Criminal Psychology

Abstract

Criminal psychology is a branch of applied psychology that studies the mental processes, motivations, and behaviour of people who commit crime, together with the psychological methods used to investigate, assess, and rehabilitate offenders. It supplies the empirical backbone for several justice-system practices: offender profiling, the evaluation of eyewitness testimony, the conduct of interrogations, and the prediction of reoffending. This article traces the field from its discredited nineteenth-century origins to its current evidence-based form. It sets out the leading theories of criminal behaviour, examines what research shows about offender profiling and its limits, reviews the fragility of eyewitness memory and the mechanics of false confession, and surveys the actuarial assessment of risk and the models that guide rehabilitation. Three demonstrations let the reader reproduce the misinformation effect, weigh the predictive value of a risk instrument, and trace the confidence–accuracy relationship.

Keywords: criminal psychology, offender profiling, eyewitness testimony

Crime has drawn psychological explanation for as long as there has been psychology, but the discipline that now goes by criminal psychology bears little resemblance to its origins. It began as an attempt to read criminality off the body and has become a set of empirical practices — measuring memory, modelling risk, testing the validity of investigative techniques — that are held to the same evidential standards as the rest of the field. Because MeSH files it beneath applied psychology, it is worth being clear at the outset what kind of applied psychology it is: not a theory of why crime exists, but the application of psychological science to the questions the justice system actually asks — who did it, can this witness be believed, is this confession real, and how likely is this person to offend again.

What Criminal Psychology Is

Criminal psychology is the study of the psychological factors involved in criminal behaviour and of the application of psychology to the detection, prosecution, and rehabilitation of offenders. It is often confused with forensic psychology, and the two overlap heavily, but the emphasis differs: criminal psychology centres on the offender — the cognition, personality, and development that bear on why crime occurs and how offending can be predicted and reduced — while forensic psychology denotes the broader application of psychology to all legal settings, including competency evaluation, civil litigation, and expert testimony. In practice the same researchers move between them, and the descriptor beneath which MeSH files this topic, applied psychology, is the common parent of both.

Three commitments distinguish the modern field from its predecessors. It is empirical: claims about offenders, witnesses, and interrogation are tested against data rather than asserted from intuition or authority. It is probabilistic: it deals in risk and likelihood, not in the fixed criminal types that earlier theories posited. And it is applied: its questions are set by the justice system, so its findings are judged not only by explanatory power but by whether they improve identification, testimony, interrogation, and rehabilitation (Eysenck, 1996).

Types of Criminal Psychology

MeSH treats criminal psychology as a narrower kind of applied psychology, its parent descriptor, and files one narrower topic beneath it in turn. Such a classification is an indexing hierarchy rather than a theory of the field: it records how the literature is catalogued, and its single child is best read as one specialized area of practice rather than an exhaustive taxonomy of everything criminal psychologists do. The subareas that follow — profiling, eyewitness research, interrogation, risk assessment — cut across it and are treated in their own sections below.

Theories of Criminal Behaviour

The field's origin is also its standing warning. In the 1870s the Italian physician Cesare Lombroso advanced criminal anthropology, the claim that criminals were an atavistic biological type identifiable by physical stigmata — skull shape, facial features, and the like. The theory was influential, empirically worthless, and pernicious, and it collapsed once its comparisons were made with proper controls. Its legacy is cautionary: it is the example against which the field now defines the difference between a testable psychological hypothesis and a prejudice dressed as science.

Twentieth-century theory replaced fixed types with dimensions and development. Hans Eysenck proposed that the traits underlying his personality model — extraversion, neuroticism, and psychoticism — are linked to differences in conditionability and arousal that make some people slower to acquire the socialized inhibitions on which law-abiding behaviour depends (Eysenck, 1996). A distinct and clinically central construct is psychopathy, operationalized by Robert Hare's Psychopathy Checklist–Revised (PCL-R) as a constellation of interpersonal, affective, and behavioural features — superficial charm, callousness, impulsivity — that predicts persistent and violent offending better than most single variables (Hare & Neumann, 2008).

A further shift located the roots of crime not in a fixed type or single trait but in development over the life course. Longitudinal studies that follow the same individuals from childhood — most notably David Farrington's Cambridge Study in Delinquent Development — established that a small number of persistent offenders account for a disproportionate share of crime and that early risk factors forecast later offending. Terrie Moffitt's influential developmental taxonomy sharpened this into two distinct pathways: a large group of adolescence-limited offenders whose antisocial behaviour is confined to the teenage years, and a small group of life-course-persistent offenders whose difficulties begin early and endure, reflecting the interaction of childhood neuropsychological deficits with an adverse rearing environment (Moffitt, 1993). More recently, neurocriminology has sought the biological correlates of antisocial behaviour in brain structure and function, autonomic reactivity, and their interaction with the social environment, while being careful not to repeat Lombroso's error of reading destiny off biology (Glenn & Raine, 2014).

Offender Profiling and Investigative Psychology

Offender profiling — inferring an unknown offender's characteristics from the features of a crime — is the field's most public activity and one of its most cautionary. The popular image derives from the intuitive, experience-based approach of early investigators; the scientific response was David Canter's investigative psychology, which sought to replace intuition with statistical models relating crime-scene actions to offender characteristics (Canter, 2004). The empirical verdict on profiling as ordinarily practised, however, has been sobering. A systematic review of the assumptions underlying profiling found weak support for the core premise that distinct offence styles map reliably onto distinct offender types, and argued that much of profiling's apparent success reflects ambiguity and hindsight rather than predictive power — the criminal profiling illusion (Snook et al., 2008).

A later meta-analysis synthesizing four decades of research reached a similarly measured conclusion: profiling can achieve above-chance accuracy on some inferences, but the effects are modest and rarely exceed what could be predicted from base rates alone (Fox & Farrington, 2018). The lesson is not that behaviour is uninformative but that its investigative value is easily overstated. A profile that seems to fit an apprehended suspect is not evidence that the profile identified the suspect, and the vividness of a good match is exactly the kind of cue that inflates confidence beyond what the data support.

Eyewitness Memory and Testimony

Nowhere has criminal psychology reshaped practice more than in the study of eyewitness memory. The foundational demonstration is Elizabeth Loftus and John Palmer's finding that the wording of a question alters what a witness reports: participants who were asked how fast cars were going when they “smashed” into each other gave higher speed estimates, and were later more likely to report broken glass that was never present, than those asked about cars that “hit” (Loftus & Palmer, 1974). The result established the misinformation effect: memory is reconstructive, and post-event information — including a single leading verb — can be absorbed into the remembered event. The applied review that followed drew out the implications for lineups, interviews, and instructions, and helped drive reforms in how identifications are conducted (Wells & Olson, 2003).

The picture is not one of uniform unreliability, and this is a point on which the field has refined its own message. A reanalysis of the confidence–accuracy relationship argued that eyewitness memory, like any measurement, is most informative under good conditions: on an initial, uncontaminated identification from a fair lineup, a witness's expressed confidence is in fact strongly diagnostic of accuracy, with high-confidence identifications proving highly accurate (Wixted et al., 2018). The threat to reliability comes less from memory itself than from the procedures that contaminate it — suggestive lineups, repeated questioning, and feedback — which is why the reform agenda targets procedure rather than dismissing witnesses.

Interrogation and False Confessions

That a person would confess to a crime they did not commit strikes most people as implausible, yet false confessions are a documented and recurring cause of wrongful conviction. The psychology behind them has two strands. The first is situational: certain interrogation techniques — prolonged questioning, the presentation of false evidence, and minimization that implies leniency — are powerful enough to produce admissions from the innocent, and a comprehensive review has catalogued these risk factors and the reforms, such as recording interrogations, that reduce them (Kassin et al., 2010). The second is dispositional: individuals differ in interrogative suggestibility, their tendency to yield to leading questions and to shift answers under pressure, a trait measurable with the scales developed by Gisli Gudjonsson.

In the most striking cases the two strands combine to corrupt memory itself. Gudjonsson has described a memory distrust syndrome in which suspects, made to doubt their own recollection, come to accept an interrogator's account and may confabulate details of an offence they did not commit (Gudjonsson, 2017). The phenomenon connects the interrogation room directly to the reconstructive memory demonstrated in the eyewitness literature: the same malleability that lets a leading verb reshape a witness's report lets a coercive interrogation reshape a suspect's.

Risk Assessment and Rehabilitation

The question the justice system most often puts to psychology is predictive: how likely is this person to offend again? For most of the twentieth century the answer came from unstructured clinical judgement, which proved little better than chance. Its replacement is actuarial risk assessment — structured instruments that combine empirically validated predictors into a score. A large meta-analysis of risk instruments found that they identify higher- and lower-risk groups with useful accuracy, but that their positive predictive value for serious violence is constrained by its low base rate: even a good instrument flags many who will not go on to offend (Fazel et al., 2012). This is a statistical fact about rare events, not a failure of the tools, and it sets a ceiling on how confidently any single prediction can be made.

Assessment is only useful if it guides action, and two frameworks dominate. The Risk–Need–Responsivity (RNR) model holds that intervention should be matched to the offender's risk level, should target the criminogenic needs that actually drive offending, and should be delivered in a way the person can respond to (Andrews et al., 2006). The Good Lives Model reframes rehabilitation more positively, arguing that reducing risk works best when it is paired with helping offenders build the capabilities to pursue legitimate goals (Ward & Brown, 2004). The two are often combined in practice: manage risk, address the needs that produce it, and give the person something to rehabilitate toward.

Interactive demonstrations

The three demonstrations below let the reader work with core findings of the field. The first reproduces the effect of question wording on eyewitness memory; the second computes the predictive value of a risk instrument as its accuracy and the base rate change; the third traces the confidence–accuracy relationship for an initial eyewitness identification.

Worked Example

Consider the risk-assessment model in the second demonstration, which makes concrete why a “high-risk” flag is not a prediction that a person will offend. The reasoning is Bayesian. Let the base rate of serious violence in the assessed population be p = 0.10 — that is, 1 in 10 will go on to offend. Let the instrument have a sensitivity (true-positive rate) of 0.80, correctly flagging 80% of those who will offend, and a specificity (true-negative rate) of 0.75, correctly clearing 75% of those who will not — so its false-positive rate is 0.25.

The positive predictive value — the probability that a flagged person actually offends — is PPV = (p × sensitivity) / [p × sensitivity + (1 − p) × false-positive rate]. Substituting: PPV = (0.10 × 0.80) / (0.10 × 0.80 + 0.90 × 0.25) = 0.080 / (0.080 + 0.225) = 0.080 / 0.305 = 0.262. So of every 100 people the instrument flags as high-risk, only about 26 will actually offend — and 74 will not, despite the flag. The negative predictive value is far higher: NPV = (0.90 × 0.75) / (0.90 × 0.75 + 0.10 × 0.20) = 0.675 / 0.695 = 0.971, so a person cleared by the instrument is very unlikely to offend.

The asymmetry is the whole point. A good instrument is excellent at identifying who is safe and only modest at identifying who is dangerous, because danger is rare and rarity dilutes the predictive value of any positive signal (Fazel et al., 2012). The demonstration recomputes both values live as the base rate and accuracy are changed; lowering the base rate drives the positive predictive value down further, which is why individual violence prediction remains inherently uncertain even with validated tools. These figures are illustrative of the statistical structure of prediction, not estimates from any single instrument.

Figure 1

Discussion

Read as a whole, criminal psychology is a discipline that has learned to distrust its own intuitions. Each of its major findings is, at bottom, a correction of a confident and wrong folk belief: that memory works like a recording, that an innocent person would never confess, that an expert can read an offender's identity from a crime scene, that a clinician can tell who is dangerous. In each case controlled research replaced a plausible intuition with a more accurate and more modest account, and in each case the accurate account has been slow to displace the intuitive one in courtrooms and interrogation rooms (Wells & Olson, 2003).

Two cautions temper the field's authority. The first is the gap between group-level findings and individual cases: risk instruments and confidence–accuracy curves describe populations, and applying them to a single defendant requires care about exactly the base-rate reasoning the worked example illustrates (Fazel et al., 2012). The second is the field's history: because criminal psychology has been used to justify prejudice before, its biological and predictive branches carry a special obligation to distinguish what the evidence supports from what it is wished to support. The strongest reading is that psychology has genuinely improved how the justice system handles memory, interrogation, and risk — while the temptation to claim more certainty than the data allow remains the field's characteristic danger (Glenn & Raine, 2014).

Current Directions

Three lines of work are currently active. The first is the reconciliation of the eyewitness literature: the once-blanket message that eyewitnesses are unreliable has given way to a more precise account distinguishing the malleability introduced by procedure from the genuine diagnostic value of an uncontaminated identification, with continuing debate over how far the optimistic reanalysis generalizes to field conditions (Wixted et al., 2018). The second is the maturing evidence base on profiling, where large-scale synthesis has replaced anecdote and set realistic expectations for what behavioural inference can and cannot deliver (Fox & Farrington, 2018).

The third is neurocriminology and its ethics. As imaging and physiological measures identify correlates of antisocial behaviour, the field faces sharpening questions about how such findings should — and should not — enter decisions about culpability, sentencing, and prevention, and how to pursue biological explanation without reviving the deterministic errors of its own past (Glenn & Raine, 2014). Across all three, the open problem is the same: how to move validated group-level science responsibly into decisions that are always, in the end, about an individual.

Common Misconceptions

“A profiler can identify an offender from the crime scene, as on television.”

Systematic review finds the core assumptions of profiling weakly supported and its accuracy rarely exceeding base rates; much of its apparent success reflects ambiguity and hindsight, the criminal profiling illusion (Snook et al., 2008).

“Eyewitness memory is simply unreliable.”

Memory is reconstructive and easily contaminated, but an initial, high-confidence identification from a fair, uncontaminated lineup is in fact strongly diagnostic of accuracy; the problem lies chiefly in suggestive procedure (Wixted et al., 2018).

“Innocent people never confess.”

False confessions are a documented cause of wrongful conviction, produced by coercive interrogation techniques interacting with individual suggestibility, and in extreme cases with a distrust of one's own memory (Kassin et al., 2010).

“A high-risk score means a person will reoffend.”

Because serious violence is rare, even a validated instrument's positive predictive value is modest: most people it flags will not offend, though those it clears are very unlikely to (Fazel et al., 2012).

Glossary

Actuarial risk assessment.

The prediction of reoffending using structured instruments that combine empirically validated predictors into a score, shown to outperform unstructured clinical judgement at the group level.

Base rate.

The prior probability of an outcome in a population; the low base rate of serious violence is what limits the positive predictive value of any risk instrument.

Criminal anthropology.

Lombroso's discredited nineteenth-century theory that criminals form an atavistic biological type identifiable by physical stigmata; now cited as a cautionary example of prejudice presented as science.

Criminal psychology.

A branch of applied psychology studying the mental processes and behaviour of offenders and the psychological methods used to investigate, assess, and rehabilitate them.

Criminogenic needs.

The dynamic, changeable risk factors that actually drive offending (such as antisocial attitudes and associates) and that effective rehabilitation targets.

Developmental taxonomy.

Moffitt's distinction between adolescence-limited offenders, whose antisocial behaviour is confined to the teenage years, and life-course-persistent offenders, whose difficulties begin early and endure.

False confession.

An admission of guilt by an innocent person, produced by coercive interrogation techniques interacting with individual suggestibility, and a documented cause of wrongful conviction.

Forensic psychology.

The broader application of psychology to legal settings, encompassing competency and civil evaluation and expert testimony as well as the offender-focused work of criminal psychology.

Good Lives Model.

A rehabilitation framework holding that risk is best reduced by helping offenders build the capabilities to pursue legitimate personal goals, complementing risk-focused approaches.

Interrogative suggestibility.

An individual's tendency to yield to leading questions and to shift answers under interpersonal pressure, measurable with the Gudjonsson Suggestibility Scales.

Investigative psychology.

Canter's empirical, statistical approach to offender profiling, which seeks to replace investigative intuition with validated relationships between crime-scene actions and offender characteristics.

Leading question.

A question whose wording presupposes or suggests a particular answer; in the eyewitness literature, a single leading verb can alter what a witness subsequently reports.

Lie detection.

The assessment of the truthfulness of statements through physiological, behavioural, or cognitive means; the narrower topic MeSH files beneath criminal psychology.

Misinformation effect.

The incorporation of misleading post-event information into memory of an event, demonstrated by Loftus and Palmer's leading-verb experiment and central to the fragility of eyewitness testimony.

Neurocriminology.

The study of the biological, neural, and psychophysiological correlates of antisocial and criminal behaviour and their interaction with the social environment.

Offender profiling.

The inference of an unknown offender's characteristics from features of a crime; empirically found to add little beyond base rates as ordinarily practised.

Positive predictive value.

The probability that a person flagged by a risk instrument actually offends; kept modest for serious violence by its low base rate even when the instrument is accurate.

Psychopathy.

A clinical construct combining interpersonal, affective, and behavioural features (superficial charm, callousness, impulsivity), operationalized by Hare's PCL-R and among the strongest single predictors of persistent offending.

Risk–Need–Responsivity (RNR) model.

A rehabilitation framework matching intervention intensity to risk, targeting criminogenic needs, and delivering treatment in a form the offender can respond to.

Key Researchers

David V. Canter. British psychologist who founded investigative psychology and put offender profiling on an empirical, statistical footing. Wikipedia

Hans J. Eysenck (1916–1997). German-British psychologist whose biologically grounded personality theory framed an influential dispositional account of criminal behaviour through differences in conditionability and arousal. Wikipedia · Wikidata

David P. Farrington (1944–2024). British criminologist who directed the Cambridge Study in Delinquent Development, a landmark longitudinal study of the developmental origins of offending. Wikipedia · Wikidata

Gisli H. Gudjonsson. Icelandic-British forensic psychologist who developed the Gudjonsson Suggestibility Scales and the concept of interrogative suggestibility, and is a leading authority on false confessions. ORCID · Wikipedia · Google Scholar

Robert D. Hare. Canadian forensic psychologist who developed the Psychopathy Checklist–Revised (PCL-R), the standard clinical instrument for assessing psychopathy. Wikipedia · Google Scholar · Faculty page

Elizabeth F. Loftus. American cognitive psychologist whose experiments on the misinformation effect and eyewitness memory reshaped the psychology of testimony and false memory. Wikipedia · Google Scholar · Faculty page

Cesare Lombroso (1835–1909). Italian physician whose theory of the “born criminal” founded criminal anthropology; now cited as the discredited starting point against which scientific criminal psychology defined itself. Wikipedia · Wikidata

Adrian Raine. British-American neurocriminologist whose brain-imaging and psychophysiological studies established biological and neural correlates of antisocial and violent behaviour. Wikipedia · Google Scholar · Faculty page

Frequently Asked Questions

What is criminal psychology in simple terms?
It is the application of psychological science to crime: understanding why people offend, and using psychology to improve how the justice system investigates crime, evaluates witnesses, conducts interrogations, and predicts and reduces reoffending. It is a branch of applied psychology focused on the offender (Eysenck, 1996).

How is criminal psychology different from forensic psychology?
The two overlap heavily. Criminal psychology centres on the offender and the causes and prediction of crime; forensic psychology is the broader application of psychology to all legal settings, including civil litigation, competency evaluation, and expert testimony.

Does offender profiling actually work?
Only weakly. Systematic review and meta-analysis find that profiling as ordinarily practised achieves at best modest, above-chance accuracy that rarely exceeds what base rates would predict, and that its apparent successes owe much to ambiguity and hindsight (Fox & Farrington, 2018).

Why is eyewitness testimony considered unreliable?
Because memory is reconstructive: post-event information, including a single leading question, can be absorbed into the remembered event, as Loftus and Palmer showed. But an initial, high-confidence identification from a fair lineup is actually quite accurate; the danger is suggestive procedure (Loftus & Palmer, 1974).

Why would an innocent person confess to a crime?
Coercive interrogation techniques — prolonged questioning, false evidence, and implied leniency — can produce false confessions, especially in suggestible individuals, and in extreme cases can make suspects distrust their own memory (Kassin et al., 2010).

What is psychopathy, and how is it measured?
Psychopathy is a constellation of interpersonal, affective, and behavioural traits — charm, callousness, impulsivity — measured by Hare's Psychopathy Checklist–Revised. It is one of the strongest single predictors of persistent and violent offending (Hare & Neumann, 2008).

Can psychologists predict who will commit violence?
Only imperfectly. Validated risk instruments predict group-level reoffending better than clinical judgement, but the low base rate of serious violence keeps the positive predictive value of any single prediction modest (Fazel et al., 2012).

What actually reduces reoffending?
Rehabilitation that matches intensity to risk, targets the criminogenic needs driving offending, and is delivered so the person can respond to it — the Risk–Need–Responsivity model — often paired with helping offenders build a viable, legitimate life (Andrews et al., 2006).

References