Participants

Twenty-four (21 right handed and 3 left handed) male participants aged 19–60 years (mean=30.04, s.d.=11.34) were recruited from University of Manchester ancillary staff and students. Specifically, participants were recruited using adverts placed in the University staff news letter and by approaching portering staff in each building. Students were recruited by targeting university sports teams (i.e. rugby teams) with the hypothesis that participants drawn from these populations may show higher levels of subclinical psychopathy spectrum personality traits. The majority (n=22) of the sample were White with the remaining participants of Asian ethnicity. The mean IQ of the sample measured using the National Adult Reading Test ^{Reference Nelson23} was 113.87 (s.d.=7.60, range 96–128). The study was approved by the University of Manchester research ethics committee and participants gave written informed consent for participation in the study.

Measurement of psychopathic personality traits

Psychopathic personality traits were assessed using the Psychopathic Personality Inventory. ^{Reference Lilienfeld and Andrews19} The PPI is a 187-item self-report questionnaire with a total score and 8 sub-scales designed to measure psychopathic personality traits in a dimensional manner.

These include:

1. (a) Machiavellian egocentricity which is characterised by ‘looking out for one's own interests before others’;

2. (b) social potency, or the ‘ability to be charming and influence others’;

3. (c) coldheartedness is the ‘propensity towards callousness, guiltlessness, and unsentimentality’;

4. (d) carefree non-planfulness, is the ‘non-planning component of impulsivity’;

5. (e) fearlessness, is the ‘absence of anxiety and harm concerning eagerness to take risks’;

6. (f) blame externalisation, is the ‘tendency to view others as source of problems’;

7. (g) impulsive non-conformity, is the ‘reckless lack of concern for social mores’;

8. (h) stress immunity, is the ‘absence of marked reactions to otherwise anxiety provoking events’ (pp. 500–2). ^{Reference Lilienfeld and Andrews19}

Each item consists of a statement to which participants must indicate how accurately it applies to them using a 4-point scale ranging from 1 ‘false’ to 4 ‘true’. The PPI has been shown to have good convergent and discriminant validity in both community and criminal samples. ^{Reference Lilienfeld and Andrews19,Reference Nelson23–Reference Patrick, Edens, Poythress, Lilienfeld and Benning27} In particular, it shows good criterion related validity when compared with structured, collaboratively rated clinical assessments of psychopathy such as the PCL–R. ^{Reference Hare18,Reference Poythress, Edens and Lilienfeld24} The PPI scores for the sample are shown in Table 1. The mean total PPI score for the sample was lower than that reported by Lilienfeld et al (S. Lilienfeld, personal communication, 2008), for a large sample of substance misusing male prisoners (see online supplement for details). However, individuals in the present sample did show total scores at or above the criminal mean and the fearlessness, social potency, coldheartedness and stress immunity scores for the present sample were remarkably similar to those reported for the criminal population. Additional figures demonstrating the sample distribution of scores on each sub-scale can be found in the online Fig. DS1.

Table 1 The Psychopathic Personality Inventory scores for the sample

	Items, n	Mean	s.d.	Range
Total score	163	364.71	35.03	313.00–442.00
Machiavellian egocentricity	30	61.75	10.78	46.00–84.00
Social potency	24	64.08	10.01	46.00–79.00
Fearlessness	19	49.42	8.92	34.00–66.00
Coldheartedness	21	45.50	7.32	36.00–59.00
Impulsive non-conformity	17	36.70	7.75	24.00–54.00
Blame externalisation	18	31.58	7.91	19.00–51.00
Carefree non-planfulness	20	37.79	6.78	26.00–50.00
Stress immunity	11	31.75	5.36	17.00–39.00

Deception paradigm

The deception task used in the present study was based on the task reported by Spence et al. ^{Reference Spence, Farrow, Herford, Wilkinson, Zheng and Woodruff4} Prior to scanning, participants were asked to fill in a questionnaire determining if they had performed 36 everyday acts during the current day (making the bed, taken a tablet, etc). Once in the scanner, participants were asked to lie or tell the truth about the performance of the 36 acts. In a standard ABAB block design, each participant was required to lie about the performance of each act once and tell the truth about the performance of each of the 36 acts once. Each of the 12 blocks contained 6 acts and each act was displayed visually on a screen for 5 s in the form ‘In the course of today have you… (made the bed)’. Participants were required to make a motor response on a button box in order to answer yes or no. They were instructed to lie or tell the truth depending on which prompt appeared on the screen. In order to increase task performance, participants were informed that an experimenter would be monitoring their responses in order to detect whether they were lying. Participants carried out a practice block prior to the main task. Response accuracy was calculated by comparing responses made to the truth or lie prompt during the task to the original response made in the 36-item questionnaire. Response times (seconds) were recorded for each trial and average response times during the truth and lie conditions were compared using a two-tailed paired-sample t-test. In addition, response times and response accuracy (relative to the original questionnaire items) were correlated with PPI sub-scales using Spearman's correlations.

MRI image acquisition

Images were acquired using a Philips (Eindhoven, Holland) 1.5 T Gyroscan ACS NT retrofitted with Powertrak 6000 gradients, operating at a software level 6.1.2 T₂ ^*-weighted volumes were acquired using a singleshot echo-planar imaging pulse sequence. Each volume comprised 40 contiguous axial slices, (response time (TR)/echo time (TE) 5000/40 ms, 64×64 data matrix, 3.5 mm thickness with an inplane resolution of 3×3 mm). The stimuli were rear projected onto a screen using a liquid crystal display projection system. Task administration was coupled to image acquisition using personal computer software and hardware linked to a response button.

Analysis

Imaging data were analysed using Statistical Parametric Mapping (SPM5, Friston, The Welcome Department of Cognitive Neurology, London, UK). Images were corrected for motion and then realigned with the first scan serving as a reference. The scans were then normalised into a standard stereotactic space ^{Reference Talairach and Tournoux28} using Montreal Neurological Institute templates. Images were finally smoothed with a 10 mm Gaussian filter to facilitate inter-individual averaging. After this spatial preprocessing, at an individual level, a general linear model with a delayed boxcar waveform was used to model BOLD signal changes during the task. The individual images were then combined in a random effects analysis that would allow inference to the general population using an independent samples t-test to investigate the main effect of the task. The main effect for the lie condition was the BOLD signal seen in the lie condition minus the BOLD signal seen in the truth condition. The main effect for the truth condition was the reverse subtraction.

The resulting statistical maps were thresholded at P<0.001 uncorrected with only cluster sizes of five or more contiguous voxels being reported. As the inferior (frontopolar/orbitofrontal/ventrolateral prefrontal), and superior frontal (dorsolateral/dorsomedial prefrontal) cortex are the regions in which activations are most consistently reported across different deception paradigms in the literature, we concentrated our primary analysis solely on these areas. In order to control for type I errors we applied small volume corrections ^{Reference Worsley, Marrett, Neelin, Vandai, Friston and Evans29} for family-wise error at P⩽0.05 to these a priori regions of interest. Areas of activation at the P<0.001 uncorrected level are also reported when bilateral activations were seen.

In addition, we performed an exploratory analysis of signal in those areas less consistently identified by the past literature as active during deceptive responses; anterior cingulate, caudate, insula, thalamus, temporal lobes, temporal poles, posterior cingulate and precuneus. Again, in these regions we performed small volume corrections for family-wise error.

The association between BOLD responses during the lie condition and scores on the PPI sub-scales was investigated using simple regression analysis. During this analysis, for each a priori specified brain area of interest that exhibited a significant voxel-based correlation, the BOLD signal change observed during the lie condition was extracted from SPM and used in Spearman's correlational analysis (SPSS version 14 for Windows) in order to produce confirmatory r-values. As a further protection against type I errors we have only reported results where the probability value of the associated r-value exceeded P⩽0.001 for unilateral activations and P⩽0.01 for bilateral activations.

Finally, in order to control for possible within-sample variation in age, we examined the correlation between age and BOLD responses during the deception condition. Age was found to be positively correlated with BOLD response in the temporal poles and insula and was therefore entered as a nuisance covariate in any analyses where significant BOLD responses were exhibited in these areas.

Behavioural data

Mean response accuracy was similar for the truth (91.06%, s.d.=11.66) and lie condition (91.55%, s.d.=12.21). The number of accurate responses (maximum of 36) for the lie condition showed a significant positive correlation with PPI stress immunity score (r=0.53, P=0.009).

Mean response times (seconds) were significantly slower during the lie condition compared with the truth condition (lie mean response time 2.66 s (s.d.=0.42), truth mean response time 2.56 s (s.d.=0.39); t=2.43, P=0.024).

There were no significant correlations between any of the other mean PPI sub-scales/factor scores and mean response accuracy or response time for the lie condition or truth condition.

Functional MRI data

Main effect of task

During the lie condition (relative to the truth condition) an increased BOLD response was seen in the left ventrolateral prefrontal cortex (Brodmann area (BA)47). An increased BOLD response at the P<0.001 uncorrected level was also seen in the right ventrolateral prefrontal cortex (Table 2 and online Fig. DS2).

Table 2 Main effect of the deception task

Cluster size (k)	Montreal Neurological Institute coordinates x, y, z	Z	Family-wise error, corrected probability	Anatomical area	Brodmann area
Lie–truth
42	–30, 24, −12	4.27	0.004	Left ventrolateral prefrontal cortex	47
42	51, 24, −15	3.24	0.105 a	Right ventrolateral prefrontal cortex	47
Truth–lie
34	–6, 48, −6	4.39	0.001	Left frontopolar prefrontal cortex	10
20	3, 48, −3	3.48	0.024	Right frontopolar prefrontal cortex	10
28	–9, 54, 0	3.85	0.030	Left medial superior frontal	10
25	6, 51, 0	3.29	0.088 a	Right medial superior frontal	10

a. Significant at P<0.001, uncorrected

During the truth condition (relative to the lie condition) an increased BOLD response was seen bilaterally in the frontopolar area of the prefrontal cortex (BA10) extending into the medial superior frontal cortex (Table 2).

The results remained similar when the analysis was rerun excluding six participants who achieved less than 90% overall response accuracy for the task.

Relationship between the PPI sub-scales and BOLD response during the lie condition

The results of the correlational analysis are shown in Table 3. During the lie condition (relative to the truth), fearlessness scores were negatively correlated with BOLD responses in the right orbitofrontal cortex (Fig. 1). Coldheartedness scores were negatively correlated with BOLD responses bilaterally in the temporal poles (Fig. 2). Machiavellian egocentricity scores were negatively correlated with BOLD responses bilaterally in the caudate. Social potency scores were negatively correlated with BOLD responses bilaterally in the right posterior cingulate. Stress immunity scores were negatively correlated with BOLD response bilaterally in the insula. These associations remained significant after covarying for handedness and age (where appropriate).

Fig. 1 The association between Psychopathic Personality Inventory (PPI) fearlessness and blood oxygen level dependent response in the right orbitofrontal cortex during deception.

Fig. 2 The association between Psychopathic Personality Inventory (PPI) coldheartedness and blood oxygen level dependent response in the temporal poles during deception.

Table 3 The association between Psychopathic Personality Inventory (PPI) sub-scale scores and blood oxygen level dependent (BOLD) responses in brain areas of interest

PSI scale	Cluster size (k)	Montreal Neurological Institute coordinates x, y, z	Z	Family-wise error, probability	Anatomical area	Brodmann area	Spearman's r
Fearlessness	5	3, 21, −12	3.87	0.007	Right orbitofrontal	11	–0.74, P<0.001
	33	12, 42, −9	3.53	0.021	Right orbitofrontal	11	–0.70, P<0.001
Coldheartedness	61	–45, 12, −24	4.21	0.005	Left temporal pole a	38	–0.75, P<0.001
	20	33, 18, −33	3.47	0.051	Right temporal pole a	38	–0.61, P=0.001
Machiavellian egocentricity	16	18, 0, 24	4.07	0.005	Right caudate	n/a	–0.75, P<0.001
	9	–3, 6, −6	3.53	0.030	Left caudate	n/a	–0.65, P=0.001
Social potency	10	3, −42, 15	4.00	0.003	Right posterior cingulate	29	–0.67, P<0.001
Stress immunity	13	–39, −18, 21	3.70	0.035	Left insula a	n/a	–0.54, P=0.006
	31	33, −18, 6	3.89	0.019	Right insula a	n/a	–0.75, P<0.001

a. Result remained significant after covarying for age

There were no significant associations between BOLD responses during the lie condition in any of the brain areas of interest and the PPI total score or impulsive non-conformity, carefree non-planfullness and blame externalisation sub-scale scores.

Behavioural data

Consistent with previous studies in this field ^{Reference Spence, Farrow, Herford, Wilkinson, Zheng and Woodruff4,Reference Nunez, Casey, Egner, Hare and Hirsch9} we found that the behavioural data indicated that mean response times were significantly longer for the lie compared with the truth condition but there was no effect of personality trait scores. It is possible that the relatively small sample size may account for the lack of an observed association between response time and personality traits. However, Nunez et al ^{Reference Nunez, Casey, Egner, Hare and Hirsch9} also failed to find an association between response times during lie responses and scores on the PPI. We did find, however, that stress immunity, which reflects a lack of anxiety, was positively correlated with response accuracy in the lie condition. This would support the argument that interpersonal differences in trait anxiety can mediate behavioural and physiological responses during deception. ^{Reference Wiley30,Reference Steinbrook31} This is an area that may warrant further investigation in imaging studies of deception.

Main effect of task imaging data

Consistent with our hypothesis, we found that the lie condition (relative to the truth condition) was associated with increased BOLD responses bilaterally in the ventrolateral prefrontal cortex. This is a replication of the finding reported by Spence et al ^{Reference Spence, Farrow, Herford, Wilkinson, Zheng and Woodruff4} using the same task in a smaller sample, and as such represents the first between-laboratory replication of an fMRI deception finding (a research gap recently highlighted by Spence ^{Reference Spence32} ). This finding also supports previous studies using different deception paradigms that have reported deception related BOLD responses in inferior frontal areas. ^{Reference Lee, Ho-Ling, Chan, Yen-Bee, Fox and Gao6,Reference Luan Phan, Magalhaes, Ziemlewicz, Fitzgerald, Green and Smith8,Reference Nunez, Casey, Egner, Hare and Hirsch9,Reference Langleben, Schroeder, Maldjian, Gur, McDonald and Ragland14} As the ventrolateral prefrontal cortex has been shown to be active during a number of cognitive control paradigms ^{Reference Ridderinkhof, van den Wildenberg, Segalowitz and Carter33–Reference Dove, Manly, Epstein and Owen36} this finding also adds further weight to the argument that deception engages executive prefrontal systems in order to achieve the production of a ‘lie’ at the same time as withholding the truth. ^{Reference Spence, Hunter, Farrow, Green, Leung and Hughes1,Reference Spence, Farrow, Herford, Wilkinson, Zheng and Woodruff4} Similar to Spence et al ^{Reference Spence, Farrow, Herford, Wilkinson, Zheng and Woodruff4} we did not find any significant BOLD responses during deception in any of the other frontal areas of interest. Given that previous studies in the area ^{Reference Spence, Farrow, Herford, Wilkinson, Zheng and Woodruff4–Reference Abe, Suzuki, Mori, Itoh and Fujii13} have shown deception-related activity in the dorsolateral prefrontal cortex and orbitofrontal cortex, it is possible that between-study deception paradigm and sample size differences may account for the current lack of findings in these areas.

We also found that the truth condition was associated with increased BOLD responses bilaterally in the frontopolar cortex. In a recent review of frontopolar function, Koechlin et al ^{Reference Koechlin and Hyafil37} suggest that during decision-making tasks, lateral inferior frontal regions inhibit frontopolar regions in order to switch to and maintain a given response set. They also propose that frontopolar regions are able to store a previous response set in a back-up buffer in order to reinstate it following a reduction in top-down inhibition. In the present study, during the lie condition, the lateral inferior frontal cortex may have been exerting a strong inhibitory influence on the frontopolar regions in order to override the alternative truthful response set and switch to the deceitful set. In the truth condition, the frontopolar regions may have shown enhanced activation while accessing and returning to the use of the truthful response set. It is possible, therefore, that the underlying ‘task-switching’ nature of the deception paradigm used in the current study may be largely responsible for the activations seen during the truth condition. The majority of published studies in this area do not specifically examine BOLD responses during the truthful condition; however, of those that do, few reported any areas of activation during truthful responding. ^{Reference Spence, Hunter, Farrow, Green, Leung and Hughes1,Reference Luan Phan, Magalhaes, Ziemlewicz, Fitzgerald, Green and Smith8–Reference Kozel, Padgett and George10} Despite these negative findings, Langleben et al ^{Reference Langleben, Loughead, Bilker, Ruparel, Childress and Busch7} reported truth-related activations in the left medial frontal gyrus (BA46). The differences in findings may not only reflect inter-study differences in methodology, but also, in this case in particular, inter-laboratory differences in the ability to accurately image the frontopolar region without a large degree of airspace related signal drop out.

Relationship between personality factors and imaging findings

Psychopathy as a construct is generally considered to be characterised by high levels of callous unemotional traits and these traits are believed to be related to dysfunction in the limbic (amygdala) striatal prefrontal circuitry. ^{Reference Blair38} The ability to measure key components of the psychopathy construct in less pathological samples allows us to postulate on the potential neuropathology of this disorder in clinical samples. The present study found evidence that BOLD responses during deception in a number of brain areas of interest were correlated with some, but not all, psychopathic personality trait scores. Specifically focusing on the prefrontal cortex, we found inverse correlations between fearlessness scores and BOLD responses in the right orbitofrontal lobe. This suggests that those with low levels of fear and harm avoidance may find it easier to lie and therefore do not activate the orbitofrontal cortex which is a key neural structure implicated in behavioural inhibition in a number of previous imaging studies. ^{Reference Ridderinkhof, van den Wildenberg, Segalowitz and Carter33,Reference Horn, Dolan, Elliot, Deakin and Woodruff39} Our own work on the neuropsychology of behavioural inhibition in samples of offenders with antisocial personality disorder suggests that this clinical group are significantly impaired on tasks probing orbitofrontal cortex function compared with healthy controls. ^{Reference Dolan and Park40} Given reports of increased white matter volume in some of the prefrontal subregions implicated in deception-related neurocircuitry in those who lie, cheat and manipulate others, ^{Reference Yang, Raine, Lencz, Bihrle, Lacasse and Colletti16,Reference Yang, Raine, Narr, Lencz, LaCasse and Colletti17} this is a brain area that warrants more detailed investigation.

In our exploratory analysis, we found that coldheartedness was inversely correlated with temporal pole BOLD responses. As the temporal pole has been found to be active during theory of mind tasks, ^{Reference Olson, Plotzker and Ezzyat41} our findings highlight the importance of looking at the role of callous unemotional traits and theory of mind/mentalising ability in tasks assessing the cognitive elements that may be involved in the deception and manipulation of others. Nunez et al ^{Reference Nunez, Casey, Egner, Hare and Hirsch9} found that coldheartedness assessed using the PPI was negatively correlated with BOLD responses in the posterior cingulate and precuneus cortices. In the present study, BOLD signal in the posterior cingulate was inversely associated with PPI social potency scores. Differences between studies may reflect the gender differences in the nature of the samples studied. Base rates of psychopathic traits are lower in female populations, ^{Reference Grann42} and female populations exhibit lower scores on some (but not all) specific symptoms such as callousness/lack of empathy on the PCL–R, ^{Reference Grann42} PPI stress immunity, PPI social potency and on a factor similar to PPI coldheartedness. ^{Reference Maesschalck, Vertommen and Hooghe43} In addition, there is some indication of a gender difference in the bio-behavioural correlates of psychopathy, with only male cohorts exhibiting a lack of physiological reactivity to aversive stimuli ^{Reference Justus and Finn44} and stress. ^{Reference O'Leary, Loney and Eckel45}

In this study we also found that lower stress immunity scores (i.e. more anxiety/stress) were associated with greater BOLD responses in the bilateral insula. As the insula has been shown to be involved in error processing during Go/No Go tasks, ^{Reference Hester, Fassbender and Garavan46} it is possible that stress immunity (which may relate to vigilance) may influence the function of the neural circuitry involved in the processing and monitoring of errors. Although at least some degree of anxiety is needed to engage error processing circuitry appropriately, it is possible that high levels may impair accuracy performance on behavioural tasks and moderate neural responses in imaging studies. Further studies are required to investigate the significance of both trait and state anxiety in deception-related brain activation patterns in those with antisocial and deceptive personality traits.

We found that higher Machiavellian egocentricity scores were associated with reduced BOLD responses in the bilateral caudate. Although there is a limited literature to compare the findings of the present study, decreased caudate activity appears to be associated with higher scores on the interpersonal (deceptive/superficial/grandiose) component of psychopathy, ^{Reference Soderstrom, Hultin, Tullberg, Wikkelso, Ekholm and Forsman47} indicating that people with these and related personality traits, such as Machiavellian egocentricity, show reduced activation of caudate regions which are a key component of the subcortical straital network. A common finding across species and methodologies is the involvement of the striatum, the input structure of the basal ganglia, in a circuit responsible for mediating goal-directed behaviour. ^{Reference Delgado48} In functional imaging studies, the caudate has been shown to be involved in the inhibition of both motor and mental responses ^{Reference Mink49–Reference Durston, Davidson, Thomas, Worden, Tottenham and Martinez55} and appears to be specifically involved in the mediation of arousal. ^{Reference Brown, Manuck, Flory and Hariri56}

Overall, our findings fit with the previous literature suggesting that simple deception tasks activate prefrontal regions implicated in behavioural restraint and conflict monitoring and that lying results in greater activation than truthful responding. Our findings also tentatively suggest that specific personality traits may have a modulating effect on brain responses to deception tasks and that future studies examining brain activation during deception in offenders with and without psychopathic traits may be of value in understanding the neuropathology of psychopathy and antisocial personality disorder.

Limitations

Although the results of the present study are suggestive of an association between psychopathic traits and the neural processes involved in deception, there are a number of limitations that need to be taken into consideration. The deception paradigm used is highly constrained and similar in terms of cognitive demands to a Go/No Go test of behavioural inhibition. Recent imaging studies of deception are utilising more complex ecologically valid paradigms. For example, a recent study by Abe et al ^{Reference Abe, Suzuki, Mori, Itoh and Fujii13} has demonstrated the involvement of the amygdala in a deception paradigm with a social component and it is possible that this neural response may also show some relationship with the callous/unemotional aspects of psychopathy. In addition, the use of the PPI in the present study may have limited the measurement of the core deceitful/manipulative components of psychopathy. The majority of the PPI items focus on impulsive/antisocial or fearless/dominant traits, ^{Reference Benning, Patrick, Hicks, Blonigen and Krueger26} a more selective use of multiple measures of deception, such as those used by Yang et al, ^{Reference Yang, Raine, Lencz, Bihrle, Lacasse and Colletti16,Reference Yang, Raine, Narr, Lencz, LaCasse and Colletti17} may have produced more specific neural correlates of a deceitful personality type.