Data leakage in deep learning studies of translational EEG.

A growing number of studies apply deep neural networks (DNNs) to recordings of human electroencephalography (EEG) to identify a range of disorders. In many studies, EEG recordings are split into segments, and each segment is randomly assigned to the training or test set. As a consequence, data from individual subjects appears in both the training and the test set. Could high test-set accuracy reflect data leakage from subject-specific patterns in the data, rather than patterns that identify a disease? We address this question by testing the performance of DNN classifiers using segment-based holdout (in which segments from one subject can appear in both the training and test set), and comparing this to their performance using subject-based holdout (where all segments from one subject appear exclusively in either the training set or the test set). In two datasets (one classifying Alzheimer's disease, and the other classifying epileptic seizures), we find that performance on previously-unseen subjects is strongly overestimated when models are trained using segment-based holdout. Finally, we survey the literature and find that the majority of translational DNN-EEG studies use segment-based holdout. Most published DNN-EEG studies may dramatically overestimate their classification performance on new subjects.

Moral conviction and metacognitive ability shape multiple stages of information processing during social decision-making.

Moral conviction has the potential to inspire activism and change, but can also instigate divisiveness and great harm. Moreover, attitudes held with moral conviction are experienced as universal objective truths and are less likely to demonstrate social conformity. Some evidence suggests that holding strong moral views may be a consequence of a cognitive style which includes lower metacognitive sensitivity. This study combined measures of attitudes, moral convictions, psychophysics, and electroencephalography (EEG) to determine how metacognition and moralization influence different stages of information processing during social decision-making, as well as the neural mechanisms underlying these effects. First, 170 participants were asked to indicate their attitudes and moral convictions about specific social issues that are currently important to US college students. Later, a subset of 81 participants completed a perceptual confidence task to assess their metacognition. Then, they underwent EEG while evaluating photographs of protests ostensibly about the same social issues. Prior to each photo, participants were provided with statistical information about social consensus from their peers for each issue. As predicted, stronger moral convictions were associated with lower social conformity. Furthermore, this effect was more pronounced in individuals with lower metacognitive ability. Moralized content was prioritized in multiple stages of information processing, including both early automatic attention and later deliberative stages, with medial frontal negativity (MFN) and early posterior negativity (EPN) predicting reduced conformity. Changes in alpha and high beta power indicated increased attention and engagement for moralized content. Together, results from these complementary methods and levels of analysis both converge to indicate that moralization alters sensitivity to social influence by shifting neural responses to efficiently prioritize information related to moralized beliefs.

Psychopathic traits modulate functional connectivity during pain perception and perspective-taking in female inmates.

The ability to share and understand the distress of others is critical for successful social interactions and is a fundamental building block of morality. Psychopathy is a personality disorder that includes lack of empathy and concern for others. In the present study, functional MRI was used to examine neural responses and functional connectivity associated with empathy and affective perspective-taking in female inmates (N = 109) with various levels of psychopathic traits, as measured with Psychopathy Checklist-Revised (PCL-R). Participants viewed hands and feet in painful or non-painful situations while adopting a first person or third person perspective. All participants demonstrated robust neural responses in anterior insula (aINS), anterior cingulate (ACC), temporoparietal junction (TPJ) and supplementary motor area (SMA) when viewing pain, both during imagine-self and imagine-other blocks. Psychopathy shifted the functional connectivity seeded in core nodes of the salience and social cognition networks. Perceiving stimuli depicting somatic pain led to decreased functional coupling from right temporoparietal junction to superior temporal sulcus, which correlated with scores on PCL-R Factor 1 (Affective/Interpersonal). In contrast, connectivity from right insula to precuneus increased with Factor 2 (Lifestyle/Antisocial) scores. When adopting a third-person perspective, psychopathic traits modulated connectivity from the social cognition network, but not the salience network, with Factor 1 scores associated with increased connectivity to sensorimotor cortex and temporal pole, while Factor 2 scores were associated with decreased connectivity with ACC/SMA and inferior frontal gyrus. Overall, these results demonstrate that psychopathic traits in incarcerated females are associated with atypical functional connectivity within the salience network during pain-empathy processing and within the social cognition network during affective perspective-taking.

Neural responses to morally laden interactions in female inmates with psychopathy.

Much of social cognition requires making inferences about the mental and emotional states of others. Moreover, understanding the emotions of others is an important foundation for moral decision-making. Psychopathy is associated with both aberrant emotional understanding and atypical hemodynamic responses when viewing and evaluating morally laden social interactions. In the present functional MRI study, female inmates (N = 107) were asked to evaluate the likely emotional state of either the recipient or the initiator of harmful or helpful interactions. Psychopathy was assessed with the Psychopathy Checklist-Revised (PCL-R). PCL-R scores were not associated with differences in confidence or accuracy ratings. However, psychopathy scores were significantly related to increased hemodynamic response in right dorsolateral prefrontal cortex when viewing harmful interactions and decreased functional connectivity from right amygdala to inferior parietal cortex and insula, and from temporal parietal junction to dorsomedial prefrontal cortex. Overall, this work indicates that in females, psychopathy is associated with normal behavioral accuracy and confidence but alterations in neural network activity during moral decision-making.

EEG distinguishes heroic narratives in ISIS online video propaganda.

The Islamic State (ISIS) was uniquely effective among extremist groups in the Middle East at recruiting Westerners. A major way ISIS accomplished this was by adopting Hollywood-style narrative structures for their propaganda videos. In particular, ISIS utilized a heroic martyr narrative, which focuses on an individual's personal glory and empowerment, in addition to traditional social martyr narratives, which emphasize duty to kindred and religion. The current work presented adult participants (n = 238) video clips from ISIS propaganda which utilized either heroic or social martyr narratives and collected behavioral measures of appeal, narrative transportation, and psychological dispositions (egoism and empathy) associated with attraction to terrorism. Narrative transportation and the interaction between egoism and empathy predicted video recruitment appeal. A subset of adults (n = 80) underwent electroencephalographic (EEG) measurements while watching a subset of the video-clips. Complementary univariate and multivariate techniques characterized spectral power density differences when perceiving the different types of narratives. Heroic videos show increased beta power over frontal sites, and globally increased alpha. In contrast, social narratives showed greater frontal theta, an index of negative feedback and emotion regulation. The results provide strong evidence that ISIS heroic narratives are specifically processed, and appeal to psychological predispositions distinctly from other recruitment narratives.

The Dark Side of Morality - Neural Mechanisms Underpinning Moral Convictions and Support for Violence.

People are motivated by shared social values that, when held with moral conviction, can serve as compelling mandates capable of facilitating support for ideological violence. The current study examined this dark side of morality by identifying specific cognitive and neural mechanisms associated with beliefs about the appropriateness of sociopolitical violence, and determining the extent to which the engagement of these mechanisms was predicted by moral convictions. Participants reported their moral convictions about a variety of sociopolitical issues prior to undergoing functional MRI scanning. During scanning, they were asked to evaluate the appropriateness of violent protests that were ostensibly congruent or incongruent with their views about sociopolitical issues. Complementary univariate and multivariate analytical strategies comparing neural responses to congruent and incongruent violence identified neural mechanisms implicated in processing salience and in the encoding of subjective value. As predicted, neuro-hemodynamic response was modulated parametrically by individuals' beliefs about the appropriateness of congruent relative to incongruent sociopolitical violence in ventromedial prefrontal cortex, and by moral conviction in ventral striatum. Overall moral conviction was predicted by neural response to congruent relative to incongruent violence in amygdala. Together, these findings indicate that moral conviction about sociopolitical issues serves to increase their subjective value, overriding natural aversion to interpersonal harm.

Me first: Neural representations of fairness during three-party interactions.

One hallmark of human morality is a deep sense of fairness. People are motivated by both self-interest and a concern for the welfare of others. However, it remains unclear whether these motivations rely on similar neural computations, and the extent to which such computations influence social decision-making when self-fairness and other-fairness motivations compete. In this study, two groups of participants engaged in the role of responder in a three-party Ultimatum Game while being scanned with functional MRI (N = 32) or while undergoing high-density electroencephalography (N = 40). In both studies, participants accepted more OtherFair offers when they themselves received fair offers. Though SelfFairness was reliably decoded from scalp voltages by 170 ms, and from hemodynamic responses in right insula and dorsolateral prefrontal cortex, there was no overlap between neural representations of fairness for self and for other. Distinct neural computations and mechanisms seem to be involved when making decisions about fairness in three-party contexts, which are anchored in an egocentric, self-serving bias.

Conflicting influences of justice motivations on moral judgments.

Humans are motivated by justice concerns, yet vary in their reactions to observing or experiencing injustice. At a proximate level, approach and avoidance represent core fundamental motivational systems which have been proposed to be involved in two independent moral systems: a prescriptive system responsive to obligations ("shoulds") and a proscriptive system concerned with prohibitions ("should nots"). It is unclear whether these motivational systems or personal involvement better explain the influence of justice dispositions on moral judgments. To clarify this theoretical argument, two experiments examined how dispositional self-oriented and other-oriented justice sensitivity influence condemnation of prescriptive and proscriptive violations while manipulating perspective between-subject or within-subject. Participants rated the permissibility of everyday moral transgressions, from the perspective of the victim or a neutral observer. In Study 1 (n = 235), self-oriented dispositions increased and other-oriented dispositions decreased the probability of rating prescriptive violations as permissible. Study 2 (n = 468) replicated the effect of other-oriented justice sensitivity. Overall, these results bridge the gap between motivational systems and self-other processing. They suggest that justice sensitivity is better conceptualised as distinct motivations which can exert opposing influences on moral decision-making. Moreover, personal involvement may not be as important as individual justice motivations for explaining everyday moral decision-making.

The neuroscience of morality and social decision-making.

Across cultures humans care deeply about morality and create institutions, such as criminal courts, to enforce social norms. In such contexts, judges and juries engage in complex social decision-making to ascertain a defendant's capacity, blameworthiness, and culpability. Cognitive neuroscience investigations have begun to reveal the distributed neural networks which interact to implement moral judgment and social decision-making, including systems for reward learning, valuation, mental state understanding, and salience processing. These processes are fundamental to morality, and their underlying neural mechanisms are influenced by individual differences in empathy, caring and justice sensitivity. This new knowledge has important implication in legal settings for understanding how triers of fact reason. Moreover, recent work demonstrates how disruptions within the social decision-making network facilitate immoral behavior, as in the case of psychopathy. Incorporating neuroscientific methods with psychology and clinical neuroscience has the potential to improve predictions of recidivism, future dangerousness, and responsivity to particular forms of rehabilitation.

The Emerging Social Neuroscience of Justice Motivation.

Humans from a very early age are deeply sensitive to issues of justice and fairness, both in their own lives and in the lives of others. Most people are highly motivated to pursue justice and condemn injustice. Where does this concern for justice come from? Here we integrate findings in evolution, development, psychology, behavioral economics, and social neuroscience to highlight multiple potential drivers of justice motivation. We argue that justice motivation arises from complementary rapid heuristics and deliberation, each utilizing distinct and interacting neural circuitry. This framework is useful for explaining observed symmetries and asymmetries in responses to experiencing or observing injustice and may help to explain why individuals vary in their responses to injustice.

Callous traits in children with and without conduct problems predict reduced connectivity when viewing harm to others.

The presence of elevated callous unemotional (CU) traits seems to designate a distinct group of children and adolescents with serious conduct problems. However, the extent to which CU traits impact the aversive reaction to harm is still a contentious issue. Here, we examined the effective connectivity seeded in the anterior insula and anterior cingulate cortex in a large number of children (N = 123, age 9-11, 60 females) with various levels of conduct disorder (CD) symptoms in response to visual stimuli depicting other people being physically injured. Perceiving others being harmed was associated with increased hemodynamic activity in the left amygdala and right temporoparietal junction (rTPJ). Children with higher callous traits showed less functional connectivity seeded in anterior cingulate with left amygdala and anterior insula. Conversely, CD symptoms were positively related to connectivity of insula with rTPJ. Overall, these results suggest that callousness is marked by the disruption of widespread cortical networks responsible for detecting and appropriately responding to important environmental cues, such as the distress of others.

Empathy and motivation for justice: Cognitive empathy and concern, but not emotional empathy, predict sensitivity to injustice for others.

Why do people tend to care for upholding principles of justice? This study examined the association between individual differences in the affective, motivational and cognitive components of empathy, sensitivity to justice, and psychopathy in participants (N 265) who were also asked to rate the permissibility of everyday moral situations that pit personal benefit against moral standards of justice. Counter to common sense, emotional empathy was not associated with sensitivity to injustice for others. Rather, individual differences in cognitive empathy and empathic concern predicted sensitivity to justice for others, as well as the endorsement of moral rules. Psychopathy coldheartedness scores were inversely associated with motivation for justice. Moreover, hierarchical multiple linear regression analysis revealed that self-focused and other-focused orientations toward justice had opposing influences on the permissibility of moral judgments. High scores on psychopathy were associated with less moral condemnation of immoral behavior. Together, these results contribute to a better understanding of the information processing mechanisms underlying justice motivation, and may guide interventions designed to foster justice and moral behavior. In order to promote justice motivation, it may be more effective to encourage perspective taking and reasoning than emphasizing emotional sharing with the misfortune of others.

Sex differences in abnormal white matter development associated with conduct disorder in children.

Associations between white matter pathway abnormalities and antisocial personality disorder in adults are well replicated, and there is some evidence for an association of white matter abnormalities with conduct disorder (CD) in adolescents. In this study, white matter maturation using diffusion tensor imaging (DTI) was examined in 110 children aged 10.0 ± 0.8 years selected to vary widely in their numbers of CD symptoms. The results replicated age-related increases in fractional anisotropy (FA) found in previous studies. There was not a significant association between the number of CD symptoms and FA, but CD symptoms were found to be significantly associated with greater axial and radial diffusivity in a broad range of white matter tracts, particularly in girls. In complementary analyses, there were similar significant differences in axial and radial diffusivity between children who met diagnostic criteria for CD and healthy children with no symptoms of CD, particularly in girls. Brain structural abnormalities may contribute to the emergence of CD in childhood, perhaps playing a greater role in girls.

Amygdala subnuclei connectivity in response to violence reveals unique influences of individual differences in psychopathic traits in a nonforensic sample.

Atypical amygdala function and connectivity have reliably been associated with psychopathy. However, the amygdala is not a unitary structure. To examine how psychopathic traits in a nonforensic sample are linked to amygdala response to violence, this study used probabilistic tractography to classify amygdala subnuclei based on anatomical projections to and from amygdala subnuclei in a group of 43 male participants. The segmentation identified the basolateral complex (BLA; lateral, basal, and accessory basal subnuclei) and the central subnucleus (CE), which were used as seeds in a functional connectivity analysis to identify differences in neuronal coupling specific to observed violence. While a full amygdala seed showed significant connectivity only to right middle occipital gyrus, subnuclei seeds revealed unique connectivity patterns. BLA showed enhanced coupling with anterior cingulate and prefrontal regions, while CE showed increased connectivity with the brainstem, but reduced connectivity with superior parietal and precentral gyrus. Further, psychopathic personality factors were related to specific patterns of connectivity. Fearless Dominance scores on the psychopathic personality inventory predicted increased coupling between the BLA seed and sensory integration cortices, and increased connectivity between the CE seed and posterior insula. Conversely, Self-Centered Impulsivity scores were negatively correlated with coupling between BLA and ventrolateral prefrontal cortex, and Coldheartedness scores predicted increased functional connectivity between BLA and dorsal anterior cingulate cortex. Taken together, these findings demonstrate how subnuclei segmentations reveal important functional connectivity differences that are otherwise inaccessible. Such an approach yields a better understanding of amygdala dysfunction in psychopathy.

Spatiotemporal neural dynamics of moral judgment: a high-density ERP study.

Morality is a pervasive aspect of human nature across all cultures, and neuroscience investigations are necessary for identifying what computational mechanisms underpin moral cognition. The current study used high-density ERPs to examine how moral evaluations are mediated by automatic and controlled processes as well as how quickly information and causal-intentional representations can be extracted when viewing morally laden behavior. The study also explored the extent to which individual dispositions in affective and cognitive empathy as well as justice sensitivity influence the encoding of moral valence when healthy participants make moral judgments about prosocial (interpersonal assistance) and antisocial (interpersonal harm) actions. Moral judgment differences were reflected in differential amplitudes for components associated with cognitive appraisal (LPP) as well as early components associated with emotional salience (N1 and N2). Moreover, source estimation was performed to indicate potential neural generators. A posterior-to-anterior shift was observed, with current density peaks first in right inferior parietal cortex (at the temporoparietal junction), then later in medial prefrontal cortex. Cognitive empathy scores predicted behavioral ratings of blame as well as differential amplitudes in LPP and component activity at posterior sites. Overall, this study offers important insights into the temporal unfolding of moral evaluations, including when in time individual differences in empathy influence neural encoding of moral valence.

The Good, the bad, and the just: justice sensitivity predicts neural response during moral evaluation of actions performed by others.

Morality is a fundamental component of human cultures and has been defined as prescriptive norms regarding how people should treat one another, including concepts such as justice, fairness, and rights. Using fMRI, the current study examined the extent to which dispositions in justice sensitivity (i.e., how individuals react to experiences of injustice and unfairness) predict behavioral ratings of praise and blame and how they modulate the online neural response and functional connectivity when participants evaluate morally laden (good and bad) everyday actions. Justice sensitivity did not impact the neuro-hemodynamic response in the action-observation network but instead influenced higher-order computational nodes in the right temporoparietal junction (rTPJ), right dorsolateral and dorsomedial prefrontal cortex (rdlPFC, dmPFC) that process mental states understanding and maintain goal representations. Activity in these regions predicted praise and blame ratings. Further, the hemodynamic response in rTPJ showed a differentiation between good and bad actions 2 s before the response in rdlPFC. Evaluation of good actions was specifically associated with enhanced activity in dorsal striatum and increased the functional coupling between the rTPJ and the anterior cingulate cortex. Together, this study provides important knowledge in how individual differences in justice sensitivity impact neural computations that support psychological processes involved in moral judgment and mental-state reasoning.

Neural processing of dynamic emotional facial expressions in psychopaths.

Facial expressions play a critical role in social interactions by eliciting rapid responses in the observer. Failure to perceive and experience a normal range and depth of emotion seriously impact interpersonal communication and relationships. As has been demonstrated across a number of domains, abnormal emotion processing in individuals with psychopathy plays a key role in their lack of empathy. However, the neuroimaging literature is unclear as to whether deficits are specific to particular emotions such as fear and perhaps sadness. Moreover, findings are inconsistent across studies. In the current experiment, 80 incarcerated adult males scoring high, medium, and low on the Hare Psychopathy Checklist-Revised (PCL-R) underwent functional magnetic resonance imaging (fMRI) scanning while viewing dynamic facial expressions of fear, sadness, happiness, and pain. Participants who scored high on the PCL-R showed a reduction in neuro-hemodynamic response to all four categories of facial expressions in the face processing network (inferior occipital gyrus, fusiform gyrus, and superior temporal sulcus (STS)) as well as the extended network (inferior frontal gyrus and orbitofrontal cortex (OFC)), which supports a pervasive deficit across emotion domains. Unexpectedly, the response in dorsal insula to fear, sadness, and pain was greater in psychopaths than non-psychopaths. Importantly, the orbitofrontal cortex and ventromedial prefrontal cortex (vmPFC), regions critically implicated in affective and motivated behaviors, were significantly less active in individuals with psychopathy during the perception of all four emotional expressions.

More than maths and mindreading: sex differences in empathizing/systemizing covariance.

Empathizing-Systemizing theory posits a continuum of cognitive traits extending from autism into normal cognitive variation. Covariance data on empathizing and systemizing traits have alternately suggested inversely dependent, independent, and sex-dependent (one sex dependent, the other independent) structures. A total of 144 normal undergraduates (65 men, 79 women) completed the Reading the Mind in the Eyes, Embedded Figures, and Benton face recognition tests, the Autism Spectrum Quotient, and measures of digit length ratio and field of study; some also completed tests of motion coherence threshold (64) and go/no-go motor inhibition (128). Empathizing and systemizing traits were independent in women, but largely dependent in men. In men, level of systemizing skill required by field of study was directly related to social interactive and mindreading deficits; men's social impairments correlated with prolonged go/no-go response times, and men tended to apply systemizing strategies to solve problems of empathizing or global processing: rapid perceptual disembedding predicted heightened sensitivity to facial emotion. In women, level of systemizing in field was related to male-typical digit ratios and autistic superiorities in detail orientation, but not to autistic social and communicative impairments; and perceptual disembedding was related to social interactive skills but independent of facial emotion and visual motion perception.

Combining computer game-based behavioural experiments with high-density EEG and infrared gaze tracking.

Experimental paradigms are valuable insofar as the timing and other parameters of their stimuli are well specified and controlled, and insofar as they yield data relevant to the cognitive processing that occurs under ecologically valid conditions. These two goals often are at odds, since well controlled stimuli often are too repetitive to sustain subjects' motivation. Studies employing electroencephalography (EEG) are often especially sensitive to this dilemma between ecological validity and experimental control: attaining sufficient signal-to-noise in physiological averages demands large numbers of repeated trials within lengthy recording sessions, limiting the subject pool to individuals with the ability and patience to perform a set task over and over again. This constraint severely limits researchers' ability to investigate younger populations as well as clinical populations associated with heightened anxiety or attentional abnormalities. Even adult, non-clinical subjects may not be able to achieve their typical levels of performance or cognitive engagement: an unmotivated subject for whom an experimental task is little more than a chore is not the same, behaviourally, cognitively, or neurally, as a subject who is intrinsically motivated and engaged with the task. A growing body of literature demonstrates that embedding experiments within video games may provide a way between the horns of this dilemma between experimental control and ecological validity. The narrative of a game provides a more realistic context in which tasks occur, enhancing their ecological validity (Chaytor & Schmitter-Edgecombe, 2003). Moreover, this context provides motivation to complete tasks. In our game, subjects perform various missions to collect resources, fend off pirates, intercept communications or facilitate diplomatic relations. In so doing, they also perform an array of cognitive tasks, including a Posner attention-shifting paradigm (Posner, 1980), a go/no-go test of motor inhibition, a psychophysical motion coherence threshold task, the Embedded Figures Test (Witkin, 1950, 1954) and a theory-of-mind (Wimmer & Perner, 1983) task. The game software automatically registers game stimuli and subjects' actions and responses in a log file, and sends event codes to synchronise with physiological data recorders. Thus the game can be combined with physiological measures such as EEG or fMRI, and with moment-to-moment tracking of gaze. Gaze tracking can verify subjects' compliance with behavioural tasks (e.g. fixation) and overt attention to experimental stimuli, and also physiological arousal as reflected in pupil dilation (Bradley et al., 2008). At great enough sampling frequencies, gaze tracking may also help assess covert attention as reflected in microsaccades - eye movements that are too small to foveate a new object, but are as rapid in onset and have the same relationship between angular distance and peak velocity as do saccades that traverse greater distances. The distribution of directions of microsaccades correlates with the (otherwise) covert direction of attention (Hafed & Clark, 2002).