International Journal of Psychophysiology, 13(1992) 129-136

Back to Lie Detection Home Page 129 electrodiff.doc

INTPSY 00405

Electrodermal differentiation of deception: potentially confounding and influencing factors Alex Vincent and John J. Furedy Department of Psychology, University of Toronto, Toronto (Canada) (Accepted 18 June 1992)

Key words: Electrodermal response; SCR; VL; Differentiation-of-Deception; Retrieval-difficulty; Interference

The demonstration of deception as a psychological process requires a comparison of physiological responding to questions answered honestly or deceptively, under conditions which differ only with respect to deception. Such electrodermal (skin conductance response (SCR) differentiation was recently reported in the Differentiation-of-Deception Paradigm. The present study had two empirical goals: (a) to assess the possible confounding role of retrieval-difficulty and novelty in producing the differentiation effect; (b) to ascertain any influencing effects of a memorial ('cumulative' mental load) and two motivational (Monetary-Incentive and Ego-Involvement) factors on the electrodermal differentiation phenomenon and on overall responding. In addition to the basic Deceptive vs. Honest manipulation of the Differentiation-of-Deception Paradigm, the present study varied, within 60 subjects, Question Type (easily retrieved Autobiographical vs. more difficult-to-retrieve Biographical). The two two-level motivational factors were varied between subjects. Finally, to assess the confounding issue, voice latency (VL), known to be sensitive to retrieval-difficulty, was measured in addition to SCR. SCRs to deceptive answers exceeded those to the honestly- answered questions, demonstrating the differentiation phenomenon. Results showed that although VL and SCR was significantly greater to Biographical than to Autobiographical questions, the differentiation effect emerged only in the SCR and not in VL, which suggests that memorial difficulty does not confound the electrodermal differentiation effect.

ELECTRODERMAL DIFFERENTIATION OF procedures) controls for two potentially impor- DECEPTION: POTENTIALLY CONFOUND- tant sources of confounding: differential question ING AND INFLUENCING FACTORS significance and differential question frequency. However, a third source of potential confound Recently, experiments have attempted to di- within the paradigm is that deceptive answers rectly differentiate deception as a psychophysio- may be more difficult to retrieve from memory logical process (see, e.g., Furedy et al., 1988, than honest answers, and this difference may be 1991). These experiments utilized the Differenti- the cause of the reported SCR difference. The ation-of-Deception Paradigm which uses neutral obtained electrodermal (skin conductance re- autobiographical questions where half are an- sponse; SCR) Deceptive > Honest outcome may swered deceptively and half honestly. The DDP be due solely to this cognitive memorial factor (unlike commonly used detection-of-deception rather than to the deceptive vs. honest manipulation. Consequently, the Differentiation Paradigm may not be assessing the process of deception at all, but rather this retrieval difficulty difference. Correspondence to: J.J. Furedy, Department of Psychology, The Furedy et al. (1988) study dealt with this University of Toronto, Toronto, Ontario, Canada, M5S 1A1. retrieval-difficulty confound by providing proce- 130 dural modifications that minimized this memorial vity of the measure yielding the null result needs factor's influence by minimizing the retrieval to be independently demonstrated within the component of the task through the use of exten- same experiment. If the VL dependent variable sively rehearsed autobiographical questions. To proved to be sensitive to the Biographical/Auto- the extent that this method of controlling for the biographical manipulation, then any null (i.e., memorial, retrieval-difficulty factor is successful, Deceptive = Honest) VL outcome would be in- it is possible to infer, on the basis of an SCR terpretable. Deceptive > Honest outcome, that the process of Another, more subtle, confounding possibility deception per se has been differentiated. is differential novelty. Giving a deceptive re- The primary purpose of the present differenti- sponse to a question like, 'What is your name?' ation experiment was to provide more informa- involves greater novelty than giving the more tion on two potential confounds: memorial re- familiar honest answer, if only because the honest trieval difficulty and novelty. The problem with answer is more frequently given. The electroder- the minimization method adopted by Furedy et mal SCR is a well-known component of the ori- al. (1988) to deal with the memorial confound, is enting reaction (OR) which is enhanced by nov- that it is arguable that even under those mini- elty (see Sokolov, 1963). Consequently, any De- mized retrieval-difficulty conditions, deceptive ceptive > Honest result may not be due to decep- items are more difficult to retrieve from memory tion, but to the greater novelty of the deceptive than honest ones. In the present study, we sought answers. However, since this novelty difference to assess, rather than to simply minimize, the between deceptive and honest items applies solely operation of the retrieval-difficulty factor in the to Autobiographical answers, the novelty account differentiation experiment by directly manipula- predicts an interaction between the Autobio- ting retrieval difficulty, by using more difficult-to- graphical/ Biographical and Honest/ Deceptive retrieve biographical questions in addition to factors, with the Deceptive > Honest difference autobiographical ones. The success of this manip- being present solely within the Autobiographical ulation would have to be verified by a dependent condition. The design of this study allowed for a measure sensitive to this type of manipulation. test of this novelty-confound-derived prediction. Therefore, voice latency (VL) was added as a The second major purpose of the study was to response variable. VL is used as a measure of assess the influence of one memorial and two effort associated with retrieval difficulty and this motivational factors on the basic Differentiation- usage is supported by an extensive literature (e.g., of-Deception phenomenon. A factor would be Kahneman et al., 1969; Porges, 1972). VL was found to have an influencing effect on differenti- expected to increase with retrieval difficulty. Sup- ation if there emerged an interaction between the port for VL's sensitivity can be demonstrated by factor and the type of answer provided (Decep- different latencies to these two different types of tive vs. Honest). Due to the fact that the basic questions which vary on retrieval difficulty. Since Differentiation-of-Deception phenomenon has retrieval difficulty is presumed to not confound only recently been isolated (Furedy et al., 1988), the differentiation effect, there is no empirical the direction of influence on the differentiation basis for expecting a Deceptive > Honest result (SCR to Deceptive > SCR to Honest) effect, if in VL in the present paradigm. Accordingly, if any, of these factors cannot be predicted. How- the differentiation result were obtained with the ever, the direction of influence of the memorial SCR but not with VL, this could permit the and the two motivational factors on SCR itself is inference that the SCR Deceptive > Honest re- predictable from previous research. sult was not due to a retrieval-difficulty differ- The potential source of memorial influence is ence between the deceptive and honest re- 'cumulative' mental load, which can be assumed sponses. However, the inference rests on the ac- to increase as questions are presented during the ceptance of a null (VL) result, and in such cases, experiment. This increase in load would be caused as detailed elsewhere (Furedy, 1978), the sensiti- by interference among the various responses, 131 leading to an increase in retrieval difficulty. That recorded the signal from Beckman Ag/AgCl this source of influence seems to affect the elec- electrodes (1 cm in diameter), in conjunction with trodermal responding in the Differentiation Beckman NaCl electrode paste, attached to the Paradigm, is suggested by the finding that, in subject's volar surface distal phalanges of the both previous differentiation experiments (Fur- index and major digits of the right hand. VL was edy et al, 1988, 1991), SCR did not show the recorded concurrently through the use of a Sony customary habituation over trials. It is an empiri- high condenser microphone, attached to the sub- cal question, however, whether Deceptive > ject's collar, interfaced with a CP/M Horizon Honest electrodermal differentiation significantly computer. VL was measured to the nearest hun- decreases or increases over the course of the dredth of a second. interview; in the previous studies (Furedy et al., Prior to the attachment of electrodes, subjects 1988, 1991), such differentiation was seen to re- were taken to a room adjoining the one used for main constant over trials. the physiological and electronic recordings, where The first of the two types of motivational fac- they were provided with written instructions con- tors which were varied between subjects was Per- taining the motivational manipulation. formance Incentive, with a monetary reward be- A 2 X 2 between-subject factorial design was ing provided for successful physiological decep- used for the two types of motivational factors. tion. This factor has been commonly manipulated The first factor was Ego-Involvement instructions in the social psychophysiological literature (e.g., with two levels being induced, which was crossed Blumenthal et al., 1983; Tranel, 1983). The other with Performance-Incentive. Each cell in the re- two-level motivational factor was Ego-Involve- sulting matrix contained 15 subjects. Subjects in ment, with the subjects within the higher-level the high Performance Incentive condition were condition being asked to try to act like a guilty offered $5, $10 or $25 contingent upon their suspect. This manipulation is less common and success (low, moderate and high, respectively) at less standardized than the Performance Incentive physiological deception, indexed by the mean manipulation, so the instructions for the high-level electrodermal result through inspection. Greater Ego-Involvement condition have been repro- SCRs to Deceptive responses (21 subjects) was duced in Appendix A. As discussed previously considered to be low success, leading to a contin- the specific nature of any interaction induced gent reward of $5. No clear difference in SCR to through the influence of either of the motiva- Deceptive and Honest responses (seven subjects) tional factors cannot be predicted. However, was considered moderate success, leading to a higher motivation levels were expected to lead to $10 reward. Finally, greater SCRs to Honest re- greater overall electrodermal responding. sponses (two subjects) was considered to be high success leading to a $25 reward. Instructions for all subjects was to appear honest while answering METHOD all questions. Subjects in the high Ego-Involve- Subjects ment condition were provided with the above 60 volunteer subjects (32 female) were re- instructions along with additional statements re- cruited through advertisements posted through- questing them to try to act like a guilty suspect out the University of Toronto. Ages ranged from would during the experiment. Instructions for all 18-24 years. subjects as well as for the Ego-Involvement motivational manipulation can be found in Appendix Apparatus and procedure A. The same recording apparatus for measuring The question material presented, comprised SCR was used as detailed in Furedy et al. (1988, two different Question-Type lists crossed with pp. 684-685). Briefly, a Narco Model 4-A physio- two Response Types. These two factors were graph, running at a chart speed of 1 cm/s, cou- varied in a within subjects fashion. The first list pled with a Coulburn Instruments preamplifier contained 20 Autobiographical questions. These 132 questions were of the type, 'Tell me the city in RESULTS which you were born.' Subjects were instructed to provide deceptive responses to half of the ques- The electrodermal SCR was defined as any tions and honest responses to the remaining half. response that was initiated, by the appearance of These questions were presented by one of the an inflection point, within 1-5 s. following ques- authors (A.V.) and were rehearsed in the same tion onset. Magnitude, measured in mm and then order for all subjects. Responses were rehearsed converted to uS, was expressed as the difference until the subject felt confident of being able to between the inflection point and the apex follow- reproduce the responses during the interview. ing response onset contained within the 1-5 s. Once through the list was typically enough for latency window. The absence of an SCR was subjects to feel capable of accurately reproducing defined as a change of less than 0.5 mm on the the responses. recording chart and was scored as a zero re- The second list of 10 Biographical questions, sponse. To reduce between-subject variability due as well as the responses was provided by M.K. to individual differences in electrodermal respon- This was done to increase the saliency between sivity, scores were range corrected by dividing the the lists and to decrease the potential for inter- values by each subject's largest SCR. This trans- ference between the lists. This list was a subset of formation yielded SCR values ranging from zero the previous list, but the subject of the question to unity and were more normally distributed than was modified from the participant to the inter- the raw SCRs. Inferential statistics were calcu- viewer (e.g., 'Tell me the city in which I was lated for the untransformed SCR scores and the born'). The responses were provided to the sub- patterns of results were similar to those reported jects, along with a statement of whether the re- for the range-corrected scores except that, as was sponse was honest or deceptive. This second list expected, the between-subject effects were more was also rehearsed until the subject could provide clearly discerned in the transformed SCR values. the answers accurately. Once again once through Consequently, for the statistical tests reported the list was usually sufficient. Response accuracy only the range-corrected values will be discussed. was greater than 98%. The VL measure was the time from question Subsequent to electrode placement, subjects offset to the first utterance of a response, to the were given a 5-min adaption period, during which nearest hundredth of a second (1/100 s). The spontaneous SCR rate was measured. Subjects inferential statistics calculated were ANOVAs on were then asked, by M.K., the same 30 questions, range corrected SCRs and voice latencies with an at the rate of one every 30 s in a random order. a level of 0.05 set for statistical significance. All These orders were determined for each subject to repeated measures analyses were tested for sig- make it possible to generate six sequential five- nificance using the Geisser-Greenhouse correc- trial blocks such that, in each block, questions tion method. requiring deceptive and honest answers appeared at least once from either list. At the end of the question's presentation a timer was activated by the interviewer by depressing a microswitch, and 1 As a check on the effectiveness of the motivational deactivated by the subject's first audible utter- manipulations, before being debriefed and dismissed, all ance of the answer into the microphone. Follow- subjects were asked to rate 'how motivated' they were 'to deceive while avoiding detection both physiologically ing the presentation of the questions, there was a and behaviorally' on a 1-10 Likert scale measuring the second 5-min period, during which spontaneous amount of motivation. The anchors provided were not very SCR rate was again recorded. The frequencies of [1] and a great deal [10]. This self-report measure was the spontaneous SCRs was insensitive to the expected to order the four groups as follows: Performance Incentive + High Ego-Involvement (highest within-subject pre/post interview factor and to level), Performance Incentive or High Ego-Involvement the between-subjects motivational factors and will (moderate), no motivation (lowest level). No significant therefore not be commented on further *. differences among the groups emerged. 133

Fig. 1 (top plot) shows the basic electrodermal differentiation of deception effect over trial blocks. The sole significant effects were those due to the Response given (Deceptive > Honest), F(l,15) = 5.07, MS = 0.009 and Trials, with an increase in SCR occurring from Block 2 onward, F(4,275) = 3.51, MS = 0.048 (e = 0.89); the linear trend statistic was F(l,280) = 39.75. A similar analysis of the VL data revealed solely a significant Trials effect, F(4,200) = 3.65, MS = 0.045 (e = 0.85); the Response effect was not in the expected direction, with deceptive (x = 0.737) having a faster VL than honest (x = 0.739). Fig. 1 also shows (bottom plot) this VL effect along with the mean transformed SCR scores, and it is clear that both SCR and VL begin their almost parallel ascent at the second Trial block. As a descriptive statistical index of the correspon- Fig. 2. Mean SCR (range corrected) as a function of Perfor- dence between the two functions, the correlation mance Incentive and Ego-Involvement. of these mean scores was computed across the six blocks to be 0.996.

The next ANOVAs performed were 4-way split-plot designs with Performance Incentive (Money vs. No-money) and Ego-Involvement (Low vs. High) as between-subject factors and Question type (Biographical vs. Autobiographi- cal) and Response type (Honest vs. Deceptive) as within-subject factors. The ANOVA performed on the electrodermal SCR revealed a significant effect due to Ego-In- volvement, F(l,60) = 8.61, MS = 39.58. There was no effect discerned due to Performance Incen- tive, F(l,60) = 2.88. The Ego-Involvement effect may be seen in Fig. 2. The electrodermal data in Fig. 2 also suggest an interaction between the two type of motivational factors, but this interaction effect did not approach significance, E(l,60) = 2.67, P>0.05. Of the two previously assessed within-subject effects, that due to Question type was significant, F(l,60) = 4.70, MS = 0.036, with larger SCRs to Biographical than Autobiographical questions. The Deceptive > Honest effect which was reliable in the previous analysis (see Fig. 1 and text) emerged in this ANOVA as marginally significant, F(l,60) = 3.66, MS = 0.029, P<0.07. No other main effects or interactions were signifi- Fig. 1. Top panel: mean SCRs (range corrected) to Honest cant. In particular, the interaction between Ques- and Deceptive questions over six 5-trial Blocks. Bottom panel: tion Type and Response Type predicted by the mean SCRs and Voice Latencies over Trial Block novelty-confound account was not significant and the trends were contrary to the prediction, with the Deceptive > Honest difference in the Autobi- 134 ographical items (0.019) being less than in the trieval-difficulty effect induced by increased men- Biographical items (0.048). tal load. As to the interpretation of the direction The ANOVA performed on the VL data of the significant SCR difference, it is possible demonstrated only a single reliable and very that including biographical items resulted in sub- marked effect due to Question Type, F(l,44) = jects being primarily concerned about their ability 44.78, MS = 1.22. Consistent with the SCR re- to recall the appropriate responses to the ques- sults VL was longer for Biographical (x = 836 ms) tions and that this concern was handled by in- than for the Autobiographical (x = 674 ms) ques- structions that focused on the deception rather tions. than on the memorial aspect of the study. The motivational factors failed to interact with the Deceptive > Honest effect, but there was one DISCUSSION significant effect on SCR (see Fig. 2), namely that, contrary to expectations, instructions to act The results were clear with regard to the pri- the part of the guilty suspect (Ego-Involvement 2) mary purpose of the study, assessment of the lowered SCR magnitude. Such a counterintuitive memorial-retrieval-difficulty and novelty con- result from this motivational factor suggests that founds in the Differentiation-of-Deception phe- experimenters need to reexamine these sorts of nomenon. The VL dependent variable's sensitiv- manipulations. In this study almost all subjects ity to retrieval difficulty was evident through the indicated, during an informal post-experimental highly significant Question-Type effect (Bio- interview, that they were maximally motivated to graphical > Autobiographical), but only the elec- 'beat' both machine and interviewer. In other trodermal SCR demonstrated the Deceptive > studies monetary incentives had no effect on per- Honest effect. In the case of VL the trend was formance and the no-incentive group was as- opposite to that predicted from the memorial-re- sessed to be at ceiling levels for motivation (e.g., trieval-difficulty account, with latencies being Beijk, 1980). It may be worthwhile to consider shorter for deceptive responses. Consequently, manipulating motivation by attempting to de- the results suggest that cognitive retrieval diffi- crease rather than to increase its level, which is a culty does not account for the electrodermal De- strategy that has rarely been reported in the ceptive > Honest result obtained here and previ- literature. However, it is interesting to note that, ously (Furedy et al., 1988, 1991). Retrieval-diffi- in terms of sheer reactive sensitivity to motiva- culty is not a confound in the Differentiation tional manipulations, i.e., the degree to which a Paradigm. Similarly, in the SCR not only did the measure produces significant differences (for de- Question-Type X Answer-Type interaction pre- tails, see Furedy and Heslegrave, 1984), the SCR dicted by the novelty account fail to emerge, the was superior both to the introspective self-report trend was in the opposite direction. Therefore, measure (intended as a manipulation check) and novelty does not account for the Deceptive > to the cognitively-based behavioral VL measure. Honest SCR result. Although comparisons between experiments The second purpose of the study was to ana- need to be interpreted with particular caution, it lyze the influencing effects of memorial mental is still relevant to note that, compared to the load and two motivational factors. There was previous Differentiation-of-Deception studies clear evidence that mental load was influencing (Furedy et al., 1988, 1991), in the present study both SCR which significantly increased over trials the electrodermal Deceptive > Honest effect in a way that is opposite to the usual habituation observed in this response, and on VL (see Fig. 1). In addition, the striking similarity in the functions 2 The high Ego-Involvement group did not receive maximally for the two measures (see Fig. 1, bottom), as well involving instructions of the type: ... high intelligence people as the fact that both began to increase only at the can beat the lie detector. This kind of instruction was consid - second trial block, suggests a 'cumulative' re- ered to be not ethically permissible. 135

TABLE I this emotional interpretation is that questions of Differentiation (SCR deceptive > SCR honest) effect strength greater personal relevance to subjects should pro- duce greater Differentiation than those with less Reported study personal information for each subject. For an Present study 1988 1991 accurate assessment of individual personal in- n 60 32 32 volvement, it would be desirable to obtain subjec- Effect tive ratings of such involvement. size (%) 30.8 34.7 36.1 F value 5.07 10.76 10.12 APPENDIX A emerged with a smaller percentage of variance accounted for (see Table I). The sole difference Subjects in all motivational conditions received between the present and previous studies is the the following statement: addition of 10 Biographical questions. The Auto- "The task in this experiment is to successfully biographical questions and the sample used were deceive the interviewer. When you are asked to the same. Consequently, these between-experi- answer the questions dishonestly you must do so ment comparisons suggest that increasing mental while appearing honest. We would request that load through the addition of Biographical items you appear honest at all times when answering reduces the Differentiation effect. From the per- the questions posed. The interviewer will attempt spective of maximizing the Differentiation-of-De- to detect deception through your behaviour. ception effect it appears that solely Autobio- Please breathe normally throughout the inter- graphical questions should be presented. view". In conclusion, the Differentiation-of-Decep- Additional instructions for the high Ego-Involve- tion effect is not caused by retrieval-difficulty ment group were the following: differences between Deceptive and Honest re- "As you can well imagine, guilty suspects are sponses. In fact, latency to Deceptive responses extremely motivated to lie and to avoid being was found to be slightly shorter than to Honest caught. Following the conclusion of a polygraph ones. In addition, VL was found to be sensitive to test a guilty person may be released because retrieval-difficulty differences manipulated he/she was able to control his or her emotional through questions varying on the retrieval-diffi- responses during questioning. The converse is culty dimension. The expected electrodermal Dif- that an innocent suspect may be convicted be- ferentiation-of-Deception effect was replicated, cause that person failed the polygraph. In the but VL was found to be insensitive to this differ- present study we would like for you to behave as ence. Yet VL was highly sensitive to a retrieval- the guilty suspect; you will be asked to lie while difficulty difference between Autobiographical being extremely motivated to avoid being de- and Biographical questions. tected. It is essential for this study that all of our Consequently, these results suggest that since participants be as motivated as possible to avoid the memorial retrieval-difficulty or novelty fac- being detected when they are instructed to an- tors do not contribute to the reported Differenti- swer questions dishonestly. Your highly moti- ation effect (SCR to Honest > SCR to Deceptive), vated participation could help in the understand- an emotional factor may be inferred. This emo- ing of the psychological processes that underlie tional factor does not seem to be influenced by detection of deception." monetary incentive or ego-involvement. It might be hypothesized that this emotional factor is based on guilt associated with deception. It is of interest ACKNOWLEDGEMENTS that this sort of emotional effect emerged in the present and previously reported studies with such This research was supported by grants to J.J.F. relatively neutral items. A prediction based on by the National Sciences and Engineering Re- 136 search Council (NSERC) of Canada. We are in- Furedy, J.J. (1986) Lie detection as psychophysiological differ- debted to Magnus Kristjansson (M.K.) for aiding entiation: some fine lines. In M. Coles, E. Donchin and S.W. Porges (Eds.), Psychophysiology: Systems, processes in the running of this study as described in the and applications-a handbook, pp. 683-701, Guilford Press, Methods section, and to Gershon Ben-Shakhar New York. and an anonymous reviewer for comments on an Furedy, J.J. and Heslegrave, R.J. (1984) The relative sensitivi- earlier version of the paper. ties of heart-rate and T-wave amplitude to stress: Com- ments on and some alternative interpretations of Penzien et al.'s results. Biol Psychol, 19; 55-61. Furedy, J.J., Davis, C. and Gurevich, M. (1988) Differentia- REFERENCES tion of deception as a psychological process: a psychophysiological approach. Psychophysiology 25; 683-688. Beijk, J. (1980) Experimental and procedural influences on Furedy, J.J., Posner, R.T. and Vincent. A. (1991) Electroder- differential electrodermal activity. Psychophysiology, 17; mal differentiation of deception: perceived accuracy and 274-278. perceived memorial content manipulations. Int. J. Psy- Ben-Shakhar, G. and Furedy, J.J. (1990) Theories and Applica- chophysioi, 11; 91-97. tions in the Detection of Deception, Springer Verlag, New Kahneman, D., Tursky, B., Shapiro, D. and Crider, A. (1969) York. Pupillary, heart rate and skin resistance changes during a Blumenthal, J.A., Lane, J.D., Williams, R.B., McKee, D.C., mental task. / Exp. Psychol., 79; 164-167. Haney, T.H. and White, A. (1983) Effect of task incentive Porges, S.W. (1972) Heart rate variability and deceleration as on cardiovascular response in Type A and Type B individ- indexes of reaction time. /. Exp. Psychology, 92; 103-110. uals. Psychophysiology, 20; 63-70. Sokolov, E.N. (1963) Perception and the conditioned reflex, Furedy, J.J. (1978) Negative Results: Abolish the name but Macmillan, New York. honour the same. In J.P. Sutcliffe (Ed.), Conceptual analy- Tranel, D.T. (1983) The effects of monetary incentive and sis and methods in psychology, pp. 169-180, Sydney Uni- frustration on heart rate and electrodermal activity. Psy- versity Press, Sydney. chophysiology, 20; 652-657.