic Medicine, 1960, 22, 204-210.
STERNBACH, R. A., GUSTAFSON, L. A., & COLIER, R. L. Don't trust the lie detector. Harvard Business Review, 1962, 40, 127-134.
STEWART, M. A., STERN, J. A., WINOKUR, G., & FREDMAN, S. An analysis of GSR conditioning. Psychological Review, 1961, 68, 60-67.
SUMMERS, W. G. Science can get the confession. Fordham Law Revue, 1939, 8, 334-354.
SUTCLIFFE, J. P. A general method of analysis of frequency data for multiple classification designs. Psychological Bulletin, 1957, 54, 134-137.
SYZ, H. C. Observations on the unreliability of subjective reports of emotional reactions. British Journal of Psychology, 1926, 17, 119-126.
THACKRAY, R. I., & ORNE, M. T. Methodological studies in detection of deception. U.S. Army Medical Research and Development Command Research Rep., filed with the Defense Documentation Center, AD-645102, January 1967.
THACKRAY, R. I., & ORNE, M. T. A comparison of physiological indices in detection of deception. Psychophysiology, 1968, 4, 329-339. (a)
THACKRAY, R. I., & ORNE, M. T. Effects of the type of stimulus employed and the level of subject awareness on the detection of deception. Journal of Applied Psychology, 1968, 52, 234-239. (b)
THIESEN, J. W. Effects of certain forms of emotion on the normal electroencephalogram. Archives of Psychology, New York, 1943, 40 (No. 285).
TROUTON, D., & EYSENCK, H. J. The effects of drugs on behaviour. In H. J. EYSENCK (Ed.), Handbook of abnormal psychology: An experimental approach. (Ch. 17.) New York: Basic Books, 1961.
TROVILLO, P. V. A history of lie detection. Journal of Criminal Law and Criminology, 1939, 29, 848-881.
TROVILLO, P. V. Deception test criteria: How can one determine truth and falsehood from polygraph records. Journal of Criminal Law and Criminology, 1942, 33, 338-358.
VAN BUSKIRK, D., & MARCUSE, F. L. The nature of errors in experimental lie detection. Journal of Experimental Psychology, 1954, 47, 187-190.
VIOLANTE, R., & Ross, S. A. Research in interrogation procedures. Office of Naval Research, Rep. No. 707-65, filed with Defense Documentation Center, AD-467624, October 1964.
WATSON, P. D, & KANTER, S. S. Some influences of an experimental situation on the psychotherapeutic process: A report, based on 44 treatment interviews, of the reactions of a patient and therapist to observation, recording, and physiological measurement. Psychosomatic Medicine, 1956, 18, 457-470.
785 ON THE DETECTION OF DECEPTION
WENGER, M. A., CLEMENS, T. L., COLEMAN, M. A., CULLEN, T. D., & ENGEL, B. T. Autonomic response specificity. Psychosomatic Medicine, 1961, 23, 185-193.
WERTHEIMER, M. Experimentelle Untersuchungen zur Tatbestandsdiagnostik. Archiv fur die gesamte Psychologie, 1905, 6, 59-131.
WERTHEIMER, M., & KLEIN, J. Psychologische Tatbestandsdiagnostik. Archiv fur Kriminalanthropologie und Kriminalistik, 1904, 15, 72-113.
WILLIAMS, A., Jr. Some psychological correlates of the electroencephalogram. Archives of Psychology, New York, 1939, 34(No. 240).
WILLIAMS, J. G. L., & WILLIAMS, B. Arterial pulse wave velocity as a psychophysiological measure. Psychosomatic Medicine, 1965, 27, 408-414.
WINTER, J. E. A comparison of the cardio-pneumopsychograph and association methods in the detection of lying in cases of theft among college students. Journal of Applied Psychology, 1936, 20, 243-248.
WOODMANSEE, J. J. Methodological problems in pupillographic experiments. Proceedings of the 74th Annual Convention of the American Psychological Association, 1966, 133-134.
WOODWORTH, R. S., & SCHLOSBERG, H. Experimental psychology. (Rev. ed.) New York: Holt, Rinehart and Winston, 1954.
YATES, A. Abnormalities of psychomotor functions. In H. J. EYSENCK (Ed.), Handbook of abnormal psychology: An experimental approach. (Ch. 2.) New York: Basic Books, 1961.
YERKES, R. M., & BERRY, C. S. The association reaction method of mental diagnosis (Tatbestandsdiagnostik). American Journal of Psychology, 1909, 20, 22-37.
Table 19.1 (p. 753) (reprinted after an article by Gustafson, L.A., & Orne, M.T. Effects of perceived role and role success on the detection of deception. Journal of Applied Psycholgy, 1965, 49, 412-417.) is reproduced here with the kind permission of the American Psychological Association © 1965. No further reproduction or distribution of this table is permitted without written permission of the publisher.
Table 19.2 (p. 754) (reprinted after an article by Gustafson, L.A., & Orne, M.T. Effects of perceived role and role success on the detection of deception. Journal of Applied Psychology, 1965, 49, 412-417.) is reproduced here with the kind permission of the American Psychological Association © 1965. No further reproduction or distribution of this table is permitted without written permmission of the publisher.
.