The brain adapts to dishonesty

Journal name:
Nature Neuroscience
Volume:
19,
Pages:
1727–1732
Year published:
DOI:
doi:10.1038/nn.4426
Received
Accepted
Published online

Abstract

Dishonesty is an integral part of our social world, influencing domains ranging from finance and politics to personal relationships. Anecdotally, digressions from a moral code are often described as a series of small breaches that grow over time. Here we provide empirical evidence for a gradual escalation of self-serving dishonesty and reveal a neural mechanism supporting it. Behaviorally, we show that the extent to which participants engage in self-serving dishonesty increases with repetition. Using functional MRI, we show that signal reduction in the amygdala is sensitive to the history of dishonest behavior, consistent with adaptation. Critically, the extent of reduced amygdala sensitivity to dishonesty on a present decision relative to the previous one predicts the magnitude of escalation of self-serving dishonesty on the next decision. The findings uncover a biological mechanism that supports a 'slippery slope': what begins as small acts of dishonesty can escalate into larger transgressions.

At a glance

Figures

  1. Procedure and dishonesty escalation.
    Figure 1: Procedure and dishonesty escalation.

    (a) Participants were presented with pictures of glass jars containing different amounts of one-penny coins and entered their advice regarding how much money is in the jar. They were led to believe this advice was relayed to the Estimator via connected computers. (b) The magnitude of dishonesty increased over the course of the block when it benefited the self but not otherwise. This was evident when regressing dishonesty on trial number for each participant (n = 55) for each condition and then entering these escalation betas in a linear model across participants controlling for starting dishonesty and study (Self-serving–Other-harming: t52 = 3.38, P = 0.001; Self-serving–Other-serving: t52 = 2.25, P = 0.03; Self-harming–Other-serving: t52 = −0.27, P = 0.79). Dishonesty escalation was greater when it benefited the self than not (Self-serving–Other-harming vs. Self-harming–Other-serving: F1,51 = 8.80, P = 0.005; Self-serving–Other-serving vs. Self-harming–Other-serving: F1,51 = 4.61, P = 0.037; Self-serving–Other-harming vs. Self-serving–Other-serving: F1,51 = 0.183, P = 0.670; statistics reported for ANCOVAs on regression coefficients with condition as a two-level repeated factor, controlling for study and initial levels of dishonesty). (ce) Averaging mean dishonesty across participants on every trial and correlating with trial number (N = 60 trials) in each condition also revealed significant escalation when dishonesty was self-serving but not otherwise (Self-serving–Other-harming: r58 = 0.66, P < 0.001; Self-serving–Other-serving: r58 = 0.83, P < 0.001; Self-harming–Other-serving: r58 = −0.23, P = 0.08). Error bars represent s.e.m.; n.s., nonsignificant; *P < 0.05.

  2. Replication and extension study.
    Figure 2: Replication and extension study.

    (a) Initial level of dishonesty across participants (n = 25) was greater than 0 when Other-serving (t24 = 3.11, P = 0.01, one-sample t-test versus 0) but not when Self-serving (t24 = 0.92, P = 0.37, one-sample t-test vs. 0) and did not differ between conditions (F1,24 = 1.24, P = 0.28, repeated-measures ANOVA). (b) Mean dishonesty over the course of the block was greater than zero when Other-serving (t24 = 2.11, P = 0.046, one-sample t-test vs. 0) and when Self-serving (t24 = 3.78, P = 0.001, one-sample t-test vs. 0) and did not differ across conditions (F1,24 = 3.50, P = 0.07, repeated-measures ANCOVA). (c) Escalation of dishonesty was significant when Self-serving (analysis conducted as in Fig. 1b; t23 = 4.53, P < 0.001) but not when Other-serving (t23 = 1.62, P = 0.12) and greater in the former condition than the latter (F1,20 = 7.55, P = 0.01, repeated-measures ANCOVA, controlling for starting dishonesty). Error bars represent s.e.m.; *P < 0.05; n.s., nonsignificant; n.s.t., nonsignificant trend.

  3. Reduction in sensitivity to dishonesty over time.
    Figure 3: Reduction in sensitivity to dishonesty over time.

    (a) Parameter estimates of time-weighted dishonesty were averaged across all voxels in an ROI generated from Neurosynth based on a meta-analysis of 11,406 studies, reflecting P(Emotion|Activation). Higher z-scores (lighter colors) indicate higher likelihood that the term 'emotion' was used in a study given that a voxel was activated (see ref. 17 for method details), suggesting stronger selective association between that region and emotion17. ROI predominantly consists of bilateral amygdala. (b) A significant positive effect of time-weighted dishonesty was revealed when dishonesty was Self-serving–Other-harming (t24 = 2.36, P = 0.027, one-sample t-test vs. 0) but not when it was Self-harming–Other-serving (trend in opposite direction: t24 = −1.93, P = 0.066, one-sample t-test vs. 0), with the former significantly larger than the latter (t24 = 2.96, P = 0.007, paired-sample t-test). Repeating this analysis restricting the ROI to voxels in the anatomically defined amygdala revealed the same pattern of results (Supplementary Fig. 2a). n = 25; error bars represent s.e.m.; *P < 0.05; n.s., nonsignificant.

  4. Reduction in brain response to dishonesty predicts its escalation.
    Figure 4: Reduction in brain response to dishonesty predicts its escalation.

    (a) For each participant, in each condition, reduction in BOLD response to one unit of dishonesty on a current trial relative to the previous (extracted across the ROI displayed in Fig. 3a) is related to escalation in dishonesty on the next trial relative to the current trial. (b) Example participant shown for dishonesty in the Self-serving–Other-harming condition. (c) Across participants, these betas revealed a significant positive effect when dishonesty was Self-serving–Other-harming (t24 = 2.48, P = 0.021, one-sample t-test vs. 0) but not when it was Self-harming–Other-serving (t24 = −1.53, P = 0.14, one-sample t-test vs. 0) with the former betas larger than the latter (t24 = 2.82, P = 0.01, paired-sample t-test). Repeating this analysis on an ROI restricted to voxels in the anatomically defined amygdala revealed the same pattern of results (Supplementary Fig. 2b). n = 25; error bars represent s.e.m.; *P < 0.05; n.s., nonsignificant.

  5. Mean dishonesty and starting dishonesty
    Supplementary Fig. 1: Mean dishonesty and starting dishonesty

    (a) Mean dishonesty over the course of the block (Self-Serving-Other-Harming: t54 = 5.49, p<0.001; Self-Serving-Other-Serving: t54 = 6.16, p<0.001; Self-Harming-Other-Serving: t54 = -0.39, p=0.70 - one sample ttests vs 0; comparisons between conditions: Self-Serving-Other-Harming vs Self-Harming-Other-Serving: F1,53 = 26.44, p<0.001; Self-Serving-Other-Serving vs Self-Harming-Other-Serving: F1,53 = 39.67 p<0.0001; Self-Serving-Other-Serving vs Self-Serving-Other-Harming: F1,53 = 7.72, p=0.008 - statistics reported for separate 2 way ANOVAs with condition as a 2 level repeated factor controlling for study). (b) Starting dishonesty (Self-Serving-Other-Harming: t54 = 4.48, p<0.001; Self-Serving-Other-Serving: t54 = 5.56, p<0.001; Self-Harming-Other-Serving: t54 = 1.69, p=0.097, one sample ttests vs 0; Self-Serving-Other-Harming vs Self-Harming-Other-Serving condition: F1,53 = 7.74, p=0.007; Self-Serving-Other-Serving vs Self-Harming-Other-Serving condition: F1,53 = 21.99, p<0.001; F1,53 = 8.84, p=0.004, 2 way repeated measure ANOVAs controlling for study). N=55.

    Error bars represent standard error of the mean.

    *p<0.05, n.s. = non-significant

  6. Voxels restricted to bilateral amygdala
    Supplementary Fig. 2: Voxels restricted to bilateral amygdala

    fMRI analysis was repeated on voxels in our ROI that were restricted to the anatomically defined amygdala. This analysis generated similar results as those portrayed in Figures 3, 4. Specifically, (a) time-weighted dishonesty regressor positively correlated with BOLD response when dishonesty was Self-Serving-Other-Harming (t24 = 2.68, p=0.01, one sample ttest vs 0) but not when it was Self-Harming-Other-Serving (t24 = -0.9, p=0.38, one sample ttest vs 0) with the former parameter betas significantly greater than the latter (t24 = 2.60, p=0.02, paired sample ttest). (b) Reduction in BOLD response to one unit dishonesty on a current trial relative to the last predicted dishonesty escalation on next trial relative to current trial, when dishonesty was Self-Serving-Other-Harming (t24 = 2.30, p=0.03, one sample ttest vs 0) but not when Self-Harming-Other-Serving (trend in opposite direction t24 = -1.77, p=0.09, one sample ttest vs 0) with the former betas significantly larger than the latter (t24 = 3.05, p=0.01, paired sample ttest). N=25.

    Error bars represent standard error of the mean.

    *p<0.05, n.s. = non-significant

  7. Awareness of dishonesty
    Supplementary Fig. 3: Awareness of dishonesty

    Fifteen participants, who completed the task outside of the scanner in testing cubicles in Experiment 1, were also asked immediately after the Self-Serving-Other-Harming block of trials to estimate the magnitude by which they gave advice over and above what they actually thought was in the jar on the last trial, as well as on average throughout that block (order of these two questions was counterbalanced). Participants did not know in advance that they would be asked to do this. Comparing participants’ self-reports to their actual dishonesty revealed no significant differences on the last trial (t14 = 0.04, p>0.95, one sample ttest against a test value of 0) and on average (t14 = 1.29, p=0.22, one sample ttest against a test value of 0).

    n.s. = non significant

  8. Baseline accuracy
    Supplementary Fig. 4: Baseline accuracy

    (a) Mean absolute error in the baseline condition (t54 = 19.06, p<0.001, n=55, one sample ttest vs 0). (b) Parameters show that errors were not changing with time in the baseline condition (regressing error rate over the 60 trials on trial number for each participant revealed no change over time: t54 = 0.67, p=0.51, n=55, one sample ttest vs 0). As participants did not receive any feedback it is not surprising that they did not show significant improvement over the course of the block.

    Error bars represent standard error of the mean.

    *p<0.05

  9. Reaction time
    Supplementary Fig. 5: Reaction time

    RTs were slower in the Self-Harming-Other-Serving condition than the Self-Serving-Other-Serving condition (t54 = -2.19, p=0.03, n=55, paired sample ttest). Note, there was no correlation between mean RT and escalation of dishonesty in any of the conditions (all P > 0.3).

    Error bars represent standard error of the mean.

    *p<0.05

References

  1. Maremont, M. Anatomy of the Kurzweil fraud. Bus. Week 16 Sept. 1996.
  2. Kirchner, B. The Bernard Madoff Investment Scam (FT Press, 2010).
  3. McLean, B. & Elkind, P. The Smartest Guys in the Room: The Amazing Rise and Scandalous Fall of Enron (Penguin, 2013).
  4. Stapel, D. Ontsporing (Prometheus Amsterdam, 2012).
  5. Graham, C., Litan, R.E. & Sukhtankar, S. The Bigger They Are, The Harder They Fall: An Estimate of the Costs of the Crisis in Corporate Governance. (The Brookings Institution, 2002). Available at: http://www.brookings.edu/research/papers/2002/07/22corporategovernance-graham.
  6. Mauro, P. Corruption and growth. Q. J. Econ. 110, 681712 (1995).
  7. Tanzi, V. & Davoodi, H. Corruption, Public Investment, and Growth (Springer, 1998).
  8. Heyneman, S.P., Anderson, K.H. & Nuraliyeva, N. The cost of corruption in higher education. Comp. Educ. Rev. 52, 125 (2008).
  9. Peterson, C. Deception in intimate relationships. Int. J. Psychol. 31, 279288 (1996).
  10. DePaulo, B.M., Kashy, D.A., Kirkendol, S.E., Wyer, M.M. & Epstein, J.A. Lying in everyday life. J. Pers. Soc. Psychol. 70, 979995 (1996).
  11. Gamer, M., Rill, H.-G., Vossel, G. & Gödert, H.W. Psychophysiological and vocal measures in the detection of guilty knowledge. Int. J. Psychophysiol. 60, 7687 (2006).
  12. Abe, N., Suzuki, M., Mori, E., Itoh, M. & Fujii, T. Deceiving others: distinct neural responses of the prefrontal cortex and amygdala in simple fabrication and deception with social interactions. J. Cogn. Neurosci. 19, 287295 (2007).
  13. Schachter, S. & Latané, B. Crime, cognition, and the autonomic nervous system. In Nebr. Symp. Motiv. 12 (ed. Levine, D.) 221275 (1964).
  14. Breiter, H.C. et al. Response and habituation of the human amygdala during visual processing of facial expression. Neuron 17, 875887 (1996).
  15. Ishai, A., Pessoa, L., Bikle, P.C. & Ungerleider, L.G. Repetition suppression of faces is modulated by emotion. Proc. Natl. Acad. Sci. USA 101, 98279832 (2004).
  16. Denny, B.T. et al. Insula-amygdala functional connectivity is correlated with habituation to repeated negative images. Soc. Cogn. Affect. Neurosci. 9, 16601667 (2014).
  17. Yarkoni, T., Poldrack, R.A., Nichols, T.E., Van Essen, D.C. & Wager, T.D. Large-scale automated synthesis of human functional neuroimaging data. Nat. Methods 8, 665670 (2011).
  18. Lane, R.D. et al. Neuroanatomical correlates of pleasant and unpleasant emotion. Neuropsychologia 35, 14371444 (1997).
  19. Zald, D.H. & Pardo, J.V. Emotion, olfaction, and the human amygdala: amygdala activation during aversive olfactory stimulation. Proc. Natl. Acad. Sci. USA 94, 41194124 (1997).
  20. Schneider, F. et al. Differential amygdala activation in schizophrenia during sadness. Schizophr. Res. 34, 133142 (1998).
  21. Ketter, T.A. et al. Anterior paralimbic mediation of procaine-induced emotional and psychosensory experiences. Arch. Gen. Psychiatry 53, 5969 (1996).
  22. Irwin, W. et al. Human amygdala activation detected with echo-planar functional magnetic resonance imaging. Neuroreport 7, 17651769 (1996).
  23. Ledoux, J. The Emotional Brain: the Mysterious Underpinnings of Emotional Life (Simon & Schuster, 1998).
  24. Cain, D.M., Loewenstein, G. & Moore, D.A. The dirt on coming clean: perverse effects of disclosing conflicts of interest. J. Legal Stud. 34, 125 (2005).
  25. Costa, V.D., Lang, P.J., Sabatinelli, D., Versace, F. & Bradley, M.M. Emotional imagery: assessing pleasure and arousal in the brain's reward circuitry. Hum. Brain Mapp. 31, 14461457 (2010).
  26. Knutson, B., Adams, C.M., Fong, G.W. & Hommer, D. Anticipation of increasing monetary reward selectively recruits nucleus accumbens. J. Neurosci. 21, RC159 (2001).
  27. MacDonald, A.W. III, Cohen, J.D., Stenger, V.A. & Carter, C.S. Dissociating the role of the dorsolateral prefrontal and anterior cingulate cortex in cognitive control. Science 288, 18351838 (2000).
  28. Miller, E.K. & Cohen, J.D. An integrative theory of prefrontal cortex function. Annu. Rev. Neurosci. 24, 167202 (2001).
  29. Greene, J.D. & Paxton, J.M. Patterns of neural activity associated with honest and dishonest moral decisions. Proc. Natl. Acad. Sci. USA 106, 1250612511 (2009).
  30. Zhu, L. et al. Damage to dorsolateral prefrontal cortex affects tradeoffs between honesty and self-interest. Nat. Neurosci. 17, 13191321 (2014).
  31. Langleben, D.D. et al. Telling truth from lie in individual subjects with fast event-related fMRI. Hum. Brain Mapp. 26, 262272 (2005).
  32. Mohamed, F.B. et al. Brain mapping of deception and truth telling about an ecologically valid situation: functional MR imaging and polygraph investigation–initial experience. Radiology 238, 679688 (2006).
  33. Chang, L.J., Gianaros, P.J., Manuck, S.B., Krishnan, A. & Wager, T.D. A sensitive and specific neural signature for picture-induced negative affect. PLoS Biol. 13, e1002180 (2015).
  34. Winnett, R. & Rayner, G. No Expenses Spared (Bantam Press, 2009).
  35. Welsh, D.T., Ordóñez, L.D., Snyder, D.G. & Christian, M.S. The slippery slope: how small ethical transgressions pave the way for larger future transgressions. J. Appl. Psychol. 100, 114127 (2015).
  36. Schrand, C.M. & Zechman, S.L.C. Executive overconfidence and the slippery slope to financial misreporting. J. Account. Econ. 53, 311329 (2012).
  37. LeDoux, J.E. Emotion circuits in the brain. Annu. Rev. Neurosci. 23, 155184 (2000).
  38. Phelps, E.A. Emotion and cognition: insights from studies of the human amygdala. Annu. Rev. Psychol. 57, 2753 (2006).
  39. Poldrack, R.A. Can cognitive processes be inferred from neuroimaging data? Trends Cogn. Sci. 10, 5963 (2006).
  40. Shenhav, A. & Greene, J.D. Integrative moral judgment: dissociating the roles of the amygdala and ventromedial prefrontal cortex. J. Neurosci. 34, 47414749 (2014).
  41. Wu, D., Loke, I.C., Xu, F. & Lee, K. Neural correlates of evaluations of lying and truth-telling in different social contexts. Brain Res. 1389, 115124 (2011).
  42. Mazar, N., Amir, O. & Ariely, D. The dishonesty of honest people: a theory of self-concept maintenance. J. Mark. Res. 45, 633644 (2008).
  43. Effron, D.A., Bryan, C.J. & Murnighan, J.K. Cheating at the end to avoid regret. J. Pers. Soc. Psychol. 109, 395414 (2015).
  44. Eickhoff, S.B. et al. A new SPM toolbox for combining probabilistic cytoarchitectonic maps and functional imaging data. Neuroimage 25, 13251335 (2005).
  45. Eickhoff, S.B., Heim, S., Zilles, K. & Amunts, K. Testing anatomically specified hypotheses in functional imaging using cytoarchitectonic maps. Neuroimage 32, 570582 (2006).
  46. Charpentier, C.J., Moutsiana, C., Garrett, N. & Sharot, T. The brain's temporal dynamics from a collective decision to individual action. J. Neurosci. 34, 58165823 (2014).
  47. Edelson, M.G., Dudai, Y., Dolan, R.J. & Sharot, T. Brain substrates of recovery from misleading influence. J. Neurosci. 34, 77447753 (2014).
  48. Drueke, B. et al. Neural correlates of positive and negative performance feedback in younger and older adults. Behav. Brain Funct. 11, 17 (2015).
  49. Maldjian, J.A., Laurienti, P.J., Kraft, R.A. & Burdette, J.H. An automated method for neuroanatomic and cytoarchitectonic atlas-based interrogation of fMRI data sets. Neuroimage 19, 12331239 (2003).
  50. Slavich, G.M., Way, B.M., Eisenberger, N.I. & Taylor, S.E. Neural sensitivity to social rejection is associated with inflammatory responses to social stress. Proc. Natl. Acad. Sci. USA 107, 1481714822 (2010).

Download references

Author information

Affiliations

  1. Affective Brain Lab, Department of Experimental Psychology, University College London, London, UK.

    • Neil Garrett,
    • Stephanie C Lazzaro &
    • Tali Sharot
  2. Fuqua School of Business, Duke University, Durham, North Carolina, USA.

    • Dan Ariely

Contributions

T.S. conceived the study. N.G., S.C.L., D.A. and T.S. designed the study. N.G. collected behavioral and fMRI data. N.G. and T.S. analyzed the data. N.G. and T.S. wrote the manuscript with edits from S.C.L.

Competing financial interests

The authors declare no competing financial interests.

Corresponding authors

Correspondence to:

Author details

Supplementary information

Supplementary Figures

  1. Supplementary Figure 1: Mean dishonesty and starting dishonesty (31 KB)

    (a) Mean dishonesty over the course of the block (Self-Serving-Other-Harming: t54 = 5.49, p<0.001; Self-Serving-Other-Serving: t54 = 6.16, p<0.001; Self-Harming-Other-Serving: t54 = -0.39, p=0.70 - one sample ttests vs 0; comparisons between conditions: Self-Serving-Other-Harming vs Self-Harming-Other-Serving: F1,53 = 26.44, p<0.001; Self-Serving-Other-Serving vs Self-Harming-Other-Serving: F1,53 = 39.67 p<0.0001; Self-Serving-Other-Serving vs Self-Serving-Other-Harming: F1,53 = 7.72, p=0.008 - statistics reported for separate 2 way ANOVAs with condition as a 2 level repeated factor controlling for study). (b) Starting dishonesty (Self-Serving-Other-Harming: t54 = 4.48, p<0.001; Self-Serving-Other-Serving: t54 = 5.56, p<0.001; Self-Harming-Other-Serving: t54 = 1.69, p=0.097, one sample ttests vs 0; Self-Serving-Other-Harming vs Self-Harming-Other-Serving condition: F1,53 = 7.74, p=0.007; Self-Serving-Other-Serving vs Self-Harming-Other-Serving condition: F1,53 = 21.99, p<0.001; F1,53 = 8.84, p=0.004, 2 way repeated measure ANOVAs controlling for study). N=55.

    Error bars represent standard error of the mean.

    *p<0.05, n.s. = non-significant

  2. Supplementary Figure 2: Voxels restricted to bilateral amygdala (32 KB)

    fMRI analysis was repeated on voxels in our ROI that were restricted to the anatomically defined amygdala. This analysis generated similar results as those portrayed in Figures 3, 4. Specifically, (a) time-weighted dishonesty regressor positively correlated with BOLD response when dishonesty was Self-Serving-Other-Harming (t24 = 2.68, p=0.01, one sample ttest vs 0) but not when it was Self-Harming-Other-Serving (t24 = -0.9, p=0.38, one sample ttest vs 0) with the former parameter betas significantly greater than the latter (t24 = 2.60, p=0.02, paired sample ttest). (b) Reduction in BOLD response to one unit dishonesty on a current trial relative to the last predicted dishonesty escalation on next trial relative to current trial, when dishonesty was Self-Serving-Other-Harming (t24 = 2.30, p=0.03, one sample ttest vs 0) but not when Self-Harming-Other-Serving (trend in opposite direction t24 = -1.77, p=0.09, one sample ttest vs 0) with the former betas significantly larger than the latter (t24 = 3.05, p=0.01, paired sample ttest). N=25.

    Error bars represent standard error of the mean.

    *p<0.05, n.s. = non-significant

  3. Supplementary Figure 3: Awareness of dishonesty (29 KB)

    Fifteen participants, who completed the task outside of the scanner in testing cubicles in Experiment 1, were also asked immediately after the Self-Serving-Other-Harming block of trials to estimate the magnitude by which they gave advice over and above what they actually thought was in the jar on the last trial, as well as on average throughout that block (order of these two questions was counterbalanced). Participants did not know in advance that they would be asked to do this. Comparing participants’ self-reports to their actual dishonesty revealed no significant differences on the last trial (t14 = 0.04, p>0.95, one sample ttest against a test value of 0) and on average (t14 = 1.29, p=0.22, one sample ttest against a test value of 0).

    n.s. = non significant

  4. Supplementary Figure 4: Baseline accuracy (21 KB)

    (a) Mean absolute error in the baseline condition (t54 = 19.06, p<0.001, n=55, one sample ttest vs 0). (b) Parameters show that errors were not changing with time in the baseline condition (regressing error rate over the 60 trials on trial number for each participant revealed no change over time: t54 = 0.67, p=0.51, n=55, one sample ttest vs 0). As participants did not receive any feedback it is not surprising that they did not show significant improvement over the course of the block.

    Error bars represent standard error of the mean.

    *p<0.05

  5. Supplementary Figure 5: Reaction time (56 KB)

    RTs were slower in the Self-Harming-Other-Serving condition than the Self-Serving-Other-Serving condition (t54 = -2.19, p=0.03, n=55, paired sample ttest). Note, there was no correlation between mean RT and escalation of dishonesty in any of the conditions (all P > 0.3).

    Error bars represent standard error of the mean.

    *p<0.05

PDF files

  1. Supplementary Text and figures (729 KB)

    Supplementary Figures 1–5 and Supplementary Tables 1–4

  2. Supplementary Methods Checklist (622 KB)

Additional data