Empathy and Criminal Behavior: What Research Does and Does Not Show
Author: Ukrainian Psychological Hub · Published: October 11, 2026 · Editorial Policy
Empathy and criminal behavior are related in psychological research, but the relationship is more selective and less predictive than the familiar claim that people commit crimes because they lack empathy. Studies sometimes find lower average empathy among groups with offending histories, particularly on certain measures. Other studies find small associations, strong methodological differences, or little evidence that a person's empathy score can predict an offense. Empathy is one component of social functioning within a much larger developmental, situational, interpersonal, and institutional picture.
Two findings illustrate the central problem. A meta-analysis of 38 studies involving 6,631 participants reported larger average offender–nonoffender differences in cognitive empathy (standardized mean difference d = 0.43) than in affective empathy (d = 0.19). By contrast, a separate meta-analysis of aggression in adults found a weak overall empathy–aggression correlation (r = −0.11). These studies examine different outcomes and populations, so their findings should be interpreted together rather than treated as contradictory verdicts. See the offending meta-analysis and aggression meta-analysis.
What does empathy mean in criminal psychology?
Empathy is a family of psychological processes involved in recognizing, understanding, and responding to another person's emotional or mental state. Research ordinarily distinguishes cognitive and affective components; some frameworks add compassionate concern, motivational responses, and the ability to regulate one's own distress. The instruments used in studies do not all measure the same mixture of these processes. As a result, a statement about 'low empathy' is incomplete until the investigator specifies what was measured and in which context.
Cognitive empathy: understanding another person's perspective
Cognitive empathy refers to understanding or inferring what someone else may think or feel. Perspective-taking can help someone notice fear, humiliation, distress, intentions, or possible consequences for another person. It also can support manipulation when the observer understands another person's vulnerabilities but has motives that do not favor their welfare. Accurate social understanding is therefore neither proof of benevolence nor evidence of criminal intent.
Affective empathy: emotional responsiveness
Affective empathy refers to emotional responsiveness to another person's emotional state. Researchers distinguish sharing emotion, empathic concern, and personal distress because these reactions can lead to different behavior. A person may recognize distress without experiencing much shared emotion, feel intense distress but have difficulty responding helpfully, or care about a victim while failing to act safely in a specific situation. Each pattern has different measurement and intervention implications.
Empathy, compassion, remorse, guilt, and moral judgment
Compassion concerns another person's welfare and can motivate assistance. Guilt concerns a perceived wrong action; remorse often includes regret for harm done. Moral judgment concerns what someone believes is permissible or required. These processes interact with empathy, but they are distinct. Remorse after an offense is not an objective laboratory measure of empathy, and a restrained emotional presentation in a courtroom is not proof that a person lacks remorse.
Does low empathy cause criminal behavior?
The best-supported answer is that lower empathy can correlate with some forms of offending and aggression, but association alone does not demonstrate that it caused a particular crime. Criminal behavior is not one uniform outcome. Property offending, interpersonal violence, coercive control, fraud, sexual offending, impulsive acts, planned offenses, and breaches of regulatory law involve different opportunities, intentions, constraints, and social contexts. A psychological finding from one category cannot automatically be transferred to another.
Evidence about causation is especially limited when research compares people after a conviction with people who have not been convicted. Differences might precede offending, emerge through later experiences, reflect selection into policing and courts, or arise partly from the questionnaires themselves. Socioeconomic conditions, substance use, perceived threats, social learning, opportunity, peer networks, early experiences, emotional regulation, and institutional responses may influence both assessed empathy and observed outcomes. The early Jolliffe and Farrington synthesis found that offender–nonoffender empathy differences were sensitive to intelligence and socioeconomic status in the adjusted analyses available.
Causal interpretation requires a plausible temporal sequence, careful measurement of confounders, and ideally prospective or experimental evidence aligned with the outcome of interest. An intervention that raises a questionnaire score but has never demonstrated an effect on actual offending provides evidence of score change, not of crime prevention. The strongest claim justified by the meta-analytic literature is that empathy can be one relevant psychological correlate whose role depends on its component, population, outcome, and assessment method.
What the major meta-analyses show
The 2004 systematic review: an association with important qualifications
Jolliffe and Farrington synthesized research comparing empathy in offending and nonoffending populations. Their analysis suggested a stronger link for cognitive empathy than affective empathy, a stronger association in younger samples, and notable differences by offense category. An important limitation was reliance on questionnaire-based empathy measures. The reported association disappeared in analyses controlling for intelligence and socioeconomic status, making simple causal interpretation untenable. The review identified precisely why better prospective designs and more valid measurements were necessary. Read the journal article.
The 2014 offender–nonoffender meta-analysis: small-to-moderate group differences
Van Langen and colleagues analyzed 38 studies with 60 independent effects and 6,631 participants. The mean group difference was larger for cognitive than affective empathy. These are standardized differences between groups; they cannot tell us which individual has offended, why a particular act occurred, or whether a person will reoffend. Age, gender, study design, matching procedures, and instruments moderated associations, and conclusions differed across statistical models. A meta-analysis can estimate an average relationship while substantial overlap remains between the two groups' scores. Read the meta-analysis.
The 2014 aggression meta-analysis: weak average correlation
Vachon, Lynam, and Johnson combined 106 effects from studies of adult empathy and aggression, including community, student, and criminal samples. They found an overall correlation of approximately −0.11; the associations varied across verbal, physical, and sexual aggression. The squared correlation is only about 1% of shared variance in the simple bivariate relationship. It is not the percentage of crimes caused by empathy and it is not a clinical prediction statistic. The result shows how limited empathy alone is as an explanation of aggression across heterogeneous samples. Read the Psychological Bulletin paper.
Why these results differ without cancelling one another
The first two reviews contrast people categorized by offending history; the third examines variation in aggression. Offending and aggression overlap but are not interchangeable. Some aggressive acts never become criminal cases, and some crimes involve no interpersonal aggression. Measures also differ: knowing what another person feels, reporting empathic concern, and identifying emotion from a face produce distinct scores. Combining them under one label can obscure the mechanisms. Effect sizes are consequently meaningful only alongside the comparison groups, sample characteristics, measurement properties, and definitions of behavior.
Cognitive and affective empathy can show different patterns
The distinction matters most when researchers move from general offending groups to specific trait patterns. A 2022 meta-analysis covering 477 records and approximately 142,692 participants found that antisocial behavior groups, on average, displayed more pronounced differences in cognitive empathy, whereas higher psychopathy scores were associated more strongly with affective empathy differences. These were comparisons across research frameworks, not diagnostic rules for individual defendants. Review the meta-analysis.
The findings challenge a one-dimensional interpretation of the familiar phrase 'empathy deficit.' An individual can understand another person's fear accurately while remaining comparatively unresponsive to that fear, or can experience emotional resonance while acting under acute anger or perceived provocation. Either person's behavior depends on motivation, self-regulation, beliefs, situational pressures, opportunity, and learned habits. Being able to infer emotion, feeling emotion, and deciding how to act remain separate questions.
Empathy in violent offending and other offense categories
Interpersonal harm can involve diminished concern for the affected person, but the immediate psychological dynamics differ across acts. Some aggression is reactive to a perceived threat or provocation; some is instrumental toward a particular goal; some occurs in a context of coercion, intoxication, escalating conflict, peer reinforcement, or distorted beliefs about entitlement. A person's apparent empathy for close friends may coexist with hostility toward an out-group or disregard for a particular target. This is a hypothesis about potentially selective concern, not a validated signature of a violent offender.
Research on offenses against property, fraud, violence, and sexual victimization cannot be compressed into a universal hierarchy of empathy. Conviction-based groups differ in detected versus undetected offenses, demographic composition, severity, referral processes, treatment history, and access to assessment. Associations can also differ depending on whether participants report their own empathy or complete a performance task. Evidence from one population should be transferred only when the target population and outcome are sufficiently comparable.
Sexual offending and victim-specific empathy
General empathy, empathy toward a specific victim, recognition of harm, minimization, and offense-supportive beliefs are separate assessment targets. The 2004 offending review reported relatively weaker empathy associations in some sexual-offense comparisons than in violent-offense comparisons. A longitudinal treatment study found that victim-specific empathy and general empathy changed differently; general empathy did not show a reliable association with sexual recidivism in that sample. These data do not imply that victims' experiences are secondary. They show that the psychological process measured and the real-world outcome must be specified. Read the victim-specific empathy study.
Risk-factor reviews for sexual recidivism emphasize that a clinically interesting characteristic becomes a credible risk marker only when it demonstrates appropriate prospective prediction. A plausible theory that someone who appreciates a victim's suffering will avoid reoffending is not sufficient evidence that an empathy questionnaire is a valid recidivism instrument. See Mann, Hanson, and Thornton's review.
Young people: empathy, callous-unemotional traits, and aggression
Youth research requires developmental precision. Emotional understanding, cognitive perspective-taking, regulation, peer orientation, and moral reasoning continue to develop through childhood and adolescence. Studies that pool a child aged 9 with a teenager aged 17 may conceal meaningful developmental differences. Justice-system involvement also reflects social conditions, surveillance, opportunities, and institutional decisions, not only individual psychological characteristics.
A meta-analysis of 192 effect sizes in child and adolescent samples found relatively small-to-moderate correlations between aggression and conventional measures of general, cognitive, and emotional empathy (approximately r = −0.06 to −0.26). In the same synthesis, callous-unemotional trait measures showed stronger positive associations with aggression (around r = 0.30 to 0.37). These are group associations; neither range supplies an individual future-crime probability. Read the youth meta-analysis.
Callous-unemotional traits concern a broader pattern of limited prosocial emotional functioning. The assessment of traits, a diagnostic specifier, and a clinical disorder must be distinguished. A trait scale is not a verdict that a young person is incapable of caring, nor a prediction that they will become an adult offender. Forensic interpretation requires age-appropriate norms, contextual information, multiple sources, and caution about cultural, linguistic, and developmental measurement differences.
Empathy and psychopathy are not synonyms
Psychopathy is a multidimensional research and forensic assessment construct involving interpersonal, affective, and behavioral characteristics. It is not interchangeable with a finding of low empathy, a legal determination of dangerousness, or a synonym for violent offending. Scores derived from psychopathy measures require trained interpretation, documentation of the population for which they were validated, and attention to the consequences of their use. An affective-empathy difference may appear in some high-scoring samples; that result is not equivalent to a universal inability to understand another person's feelings.
Recent comparisons of self-report and performance-based empathy measures also illustrate how conclusions can diverge when instruments change. One 2026 study found different associations with psychopathic traits depending on whether cognitive empathy was assessed by questionnaire or behavioral task. It is an informative primary study, not proof that any single alternative task perfectly measures empathy. Read the study. For broader forensic purposes, these studies reinforce a simple requirement: state precisely what a score measures and what validation supports its interpretation.
Can someone commit a crime and still feel empathy?
Yes. A person may experience substantial affection, distress at others' suffering, or perspective-taking and nevertheless commit an unlawful act. The presence of empathy does not reliably prevent every act under pressure, and a single act does not reveal the person's stable emotional capacity. The relation between values, intentions, emotion, impulse control, perceived justification, situational incentives, and behavior is complex. Psychology can investigate that relation; it cannot infer a complete character profile from a charge or conviction.
Some acts may involve temporarily narrowed attention, a focus on an immediate goal, hostile attribution, acute fear, anger, or the perception that the other person is deserving of harm. Whether a given mechanism was operative in a particular case requires evidence. These explanations are testable possibilities, not ready-made excuses and not substitutes for legal proof of the elements of an offense.
Does having low empathy mean someone will become a criminal?
No reliable individual prediction follows from a general low-empathy label. Even where two groups differ on average, their score distributions overlap. Prediction in a specific setting requires prospective validation, clearly defined outcomes, appropriate base rates, and evidence that using the measure improves decisions beyond other available information. The published meta-analyses establish associations, not universal classification accuracy for criminal behavior.
A low score might indicate a measurement difficulty, low self-reported emotional engagement, a response style, limited insight, or a domain-specific difference. It does not establish the cause of a past offense, the probability of another offense, a diagnosis, or guilt. Conversely, a high self-reported empathy score does not establish innocence, safety, or honesty. Such scores are evidence about a particular measurement procedure, not legal findings.
Why measuring empathy in forensic settings is difficult
Questionnaires versus observed performance
Many classic studies ask individuals to agree or disagree with statements about how they usually respond to others. These responses depend on introspection, comprehension, social norms, motivation, and the interpretation of each question. A performance task may instead ask a participant to interpret emotions, infer mental states, or make choices under specified conditions. Performance tasks are themselves limited by task design, ecological validity, fatigue, language, and how accurately the task reproduces real-world interactions. Convergence between methods cannot be assumed.
Response incentives, impression management, and setting
In criminal proceedings, institutional evaluations, custody decisions, or parole contexts, the consequences attached to a score can influence how questions are understood and answered. Both impression management and adversarial framing are relevant, without assuming that a particular defendant or prisoner is being deceptive. Interviews and questionnaires should be interpreted alongside corroborated information where appropriate. The APA Specialty Guidelines for Forensic Psychology call for methods suitable to the referral question, consideration of response style, multiple data sources, and transparent communication of limitations.
Cultural fairness and population validity
Language, cultural practices concerning emotion expression, norms for eye contact and self-disclosure, educational experience, disability-related communication differences, and assessment accessibility can affect results. A measure standardized in one population is not automatically valid for a person from another community or for someone assessed through an interpreter. A fair report explains which normative data were used, what the measure has demonstrated in comparable groups, and where extrapolation remains uncertain. No visible facial expression, body-language cue, or lack of tears independently measures empathy.
Selection bias and records of offending
Research often relies on official convictions or justice-system contact. These records measure adjudicated or detected behavior under particular laws and enforcement practices. They are not an exhaustive register of harmful acts or offending in the population. Matching groups only on age or sex may leave substantial differences in income, education, neighborhood, trauma exposure, referral history, and other variables. Researchers should therefore describe who was sampled and avoid projecting a group comparison onto all people who have committed or never committed a crime.
What about empathy in victim–offender interactions?
Victim-specific concern can be ethically and therapeutically important, especially when someone is working to acknowledge harm and take responsibility. It may also be distorted by shame, defensive minimization, selective memory, self-protection, or the social pressures of treatment and litigation. Victims are never responsible for producing empathy in people who harmed them, and observable apologies should be evaluated by behavior and context rather than assumed to prove durable change.
Restorative processes and offender programs need voluntariness, safeguarding, and a clear understanding of their objectives. A participant's improved account of another person's experience is meaningful evidence of changed articulation or awareness; demonstrating sustained behavioral change, improved victim safety, or reduced recidivism requires separate outcome evaluation. The ethical priority is avoiding unnecessary pressure on victims while using methods supported for the relevant context.
Can empathy training prevent crime or reduce reoffending?
Programs can teach perspective-taking, emotion recognition, communication, reflection on consequences, and responses to interpersonal conflict. Whether improvement on these exercises leads to fewer crimes is a different empirical question. Reviews of empathy interventions with justice-involved populations report mixed results across types of offending, samples, program formats, and empathy outcomes. The literature is not strong enough to claim that generic empathy training reliably prevents recidivism. Read the intervention review and the critical clinical review.
Outcome research needs a comparison group, adequate follow-up, clearly defined offending outcomes, consideration of baseline risk and treatment needs, attention to attrition, and monitoring for unintended effects. Short-term changes in empathy questionnaires may be useful intermediate outcomes, but they are not substitutes for observed offending rates or independently measured functional change. Different interventions may be appropriate for people whose main difficulties involve hostile beliefs, impulsivity, emotion regulation, social cognition, substance-related impairment, or environmental instability.
There is a wider evidence base for multicomponent rehabilitation and for interventions matched to specific needs. This article does not evaluate every correctional treatment program: the canonical question here is whether empathy specifically explains or alters offending. The finding that empathy training alone lacks established recidivism efficacy should not be read as a claim that psychological rehabilitation generally fails.
The forensic assessment question: what can an expert actually say?
A forensic evaluator may discuss empathy if it is directly relevant to an identified referral question and the available methods can support an interpretation. A defensible report separates direct observations from test scores, test scores from inferences about psychological processes, and those inferences from conclusions relevant to the legal issue. It should give alternative explanations, sources used, cultural and assessment limits, and the degree of confidence justified by the evidence. A generic adjective such as 'empathy-deficient' is rarely a sufficiently operationalized forensic conclusion.
The APA guidelines emphasize legally relevant information, scientific foundation, impartiality, methods appropriately validated for the assessed population, and clear boundaries on conclusions. Psychological constructs do not themselves determine criminal responsibility, competency to stand trial, witness credibility, sentencing, or guilt. Those are legal questions governed by applicable standards and evidence in the jurisdiction. Empathy is not a direct test of any such legal standard.
Does empathy predict violence risk or recidivism?
A hypothesis that empathy reduces interpersonal harm is not equivalent to a validated risk assessment procedure. Risk assessment asks a prospective question about a specified outcome, time horizon, setting, and population. Useful evidence includes discrimination between people with and without the outcome, calibration of predicted probabilities, sensitivity and specificity, false-positive and false-negative rates at a defined threshold, and performance in relevant demographic and clinical subgroups. Most empathy-and-offending meta-analyses do not supply such risk-model validation.
Base rates matter. In a low-rate outcome, an apparently persuasive score threshold can classify many people as likely to offend who will not do so. Conversely, some individuals who score as empathic may commit offenses. Score differences reported as standardized mean differences or correlations cannot be converted into an individual's future-crime probability without an appropriately validated prediction model. Professional decisions should consider what evidence is actually predictive in the target context and how errors affect liberty, safety, and fairness.
Recidivism itself requires definition: a rearrest, reconviction, imprisonment, self-reported behavior, or a particular type of repeat offense is not the same endpoint. Follow-up duration and jurisdictional justice practices further alter observed rates. For this reason empathy should not be presented as a stand-alone clinical or legal risk instrument, and a risk score should never be treated as evidence that a particular offense has occurred.
What the research does not show
Research does not establish a single empathy profile shared by all people convicted of crimes. It does not show that a person's diagnosis or psychological trait explains their criminal conduct. It does not validate reading empathy from facial expressions, body language, apparent remorse, or courtroom demeanor. It does not show that empathy scales prove culpability or that low empathy equates to psychopathy. It does not establish a universal causal pathway from a childhood experience to violent offending. Finally, it does not show that making empathy scores higher necessarily reduces recidivism.
These are evidence limits, not claims that empathy is irrelevant. The best available evidence supports examining empathy as one component of social behavior, differentiating cognitive and affective processes, and asking which populations, offenses, measurement conditions, and outcomes the association applies to. Better longitudinal designs, richer behavioral measurement, transparent subgroup analyses, and independently replicated intervention studies remain important research priorities.
Mental health, empathy, and the risk of stigma
Empathy varies across individuals and situations. A person with a mental-health condition, a personality trait, an intellectual or developmental difference, or a communication difference should not be characterized as dangerous because of that fact alone. Empathy questionnaires are not diagnostic interviews, and diagnostic categories do not establish criminal conduct. Inferences about guilt or violence from a diagnosis or an atypical communication style can cause serious harm.
The distinction becomes particularly important when popular explanations collapse psychosis, personality disorders, neurodivergence, cruelty, and psychopathy into a single notion of 'lack of empathy.' These constructs have different definitions and evidence bases. For the diagnosis-specific intersection, the Hub's separate Schizophrenia and Violence: What Research Really Shows examines evidence under its established clinical ownership. The present page is about empathy's role in criminal-behavior research.
Psychological explanations and legal responsibility
In the United States, criminal laws and rules of evidence vary across states and between state and federal systems. Other jurisdictions use different definitions, procedures, and expert-evidence requirements. Empathy research is descriptive science about behavior and measurement; legal responsibility depends on legally specified elements, mental states, defenses, and findings. A court may consider relevant expert evidence under the applicable law, but a psychological trait does not itself meet or negate a legal criterion.
Explaining a risk factor also does not excuse conduct, establish the identity of an offender, or settle a person's credibility. Forensic testimony should distinguish what was observed, what was reported, what can be inferred at a group level, and what can be supported about the individual. These distinctions protect the integrity of research and the fairness of proceedings.
How readers can evaluate a claim about empathy and crime
When a headline says 'criminals lack empathy,' ask which people were studied, which offenses were counted, and which dimension of empathy was measured. Look for whether the result describes a group difference, correlation, prospective prediction, or the effect of an intervention. Check whether researchers adjusted for major confounders and whether findings were reproduced in independent groups. Quantitative estimates should be accompanied by uncertainty, sample composition, and the appropriate outcome.
When a forensic report or popular true-crime account claims to know a person's inner emotional state from a video, a confession, an interview, or a dramatic incident, ask what independent evidence supports that inference and whether competing explanations have been considered. Specific verified behavior is more informative than character labels. Strong evidence usually accumulates from multiple sources and is constrained by what each source can actually establish.
Frequently asked questions
Is lack of empathy a cause of crime?
It may contribute to some harmful decisions in particular contexts, but existing associations do not prove that low empathy causes crime. Offending has many pathways and the literature is sensitive to measurement, selection, socioeconomic differences, and offense definitions.
Do all violent offenders lack empathy?
No. Group findings are variable and individual differences are substantial. Aggression and empathy are related only weakly on average in one major adult meta-analysis. A particular offense, motive, or person requires an evidence-specific assessment.
Which matters more, cognitive or emotional empathy?
Several offender–nonoffender comparisons found a larger difference for cognitive empathy, whereas some studies of psychopathy emphasize affective empathy. The answer depends on the population, construct, instrument, and behavior being examined.
Can a person understand another person's feelings and still manipulate or harm them?
Yes. Understanding another person's mental state supplies information, while motives, moral concerns, self-control, social norms, and circumstances influence how that information is used. Cognitive understanding and caring behavior should be assessed separately.
Can empathy tests reveal whether someone is lying or guilty?
No. An empathy questionnaire measures responses under a defined procedure; it is not a validated lie detector and cannot establish whether the legal elements of a charged offense are met.
Is low empathy the same as psychopathy?
No. Psychopathy is a broader multidimensional construct. Neither a low empathy score nor a reported interpersonal difficulty is sufficient for a forensic psychopathy assessment.
Does empathy training lower reoffending rates?
Some programs report improved awareness or empathy measures, but reviews have not established a reliable, general reduction in recidivism from empathy training alone. Relevant outcomes require direct follow-up data and controlled evaluation.
How should empathy be considered in a forensic evaluation?
Only as a specifically defined and sufficiently validated piece of evidence related to the referral question. Evaluation should consider multiple information sources, response incentives, cultural and linguistic validity, the alternatives to the proposed interpretation, and the governing law.
Related Articles
References
American Psychological Association. (2013). Specialty guidelines for forensic psychology. American Psychologist, 68(1), 7–19. DOI: 10.1037/a0029889.
Day, A., Casey, S., & Gerace, A. (2010). Interventions to improve empathy awareness in sexual and violent offenders: Conceptual, empirical, and clinical issues. Aggression and Violent Behavior, 15(3), 201–208. DOI: 10.1016/j.avb.2009.12.003.
Further peer-reviewed research discussed above: Psychopathy, antisocial behavior, and empathy meta-analysis (2022); Youth aggression, empathy, and callous-unemotional traits meta-analysis (2022); General and victim-specific empathy and recidivism study; Performance-based and self-report empathy comparison (2026).
Jolliffe, D., & Farrington, D. P. (2004). Empathy and offending: A systematic review and meta-analysis. Aggression and Violent Behavior, 9(5), 441–476. DOI: 10.1016/j.avb.2003.03.001.
Mann, R. E., Hanson, R. K., & Thornton, D. (2010). Assessing risk for sexual recidivism: Some proposals on the nature of psychologically meaningful risk factors. Sexual Abuse, 22(2), 191–217. DOI: 10.1177/1079063210366039.
Trivedi-Bateman, N., & Crook, E. L. (2022). The optimal application of empathy interventions to reduce antisocial behaviour and crime: A review of the literature. Psychology, Crime & Law, 28(8), 796–819. DOI: 10.1080/1068316X.2021.1962870.
Vachon, D. D., Lynam, D. R., & Johnson, J. A. (2014). The (non)relation between empathy and aggression: Surprising results from a meta-analysis. Psychological Bulletin, 140(3), 751–773. DOI: 10.1037/a0035236.
van Langen, M. A. M., Wissink, I. B., van Vugt, E. S., van der Stouwe, T., & Stams, G. J. J. M. (2014). The relation between empathy and offending: A meta-analysis. Aggression and Violent Behavior, 19(2), 179–189. DOI: 10.1016/j.avb.2014.02.003.