Psychophysiological measures of deception detection are based on the expectation that lying may cause guilt and fear, which are, in turn, reflected in autonomic nervous system (ANS) responses. However, measures such as polygraphs have been challenged for decades because of limitations such as susceptibility to countermeasures (National Research Council, 2003), the habituation effect (Elaad & Ben-Shakhar, 1997; Verschuere, Crombez, De Clercq, & Koster, 2004), and an individual’s level of morality and ethics, with each one producing experimental variability. Hence, there is a need for alternative methods of deception detection (Mertens & Allen, 2008).

Researchers have recently begun looking at the cognitive component of lying. Leal, Vrij, Fisher, and van Hooff (2008) argued that lying is cognitively more demanding than telling the truth. As deception is involved in various aspects of cognitive activations, such as conflict control and response inhibition (Nuñez, Casey, Egner, Hare, & Hirsch, 2005), the cognitive subcomponents involved in deception have also become interesting topics in cognitive neuroscience (Abe et al., 2008). Researchers using functional magnetic resonance imaging (fMRI) to investigate deception have found more brain activation when an individual is lying compared to when they tell the truth (Langleben et al., 2002; Spence et al., 2001).

Event-related electroencephalogram (EEG) alpha desynchronization (ERD) is a reliable measure that reflects cognitive demand (Fink, Grabner, Neuper, & Neubauer, 2005). The ERD method is also sensitive in indexing different task requirements during reasoning (Neubauer & Fink, 2003) and working memory processing (Stipacek, Grabner, Neuper, Fink, & Neubauer, 2003). However, to our knowledge, no researcher has investigated whether alpha ERD can differentiate between the cognitive load induced by lying versus that induced by telling the truth.

We hypothesized that ERD would reflect the cognitive activation generated by deception, because EEG alpha power is known to desynchronize when individuals are mentally active. In addition, the relationship between EEG alpha power and several indices of cognitive performance has been studied extensively (Fink et al., 2005) and in studies on cognitive processes an increase in cognitive control during deception has been demonstrated. Therefore, we believe that an alpha ERD increase will reflect the higher cognitive activation of deception.

However, as cognitive activity while lying could differ according to the type of deception, researchers have begun discriminating between the various types. They have divided deception into various subtypes so that specific cognitive actions involved with each subtype can be observed more clearly. Many factors have been examined including lies in which the contents are created or predetermined and memorized; isolated lies (in content) and lies that fit a scenario (Ganis, Kosslyn, Stose, Thompson, & Yurgelun-Todd, 2003); lies that are spontaneously told and instructed lies (Seth, Iversen, & Edelman, 2006); and deception for events that have been experienced or not experienced (Abe et al., 2006).

In line with this previous research, we thought there could be a difference in ERD in the deception types owing to the different levels of cognitive activation attributable to each subtype. Therefore, we created four subtypes of deception by orthogonally crossing the two dimensions from the Ganis et al. (2003) and Seth et al. (2006) studies. The first dimension was defined by Seth et al.’s spontaneously chosen lie and instructed lie. The second dimension was defined by Ganis et al.’s spontaneously created lie and memorized scenario lie. Decision-making processes in the first dimension are more likely to be required in the spontaneous condition than in the instructed condition, in which participants respond as instructed. The second dimension, crossed with the instructed-spontaneous (IS) factor, was labeled memorized-created (MC). In the memorized condition, participants are asked to lie according to a memorized scenario, requiring mentally demanding memory retrieval, while in the created lies condition participants are required to create the contents of the deception.

Four types of lies were thus created: an instructed memorized (IM) lie, an instructed created (IC) lie, a spontaneous memorized (SM) lie, and a spontaneous created (SC) lie. It was expected that different patterns of emotional and cognitive responses, involving each type of deception, would be generated. The ERD measure, especially in the alpha range, has been found to correlate with perceived task complexity (Knyazev, Savostyanov, & Levin, 2006). Therefore, the perceived difficulty of each type of deception and truth may be reflected by an alpha ERD enhancement. We also examined skin conductance response (SCR), identified as the best peripheral measure for deception detection (Vandenbosch, Verschuere, Crombez, & De Clercq, 2009), to evaluate emotional responses elicited by the four types of deception.

In summary, because a decrease in alpha ERD reflects mental activation within the memory system, it was expected that the differing cognitive load of each type of deception and truthful response would be reflected in alpha band desynchronization. We hypothesized that, first, telling any type of lie will produce a greater increase in ERD than will telling the truth: second, created lies will involve response selection, which will be reflected in alpha ERD; and third, differences in the responses produced by the instructed and spontaneous factor will help in any examination of the difference between participant-derived lies and those ordered by others. We expected SCR to be more sensitive for detecting spontaneous (memorized, created) than instructed lies, but minimally effective in detecting memorized lies (instructed, spontaneous) owing to habituation.

Method

Participants

Participants were 29 undergraduate students (15 male, 14 female) with a mean age of 22.04 (SD = 3.21); they were paid US$15 for their participation.

Pretesting Procedure and Stimuli

Participants completed a questionnaire and consent form. Two participants took part in a mock crime experiment at the same time. One was instructed to play the role of an innocent person, while the other was instructed to play the role of a guilty person. Innocent participants were eyewitnesses to a crime that the guilty participants committed. They would be suspects because they were at the crime scene and therefore needed to argue their innocence. Guilty participants were instructed to claim innocence and each blame the other participant. We aimed to compare the EEG patterns elicited by truthful and deceptive responses of the guilty participants. Innocent participants were excluded from the SCR and ERD recording phase because they had no need to lie.

Guilty participants committed a mock crime in a 3D virtual environment using the computer game Grand Theft Auto (^©Rockstar Games). Participants were then asked for information about the seven main facts of the crime. They either lied or told the truth while being recorded on an EEG. Afterwards, an instructor had participants rehearse and memorize an alternative false scenario (memorized lie), while helping them restrict the scenario to be consistent with the seven facts. After a practice session, participants moved to a different room for EEG and SCR measurements.

Experimental Task

There were five experimental conditions. First, in the truth condition, participants told the truth in response to the seven questions. Second, in the memorized lie condition (IM lie), participants had to lie according to the false scenario they had memorized. Third, in the created lie condition (IC lie), participants lied without trying to be consistent with the false scenario, and were told not to devise a coherent story. In the final two conditions, participants chose whether to lie or not so that in the fourth condition, participants chose between the memorized lie (SM lie) and truth, whereas in the fifth condition, they chose between a created lie (SC lie) and truth.

Questions were each repeated three times (21 trials) in conditions 1 (truth), and 2 (IM lie), and 3 (IC lie). The same seven questions were each repeated seven times (49 trials) in conditions 4 (SM lie) and 5 (SC lie) in a random order. The same questions and conditions were not in any two successive presentations.

Electrophysiological Recording

EEG data were acquired by a Laxtha EEG monitoring device. Eight silver electrodes were placed on each participant’s scalp using electrode glue and the 10/20 system of electrode placement at F3, F4, F7, F8, P3, P4, O1, and O2. They were referenced to linked mastoids. Amplifier output was passed through a 12-bit A/D converter sampling at 256 Hz. All impedances were maintained at 5 kΩ, bandpass was 0.05-50.00 Hz, and a 60 Hz notch filter was applied. Trials in which EEG readings exceeded ±100μV were rejected as artifact contamination. An electrooculogram (EOG) was also recorded using two 8 mm diameter Ag/ AgCl electrodes placed on the vertical midline of the right eye above the eyebrow and approximately 1 cm below the lower lid. EOG data were used as a visual reference for artifact rejection. Two 6 mm diameter Ag/AgCl skin conductance electrodes were attached to the palmar sides of the second and fourth fingers of the right hand with Velcro straps. The same device used for recording EEG was used.

Data Analysis

The alpha band range was 7.5-12.5 Hz for analyzing ERD. All trials started with a 4000 ms fixation cross, followed by an auditory warning. In each trial, a 2000-4000 ms time period served as reference interval (R), and the time period between the question onset and participants’ response served as activation interval (A) when calculating ERD. Only correct answers were included in the ERD analyses. The percentage of decrease in alpha power (mV²) was calculated using the formula: %ERD = ([R – A]/R) × 100.

The maximum increase in conductance obtained from 1000-5000 ms after stimulus onset was used to measure SCR. To eliminate individual differences in responsiveness and to permit a meaningful summation of the responses of individual participants, SCRs were transformed into within-subject standard scores. The z scores were computed relative to the mean and standard deviation of the participant’s response distribution for each of the seven questions. Within-question z scores have proved to be more resistant to habituation effects and therefore more efficient (Elaad & Ben-Shakhar, 1997). However, the z score is still only a supplement that improves, but cannot change, the basic nature of SCR.

Results

In the spontaneous condition, from a total of 49 trials, the average frequency of telling the truth was 20 and lying was 29. One participant made only five truthful responses and lied 44 times, and was excluded from the analysis. All other participants made at least 10 truthful and 10 deceptive responses.

Skin Conductance z score Analysis

The IS × deception (truth, lie) analysis of variance (ANOVA) revealed a significant main effect for deception. SCR responses were greater when lying than telling the truth F(1, 28) = 6.69, p < .05.

In the instructed condition, a repeated one-way ANOVA revealed no significant difference between the instructed truth (I truth) and IM lie conditions. However, the repeated-measures ANOVA, in which the I truth condition was compared with the IC lie condition, revealed a significant difference in SCR z scores, indicating that SCR responses were greater when telling a created lie than when telling the truth (see Figure 1).

In the spontaneous condition, a spontaneous memorized (SM) deception (truth, lie) ANOVA yielded a significant main effect for MC, indicating that higher SCR responses were produced in conditions that allowed created lies compared to conditions with only memorized lies or the truth. Follow-up analysis indicated that there was no difference in SCR response between the S truth and the SM lie. However, there was a significant difference between the S truth and the SC lie F(1, 28) = 5.24, p < .05.

Although SCR was found to be a reliable measure for differentiating between created lies and truth, it revealed a weakness in differentiating truth from memorized lies in both the instructed and spontaneous conditions.

Alpha ERD Difference between Deception and Truth and the Effect of the IS Condition

A repeated measures ANOVA on the alpha ERD data with the factors of IS condition, deception (truth, lie), and electrode (F3, F4, F7, F8, P3, P4, O1, O2), yielded significant main effects for the electrode F(7, 196) = 19.89, p < .001 and deception factors F(1, 28) = 7.76, p < .01. As expected, the main effect for deception led to a greater increase in alpha ERD when lying than when being truthful. This indicated that lying is more cognitively demanding, with increased alpha desynchronization reflecting greater cortical activation.

Instructed condition. To examine differences in cortical activation between I truth and IM-IC lies, three one-way ANOVAs were used to compare the effects for I truth versus IM lie, I truth versus IC lie, and IC lie versus IM lie on ERD (see Figure 2). The ANOVAs revealed significant differences between truth and IM lie F(1, 28) = 13.22, p < .01, and between truth and IC lie F(1, 28) = 18.15, p < .01. Although IC lie produced higher cortical activation than IM lie, it was not statistically significant.

Spontaneous condition. A three-way repeated measures ANOVA was performed with the factors of MC deception, deception (truth, lie), and electrode (F3, F4, F7, F8, P3, P4, O1, O2). A significant main effect for the electrode factor F(7, 196) = 16.26, p < .01 was found, but the difference between the deception (SM lie combined with SC lie) and truth (SM truth, SC truth) factors was not significant. There was also a significant MC × deception interaction F(1, 28) = 4.33, p < .05. As depicted in Figure 2, although an SM lie produced a greater increase in ERD than did an S truth, alpha ERD tended to decrease for an SC lie.

To discover whether the process of choosing to lie or tell the truth resulted in increased cortical activation, three one-way ANOVAs were conducted, using the factors of IS dimension in the truth condition, indicated by I truth, and the S truth condition. By comparing instructed and chosen truths, we sought to confirm the effect of the decision-making process because only the spontaneous truth option would involve the cognitive activity generated by choosing. We found that I truth induced significantly weaker cortical activation than either of the spontaneous truths. The results for spontaneous memorized (SM) truth was F(7, 196) = 13.02, p < .01 and for spontaneous created (SC) truth F(7, 196) = 13.47, p < .01. These results imply that spontaneously choosing whether or not to lie could be more cognitively taxing than telling the truth or lying as instructed.

The lack of a significant difference in ERD between SM truth and SC truth, and spontaneously chosen truth against SM lie and SC lie, confirmed that an IS condition effect occurred, rather than a task-specific difference.

Interaction Effect of two Orthogonal Deception Dimensions Reflected in Alpha Band Desynchronization

A three-way repeated measures ANOVA with the factors of IS, MC, and electrode, was used to investigate how two deception dimensions worked together to produce only lie responses.

A significant main effect for the electrode factor was observed. Although a main effect was not observed for the IS and MC factors, there was an interaction between them F(1, 28) = 4.58, p < .05 (see Figure 3). This interaction effect showed that although an SM lie yielded greater ERD than an IM lie, the reverse pattern was observed when creating the content of the lie.

Discussion

To examine whether the EEG alpha frequency band could assist us to understand the neural correlates of deception, we created four types of lies by orthogonally crossing two dimensions. Before testing alpha ERD, SCR was analyzed and found to be significantly higher during deception than when telling the truth. However, when analyzed under four subtypes, SCRs for truth and deception were different only for truth and created lie conditions. There were no significant differences between truth and memorized lie conditions. This result reflects the key weakness of any physiological measure that is affected by the habituation effect (Elaad, 2008) and it should be pointed out again that the application of a z score, does not safeguard the measure from habituation. It is simply a better option. In this study, participants were prepared to lie about a prepared and practiced scenario that may have caused habituation. However, SCR differentiated between the truth and created lies in both the instructed and spontaneous conditions, demonstrating its reliability.

Alpha band desynchronization was tested to discriminate between deception and the truth. A major finding was that in the instructed conditions, ERD could be used to discriminate truth from both memorized and created lies. ERD results were thus different from those of SCR, because ERD was not affected by habituation. Also, because ERD was sensitive to memory load (Krause et al., 2000), it may have successfully discriminated between I truth and IM lies. The second major finding occurred in the spontaneous condition, in which ERD was significantly higher for SM lies than for truth. However, SC lies could not be discriminated from truth, and although not statistically significant, the ERD value was actually smaller. ERD thus differentiated between truth and IM lies, IC lies, and SM lies, but not SC lies.

When the effects of the different types of lies were compared there were no significant main effects of the MS and IS dimensions. However, the interaction between these two factors was significant. When further simple main effect analysis was conducted, lies tended to produce larger ERD in the spontaneous lie condition, of which SM lies tended to show larger ERD than SC lies, which yielded the lowest ERD values. The SC lie was the only subtype that ERD did not differentiate from truth.

This result could be considered counterintuitive because SC lies seemed to contain many cognitive components, such as decision making. There are several possible explanations. First, SC lies may have been easier than other types of lies, because they were chosen lies that did not need to fit a scenario. Therefore, participants may have told SC lies as they came to mind or when it was easier to do so. The fact that any response was admissible may have allowed participants to easily generate their lies.

Second, as predicted, SC lies may have required the highest cognitive activity among the subtypes, which would have resulted in increased alpha power. SC lies may have cognitively been the hardest task, because the spontaneous factor itself showed higher ERD than the instructed factor did in the truthful condition, which is not affected by the cognitive activity generated by lying. Also, the created deceptive response factor was higher when comparing IM lies and IC lies. The result of decreasing alpha ERD in SC lies may have come from the interaction of all the cognitive activity involved with various factors such as IS, MC, and truth- deception. However, further research is needed to find an exact causal factor.

In summary, ERD was found to be a sensitive measure for detecting deception-requiring memory load. Executive functioning, an important cognitive process for generating deception, was also reflected in ERD.

Our results have important implications. When discriminating between different types of deception, we found that it is important to consider the type of deception, as each type has unique cognitive and emotional aspects. That ERD can be used to differentiate between truth and instructed memorized lies, instructed created lies, and spontaneously memorized lies, and that SCR can be used to discriminate between spontaneously created lies and truth indicates that the combination of SCR and ERD measures has the potential to become a powerful tool for deception detection.

Abe, N., Okuda, J., Suzuki, M., Sasaki, H., Matsuda, T., Mori, E., ... Fujii, T. (2008). Neural correlates of true memory, false memory, and deception. Cerebral Cortex, 18, 2811-2819. http://doi.org/hc3

Abe, N., Suzuki, M., Tsukiura, T., Mori, E., Yamaguchi, K., Itoh, M., & Fujii, T. (2006). Dissociable roles of prefrontal and anterior cingulate cortices in deception. Cerebral Cortex, 16, 192-199. http://doi.org/hc4

Elaad, E. (2008). Effect of context and state of guilt on the detection of concealed crime information. International Journal of Psychophysiology, 71, 225-234. http://doi.org/hc6

Elaad, E., & Ben-Shakhar, G. (1997). Effects of item repetitions and variations on the efficiency of the guilty knowledge test. Psychophysiology, 34, 587-596. http://doi.org/hc7

Fink, A., Grabner, R. H., Neuper, C., & Neubauer, A. C. (2005). EEG alpha band dissociation with increasing task demands. Cognitive Brain Research, 24, 252-259. http://doi.org/hc8

Ganis, G., Kosslyn, S. M., Stose, S., Thompson, W. L., & Yurgelun-Todd, D. A. (2003). Neural correlates of different types of deception: An fMRI investigation. Cerebral Cortex, 13, 830-836. http://doi.org/hc9

Knyazev, G. G., Savostyanov, A. N., & Levin, E. A. (2006). Alpha synchronization and anxiety: Implications for inhibition vs. alertness hypotheses. International Journal of Psychophysiology, 59, 151-158. http://doi.org/hdc

Krause, C. M., Sillanmäki, L., Koivisto, M., Saarela, C., Häggqvist, A., Laine, M., & Hämäläinen, H. (2000). The effects of memory load on event-related EEG desynchronization and synchro- nization. Clinical Neurophysiology, 111, 2071-2078. http://doi.org/hdd

Langleben, D., Schroeder, L., Maldjian, J., Gur, R., McDonald, R., Ragland, J., ... Childress, A. (2002). Brain activity during stimulated deception: An event-related functional magnetic resonance study. NeuroImage, 15, 727-732. http://doi.org/hdf

Leal, S., Vrij, A., Fisher, R. P., & van Hooff, H. (2008). The time of the crime: Cognitively induced tonic arousal suppression when lying in a free recall context. Acta Psychologica, 129, 1-7. http://doi.org/hdg

Mertens, R., & Allen, J. J. B. (2008). The role of psychophysiology in forensic assessments: Deception detection, ERPs, and virtual reality mock crime scenarios. Psychophysiology, 45, 286-298. http://doi.org/hdh

National Research Council. (2003). The polygraph and lie detection. Washington, DC: National Academies Press.

Neubauer, A. C., & Fink, A. (2003). Fluid intelligence and neural efficiency: Effects of task complexity and sex. Personality and Individual Differences, 35, 811-827. http://doi.org/hdj

Nuñez, J. M., Casey, B. J., Egner, T., Hare, T., & Hirsch, J. (2005). Intentional false responding shares neural substrates with response conflict and cognitive control. NeuroImage, 25, 267-277. http://doi.org/hdk

Seth, A. K., Iversen, J. R., & Edelman, G. M. (2006). Single-trial discrimination of truthful from deceptive responses during a game of financial risk using alpha-band MEG signals. NeuroImage, 32, 465-476. http://doi.org/hdm

Spence, S. A., Farrow, T. F. D., Hereford, A. E., Wilkinson, I. D., Zheng, Y., & Woodruff, P. W. R. (2001). Behavioural and functional anatomical correlates of deception in humans. Neuroreport, 12, 2849-2853. http://doi.org/hdn

Stipacek, A., Grabner, R. H., Neuper, C., Fink, A., & Neubauer, A. C. (2003). Sensitivity of human EEG alpha band desynchronization to different working memory components and increasing levels of memory load. Neuroscience Letters, 353, 193-196. http://doi.org/hdp

Vandenbosch, K., Verschuere, B., Crombez, G., & De Clercq, A. (2009). The validity of finger pulse line length for the detection of concealed information. International Journal of Psychophysiology, 71, 118-123. http://doi.org/bdpq6t

Verschuere, B., Crombez, G., De Clercq, A., & Koster, E. H. W. (2004). Autonomic and behavioral responding to concealed information: Differentiating orienting and defensive responses. Psychophysiology, 41, 461-466. http://doi.org/hdr

This work was supported by the National Research Foundation of Korea grant funded by the government of Korea (MEST) (No. 2011-0027731).

Jang-Han Lee, Department of Psychology, Chung-Ang University, 221 Heukseok-dong, Dongjak-gu, Seoul 156-756, Republic of Korea. Email: [email protected]