The Dutch sample responding on the web-based version of this task replicated the overall pattern obtained in prior research: subjects provided fewer endorsements of personal dilemmas (M = 0.37, SD = 0.28) than of impersonal dilemmas (M = 0.75; SD = 0.26; U = 22; z = 2.01; P = 0.04; r = 0.08).
As in our larger Dutch sample, all three test groups judged impersonal cases as more permissible than personal cases (): healthy controls (U = 13.0; z = 2.69; P = 0.007; r = 0.46); non-psychopathic delinquents (U = 18.0; z = 2.32; P = 0.02; r = 0.48); psychopaths (U = 23.5; z = 1.92; P = 0.05; r = 0.52). Thus, for all four test populations, individuals are more likely to perceive up close and personal harms as less permissible than harms that come about by impersonal means, such as flipping a switch in the classic trolley problem.
Fig. 1 The left column presents the results for subjects’ judgments of all 21 dilemmas. There were no differences between healthy controls (white bar, n = 35), non-psychopathic delinquents (gray bar, n = 23), and psychopaths (hashed bar, n = 14) for (more ...)
To examine whether the groups differed on the percentage of cases in which they endorsed the action – supporting the utilitarian outcome – and more generally, test the hypothesis that psychopaths (like VMPC patients) are more utilitarian on personal scenarios, we performed a 3 (test populations) × 2 (impersonal vs
personal dilemmas) ANOVA (see also, Supplementary Information
for Bayesian analyses of the same data set, designed to test the null hypothesis of no group differences). There was, as noted above, a highly significant dilemma type effect [F
(1,69) = 20.02; P
= 0.0001; d
= 2.03], but no significant group effect [F
(2, 69) = 0.21; P
= 0.81], and a non-significant interaction [F
(2,69) = 0.22; P
= 0.80; ]. Bonferroni corrected post-hoc tests revealed no statistically significant group effect for either impersonal (P
’s > 0.18) or personal moral dilemma (P
’s > 0.41).
Evaluation of educational level demonstrated a significant difference between the groups, with offenders having lower levels of education than non-offenders, but no difference between the two groups of offenders (χ2 = 12.90; P < 0.05). More importantly, an ANCOVA demonstrated that there was no significant effect of education on judgments of either personal or impersonal dilemmas (all P’s > 0.05).
Healthy controls were generally younger than both delinquent groups. Since there was a significant age difference [F(2,69) = 9.29; P < 0.000], due to healthy controls being younger than non-psychopathic delinquents, we conducted a correlation analysis to examine whether age was related to moral responses. For both personal as well as impersonal dilemmas, there was no effect of age (r = 0.04 and −0.21 respectively with all P’s > 0.05).
Given that prior work on VMPC and FTD patients revealed a highly selective deficit within the personal dilemmas, with greater endorsements of the utilitarian outcome for other-serving (i.e. harming one for the benefit of others) than self-serving (harming one for self-benefit) personal dilemmas, we explored in greater detail the variance in responses to personal dilemmas by both delinquent groups and our controls (). An overall 2 (Self vs Other-serving) × 3 (Psychopaths, Delinquents, Controls) ANOVA revealed a statistically significant main effect for dilemma type, with subjects judging other serving cases as more permissible than self-serving [F (1, 36) = 48.52; P < 0.0001]. There was, however, no main effect for the three test populations [F (2, 36) = 0.81; P = 0.45] and nor was there a statistically significant interaction between dilemma type and test population [F (2, 36) = 1.01; P = 0.37]. Thus, psychopaths were not more likely to endorse the utilitarian outcome for other-serving, personal dilemmas.
Fig. 2 Scatterplot of subjects’ judgments (mean proportion of Yes responses) for all personal moral dilemmas, divided into self-serving (far left, first three cases) and other-serving (right, 11 cases) vignettes. Healthy controls are indicated by white (more ...)
On a scenario level, there were several dilemmas that elicited virtually complete support of the utilitarian outcome by subjects in all groups (80–100% Yes judgments) or virtually complete prohibition of this outcome (0–20% Yes judgments; ). For example, each of our test populations agreed that the actions to be taken in dilemmas 2, 3, 4 and 6 were largely impermissible, whereas those in dilemmas 13 and 14 were largely permissible; furthermore, although subjects in all three test populations were less clear about the permissibility of the action for several cases (e.g. 7, 8, and 11 in ), all clustered around the same proportion of Yes responses. Lastly, although the mean permissibility ratings for psychopaths were higher than the control populations for 8 out of 11 other-serving dilemmas, the variance in all three groups was sufficiently high to make this apparent difference non-significant. More specifically, for 3 of the 11 other-serving dilemmas, the delinquents provided a greater proportion of Yes judgments; for four of these dilemmas, the psychopaths differed from the other groups by less than 15%, leaving only four cases where the psychopaths judged the case more permissible by 20–40%. Thus, even on a case by case basis, there is no consistent pattern of judgments that is mediated by the characteristics of our study populations.
We also explored the difference in judgments within the class of other-serving cases in which sometimes, harming one to benefit many others makes the one worse off (e.g., the footbridge trolley case where pushing the man off the bridge kills him but saves five) whereas in others, harm to the one is inevitable, does not make the individual worse off, and yet benefits many others (e.g., every person in a war bunker will be killed by enemy soldiers if anyone makes noise, so if a baby starts crying, killing the crying baby doesn’t make her worse off, but saves the others); these latter cases are often described as Pareto dilemmas, and in previous work, are typically judged more permissible than non-Pareto cases where the one is made worse off (Huebner, Pettit, and Hauser, in review; Moore et al.
). Group contrasts for the Pareto cases failed to reveal a significant difference (P
’s > 0.22).
Of the 37 delinquents, PCL-R factor scores were available for 15 subjects. There was no statistically significant correlation between subjects’ moral judgments on personal dilemmas and their factor 1 (r = −0.02, P = 0.95) or 2 scores (r = −0.02; P = 0.93; A and B).
Though there is a generally agreed upon cut-off on the PCL-R diagnostic for classifying individuals as psychopaths (i.e., scores of 26 or higher), there was, as in all previous work, variation among our subjects in such scores, as well as in the nature of their criminal conviction. To assess whether such variation was related to their moral judgments, we plotted (A) each psychopath’s PCL-R score against the proportion of personal dilemmas that they endorsed, and further grouped the subjects by their type of crime. Though the sample size is too small to evaluate statistically, neither the scatter in the data shows relationship between PCL-R score and proportion of personal dilemmas endorsed, nor a clear pattern for type of crime. Similarly, there was no effect of PCL-R score or type of conviction on the proportion of utilitarian outcomes endorsed for the other-serving cases (B).
Results on the SRM-SF showed that overall there was no statistically significant difference among the groups, with psychopaths showing slightly lower SRM-SF scores (M = 276.14; SD = 33.43) than healthy controls (M = 286.03; SD = 45.15), whereas non-psychopathic offenders had slightly higher SRM-SF scores (M = 290.01; SD = 46.59) than healthy controls [F(2,69) = 0.45; P = 0.64]. None of the post-hoc tests were statistically significant (all t′s < 0.98; P > 0.34).