The ability of psychopaths to con, lie and manipulate means that their potential danger to others may not be appreciated by criminal-justice agencies, that their response to clinical assessment is often rigged3, and that researchers are blinded to underlying motives (Fig. 1). Faced with the problem of detecting socially stigmatic beliefs, psychologists have developed tasks that measure these attitudes implicitly. The IAT has successfully quantified beliefs that people may wish to disguise, such as negative views about racial groups, homosexuality or obesity. We have adapted the IAT to reveal implicit beliefs about violence in psychopathic murderers.

Figure 1
figure 1


Portrait of a serial killer: psychopathic murderers such as Ted Bundy can often project a deceptively charming persona.

We followed the general methodology of the IAT2. Briefly, uppercase words (for example, 'UGLY') are classified as being 'pleasant' or 'unpleasant', and lowercase words (for example, 'kill') are classified as 'violent' or 'peaceful', by pressing corresponding buttons. When the same response key is assigned for both the unpleasant and violent words (this is termed the congruent condition), most people find the task easy. But when pleasant and violent words share the same response key (the incongruent condition), most people find this confusing. The association between 'pleasant–unpleasant' and 'violent–peaceful' is indexed by means of the IAT effect (reaction time for the incongruent condition minus reaction time for the congruent condition).

Our sample (n = 121) was recruited from consecutive admissions to a secure therapeutic community for male offenders with personality disorders. None of the participants had a co-morbid mental illness and none was taking any illicit drug or medication. Psychopathy was measured using the Psychopathy Checklist-Revised test4; those who scored over 25 were labelled as psychopaths5. We selected the murderers in order to target the most seriously violent group. Our final groups consisted of 13 psychopathic murderers, 17 non-psychopathic murderers, 39 psychopathic other offenders and 52 non-psychopathic other offenders. There was no significant difference in intelligence quotient (IQ) among the groups.

The results from the 'violent' IAT test are shown in Fig. 2. Statistical analysis (analysis of variance) indicates that there was no significant effect due to grouping (murderer versus other offender) or psychopathy score. Crucially, the interaction between these variables was significant (P < 0.05) and remained so even when age and IQ (as assessed by the revised National Adult Reading Test6) were added as covariates. As expected, the psychopathic murderers showed a much lower IAT effect than the non-psychopathic murderers (P < 0.05). There was no significant difference in error rate related to offence type, psychopathy or interaction between these variables (Fig. 2, inset). A control IAT (flower–insect; pleasant–unpleasant) revealed no effect for psychopathy or offence type.

Figure 2: Results of the Implicit Association Test (IAT) effect (difference in reaction time between incongruent and congruent conditions; see text), plotted as a function of psychopathy score (PCL-R4) for murderers (red) and non-murderers (blue).
figure 2

Inset, IAT effect for error rates. All participants gave written informed consent. We used eight 'pleasant' and eight 'unpleasant' words9, and eight 'violent' and eight 'peaceful' words. Both the congruent and incongruent blocks consisted of 64 trials, with words presented in a pseudo-random order. Further details are available from the authors.

Our results indicate that the reduced violent-IAT effect seen in psychopathic murderers is likely to be due to their abnormal beliefs about violence, rather than to some other nonspecific effect such as poor impulse control and/or deficits in decision-making. Psychopathic murderers have diminished negative reactions to violence compared with non-psychopathic murderers and other offenders.

To our knowledge, this is the first cognitive demonstration of abnormal social beliefs in murderous psychopaths. The interaction between psychopathy and murder suggests that it is the combination of these two factors that is crucial in identifying deficient implicit beliefs about violence, resulting in a relatively rare subpopulation of very violent psychopaths. Obviously, not all murderers have similar motivations or beliefs about violence, and not all psychopaths commit murder (as illustrated by the phenomenon of the 'white-collar' psychopath7).

Our results have some testable implications. For example, non-psychopaths could become sensitized to violence after committing murder8 (hence their greater scores in the IAT). The difference between the psychopathic groups may reflect two separate, stable populations of psychopathic offenders: one with deficient social beliefs (and therefore an increased disposition towards extreme violence) and the other in which such negative beliefs are absent. If this difference can be picked up by the 'violent' IAT before an offence is committed, this test may become an important tool for distinguishing psychopaths who are likely to commit extremely violent offences from those who are not.