In a detection of deception experiment comparisons were made of the effects of 2 methods of stimulus presentation and 2 different subject tasks. The relevant-irrelevant method of stimulus presentation proved equally effective for both tasks, but the peak-of-tension method was significantly less effective where the S's task was to deceive as to the nature of guilty information possessed (guilty information paradigm) than it was where the task was to deceive as to the possession of any information (guilty person paradigm). In general, Ss found it easier to deceive in the guilty information paradigm, where they could attempt to "appear guilty" on a noncritical item and especially when they could anticipate the order of presentation of items (peak-of-tension method).
This study was designed to investigate the relative effectiveness of two different methods of stimulus presentation currently used in the detection of deception. The method most used in experimental studies presents significant items randomly interspersed with irrelevant ones without the subjects (Ss) knowing which items will be presented next. This technique, once widely used in commercial lie detection, is known as the relevantirrelevant method (RI). The method of presentation which has been most popular in recent years in commercial lie detection was described in detail by Inbau and Reid (1953) and by Lee (1953) and is called the peak-oftension method (PT). Here stimuli are presented to S in a known sequence. He knows exactly when crucial items will be presented and the technique is built upon this knowledge. The operator looks for gradual increase in tension coming to a peak at the crucial item with a sudden relaxation of tension when the item is past.
In careful analysis of our pilot study data it became clear that there may be a marked interaction between these two methods of stimulus presentation and the precise task which is required of S in order to deceive. In most laboratory studies S is given an item of information of five or six possible items and the experimenter (E) studies the differences in physiological responses to these items to detect the "guilty information" (this will be called the "guilty information paradigm").3 Previous findings by the authors indicate that successful detection depends on S's motivation to deceive (Gustafson & Orne, 1963). Careful postexperimental inquiry of the motivated Ss and analysis of the data indicate that Ss who successfully deceive most frequently do so by volitionally producing a physiological response to the wrong item rather than by suppressing the "automatic" response to the crucial one. In the laboratory situation designed to study the guilty knowledge this strategy effectively prevents detection. However, it would not work in most field situations where the
2 The authors wish to express their appreciation to Emily Carota Orne and M. J. Moskowitz for their critical comments in the preparation of this manuscript. This research was conducted at Massachusetts Mental Health Center and Harvard Medical School.
3 The similarity between the guilty information paradigm used here and the "guilty knowledge technique" used by Lykken (1960) should be pointed out. The authors acknowledge their debt to Lykken in the formulation of this paradigm. Lykken (1959) has used the guilty person paradigm, but in neither case does he point out the differences in the dynamics of the two situations.
383
384 LAWRENCE A. GUSTAFSON AND MARTIN T. ORNE
suspect's task is to convince the operator that he is innocent -- not that he is guilty of something else. This situation, which may be called the "guilty person paradigm," requires S to mask all his responses rather than voluntarily responding to an irrelevant item.
For these reasons an experimental guilty person situation was designed. Whereas in the guilty information situation S selects one card from a deck where all the cards are number cards with the task of E to determine which was chosen, in the guilty person situation, S selects a card from a deck in which there are both blank and number cards with S's task to convince E that he has drawn a blank card. In the guilty information situation it is thus possible for S to employ the strategy of creating false positives and succeed in deception while in the guilty person situation this strategy could not succeed. In the latter the S is required to appear innocent, i.e., to have drawn a blank card, and success is defined as: "If you draw a blank card, you must appear innocent -- as in fact you are. If you draw a number, you must still appear as though you had drawn a blank card, i.e., innocent. If you seem to have drawn a numbered card whether or not you did, you fail on the experiment." This now more closely approximates real life where the suspect must appear innocent to the operator in order to succeed. It does not help him to appear guilty of crimes he did not commit, nor does he serve his needs to appear guilty if he is in fact innocent.
The present study was an attempt to make a direct comparison of the detection of deception between the two paradigms of deception discussed above and using the two methods of stimulus presentation described. Specifically, the study was designed to determine the relative effectiveness in detection of the two methods of stimulus presentation (PT and RI) when used with two models of deception (guilty information paradigm and guilty person paradigm).
METHOD
Subjects. Fifty-three male undergraduate students from local universities were recruited from their school employment offices and paid for their participation. They had not previously taken part in studies of detection in deception, but many had taken part in other psychological experiments.
Procedures. The Ss were divided into two groups with 24 Ss in the guilty information group and 29 Ss in the guilty person group. (Because in the guilty person group it was possible for a S to be "innocent" on one or more trials, extra Ss were run until a group of 24 had been obtained who were guilty on both a PT and an RI trial.)
Each S in both groups received two trials using PT (peak-of-tension) presentation and two trials using RI (relevant-irrelevant) presentation of materials.
The stimulus materials consisted of packs of seven cards each bearing a different number (for the guilty person group, two of the seven cards in each pack were blank) and corresponding tapes on which E had vocally recorded the same numbers, each number being presented twice. For different trials the numbers on the cards and tapes were of different series, the series used being 22-28, 32-38, 42-48, and 52-58. For PT trials, the numbers were recorded in ascending order and repeated in descending order. For RI trials, the numbers were presented in random order and repeated in a different order.
The Guilty Information Group. Each S listened to a tape recording of instructions which informed him that his task was to try to deceive E as to the number on the card he picked at the beginning of each trial. The S was instructed to write the number he had picked on a sheet of paper which would later be sealed in an envelope to insure that he actually attended to the number. He was also told that numbers would be read to him including the number which he had selected, that these numbers would sometimes be read to him in numerical sequence and sometimes in random order, and that he would be informed at the beginning as to the method of presentation of the numbers. Both E and S knew that all cards had numbers on them.
When the instructions had been given, S was given an opportunity to ask questions about the procedure which E answered when possible. The E then attached the necessary electrodes for recording skin resistance, heart rate, respiration, and finger pulse volume. The methods for recording skin resistance were those used by Wenger, Engel, and Clemens (1957). (The other measures were included for exploratory reasons and will not be discussed further.)
All Ss then drew a card, looked at it, wrote the number down on another card, and sealed it in an envelope. The E then went into an adjoining room and turned on the polygraph and the tape recording, which told S whether the numbers would be in consecutive order or in random order. The numbers were presented at 15-second intervals. At the completion of one trial another card was drawn from a different range of numbers and the procedure was repeated.
The Guilty Person Group. Each S in this group was given taped instructions similar to those given
385 DETECTION OF DECEPTION
to the first group. Here, however, the S was informed that at times (probably on at least one trial) he would draw a blank rather than a number card and that his task on all trials was to try to convince E that he had drawn a blank rather than a number. When a blank was drawn (S was instructed), he was to write a zero on the slip of paper. Except for these instructions, each S was treated in the same manner as the first group.
The Ss in both groups were instructed that they were to say "no," as they heard each number, in such a way that their voice would not give them away. No information was given between trials to Ss concerning how successful they had been at deception. Each S was interviewed after the session to see how the S perceived the situation. It was especially important that the Ss in the guilty person group perceived their task as that of trying to make E think they had drawn a blank card and not another number .4
ANALYSIS OF THE DATA
The basic response data analyzed in this study were the skin resistance and changes in skin resistance as recorded by the GSR. As in the previous study by the authors, the primary concern in this analysis was not with the number of correct detections per se, but rather with the effect of the instructions and experimental procedures on the response to critical items relative to the response to other items.
Because of the differences in the dynamics of the RI and PT methods of presentation, different objective criteria of deception were required. In the RI trials, the magnitude of change in skin resistance was measured for each stimulus (number). The stimuli were then ranked in terms of average response magnitude, the one showing the greatest average change being ranked one, that showing the second greatest change being ranked two, etc. Once all stimuli in a trial had been ranked, the rank assigned to the critical item for that trial could be determined and used as the measure of relative responsiveness for purposes of analysis.
On PT trials, S anticipates the stimuli prior to their presentation, so that the prestimulus response level is affected by the stimulus. Thus magnitude of change in resistance from the prestimulus level is not an effective measure of responsivity. For this reason, the lowest level of skin resistance for each stimulus in a trial was taken as the criterion. The mean level for each stimulus (number) was calculated and a rank of one assigned to the stimulus producing the lowest level, etc. As in the other method, the rank assigned to the critical item in the trial was determined and used in the analysis.
Since comparisons of the relative effectiveness of RI and PT methods of stimulus presentation are intrasubject comparisons, the Wilcoxon matched-pairs signed-rank test was used here. For comparisons of the relative effectiveness of the guilty information and guilty person groups the Mann-Whitney U test was used.
As it was possible for S to draw a blank card in the guilty person paradigm on one or more trials, some Ss were not guilty on both a PT and RI trial. Therefore, it was necessary to discard such Ss in order to make intrasubject comparisons. Of the 29 Ss run, 24 were guilty on both RI and PT trials. Only these 24 are included in the analysis presented here.
RESULTS
As can be seen in Table 1, detection in the guilty person paradigm proved significantly superior to that in the guilty information paradigm (overall Mann-Whitney U = 147, n1 = n2 = 24).
386 LAWRENCE A. GUSTAFSON AND MARTIN T. ORNE
There was no significant difference in the effectiveness of the RI method of stimulus presentation between the guilty person paradigm and the guilty information paradigm. On the other hand, the PT method proved significantly less effective than the RI method in the guilty information paradigm, and significantly less effective in that paradigm than it was in the guilty person paradigm, both in the distribution of ranks assigned to the critical items and in the number of correct detections (the number of times a rank of one was assigned to a critical item). (There was no difference in the effectiveness of the PT and RI methods in the guilty person paradigm, both proving very effective.)
The use of nonparametric tests of significance, required by the nature of the data, makes it impossible to test directly for interaction effects. However, the data of Table 1 provide rather convincing evidence for an interaction between paradigms and methods of stimulus presentation as described in the preceding paragraph.
DISCUSSION
The results of this study and those of the preceding study of this series suggest several points in two major aspects of the detection of deception, the first dealing with the results of experimental studies in this area and the second dealing with the behavior of the deceiver.
In the present study, the RI method of stimulus presentation proved more effective in detection than the PT method when S was trying to deceive E as to which item of information he possessed (guilty information paradigm). The RI method and the guilty information paradigm is the combination most commonly used in laboratory studies of deception (Burtt, 1921; Ellson, Davis, Saltzman & Burke, 1952; Kubis, 1962; Lykken, 1960; Marston, 1917; Van Buskirk & Marcuse, 1954; as well as Gustafson & Orne, 1963). Thus it would appear that these studies have used the best method of stimulus presentation for the type of deception that they were studying.
However, laboratory investigations differ in a variety of ways from the field situation. One is "real life" and the other is an "experiment." Some have felt that this is an overwhelming obstacle. However, it is possible to investigate variables which will be working in both situations. The laws of behavior are not different in the two situations. Unfortunately, many of the laboratory studies do not take into account variables which are operating in a field situation and as a result use designs which are not generalizable to the field situation. This is likely to lead to false generalizations about the field situation, when in fact the problem is in the experimental design itself.
For example, the relative ineffectiveness of the PT method, as reported in this study, for the guilty information paradigm (the typical laboratory situation) would seem to make it a poor choice in the detection of deception. However, when the conditions more closely approximated the field situation, i.e., when the explicit task of S is to appear as though he did not have the information (guiltyperson paradigm), the PT method was no less effective than the RI method.
Two recent studies have used experimental situations similar to the guilty person paradigm of the present study. One of these, Kubis (1962), reports successful detections of the order of 97%o in one condition which used judgments of experienced operators. Lykken (1960) using only objective methods of classification reported 90% success in detection. Compared to the 70% success reported in laboratory studies using the guilty information paradigm (see, for example, Ellson et al., 1952), the results of these studies support the findings of the present study that the guilty person paradigm provides significantly greater ease of detection.
In general, then, it is necessary to urge caution in the comparison of laboratory studies of deception where different paradigms and methods may be used and still greater caution in the application of findings obtained in the laboratory to situations occurring in the field.
In order for deception to be detected, the responses of the deceiver to critical items must differ from his responses to noncritical items. The present study and its predecessor indicate two factors which affect this differentiation: the motivation of the deceiver and the tactics which he must use to deceive.
387 DETECTION OF DECEPTION
The differentiation of responses to critical and noncritical items can be heightened in two ways: increasing the magnitude of response to the critical items or reducing the amount of response to noncritical items (background activity). Increasing the motivation to deceive increased the response to critical items over that to noncritical items (Gustafson & Orne, 1963). Background activity was also increased; in other words, Ss showed increased responsiveness in general. This increase in general activity is probably due to the use in that study of the RI method of stimulus presentation which requires S to maintain a "set" for the critical item which can occur at any time.
In the PT method of presentation, where S can correctly anticipate the occurrence of the critical item, it would be expected that background activity would be reduced. The present study supports this expectation. There were significantly fewer responses overall during PT trials than during RI trials (Wilcoxon matched-pairs signed-ranks test: T = 112.5, n = 28, p. < .05).
As is suggested by the present study, the tactics of the deceiver vary with the paradigm and simultaneously with the means of stimulus presentation. (The interaction between paradigms and methods of presentation has been pointed out in the results of this study.) In the guilty information paradigm the deceiver can, and usually does, attempt to shift the attention of E to some noncritical item -- an attempt which is easiest when S is able to anticipate the occurrence of the items (PT method). It is this situation, as was found in this study, which makes deception most difficult to detect.
The results of the comparisons of the relative effectiveness of different paradigms and different methods suggest that it is easier for S to produce autonomic nervous system (ANS) responses (at least of small magnitude) to noncritical items than it is for S to inhibit responses to critical items, an observation commonly made at the clinical level. This is an important point for discussions of "control" of autonomic responses.
It would appear, then, that the "optimum" conditions for detection of deception would be found in a situation where S must prove that he is innocent (the guilty person model) where he is very highly motivated to deceive (heightened response to critical items), and where he "knows" exactly when he must deceive (decreased background activity with PT presentation).
Finally, the results of these studies indicate that the detection of deception is not a simple matter of asking a laboratory deceiver (or suspect) a few questions and recording his physiological response. It is an extremely complicated process which appears to be a function of a variety of psychological factors. Fortunately, however, these factors, as well as the physiological procedures themselves, are subject to experimental investigation.
REFERENCES
BURTT, H. E. The inspiration-expiration ratio during truth and falsehood. J. exp. Psychol., 1921, 4, 1-23.
ELLSON, D. G., DAVIS, R. C., SALTZMAN, I. J., & BURKE, C. J. A report on research on detection of deception. Contract Nonr 60nr-18011, 1952, Indiana University, Department of Psychology, Office of Naval Research.
GUSTAFSON, L. A., & ORNE, M. T. The effects of heightened motivation on the detection of deception. J. appl. Psychol., 1963, 47, 408-411.
INBAU, F. E., & REID, J. E. Lie detection and criminal investigation. (3rd ed.) Baltimore: Williams & Wilkins, 1953.
KUBIS, J. F. Studies in detection: Computer feasibility considerations. Technical Report No. 62-205; June 1962, Rome Air Development Center, New York; Contract AF 30(602)-2270, Project 5534; United States Air Force, Armed Services Technical Information Agency (now known as Defense Documentation Center), No. AD-284 902.
LEE, C. D. The instrumental detection of deception. Springfield, Ill.: Charles C Thomas, 1953.
LYKKEN, D. T. The GSR in the detection of guilt. J. appl. Psychol., 1959, 43, 385-388.
LYKKEN, D. T. The validity of the guilt knowledge technique: The effects of faking. J. appl. Psychol., 1960, 44, 258-262.
MARSTON, W. M. Systolic blood pressure symptoms of deception. J. exp. Psychol., 1917, 2, 117-163.
VAN BUSKIRK, D., & MARCUSE, F. L. The nature of errors in experimental lie detection. J. exp. Psychol., 1954, 47, 187-190.
WENGER, M. A., ENGEL, B. T., & CLEMENS, T. L. Studies of autonomic response patterns: Rationale and methods. Behav. Sci., 1957, 2, 216-221.
(Received September 20, 1963)