Study 1 – Lying in Front of Peers

Participants.

For this study, 34 University students were recruited. Participants with invalid data (e.g., incomplete/unclear video footages) were excluded. Valid data from 32 participants (17 males, 15 females, Mage = 22.86 years, SD = 3.89) were included in the analyses. To participate, students

were required to have normal, or corrected to normal vision and hearing, as well as normal ability of body movement and communication.

Apparatus.

The experiment, conducted by three people, took place in a university lecture room. An envelope for each participant contained an instruction sheet, an evaluation sheet, and a token. The token was a small card with one of nine possible combinations of its features based on three colors (e.g., red, blue or yellow) and three shapes (e.g., heart, square or triangle). A covariate battery including a short self-rating manipulation check questionnaire assessing levels of nervousness, perception of levels of task difficulty and motivation etc., a demographic information completion sheet (e.g., age, gender and nationality), and scales assessing the individual factors specified in Section 1.3 were included. Two high-definition video cameras were used to record behaviours. Each participant received a confectionery reward for completing the study.

Design and procedure.

In this study the independent variables of deception and degree of probability of detection were manipulated in a 2 (veracity: deceptive vs. honest) × 2 (probability level: higher vs. lower) between-subjects design and participants were randomly assigned to the four groups. Veracity

was manipulated by instructing participants to deceive or tell truth about the color and shape of the token they received. Probability levels were manipulated by whether or not asking peers of the participants to evaluate whether the participants were being deceptive about the token they had. Upon receiving an envelope containing task instructions and a token, participants were instructed not to expose the token during the experiment. Each participant stood at the front of the classroom and gave a short presentation to the audience, including their student number, name, a curious fact about themselves (e.g., “I run five miles every day”) and a description of their token containing its color and shape (e.g., “I have a blue triangle”). The content of the presentation, as related to participants themselves, was designed with the purpose of enhancing motivation to perform well by introducing self identity-related tasks (DePaulo et al., 2003).

The group ‘deceiver & evaluated’ (DE) lied about the token and were evaluated by their peers. Then the group ‘truth-teller & evaluated’ (TE) performed the task and did not lie about the token but were also evaluated by their peers. Participants in these two groups were told that the overall group ranking was to be announced later to the class members; however, this was designed purely as a mechanism to manipulate the level of probability involved in the evaluated presentations. Groups of ‘deceiver & non-evaluated’ (DN) and ‘truth-teller & non-evaluated’ (TN) participated similarly as DE and TE but were not evaluated by their peers. After their presentations, participants completed the questionnaire pack together with manipulation check questions and then received their confectionery reward for their participation. Behavioural data were recorded using video cameras and transferred into numerical data through the coding processes.

The previous research concerning behavioural indicators of deception has usually performed the analysis of variance (ANOVA) on each behavioural unit (e.g., Lancaster et al., 2013; Mann et al., 2012; Van Swol et al., 2012, etc.), so as to assess the effect of independent variables on each body movement. ANOVA is widely used given its proven ability of comparing means from different groups (e.g., high and low level of the probability, and deceive or honest conditions) simutanously, which is a favourable feature for comparative experimental studies (Field, 2005). In line with this, a series of analyses of covariance (ANCOVA), which was built on the basis of ANOVA and additionally controls the effect of convariates, were conducted within each of the four studies for analysing body movement and impression variables.

Bonferroni corrections were employed to reduce the chance of Type I errors (Field, 2005) and demographic information, including age, gender and nationality (coded into western/non- western), as well as individual variables (see reliability of scales in Table IV-4) including personality, public self-cousciousness, ability to control behaviour, and self-monitoring, were assessed as covariates. The ANCOVA tests run across all four studies indicate that the covariates of the individual factors did not consistently influence the results. Thus, only the demographic characteristics were retained in the final data analyses as a consistent set of covariates across the four studies.

Table IV-4:

Reliability for the scales

Cronbach’s alpha

Study 1 Study 2 Study 3 Study 4

Big Five Scale/ BFI Extraversion Agreeableness Conscientiousness Neuroticism Openness

Public Self-consciousness Scale Self-monitoring Scale

Ability to Control Behaviour Scale

.85 .84 .85 .82 .42 .76 .69 .67 .86 .80 .85 .83 .87 .73 .67 .78 .84 .71 .78 .79 .71 .75 .56 .69 .83 .77 .75 .84 .60 .59 .63 .76

Note. Big Five Scale was used in Study 1 & 2, whereas BFI was used in Study 3 & 4. !

In this study, ANCOVAs were conducted for all dependent variables, followed by Bonferroni pairwise comparison. Age, gender and nationality (western/non-western) were retained as covariates. Based on the coding of video data, the results of ANCOVAs for the significant dependent variables are presented in Table IV-5. By controlling for the covariates including age, gender and nationality (see details in Table IV-5), a significant effect of veracity emerged for holding of hands: F(1, 21) = 4.75, p < .05, ηp2 = .18, d = 0.83. This illustrated that hand holding was higher for liars (M = 4.74, SD = 4.07) than truth-tellers (M = 1.36, SD = 4.05). Holding of hands was classified under hand/arm movements and indicated movement reduction and moderately tense behaviour (Mehrabian, 1968). This finding supports Hypothesis 1:

Deceivers will present different amounts of observable body movements, as compared with truth- tellers.

There was a trend towards significance for hand/arm movements for the evaluation variable indicating the probability of detection: F(1, 21) = 4.29, p = .051, ηp2 _{= .17, d = -0.80.}

non-evaluation condition (M = 5.90, SD = 4.62). As decreased limb movement is one of the recognised cues to deception, this finding shows some support, although not statistically significant, for Hypothesis 5: There will be a greater extent of observable nonverbal cue(s) (either body movement(s) and/or impression(s)) exhibited in deceivers when the probability of detection is higher, compared with lower probability conditions. Nonetheless, the finding of the trend of effect for the probability of detection across veracity conditions indicates that the same effect applies for truth-tellers as well. No covariates were statistically significant in relation to the level of probability. Values of mean and standard deviation for all the analysed dependent variables are presented in Appendix 3 Table A-1.

!

Table IV-5

ANCOVA results for Study 1a

Deception Level of Probability

F df p ηp2 F df p ηp2

Movements and Covariates

Hand and arm* 1.14 1, 21 .30 .05 4.29 1, 21 .51 .17 Age 1.58 1, 21 .22 .07 1.58 1, 21 .22 .07 Gender 1.23 1, 21 .28 .06 1.23 1, 21 .28 .06 Western/Non-western 0.63 1, 21 .44 .03 0.63 1, 21 .44 .03 Hand holding** 4.75 1, 21 .04 .18 2.24 1, 21 .15 .10 Age 5.62 1, 21 .03 .21 5.62 1, 21 .03 .21 Gender 6.98 1, 21 .02 .25 6.98 1, 21 .02 .25 Western/Non-western 0.01 1, 21 .94 .00 0.01 1, 21 .94 .00

Note. aInsignificant dependent variables were not listed. ηp 2

= effect size estimate – partial eta squared. **Significant effect of variables: p < .05 states in Bold; *trend of significant effect of variables: P < .055 states in Bold.

3.3. Discussion

This study revealed that the probability of deception detection might have a similar effect as veracity (deception) in altering nonverbal behaviour (e.g., a reduction in hand movements). The finding suggests that not only the stakes (i.e. the extent of consequences of deception) as

Reprinted by permission from Springer Nature Customer Service Centre GmbH: Springer Journal of Police and Criminal Psychology, Deception in Context: Coding Nonverbal Cues, Situational Variables and Risk of Detection, Zhang, K., Frumkin, L.A., Stedmon, A. et al. © 2013 https://doi.org/10.1007/s11896-013-9127-9

found in previous research (e.g., Caso et al., 2005; Vrij, 2008), but also the probability associated with the consequences might have a magnification effect on deceivers’ behaviours. However, similar to stakes, enhancing the probability of detection can lead to misjudgment as to whether or not someone is lying, since truth-tellers under higher probabilities can present similar reduced hand movements as deceivers. In addition, it is unknown whether there are unmeasured individual differences in perceiving the levels of probability of deception detection. To further address these matters, Study 2 used a within-subjects design to compare behaviour exhibited by the same person across different conditions of veracity and the probability of deception detection. Different levels of probabilities (low, moderate, and high) were assessed to investigate if the extent of nonverbal behaviours presented by the same individual is affected by increased probability levels.

4. Study 2 – Lying During Interaction with Another Person