ArticlePDF Available

Abstract and Figures

Background: Psychopathy has repeatedly been linked to disturbed associative learning from aversive events (i.e., threat conditioning). Optimal threat conditioning requires the generation of internal representations of stimulus-outcome contingencies and the rate with which these may change. Because mental representations are imperfect, there will always be uncertainty about the accuracy of representations in the brain (i.e., representational uncertainty). However, it remains unclear 1) to what extent threat conditioning is susceptible to different types of uncertainty in representations about contingencies during the acquisition phase and 2) how representational uncertainty relates to psychopathic features. Methods: A computational model was applied to functional neuroimaging data to estimate uncertainty in representations of contingencies (CoUn) and the rate of change of contingencies (RUn), respectively, from brain activation during the acquisition phase of threat conditioning in 132 adolescents at risk of developing antisocial personality profiles. Next, the associations between these two types of representational uncertainty and psychopathy-related dimensions were examined. Results: The left and right amygdala activations were associated with CoUn, while the bilateral insula and the right amygdala were associated with RUn. Different patterns of relationships were found between psychopathic features and each type of uncertainty. Callous-unemotional traits and impulsive-irresponsible traits uniquely predicted increased CoUn, while only impulsive-irresponsible traits predicted increased RUn. Conclusions: The findings suggest that 1) the insula and amygdala differ in how these regions are affected by different types of representational uncertainty during threat conditioning and 2) CoUn and RUn have different patterns of relationships with psychopathy-related dimensions.
Content may be subject to copyright.
Accepted Manuscript
Representational uncertainty in the brain during threat conditioning and the link with
psychopathic traits
Inti A. Brazil, PhD, Christoph D. Mathys, PhD, Arne Popma, MD, PhD, Sylco S.
Hoppenbrouwers, PhD., Moran D. Cohn, MD
PII: S2451-9022(17)30089-7
DOI: 10.1016/j.bpsc.2017.04.005
Reference: BPSC 152
To appear in: Biological Psychiatry: Cognitive Neuroscience and
Neuroimaging
Received Date: 23 February 2017
Revised Date: 12 April 2017
Accepted Date: 12 April 2017
Please cite this article as: Brazil I.A., Mathys C.D., Popma A., Hoppenbrouwers S.S. & Cohn M.D.,
Representational uncertainty in the brain during threat conditioning and the link with psychopathic
traits, Biological Psychiatry: Cognitive Neuroscience and Neuroimaging (2017), doi: 10.1016/
j.bpsc.2017.04.005.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
Representational uncertainty during threat conditioning and psychopathic traits
1
Representational uncertainty in the brain during threat conditioning and the link with
psychopathic traits
Inti A. Brazil, PhD
1,2,3,4
, Christoph D. Mathys, PhD
5,6,7
, Arne Popma, MD, PhD
8,9
, Sylco S.
Hoppenbrouwers, PhD.
10
, Moran D. Cohn, MD
8
1
Donders Institute for Brain, Cognition and Behaviour, Radboud University, Nijmegen, the
Netherlands
2
Forensic Psychiatric Centre Pompestichting, Nijmegen, the Netherlands
3
Collaborative Antwerp Psychiatric Research Institute (CAPRI), University of Antwerp,
Antwerp, Belgium
4
Centre for Psychology, Behaviour, & Achievement, Faculty of Health and Life Sciences,
Coventry University, UK
5
Scuola Internazionale Superiore di Studi Avanzati (SISSA), Trieste, Italy
6
Max Planck UCL Centre for Computational Psychiatry and Ageing Research, London, United
Kingdom
7
Wellcome Trust Centre for Neuroimaging, Institute of Neurology, UCL, London, United
Kingdom
8
Department of Child and Adolescent Psychiatry, VU University Medical Center Amsterdam,
Amsterdam, the Netherlands
9
Institute for Criminal Law & Criminology, Leiden University, Leiden, the Netherlands
10
Department of Clinical Psychology, Erasmus University, Rotterdam, the Netherlands
Correspondence:
Inti A. Brazil, PhD
E-mail: I.Brazil@donders.ru.nl
Word count:
Abstract: 245, Main text: 4000, tables: 0, figures: 2, supplements: 1
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
Representational uncertainty during threat conditioning and psychopathic traits
2
Abstract
Background: Psychopathy has repeatedly been linked to disturbed associative learning from
aversive events (i.e., threat conditioning). Optimal threat conditioning requires the
generation of internal representations of stimulus-outcome contingencies and the rate with
which these may change. Because mental representations are imperfect, there will always be
uncertainty about the accuracy of representations in the brain (i.e., representational
uncertainty). However, it remains unclear i) to what extent threat conditioning is susceptible
to different types of uncertainty in representations about contingencies during the
acquisition phase and ii) how representational uncertainty relates to psychopathic features.
Methods: A computational model was applied to functional neuroimaging data to estimate
uncertainty in representations of contingencies (CoUn) and the rate of change of
contingencies (RUn), respectively, from brain activation during the acquisition phase of
threat conditioning in 132 adolescents at risk of developing antisocial personality profiles.
Next, the associations between these two types of representational uncertainty and
psychopathy-related dimensions were examined.
Results: The left and right amygdala activation were associated with CoUn, while the bilateral
insula and the right amygdala were associated with RUn. Different patterns of relationships
were found between psychopathic features and each type of uncertainty. Callous-
unemotional traits and impulsive-irresponsible traits uniquely predicted increased CoUn,
while only impulsive-irresponsible traits predicted increased RUn.
Conclusions: The findings suggest that i) the insula and amygdala differ in how these regions
are affected by different types of representational uncertainty during threat conditioning,
and ii) CoUn and RUn have different patterns of relationships with psychopathy-related
dimensions.
Keywords: threat conditioning; fear conditioning; uncertainty; representations; psychopathy;
amygdala; insula; computational modelling.
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
Representational uncertainty during threat conditioning and psychopathic traits
3
Introduction
In a dynamic world, success relies heavily on our ability to adapt our behaviour to avoid
aversive outcomes. Threats have a large impact on the modulation of our behaviour (1). One
developmental condition that has repeatedly been linked to diminished learning from
aversive experiences is psychopathy (2, 3), which encompasses callous-unemotional traits
(e.g., blunted affect, lack of empathy or remorse), a grandiose-manipulative interpersonal
style (e.g., dishonesty, superficial charm, lying, manipulation of others) and impulsive-
irresponsible behavioural tendencies (e.g., thrill-seeking, lack of impulse control) (4).
Adolescents with high levels of psychopathic traits are at increased risk for engaging in
antisocial behaviour (5) and may be more difficult to treat because of their more severe
antisocial behaviour and diminished treatment responsivity (6). The maladaptive behaviour
seen in psychopathy is thought to be strongly influenced by disturbed learning from aversive
events, such as threats (i.e., threat conditioning)(7–9), which is reflected in abnormal
physiological and brain responses in both psychopathic adults (7–9) and youths with severe
antisocial personality profiles (10, 11).
Threat conditioning is multifaceted and learning relies on interacting cognitive
computations, similar to other forms of associative learning (12, 13). Learning which stimuli
are threatening requires accurate representations of threat contingencies. In order to
maintain accurate representations of contingencies we need to update the representations
continually after each observation, also taking into account that the learned contingencies
may change. However, our observations are imperfect (14). Therefore, there will always be
some uncertainty regarding the accuracy of the cognitive representations (i.e.,
‘representational uncertainty’)
1
that are generated based on these imperfect observations
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
Representational uncertainty during threat conditioning and psychopathic traits
4
(15). Learning about contingencies involves generating representations based on our
estimates of the likelihood that contingencies will change and of the rate at which these
changes occur (13). As these representations are imperfect, there will be uncertainty about
the accuracy of the representation of contingency changes (i.e., contingency uncertainty;
CoUn) and that of the rate at which changes occur (i.e., change rate uncertainty; RUn (16)).
Empirical evidence indicates that these two types of representational uncertainties are
hierarchically related, as the rate at which changes in contingencies are perceived to occur
will influence our belief about the overall likelihood that the contingencies will change (13,
15). Furthermore, a recent study has shown that beliefs about each type of uncertainty play
key roles in modulating physiological responses to threats (17). Importantly, effective
learning requires cognitive uncertainty to be minimized (18). Therefore, it is likely that
uncertainty in the representation of contingencies may also play a role in understanding the
impairments in threat conditioning seen in psychopathy.
Threat conditioning in psychopathy has often been investigated in case-control
studies in which groups of individuals scoring high on psychopathic features are compared to
a low scoring group (e.g., (7, 11)). In one of the few threat conditioning studies employing a
dimensional approach to psychopathy, Cohn et al. (19) reported a positive relationship
between BOLD activation in the amygdala and insula (during acquisition learning) and
impulsive-irresponsible traits in at-risk adolescents, but a negative correlation between
callous-unemotional traits and BOLD activation in these regions. Other studies using
physiological measures have found a similar negative relationship between threat
conditioning and fearless dominance (a construct that overlaps with callous-unemotional
traits) but no correlation with impulsive-antisociality in undergraduates (20), and reduced
threat conditioning in adult psychopathic individuals scoring high on interpersonal-affective
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
Representational uncertainty during threat conditioning and psychopathic traits
5
deficits (21). Taken together, the evidence points towards decreased threat conditioning in
individuals high on callous-unemotional traits, while the findings are mixed for the impulsive-
antisocial features. However, these prior studies have approached threat conditioning as a
unitary form of learning, without taking the multifaceted nature of associative learning into
account. As a consequence, how psychopathy may be related to any of the various
interacting cognitive computations that subserve acquisition learning during threat
conditioning has been overlooked. Systematically studying the integrity of these
computations is essential for pinpointing the deficiencies in the threat conditioning
mechanism in psychopathy.
In this study, we examined if the threat conditioning impairments seen during the
acquisition phase in relation to psychopathy may be partly attributed to increased
uncertainty in representations of contingencies and their rate of change. Importantly, as the
uncertainty computations are latent, it is impossible to directly quantify them without
employing computational modelling approaches. A computational model was applied to the
large fMRI dataset collected by Cohn et al. (19) to quantify uncertainty in the representations
of contingency change and the rate of change in target brain areas during threat
conditioning. The right and left amygdala and insula, respectively, were chosen as regions of
interest (ROIs) because these areas: i) consistently show activation during threat conditioning
across studies (22), ii) show relatively large responses to uncertainty about threat
contingencies (23, 24), iii) are linked to learning impairments in psychopathy (25), and iv) we
aimed to maintain comparability with the very few previous studies on psychopathy
dimensions and threat conditioning. Importantly, as two types of representational
uncertainties were quantified, it was unlikely that all ROIs showed an equal amount of CoUn
and Run, respectively. To take account of this, we subsequently used Bayesian structural
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
Representational uncertainty during threat conditioning and psychopathic traits
6
equation modelling (BSEM) to examine which of the ROIs was more affected by each type of
representational uncertainty, and determined how psychopathy-related personality
dimensions uniquely predicted CoUn and RU.
Extant findings suggest that threat conditioning is reduced in individuals scoring high
on callous-unemotional features, and we hypothesized that this learning impairment should
be linked to increased representational uncertainty in these individuals. This is based on the
premise that disturbed learning should be related to a failure in reducing uncertainty in the
information processed in the brain [cf. 25]. But, given that we sought to obtain a higher level
of precision by parsing representational uncertainty into CoUn and RUn for the first time, it is
difficult to predict which type of uncertainty is related to the impairments seen in those with
elevated levels of callous-unemotional traits. The findings for the impulsive-irresponsible
dimension are mixed, but the hyper-conditioning found by Cohn et al. (19) suggest that
elevated impulsive-irresponsible traits should be linked to reduced representational
uncertainty during threat conditioning.
Methods and Materials
Participants and assessment
Participants were recruited from a Dutch cohort
of 364 adolescents who were first arrested
by the Dutch police before the age of 12, all of whom had participated in three previous
waves of a longitudinal study (26). The mean age at study entrance was 10.9 (SD=1.4) years.
All participants were assessed using the Dutch Youth Psychopathic traits Inventory (YPI)(27),
a valid and reliable 50-item self-report instrument developed to assess psychopathic traits in
juvenile community samples (28). In the current study, internal consistency of the total score
and its constituting dimensions were good to excellent: Cronbach’s alpha was .93 for the
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
Representational uncertainty during threat conditioning and psychopathic traits
7
total score and .86, .89 and .85 for callous-unemotional, grandiose-manipulative and
impulsive-irresponsible scales, respectively.
For the current study, data was re-analysed from a subsample characterized by a wide
range of externalizing risk (N=150), with an even distribution of participants with low-risk,
medium-risk, and high-risk of antisocial behaviour. Eighteen participants were excluded from
analyses due to invalid MRI data (n=9); invalid task performance (n=6); drug use prior to
scanning (n=2); and missing questionnaire data (n=1). Analyses were performed on the
remaining 132 participants (mean age=17.7, SD=1.6; see Supplement)
Procedure
This study was approved by the Institutional Review Board of the VU University Medical
Center Amsterdam (VUmc). All participants and their parents/custodians (if the participant’s
age was below 18) signed for informed consent. Participants underwent a neuroimaging
protocol in a Philips 3T Intera MRI scanner at the VUmc. All participants were instructed to
refrain from using alcohol, cannabis or psychostimulant medication for at least 24 hours
before the MRI-scan.
Threat conditioning task
A classical differential delay threat conditioning task was employed (7). Pictures of two
neutral male faces served as conditioned stimuli (CS), one of which (chosen at random during
each experiment) was consistently paired with an aversive electric unconditioned stimulus
(US) at the end of a 10s viewing period (CS+; 100% reinforcement) during the acquisition
period, while the other picture (CS-) was never followed by a US. The acquisition period,
which consisted of 8 trials of each CS, was preceded by a habituation phase, in which CSs
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
Representational uncertainty during threat conditioning and psychopathic traits
8
were presented 4 times each for 3.5s without a US and was followed by an extinction phase
in which CSs were presented 4 times each for 7s and were not followed by a US either.
fMRI protocol
T1-weighted anatomical scans (180 slices, 1mm
3
voxels, FOV 256x256mm, TR 9.0ms, TE
3.5ms), were acquired using an 8-channel SENSE head-coil. Furthermore, 400 T2*weighted
axial echo-planar images (EPI) were acquired during threat-conditioning (38 slices, 3mm
thickness, 2.29x2.29 in-plane resolution, FOV 220x220mm, TR 2300ms, TE 30ms).
Statistical analyses
Functional MRI data were processed using SPM8, including realignment, unwarping, slice-
time correction, normalization to MNI space based on the segmented anatomical scan, and
8mm FWHM smoothing. First level models included separate regressors for CS+/- and CS-
acquisition, US and rating blocks. During acquisition, the first five seconds of each trial were
modeled separately from the remainder of the trial (5s) to account for fast within-trial
habituation of threat neurocircuitry, focusing analyses on the first epoch only during
acquisition (19). Realignment parameters were also included in first-level models to account
for movement effects. Next, average neural response estimates for each ROI were extracted
using the MarsBaR toolbox for SPM (29). Following previous work (23, 30), we focused
analyses on the amygdala and insula. Amygdala and insula were anatomically defined using
the Automated Anatomic Labeling atlas (31).
Computational modeling
In order to model how activity in the ROIs responded to CS in the experiment, the activation
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
Representational uncertainty during threat conditioning and psychopathic traits
9
trajectory of the principal eigenvariate was extracted for each ROI. This trajectory was then
analyzed using the HGF (15). The HGF is a generic hierarchical Bayesian time series model,
which explains a time series as a sequence of noisy observations of a hidden state. Crucially,
the HGF also models the dynamics of change (i.e., volatility) of the hidden state explicitly and
gives one-step update equations for the evolving estimate of both the hidden state and its
volatility. This evolving estimate is a probability distribution referred to by us as a belief. In
the HGF, beliefs follow a normal distribution and are fully specified by their mean and
variance. The mean of the belief represents the most probable value of the hidden state and
the variance represents the uncertainty.
More formally, the HGF consists of a hierarchy of Gaussian random walks where the
variance of each walk is a function of the value at the next higher level. Since the variance of
its walk determines a quantity’s likelihood of changing, each value but the one at the base of
the hierarchy represents the volatility of the next lower level. In the present study, we used a
two-level HGF where the first level represents the BOLD activation in a given brain region and
the second level is the volatility of that activation (Supplement, Fig.S1). We applied the HGF
to the measured BOLD activation (represented as ݑ in the HGF model graph of supplemental
Fig.S1) to infer the latent true activation at the time of measurement and its volatility along
with the posterior uncertainty about these quantities. This resulted in two estimated belief
trajectories for each subject and ROI: that of the BOLD activation (ݔ
in the HGF hierarchy, see
Fig.S1) and that of its volatility (ݔ
in the HGF hierarchy, Fig.S1). Each of the two belief
trajectories consisted of a mean trajectory (ߤ
regarding ݔ
and ߤ
regarding ݔ
) and a
variance (i.e., uncertainty) trajectory (ߪ
and ߪ
). Crucially, this means that every observed
update implies a certain level of uncertainty, and by observing the evolution of the BOLD
signal in the ROIs during threat conditioning, we can infer the evolution of the uncertainty
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
Representational uncertainty during threat conditioning and psychopathic traits
10
the brain has about the quantity encoded by the activity in each ROI. Estimation was
performed using the HGF Toolbox v3.0 (https://www.tnu.ethz.ch/en/software/tapas.html)
and fits could be obtained for all subjects and ROIs. All HGF modeling was performed before
the BSEM described below and all of the hyper-parameters of the HGF analysis (number of
levels, noise level, regularizing priors) were chosen only in terms of this analysis.
Bayesian structural equation modeling
To assess which brain areas were substantially affected by representational uncertainty
during threat conditioning in a data-driven fashion we employed BSEM to determine i) which
ROIs were involved in coding CoUn and RUn and ii) which psychopathy-related dimensions
predicted each type of representational uncertainty. First, mean uncertainty estimates for
learning were created by averaging and subtracting the uncertainty trajectories of CS- trials
from that of the CS+ trials in each ROI, yielding 4 average scores representing the ‘additional
uncertainty’ found in the CS+ relative to the CS- condition for CoUn and Run, respectively. In
particular, the CoUn score was the mean of the uncertainty (i.e., variance ߪ
) at the bottom
level of the HGF hierarchy during the CS+ condition minus the mean of the same uncertainty
ߪ
during the CS- condition. Correspondingly, the RUn score was the mean of the uncertainty
(i.e., variance ߪ
) at the second level of the HGF hierarchy during the CS+ condition minus
the mean of the same uncertainty ߪ
during the CS- condition. Note that this subtraction
method is the common approach to obtaining BOLD responses reflecting conditioning, but
that we only used the uncertainty estimate derived from the BOLD signal instead of the raw
BOLD signal. Next, a BSEM was built in which the 4 average CoUn estimates were loaded on a
latent factor, which was regressed on the scores of the callous-unemotional, grandiose-
manipulative and impulsive-irresponsible subscales of the YPI. The same procedure was
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
Representational uncertainty during threat conditioning and psychopathic traits
11
followed for the estimates of RUn.
The analyses were conducted in Mplus v7.4 (32) using a Bayesian estimator (PX1)
with five Markov chain Monte Carlo (MCMC) chains and 100.000 iterations. The first half of
the iterations was discarded (i.e., burn-in trials) and model fit was determined using different
indexes for Bayesian testing: i) a Chi-Square test for Posterior Predictive Checking and ii) the
Posterior Predictive P-value (PPP-value). Convergence of the MCMC chains was established
with Gelman-Rubin’s potential scale reduction (PSR) factor (33). In general, a good fit is
indicated by a 95% credibility interval (CI) for the Chi-square posterior predictive check that
includes the value 0, the PPP-value should approach the value 0.50 and convergence is
achieved when the PSR is below 1.05 (32). Significance of the regression weights was
determined based on the 95% CIs of the Bayesian posterior distributions and variables with
95% CIs not containing the value 0 were considered as significant.
Results
After creating the mean difference scores between the CS+ and CS- conditions for CoUn and
Run, respectively, the mean difference scores were used as outcome variables in the BSEMs.
The first BSEM included each subject’s mean CoUn estimates in the right and left amygdala
and insula, respectively, loaded on a latent factor that was in turn regressed on the YPI
subscales. This model had very good fit (95% CI -19.2225.35, PPP=0.38), and indicated that
the left and right amygdala loaded significantly on the latent factor capturing CoUn and that
callous-unemotional (β=.45, 95% CI 0.210.72) and impulsive-irresponsible scales were
significant positive predictors (β=.27, 95% CI 0.050.51) of the latent factor (Figure 1). The
grandiose-manipulative scale was not a significant predictor (β=.10, 95% CI -0.150.35). This
procedure was repeated for change rate uncertainty in the 4 ROIs, loaded on a latent factor
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
Representational uncertainty during threat conditioning and psychopathic traits
12
measuring RUn that was regressed on the 3 YPI scales. This model also had a very good fit
(95% CI -27.5217.17, PPP=0.67). The results showed that the left and right insula and the
right amygdala loaded significantly on the latent factor and that the impulsive-irresponsible
scale predicted the latent variable for RUn (β=.29, 95% CI 0.070.49) (Figure 2). The callous-
unemotional (β=-.02, 95% CI -0.250.21) and grandiose-manipulative (β= -.12, 95% CI -
0.350.12) scales were not significant predictors.
Discussion
This study is the first to provide direct quantifications of different types of representational
uncertainty in the brain during threat conditioning in a large sample of adolescents at risk of
developing persistent antisocial behavior. The findings show that a significant level of CoUn
can be found in the left and right amygdala during threat conditioning, while the right
amygdala, left and right insula are more responsive to RUn. The mapping of these brain
responses onto different dimensions of psychopathy indicated that both callous-unemotional
and impulsive-irresponsible traits were uniquely related to CoUn, while only impulsive-
irresponsible features were positively linked to RUn (Figures 1-2).
Our results in an at-risk population are in line with those obtained in healthy samples
indicating that the amygdala (23, 34) and the insula (24) are involved in processing
uncertainty related to threat contingencies. Our results significantly advance this knowledge
by specifying how activation in the insula and amygdala is related to uncertainty about the
accuracy of different aspects of contingency representation. The amygdala seems to be
particularly sensitive to CoUn during threat conditioning, but also encodes a relatively high
amount of RUn together with the insula. Given the hierarchical relationship between RUn
and CoUn in our computational framework (13, 15), these results converge with the
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
Representational uncertainty during threat conditioning and psychopathic traits
13
suggestion that the insula passes information concerning aversive stimuli to the amygdala
during threat conditioning (35). One tentative interpretation of the general pattern of results
is that the interaction between the insula and the amygdala during threat conditioning might
reflect a circuit in which the insula is primarily responsive to RUn and the information is then
relayed to the right amygdala, which would then function as a point of entry, to inform
computations taking place in the right and left amygdala that are related to the estimation of
the likelihood that contingences changes occur. This proposal would also be consistent with
the general notion that the insula relays somatosensory information to other areas to initiate
adaptive responses (36).
Regarding the link with psychopathy, callous-unemotional and impulsive-irresponsible
traits predicted increased CoUn in the amygdala. Because associative learning relies on
successful reduction of uncertainty (15), heightened levels of uncertainty in the cognitive
computations that are engaged should ultimately lead to reduced learning (18). In
agreement with this prediction, Cohn et al. (19) reported a negative relationship between
BOLD activation and callous-unemotional traits during threat conditioning in the dataset
used in the present study, pointing towards mechanistic disturbances in threat conditioning
(20, 21). Our findings further specify one aspect of the mechanism that seems impaired in
individuals with high levels of callous-unemotional traits, who seem to form more uncertain
(i.e., less accurate) representations of contingency changes in the amygdala, while
representations of change rate in the insula are relatively accurate.
Cohn et al. [18] also found a positive unique correlation between impulsive-
irresponsible traits and BOLD signal during threat conditioning. This positive relationship,
indicating enhanced learning from threats in those with increased levels of impulsive-
irresponsible traits, was expected to be related to reduced representational uncertainty (i.e.,
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
Representational uncertainty during threat conditioning and psychopathic traits
14
more accurate representations) in the present study. Instead, we found that impulsive-
irresponsible traits were positively linked to increased RUn and CoUn in contingency
representations. Given these results, it seems plausible that there is a broader deficiency in
forming representations concerning change in individuals scoring high on impulsive-
irresponsible traits during threat conditioning, while the deficiency is limited to CoUn in
those scoring high on callous-unemotional traits. However, representational uncertainty does
not seem to be sufficient for explaining the enhanced threat conditioning seen with elevated
impulsive-irresponsible traits (19). Therefore, it should be considered whether an additional
mechanism could be interacting with the threat conditioning processes in this sample of at-
risk adolescents. One possibility is that the insula and amygdala are hyper-sensitive to
aversive information in impulsive-irresponsible individuals. This interpretation builds on the
previous finding that higher perceived uncertainty sensitizes the insula and amygdala to
aversive information (23, 24, 34), presumably leading to exaggerated aversive responding in
these regions. Thus, more representational uncertainty during aversive learning may be
interacting with a bias toward exaggerated affective responding in the amygdala in
individuals with high levels of irresponsible-impulsive features, while this bias may not be
present in individuals with elevated callous-unemotional traits. One tentative prediction that
follows is that impulsive-irresponsible individuals should also show exaggerated responses
during extinction learning, as the hyper-sensitization of the amygdala after threat
conditioning combined with excessive representational uncertainty should interfere with the
unlearning of contingencies during extinction. Obviously, this proposal is made with care as
we did not quantify amygdala bias in the present study and also because threat extinction
has yet to be studied using a dimensional approach to psychopathy instead of group
comparisons [11,37].
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
Representational uncertainty during threat conditioning and psychopathic traits
15
Interestingly, our findings also suggest that representational uncertainty could play an
important role in explaining other learning impairments found in antisocial populations, such
as disturbed passive avoidance (38, 39) and reversal learning (40, 41). During reversal
learning, for example, we learn that a stimulus previously associated with reward now may
lead to negative outcomes, which requires us adapt our beliefs and behaviour to avoid
negative consequences. Importantly, the change in stimulus-outcome associations introduces
representational uncertainty about the contingencies and their rate of change, which affects
how well and how fast we learn the new contingencies. From this perspective, the reversal
learning impairment found in children (41) and adults (3, 40, 42) with psychopathic
tendencies could reflect sub-optimal management of representational uncertainty, similar to
what we found in the present study. In line with this prediction, Budhani and Blair (41) found
that the reversal learning impairment in boys with psychopathic tendencies got worse as the
saliency of the contingency changes decreased. Such a reduction of saliency increases
ambiguity and uncertainty about the contingencies, so their results suggest that increased
CoUn may play a significant role in the reversal impairment. Future studies should try to
confirm this expectation and determine the impact of representational uncertainty on
reversal learning.
One critique to the current study could be that our task did not include a
manipulation of uncertainty, as the aversive stimulus was always associated with the same
neutral face during learning, which would make it clear when to expect the aversive stimulus.
This argument stems from the issue that quantifying the impact of uncertainty on threat
conditioning is not possible using traditional analytical approaches, thus requiring the
manipulation of uncertainty through the experimental task in blocked designs. However, the
computational model used in the present study overcomes this limitation, as it directly
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
Representational uncertainty during threat conditioning and psychopathic traits
16
quantifies the level of uncertainty about contingencies on a trial-by-trial basis. Therefore, the
impact of uncertainty on the cognitive operations involved in learning can be measured
without introducing drastic changes in contingencies through task design. Another potential
issue is the possibility that the increased RUn in the insula found in those with elevated
impulsive-irresponsible tendencies could be driven by a more general impairment in the
integration of multiple sources of information used to generate contingency representations
during threat conditioning (43, 44). The insula can be seen as a processing hub implicated in
various cognitive operations that include representations of pain, contextual appraisal and
general uncertainty (36). Thus, it is possible that increased RUn is a consequence of
inaccurate lower level representations in the insula, such as those pertaining to the pain
stimulus. Such an account would also be in line with studies on reinforcement-based
decision-making in adolescents with conduct problems, who show impairments in generating
accurate representations of expected value in the insula (39, 45). Together, these caveats and
novel hypotheses highlight the need for further studies that focus on the interaction
between personality dimensions, representational uncertainty, and their neural correlates
during threat conditioning and other forms of associative learning.
In conclusion, the present study is the first to directly quantify different kinds of
representational uncertainty during threat conditioning in an at-risk sample of adolescents.
The results highlight the importance of examining how uncertainty in cognitive
representations may be key to understanding some of the maladaptive characteristics often
linked to psychopathy. A more precise understanding of the various interacting cognitive
computations involved in maladaptive learning may lead to better, (neuro)biology-oriented
diagnostics and the development of targeted treatment approaches in various conditions
showing disturbed associative learning (46, 47).
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
Representational uncertainty during threat conditioning and psychopathic traits
17
Acknowledgements
MDC, AP and IAB were supported by a Mosaic (017.007.022), Brain and Cognition
(056.23.010) and VENI Grant (451.15.014), respectively, from Netherlands Organization for
Scientific Research. The authors thank Professor Essi Viding for her valuable input.
Financial Disclosures
All authors report no biomedical financial interests or potential conflicts of interest.
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
Representational uncertainty during threat conditioning and psychopathic traits
18
References
1. LeDoux JE (2014): Coming to terms with fear. Proceedings of the National Academy of
Sciences. 111: 2871–2878.
2. Blair RJR, Mitchell DGV, Leonard A, Budhani S, Peschardt KS, Newman C (2004): Passive
avoidance learning in individuals with psychopathy: modulation by reward but not by
punishment. Personality and Individual Differences. 37: 1179–1192.
3. Gregory S, Blair RJR, Ffytche DH, Simmons A, Kumari V, Hodgins S, Blackwood N (2015):
Punishment and the psychopath: an fMRI investigation of reinforcement learning in
violent antisocial personality disordered men. The Lancet Psychiatry. 2: 153–160.
4. Andershed HA, Kerr M, Stattin H, Levander S (2002): Psychopathic traits in non-referred
youths: A new assessment tool. .
5. Hawes SW, Byrd AL, Waller R, Lynam DR, Pardini DA (2016): Late childhood interpersonal
callousness and conduct problem trajectories interact to predict adult psychopathy.
Journal of Child Psychology and Psychiatry. n/a-n/a.
6. Wilkinson S, Waller R, Viding E (2016): Practitioner Review: Involving young people with
callous unemotional traits in treatment – does it work? A systematic review. Journal
of Child Psychology and Psychiatry. 57: 552–565.
7. Birbaumer N, Veit R, Lotze M, Erb M, Hermann C, Grodd W, Flor H (2005): Deficient fear
conditioning in psychopathy: a functional magnetic resonance imaging study. Archives
of General Psychiatry. 62: 799–805.
8. Flor H, Birbaumer N, Hermann C, Ziegler S, Patrick CJ (2002): Aversive Pavlovian
conditioning in psychopaths: peripheral and central correlates. Psychophysiology. 39:
505–18.
9. Hoppenbrouwers SS, Bulten BH, Brazil IA (2016): Parsing Fear: A Reassessment of the
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
Representational uncertainty during threat conditioning and psychopathic traits
19
Evidence for Fear Deficits in Psychopathy. Psychological Bulletin. 142: 573–600.
10. Fairchild G, Stobbe Y, van Goozen SHM, Calder AJ, Goodyer IM (2010): Facial Expression
Recognition, Fear Conditioning, and Startle Modulation in Female Subjects with
Conduct Disorder. Biological Psychiatry. 68: 272–279.
11. Fairchild G, Van Goozen SH, Stollery SJ, Goodyer IM (2008): Fear Conditioning and
Affective Modulation of the Startle Reflex in Male Adolescents with Early-Onset or
Adolescence-Onset Conduct Disorder and Healthy Control Subjects. Biological
Psychiatry. 63: 279–285.
12. Behrens TEJ, Hunt LT, Rushworth MFS (2009): The computation of social behavior.
Science. 324: 1160–1164.
13. Mathys C, Daunizeau J, Friston KJ, Stephan KE (2011): A Bayesian foundation for
individual learning under uncertainty. Frontiers in human neuroscience. 5: 39.
14. Jazayeri M, Movshon JA (2006): Optimal representation of sensory information by neural
populations. Nat Neurosci. 9: 690–696.
15. Mathys CD, Lomakina EI, Daunizeau J, Iglesias S, Brodersen KH, Friston KJ, Stephan KE
(2014): Uncertainty in perception and the Hierarchical Gaussian Filter. Frontiers in
Human Neuroscience. 8: 825.
16. Behrens TEJ, Woolrich MW, Walton ME, Rushworth MFS (2007): Learning the value of
information in an uncertain world. Nature Neuroscience. 10: 1214–1221.
17. de Berker AO, Rutledge RB, Mathys C, Marshall L, Cross GF, Dolan RJ, Bestmann S (2016):
Computations of uncertainty mediate acute stress responses in humans. Nat
Commun. 7. Retrieved from http://dx.doi.org/10.1038/ncomms10996.
18. Cohn MD, Popma A, van den Brink W, Pape LE, Kindt M, van Domburgh L, et al. (2013):
Fear conditioning, persistence of disruptive behavior and psychopathic traits: an fMRI
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
Representational uncertainty during threat conditioning and psychopathic traits
20
study. Transl Psychiatry. 3: e319.
19. Lopez R, Poy R, Patrick CJ, Molto J (2013): Deficient fear conditioning and self-reported
psychopathy: the role of fearless dominance. Psychophysiology. 50: 210–8.
20. Veit R, Konicar L, Klinzing JG, Barth B, Yilmaz O, Birbaumer N (2013): Deficient fear
conditioning in psychopathy as a function of interpersonal and affective disturbances.
Front Hum Neurosci. 7: 706.
21. Sehlmeyer C, Schöning S, Zwitserlood P, Pfleiderer B, Kircher T, Arolt V, Konrad C (2009):
Human fear conditioning and extinction in neuroimaging: a systematic review. PloS
one. 4: e5865.
22. Herry C, Bach DR, Esposito F, Di Salle F, Perrig WJ, Scheffler K, et al. (2007): Processing of
Temporal Unpredictability in Human and Animal Amygdala. The Journal of
Neuroscience. 27: 5958–5966.
23. Sarinopoulos I, Grupe DW, Mackiewicz KL, Herrington JD, Lor M, Steege EE, Nitschke JB
(2009): Uncertainty during Anticipation Modulates Neural Responses to Aversion in
Human Insula and Amygdala. Cerebral Cortex. . doi: 10.1093/cercor/bhp155.
24. Blair RJR (2013): The neurobiology of psychopathic traits in youths. Nature Reviews
Neuroscience. 14: 786–799.
25. Friston K (2010): The free-energy principle: a unified brain theory? Nature Reviews
Neuroscience. 11: 127–138.
26. van Domburgh L, Vermeiren R, Blokland AAJ, Doreleijers TAH (2009): Delinquent
Development in Dutch Childhood Arrestees: Developmental Trajectories, Risk Factors
and Co-morbidity with Adverse Outcomes during Adolescence. J Abnorm Child
Psychol. 37: 93–105.
27. van Baardewijk Y, Stegge H, Andershed H, Thomaes S, Scholte E, Vermeiren R (2008):
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
Representational uncertainty during threat conditioning and psychopathic traits
21
Measuring psychopathic traits in children through self-report. The development of
the Youth Psychopathic traits Inventory—Child Version. International Journal of Law
and Psychiatry. 31: 199–209.
28. Andershed H, Hodgins S, Tengström A (2007): Convergent Validity of the Youth
Psychopathic Traits Inventory (YPI): Association With the Psychopathy Checklist: Youth
Version (PCL:YV). Assessment. 14: 144–154.
29. Brett M, Anton J, Valabregue R, Poline JB (2002): Region of interest analysis using the
MarsBar toolbox for SPM 99. Neuroimage. 16: S497.
30. Phelps EA, O’Connor KJ, Gatenby JC, Gore JC, Grillon C, Davis M (2001): Activation of the
left amygdala to a cognitive representation of fear. Nat Neurosci. 4: 437–441.
31. Tzourio-Mazoyer N, Landeau B, Papathanassiou D, Crivello F, Etard O, Delcroix N, et al.
(2002): Automated Anatomical Labeling of Activations in SPM Using a Macroscopic
Anatomical Parcellation of the MNI MRI Single-Subject Brain. NeuroImage. 15: 273–
289.
32. Muthén LK, Muthén BO (2012): Mplus. The comprehensive modelling program for
applied researchers: User’s guide. 5.
33. Gelman A, Rubin DB (1992): Inference from iterative simulation using multiple
sequences. Statistical science. 457–472.
34. Whalen PJ (2007): The uncertainty of it all. Trends in Cognitive Sciences. 11: 499–500.
35. Hartley CA, Fischl B, Phelps EA (2011): Brain Structure Correlates of Individual Differences
in the Acquisition and Inhibition of Conditioned Fear. Cerebral Cortex. 21: 1954–1962.
36. Singer T, Critchley HD, Preuschoff K (2009): A common role of insula in feelings, empathy
and uncertainty. Trends in Cognitive Sciences. 13: 334–340.
37. Cohn MD, van Lith K, Kindt M, Pape LE, Doreleijers TAH, van den Brink W, et al. (2015):
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
Representational uncertainty during threat conditioning and psychopathic traits
22
Fear extinction, persistent disruptive behavior and psychopathic traits: fMRI in late
adolescence. Social Cognitive and Affective Neuroscience. . doi: 10.1093/scan/nsv067.
38. Newman JP, Patterson CM, Howland EW, Nichols SL (1990): Passive avoidance in
psychopaths: The effects of reward. Personality and Individual Differences. 11: 1101–
1114.
39. White SF, Pope K, Sinclair S, Fowler KA, Brislin SJ, Williams WC, et al. (2013): Disrupted
expected value and prediction error signaling in youths with disruptive behavior
disorders during a passive avoidance task. American Journal of Psychiatry. 170: 315–
323.
40. Brazil IA, Maes JHR, Scheper I, Bulten BH, Kessels RPC, Verkes R, de Bruijn ERA (2013):
Reversal deficits in individuals with psychopathy in explicit but not implicit learning
conditions. Journal of psychiatry & neuroscience. 38: E13–E20.
41. Budhani S, Blair RJR (2004): Response reversal and children with psychopathic
tendencies: success is a function of salience of contingency change. Journal of Child
Psychology and Psychiatry. 46: 972–981.
42. Budhani S, Richell RA, Blair RJR (2006): Impaired reversal but intact acquisition:
probabilistic response reversal deficits in adult individuals with psychopathy. Journal
of Abnormal Psychology. 115: 552–558.
43. Hoppenbrouwers SS, Van der Stigchel S, Sergiou CS, Theeuwes J (2016): Top-down
attention and selection history in psychopathy: Evidence from a community sample.
Journal of Abnormal Psychology. 125: 435–441.
44. Hoppenbrouwers SS, Van der Stigchel S, Slotboom J, Dalmaijer ES, Theeuwes J (2015):
Disentangling attentional deficits in psychopathy using visual search: Failures in the
use of contextual information. Personality and Individual Differences. 86: 132–138.
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
Representational uncertainty during threat conditioning and psychopathic traits
23
45. White SF, Tyler PM, Erway AK, Botkin ML, Kolli V, Meffert H, et al. (2016): Dysfunctional
representation of expected value is associated with reinforcement-based decision-
making deficits in adolescents with conduct problems. Journal of Child Psychology
and Psychiatry. 57: 938–946.
46. Stephan KE, Iglesias S, Heinzle J, Diaconescu AO (2015): Translational Perspectives for
Computational Neuroimaging. Neuron. 87: 716–732.
47. Brazil IA, van Dongen JDM, Maes JHR, Mars RB, Baskin-Sommers AR (n.d.): Classification
and treatment of antisocial individuals: From behavior to biocognition. Neuroscience
& Biobehavioral Reviews. . doi: 10.1016/j.neubiorev.2016.10.010.
48. Kuhnen CM, Knutson B (2005): The Neural Basis of Financial Risk Taking. Neuron. 47:
763–770.
49. Liu X, Powell DK, Wang H, Gold BT, Corbly CR, Joseph JE (2007): Functional Dissociation in
Frontal and Striatal Areas for Processing of Positive and Negative Reward Information.
J Neurosci. 27: 4587–4597.
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
Representational uncertainty during threat conditioning and psychopathic traits
24
Footnote
1
Note that the operationalization of uncertainty used in this study differs from other
definitions of the term. Neuroimaging studies have primarily focused on how uncertainty
due to unsure outcomes is processed in the brain (e.g., (48, 49)), while we examined the
uncertainty about the accuracy of cognitive representations of change. In other words, our
definition of uncertainty in this study refers to the inaccuracy in representations that are
formed, which is driven by imperfect observations in addition to uncertain outcomes.
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
Representational uncertainty during threat conditioning and psychopathic traits
25
Figures
Figure 1. Structural Equation model depicting the relationships between the YPI scales
measuring callous-unemotional (YPI-CU), grandiose-manipulative (YPI-GM), impulsive-
irresponsible (YPI-II) traits and a latent factor representing the amount of perceptual
uncertainty concerning contingencies (CoUn), as estimated from the BOLD signal trajectories.
Only the left amygdala (L-Amy) and the right amygdala (R-Amy) loaded significantly on this
latent factor. To increase readability, non-significant loadings are not depicted. Solid arrows
represent significant unique correlations, dashed arrows represent non-significant effects,
and dotted arrows represent factor loadings.
Figure 2. Structural Equation model depicting the relationships between the YPI scales
measuring callous-unemotional (YPI-CU), grandiose-manipulative (YPI-GM), impulsive-
irresponsible (YPI-II) traits and a latent factor representing perceptual uncertainty concerning
change rate of contingencies (RUn), as estimated from the BOLD signal trajectories. The left
insula (L-Ins), right insula (R-Ins) and right amygdala (R-Amy) loaded significantly on this
latent factor. To increase readability, non-significant loadings are not depicted. Solid arrows
represent significant unique correlations, dashed arrows represent non-significant effects,
and dotted arrows represent factor loadings.
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
!"#$%&
!"#$'(
!"#$##
')*+,*-.*/01
(*/.2+3,*+0
4$560
7$560
899:
8;<9
89:9
8:<=
8==;
8>?<
8@<A
8>=<
B,-C2.1:8
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
!"#$%&
!"#$'(
!"#$##
')*+,-.
/*0-
(+1-20*3+04
/$564
/$#+7
899:
8;<9
89:9
$8::;
$8<==
8=>?
89@@
8>;;
A$#+7
8B9C
D3,E2-.=8
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
Brazil et al. Supplement
1
Representational Uncertainty in the Brain During Threat Conditioning
and the Link With Psychopathic Traits
Supplemental Information
Additional information about subjects (1)
The 132 subjects (86% male) consisted of individuals from 3 groups, with an even distribution of
subjects with low-risk, medium-risk, and high-risk of antisocial behavior. These subsamples
were selected in order to increase the likelihood that the scanning sample would reflect the full
spectrum of behavioral functioning. There was a i) low-risk group without a diagnosis of
Disruptive Behavior Disorder (DBD), Oppositional Defiant Disorder or Conduct Disorder in the
three previous waves according to the DISC-IV and below median scores on aggression (RPQ)
and callous-unemotional traits (YPI) in the 3 previous waves (n=34 out of the original sub-
sample of 110); ii) a medium-risk group with above-median scores on aggression and callous-
unemotional traits in the previous waves but no previous diagnosis of DBD (n=52 out of 174);
and, iii) a high-risk group with a previous DBD diagnosis (n=46 out of 80).
Computational model
The nature of each trajectory was determined by five estimated parameters (see Table S1
below). Those were the initial variances of the two trajectories, their volatilities, and the initial
value of the activation trajectory. Since the measurement scale of volatility was arbitrary in our
case, the initial value of the volatility trajectory could be set to 1 without loss of generality (for
details see Appendix F in the publication by (2)).
Estimation of the trajectories took place by finding the maximum-a-posteriori (MAP)
value of the HGF parameters that govern them using the Broyden-Fletcher-Goldfarb-Shanno
(BFGS) algorithm. MAP estimation entails maximizing the log-likelihood under regularization by
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
Brazil et al. Supplement
2
appropriate priors (see Table S1). Parameters were estimated in logarithmic space if they had a
lower bound such that Gaussian priors could be chosen in all cases.
Table S1. Initial values of the estimated parameters.
Parameter Estimation Space Mean Variance
Initial value of BOLD
activation trajectory Native 0 log(3)=1.1
Initial variance of BOLD
activation trajectory Logarithmic 1 1
Initial variance of phasic
volatility trajectory Logarithmic 0.1 1
Tonic volatility of BOLD
activation Native -1 1
Meta-volatility Native -4 16
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
Brazil et al. Supplement
3
Figure S1. Graphical depiction of the HGF model used. 𝑘: observation index. 𝑢: observed
BOLD signal. 𝑥1: true hidden neuronal activation level. 𝑥2: volatility of activation level. 𝜔1: tonic
volatility of 𝑥1. 𝜔2: tonic volatility of 𝑥2. Observation noise: 𝛼=log 2. Hexagons represent
quantities that depend on their own previous state, diamonds represent quantities that do not
depend on their own previous state, and circles represent constants.
MANUS CRIP T
ACCEP TED
ACCEPTED MANUSCRIPT
Brazil et al. Supplement
4
Supplemental References
1. Cohn MD, et al. (2013) Fear conditioning, persistence of disruptive behavior and
psychopathic traits: an fMRI study. Transl Psychiatry 3:e319.
2. Mathys CD, et al. (2014) Uncertainty in perception and the Hierarchical Gaussian Filter.
Frontiers in Human Neuroscience 8:825.
... In most individuals, the prospect of inflicting harm is enough to prevent behaviors resulting in such outcomes (Blair, 2007(Blair, , 2013. However, research on psychopathy in both offender and non-offender samples has shown that the processing of various types of negative outcomes is disrupted in individuals with relatively high levels of psychopathic traits (von Borries et al., 2010;Brazil et al., 2017), ultimately affecting how they learn from social encounters (Blair, 2007(Blair, , 2013Brazil et al., 2011). People with elevated levels of psychopathy exhibit multiple impairments in moral decision-making (Driessen et al., 2021), including greater utilitarian thinking on sacrificial moral dilemmas, and a tendency to pursue personal advantage even when it causes pain to others in some way (Pletti et al., 2017). ...
... The first half of the iterations were considered burnin trials to train the model and were excluded. Model fit was determined based on multiple Bayesian indexes: (i) The posterior predictive p-value (PPP value), which should approach the value 0.5, (ii) a posterior predictive check using χ 2 testing, which indicates good model fit when the 95% CI interval of the χ 2 test includes the value 0, and (iii) convergence of the MCMC chains based on the proportional scale reduction (PSR) factor, which should have a value <1.05 (Brazil et al., 2017;Muthén and Muthén, 2017). The significance of the regression weights within the model was determined based on the 95% CI, which should not contain the value of exactly 0. Both direct and indirect effects were tested. ...
... Thus, it seems plausible that individuals scoring higher on the lifestyle facet are systematically biased towards favoring positive outcomes with high hedonic value while downplaying the subjective value of aversive outcomes, including distress. If true, the bias could help explain findings linking psychopathy to reduced learning from painful outcomes (e.g., Hare, 1965;Birbaumer et al., 2005;Brazil et al., 2017). However, this claim should be treated with caution, as it still remains to be determined how the trade-off between subjective valuation of reward and pain affect learning and decision-making in psychopathy. ...
Article
Full-text available
Psychopathy is a multifaceted personality construct entailing interpersonal-affective disturbances, antisocial traits, and a tendency to lead an erratic lifestyle. Elevated levels of psychopathic traits have been linked to having an altered experience of pain, reduced responsivity to distress in others, and making poor moral choices that bring harm to others. In the context of moral decision-making, it is possible that the capacity to estimate the distress felt by others is linked to a limitation in the first-hand experience of distress, as the presence of psychopathic traits increases. We employed a model-based approach in a non-offender sample ( n = 174) to investigate whether pain-related distress mediated the links between facets of psychopathy and estimates of the pain distress potentially experienced by others. Participants judged the permissibility of moral dilemmas and rated how much pain distress they would experience while making such judgements, as well as how much pain distress they believed the “victims” would feel as a result of the moral choice made by the participant. We found that ratings of own pain distress predicted beliefs about the distress others may experience, and elevated scores on the lifestyle facet of psychopathy uniquely predicted lower estimates of own pain distress. Furthermore, own pain distress mediated the relationship between the lifestyle facet and beliefs about others’ distress. Finally, exploratory zero-order correlation analyses revealed that ratings of own pain distress decreased as the scores on multiple psychopathic traits increased. Only the lifestyle facet correlated in the negative direction with beliefs about others’ distress. Taken together, our findings suggest that beliefs about how much pain distress others may experience is indeed mediated by own pain distress, and that the tendency to lead an erratic lifestyle is linked to alterations in this mechanism.
... All else equal, when volatility is higher, the organism is more uncertain about the cue's value (because the true value will on average have fluctuated more following each observation), and so the learning rate (the reliance on each new outcome) should be higher. A series of experiments have reported behavioral and neural signatures of these effects of volatility enhancing learning rate, and also their disruption in relation to psychiatric symptoms 6,8,10,[13][14][15][16][17][18][19][20][21][22][23] . Conversely, stochasticity also affects the learning rate, but in the opposite direction: all else equal, when individual outcomes are more stochastic (larger stochasticity), they are less informative about the cue's true value and the learning rate, in turn, should be smaller. ...
... The observation that experienced noise can, to a first approximation, be explained by either volatility or stochasticity-and that these effects might be confused, either by experimenters or by learners-has implications. First, previous work apparently showing variation in volatility processing in different groups, such as various psychiatric patients 14,16,17,19,21,22,27,28 (using a model and tasks that do not vary stochasticity), might instead reflect misidentified abnormalities in processing stochasticity. We suggest that future research should test both the dimensions of learning explicitly. ...
... Our work builds most directly on a rich line of theoretical and experimental work on the relationship between the volatility and learning rates 6,8,13,15,27,68,69 . There have been numerous reports of volatility effects on healthy and disordered behavioral and neural responses, often using a two-level manipulation of volatility like that from Fig. 5a 6,8,10,[13][14][15][16][17][18][19][20][21][22][23]38 . Our modeling suggests that it will be informative to drill deeper into these effects by augmenting this task to cross this manipulation with stochasticity so as more clearly to differentiate these two potential contributors 24 . ...
Article
Full-text available
Previous research has stressed the importance of uncertainty for controlling the speed of learning, and how such control depends on the learner inferring the noise properties of the environment, especially volatility: the speed of change. However, learning rates are jointly determined by the comparison between volatility and a second factor, moment-to-moment stochasticity. Yet much previous research has focused on simplified cases corresponding to estimation of either factor alone. Here, we introduce a learning model, in which both factors are learned simultaneously from experience, and use the model to simulate human and animal data across many seemingly disparate neuroscientific and behavioral phenomena. By considering the full problem of joint estimation, we highlight a set of previously unappreciated issues, arising from the mutual interdependence of inference about volatility and stochasticity. This interdependence complicates and enriches the interpretation of previous results, such as pathological learning in individuals with anxiety and following amygdala damage.
... All else equal, when volatility is higher, the organism is more uncertain about the cue's value (because the true value will on average have fluctuated more following each observation), and so the learning rate (the reliance on each new outcome) should be higher. A series of experiments have reported behavioral and neural signatures of these volatility effects, and also their disruption in relation to psychiatric symptoms (Behrens et al., 2007;Brazil et al., 2017;Browning et al., 2015;Cole et al., 2020;Deserno et al., 2020;Diaconescu et al., 2020;Farashahi et al., 2017;Iglesias et al., 2013;Katthagen et al., 2018;Lawson et al., 2017;Paliwal et al., 2019;Piray et al., 2019;Powers et al., 2017;Soltani and Izquierdo, 2019). However, volatility is only one of two noise parameters in the underlying Kalman filter; the second is unpredictability, which controls how noisy are each of the outcomes (the width of the likelihood) individually. ...
... Our work builds directly on a rich line of theoretical and experimental work on the relationship between volatility and learning rates (Behrens et al., 2007;de Berker et al., 2016;Browning et al., 2015;Diaconescu et al., 2014;Farashahi et al., 2017;Iglesias et al., 2013;Khorsand and Soltani, 2017). There have been numerous reports of volatility effects on healthy and disordered behavioral and neural responses, often using a two-level manipulation of volatility like that from Figure 2 (Behrens et al., 2007;Brazil et al., 2017;Browning et al., 2015;Cole et al., 2020;Deserno et al., 2020;Diaconescu et al., 2020;Farashahi et al., 2017;Iglesias et al., 2013;Katthagen et al., 2018;Lawson et al., 2017;Paliwal et al., 2019;Piray et al., 2019;Powers et al., 2017;Pulcu and Browning, 2017;Soltani and Izquierdo, 2019). Our modeling suggests that it will be informative to drill deeper into these effects by augmenting this task to cross this manipulation with unpredictability so as more clearly to differentiate these two potential contributors. ...
... Abnormalities in uncertainty and inference, broadly, have been hypothesized to play a role in numerous disorders, including especially anxiety and schizophrenia. Abnormalities in volatility-related learning adjustments have also been reported in patients or people reporting symptoms of several mental illnesses (Brazil et al., 2017;Browning et al., 2015;Cole et al., 2020;Deserno et al., 2020;Diaconescu et al., 2020;Katthagen et al., 2018;Lawson et al., 2017;Paliwal et al., 2019;Piray et al., 2019;Powers et al., 2017;Pulcu and Browning, 2017). The current model provides a more detailed framework for better dissecting these effects, though this will ideally require a new generation of experiments manipulating both factors. ...
Preprint
Full-text available
Influential research in computational neuroscience has stressed the importance of uncertainty for controlling the speed of learning, and of volatility, i.e. the inferred rate of change, in this process. Here, we investigate a neglected feature of these models: learning rates are jointly determined by the comparison between volatility and a second factor, unpredictability, which reflects moment-to-moment stochasticity. Like volatility, unpredictability can vary and must be estimated by the learner, but much previous research has focused on estimation of volatility while unpredictability is assumed fixed and known. We introduce a new learning model, in which both factors are learned from experience. We show evidence from behavioral neuroscience that the brain distinguishes these two factors and adjusts the learning rate accordingly. The model highlights the interdependency in inferences about volatility and unpredictability, which leads it to paradoxical compensatory behaviors if inference about either factor is damaged. This provides a novel mechanism for understanding pathological learning in amygdala damage and anxiety disorders.
... Psychopathic individuals might take more time to fully process the faces as a result of the exaggerated bottleneck, at the expense of allocating sufficient resources to properly process the threatening shock that follows. Such an impairment would make it more challenging to accurately extract and accumulate all the information required to learn about how the face and the shock may be associated [85], ultimately resulting in diminished learning from threats. ...
Article
The psychopath has long captured the imagination. A name such as Ted Bundy evokes a morbid curiosity. The crimes committed by Bundy are so cruel that it is hard to imagine how someone could do such things. In this review we discuss evidence that exaggeration in an attention bottleneck is one mechanism that makes it possible for psychopathic individuals to be adept at focusing on a single stimulus feature or goal but struggle to process multiple streams of information simultaneously. This exaggeration may partly explain the behavioral, affective, and social deficits that are apparent among psychopathic individuals. Further research on this attentional mechanism may promote a science that adequately captures the complexity of psychopathic behavior and offers new avenues for intervention.
... The HGF can also be applied to other time series than behavioral ones. In an example from TN/CP, Brazil et al. (219) applied an HGF directly to BOLD signal time series from an fMRI experiment. ...
Article
Full-text available
Psychiatry faces fundamental challenges with regard to mechanistically guided differential diagnosis, as well as prediction of clinical trajectories and treatment response of individual patients. This has motivated the genesis of two closely intertwined fields: (i) Translational Neuromodeling (TN), which develops “computational assays” for inferring patient-specific disease processes from neuroimaging, electrophysiological, and behavioral data; and (ii) Computational Psychiatry (CP), with the goal of incorporating computational assays into clinical decision making in everyday practice. In order to serve as objective and reliable tools for clinical routine, computational assays require end-to-end pipelines from raw data (input) to clinically useful information (output). While these are yet to be established in clinical practice, individual components of this general end-to-end pipeline are being developed and made openly available for community use. In this paper, we present the Translational Algorithms for Psychiatry-Advancing Science (TAPAS) software package, an open-source collection of building blocks for computational assays in psychiatry. Collectively, the tools in TAPAS presently cover several important aspects of the desired end-to-end pipeline, including: (i) tailored experimental designs and optimization of measurement strategy prior to data acquisition, (ii) quality control during data acquisition, and (iii) artifact correction, statistical inference, and clinical application after data acquisition. Here, we review the different tools within TAPAS and illustrate how these may help provide a deeper understanding of neural and cognitive mechanisms of disease, with the ultimate goal of establishing automatized pipelines for predictions about individual patients. We hope that the openly available tools in TAPAS will contribute to the further development of TN/CP and facilitate the translation of advances in computational neuroscience into clinically relevant computational assays.
... They present evidence that when encountering a more volatile set of associations, people will adjust by learning faster, effectively adjusting to using a narrower set of recent experiences to make their decisions about the likelihood of an upcoming event. Similarly, Brazil, Mathys, Popma, Hoppenbrouwers, and Cohn (2017) hypothesised that learners develop mental representations not only of the contingencies between events but also their uncertainty about these representations and hypothesise that differences in representational uncertainty may explain some of the characteristics of psychopathy. ...
Article
Full-text available
Psychopathic traits and the childhood analogue, callous-unemotional traits, have been severely neglected by the research field in terms of mechanistic, falsifiable accounts. This is surprising given that some of the core symptoms of the disorder point towards problems with basic components of associative learning. In this manuscript we describe a new mechanistic account that is concordant with current cognitive theories of psychopathic traits and is also able to replicate previous empirical data. The mechanism we describe is one of individual differences in an index we have called, “learning window width”. Here we show how variation in this index would result in different outcome expectations which, in turn, would lead to differences in behaviour. The proposed mechanism is intuitive and simple with easily calculated behavioural implications. Our hope is that this model will stimulate discussion and the use of mechanistic and computational accounts to improve our understanding in this area of research.
... The HGF can also be applied to other time series than behavioral ones. In an example from TN/CP, Brazil et al. (2017) applied an HGF directly to BOLD signal time series from an fMRI experiment. late Action (SERIA) model represents a computational model of an agent's behavior during the antisaccade task by modeling early reflexive and late intentional eye movement via two interacting race-to-threshold processes. ...
Preprint
Full-text available
Psychiatry faces fundamental challenges with regard to mechanistically guided differential diagnosis, as well as prediction of clinical trajectories and treatment response of individual patients. This has motivated the genesis of two closely intertwined fields: (i) Translational Neuromodeling (TN), which develops "computational assays" for inferring patient-specific disease processes from neuroimaging, electrophysiological, and behavioral data; and (ii) Computational Psychiatry (CP), with the goal of incorporating computational assays into clinical decision making in everyday practice. In order to serve as objective and reliable tools for clinical routine, computational assays require end-to-end pipelines from raw data (input) to clinically useful information (output). While these are yet to be established in clinical practice, individual components of this general end-to-end pipeline are being developed and made openly available for community use. In this paper, we present the Translational Algorithms for Psychiatry-Advancing Science (TAPAS) software package, an open-source collection of building blocks for computational assays in psychiatry. Collectively, the tools in TAPAS presently cover several important aspects of the desired end-to-end pipeline, including: (i) tailored experimental designs and optimization of measurement strategy prior to data acquisition, (ii) quality control during data acquisition, and (iii) artifact correction, statistical inference, and clinical application after data acquisition. Here, we review the different tools within TAPAS and illustrate how these may help provide a deeper understanding of neural and cognitive mechanisms of disease, with the ultimate goal of establishing automatized pipelines for predictions about individual patients. We hope that the openly available tools in TAPAS will contribute to the further development of TN/CP and facilitate the translation of advances in computational neuroscience into clinically relevant computational assays.
... Thus, the present AI findings may reflect diminished capacity to represent uncertain information in social scenes in psychopathic individuals. Future studies may use computational modeling as in Brazil et al. (2017) to determine how AI represents uncertain information during affective perspective-taking in psychopathy. Relatedly, the present findings may reflect disrupted processing of threat cues in psychopathy. ...
Article
Full-text available
Psychopathic individuals are notorious for their callous disregard for others' emotions. Prior research has linked psychopathy to deficits in affective mechanisms underlying empathy (e.g., affective sharing), yet research relating psychopathy to cognitive mechanisms underlying empathy (e.g., affective perspective-taking and Theory of Mind) requires further clarification. To elucidate the neurobiology of cognitive mechanisms of empathy in psychopathy, we administered an fMRI task and tested for global as well as emotion-specific deficits in affective perspective-taking. Adult male incarcerated offenders (N=94) viewed images of two people interacting, with one individual's face obscured by a shape. Participants were cued to either identify the emotion of the obscured individual or identify the shape from one of two emotion or shape choices presented on each trial. Target emotions included anger, fear, happiness, sadness, and neutral. Contrary to predictions, psychopathy was unrelated to neural activity in the Affective Perspective-taking > Shape contrast. In line with predictions, psychopathy was negatively related to task accuracy during affective perspective-taking for fear, happiness, and sadness. Psychopathy was related to reduced hemodynamic activity exclusively during fear perspective-taking in several areas: left anterior insula extending into posterior orbitofrontal cortex, right precuneus, left superior parietal lobule, and left superior occipital cortex. Although much prior research has emphasized psychopathy-related abnormalities in affective mechanisms mediating empathy, current results add to growing evidence of psychopathy-related abnormalities in a cognitive mechanism related to empathy. These findings highlight brain regions that are hypoactive in psychopathy when explicitly processing another's fear.
... This allows researchers to discover and investigate latent variables that might be responsible for producing the observed data from brain studies and during behavioral performance. In this regard, a recent study indicated that psychopathic features are related to less accurate estimates about the relationship between the imperative stimulus and punishment during a conditioning task (Brazil, Mathys, Popma, Hoppenbrouwers, & Cohn, 2017). More specifically, the latter study pointed out that inaccuracy (i.e., "uncertainty") in different types of representations related to the perceived changeability of the stimulus-outcome contingencies during threat conditioning were differentially correlated with different psychopathic traits. ...
Article
Full-text available
Effective and specifically targeted social and therapeutic responses for antisocial personality disorders and psychopathy are scarce. Some authors maintain that this scarcity should be overcome by revising current syndrome-based classifications of these conditions and devising better biocognitive classifications of antisocial individuals. The inspiration for the latter classifications has been embedded in the Research Domain Criteria (RDoC) approach. RDoC-type approaches to psychiatric research aim at transforming diagnosis, provide valid measures of disorders, aid clinical practice, and improve health outcomes by integrating the data on the genetic, neural, cognitive, and affective systems underlying psychiatric conditions. In the first part of the article, we discuss the benefits of such approaches compared with the dominant syndrome-based approaches and review recent attempts at building biocognitive classifications of antisocial individuals. Other researchers, however, have objected that biocognitive approaches in psychiatry are committed to an untenable form of explanatory reductionism. Explanatory reductionism is the view that psychological disorders can be exclusively categorized and explained in terms of their biological causes. In the second part of the article, we argue that RDoC-like approaches need not be associated with explanatory reductionism. Moreover, we argue how this is the case for a specific biocognitive approach to classifying antisocial individuals.
Chapter
Uncertainty arises because of ignorance or the absence of information in any situation. Probability, risk, and randomness are concepts closely related to uncertainty that significantly impact the decision-making process. In this chapter, we discuss the concept and types of uncertainty and the role of uncertainty in decision making. The continuum of pure uncertainty to certainty represents different models of this dimension. We present different types of uncertainty in the context of the function, ethics, and stimulus–response outcome rules. A decision maker handles vagueness in a problem situation with his unique cognitive process and orientation. The ongoing discoveries in the neuroscience field give a better comprehension of brain mechanisms involved in decision making under uncertainty. This chapter additionally endeavours to address the limitations faced in decision making resulting from uncertainty and different approaches to managing such a circumstance. Notwithstanding, uncertainty does not only count for hindering the effective decision-making process. In the last section, the implication for decision making under uncertainty is presented.
Article
Full-text available
Background Studies have demonstrated a robust association between interpersonal callousness (IC) and the development of severe and chronic conduct problems (CP) in youth. Although children exhibiting IC are also believed to be at particularly high risk for developing psychopathic personality features in adulthood, there is little longitudinal evidence supporting this assumption, particularly after controlling for co-occuring CP severity. Methods This study used data collected on a longitudinal cohort of boys (n=508), with an oversampling of youth exhibiting elevated conduct problems. Analyses examined the unique and interactive association between latent growth curve trajectories of IC and CP assessed bi-annually from late childhood to early adolescence (similar to ages 10-13) and features of psychopathy in early adulthood (age similar to 24) assessed using the Psychopathy Checklist - Short Version (PCL:SV; Hart, Cox, & Hare, 1995). ResultsGrowth curve analysis indicated that initial levels of IC and CP in childhood (similar to age 10 intercept) both uniquely predicted the development of the interpersonal/affective features of adult psychopathy, and boys with a combination of high initial levels of IC and CP were at particularly high risk for developing the impulsive/antisocial features of the disorder. Boys who exhibited systematic increases in CP from late childhood to early adolescence also demonstrated higher adult psychopathy scores, but changes in IC across this developmental period did not significantly add to the prediction of adult psychopathy. Conclusions Findings highlight the importance of developing targeted interventions for boys exhibiting severe IC and CP in childhood, as they appear to be at high risk for developing adult psychopathic features.
Article
Full-text available
The effects of stress are frequently studied, yet its proximal causes remain unclear. Here we demonstrate that subjective estimates of uncertainty predict the dynamics of subjective and physiological stress responses. Subjects learned a probabilistic mapping between visual stimuli and electric shocks. Salivary cortisol confirmed that our stressor elicited changes in endocrine activity. Using a hierarchical Bayesian learning model, we quantified the relationship between the different forms of subjective task uncertainty and acute stress responses. Subjective stress, pupil diameter and skin conductance all tracked the evolution of irreducible uncertainty. We observed a coupling between emotional and somatic state, with subjective and physiological tuning to uncertainty tightly correlated. Furthermore, the uncertainty tuning of subjective and physiological stress predicted individual task performance, consistent with an adaptive role for stress in learning under uncertain threat. Our finding that stress responses are tuned to environmental uncertainty provides new insight into their generation and likely adaptive function.
Article
Antisocial behavior is a heterogeneous construct that can be divided into subtypes, such as antisocial personality and psychopathy. The adverse consequences of antisocial behavior produce great burden for the perpetrators, victims, family members, and for society at-large. The pervasiveness of antisocial behavior highlights the importance of precisely characterizing subtypes of antisocial individuals and identifying specific factors that are etiologically related to such behaviors to inform the development of targeted treatments. The goals of the current review are (1) to briefly summarize research on the operationalization and assessment of antisocial personality and psychopathy; (2) to provide an overview of several existing treatments with the potential to influence antisocial personality and psychopathy; and (3) to present a procedure for integrating and using biological and cognitive measures as starting points to more precisely characterize and treat these individuals. A focus on integrating factors at multiple levels of analysis will uncover person-specific characteristics and highlight potential targets for treatment to alleviate the burden caused by antisocial behavior.
Article
Background: Previous work has shown that patients with conduct problems (CP) show impairments in reinforcement-based decision-making. However, studies with patients have not previously demonstrated any relationships between impairment in any of the neurocomputations underpinning reinforcement-based decision-making and specific symptom sets [e.g. level of CP and/or callous-unemotional (CU) traits]. Methods: Seventy-two youths [20 female, mean age = 13.81 (SD = 2.14), mean IQ = 102.34 (SD = 10.99)] from a residential treatment program and the community completed a passive avoidance task while undergoing functional MRI. Results: Greater levels of CP were associated with poorer task performance. Reduced representation of expected values (EV) when making avoidance responses within bilateral anterior insula cortex/inferior frontal gyrus (AIC/iFG) and striatum was associated with greater levels of CP but not CU traits. Conclusions: The current data indicate that difficulties in the use of value information to motivate decisions to avoid suboptimal choices are associated with increased levels of CP (though not severity of CU traits). Moreover, they account for the behavioral deficits observed during reinforcement-based decision-making in youth with CP. In short, an individual's relative failure to utilize value information within AIC/iFG to avoid bad choices is associated with elevated levels of CP.
Article
Psychopathy is a personality disorder characterized by interpersonal manipulation and callousness, and reckless and impulsive antisocial behavior. It is often seen as a disorder in which profound emotional disturbances lead to antisocial behavior. A lack of fear in particular has been proposed as an etiologically salient factor. In this review, we employ a conceptual model in which fear is parsed into separate subcomponents. Important historical conceptualizations of psychopathy, the neuroscientific and empirical evidence for fear deficits in psychopathy are compared against this model. The empirical evidence is also subjected to a meta-analysis. We conclude that most studies have used the term "fear" generically, amassing different methods and levels of measurement under the umbrella term "fear." Unlike earlier claims that psychopathy is related to general fearlessness, we show there is evidence that psychopathic individuals have deficits in threat detection and responsivity, but that the evidence for reduced subjective experience of fear in psychopathy is far less compelling. (PsycINFO Database Record
Article
Psychopathy is a severe personality disorder, the core of which pertains to callousness, an entitled and grandiose interpersonal style often accompanied by impulsive and reckless endangerment of oneself and others. The response modulation theory of psychopathy states that psychopathic individuals have difficulty modulating top-down attention to incorporate bottom-up stimuli that may signal important information but are irrelevant to current goals. However, it remains unclear which particular aspects of attention are impaired in psychopathy. Here, we used 2 visual search tasks that selectively tap into bottom-up and top-down attention. In addition, we also looked at intertrial priming, which reflects a separate class of processes that influence attention (i.e., selection history). The research group consisted of 65 participants that were recruited from the community. Psychopathic traits were measured with the Psychopathic Personality Inventory (PPI; Uzieblo, Verschuere, & Crombez, 2007). We found that bottom-up attention was unrelated to psychopathic traits, whereas elevated psychopathic traits were related to deficits in the use of cues to facilitate top-down attention. Further, participants with elevated psychopathic traits were more strongly influenced by their previous response to the target. These results show that attentional deficits in psychopathy are largely confined to top-down attention and selection history. (PsycINFO Database Record
Article
Background: Children and adolescents with callous unemotional (CU) traits are at risk of severe and persistent antisocial behavior. It is commonly assumed that these children are difficult to treat but it has been proposed that they may benefit from being involved in interventions that go beyond typical parent training programs. Aim: This systematic review sought to answer two previously unanswered questions: do interventions involving young people reduce levels of CU traits? Do CU traits predict the effectiveness of interventions for antisocial behavior involving young people? Method: Studies were included that adopted an randomized controlled trial, controlled or open trial design and that had examined whether treatment was related to reductions in CU traits or whether CU traits predicted or moderated treatment effectiveness. Results: Treatments used a range of approaches, including behavioral therapy, emotion recognition training, and multimodal interventions. 4/7 studies reported reductions in CU traits following treatment. There was a mixed pattern of findings in 15 studies that examined whether CU traits predicted treatment outcomes following interventions for antisocial behavior. In 7/15 studies, CU traits were associated with worse outcomes, although three of these studies did not provide data on baseline antisocial behavior, making it difficult to evaluate whether children with high CU traits had shown improvements relative to their own behavioral baseline, despite having the worst behavioral outcomes overall. CU traits did not predict outcomes in 7/15 studies. Finally, a single study reported that CU traits predicted an overall increased response to treatment. Conclusions: Overall, the evidence supports the idea that children with CU traits do show reductions in both their CU traits and their antisocial behavior, but typically begin treatment with poorer premorbid functioning and can still end with higher levels of antisocial behavior. However, there is considerable scope to build on the current evidence base.