A twelve-year-old girl comes back from an errand to find her family shot in their flat by men with badges. She keeps walking down the corridor, past her own door, and knocks on the neighbor's. Inside is a contract killer with a houseplant and no social life. Within days she is asking him, in plain language and with a specific vocabulary she has evidently thought about, to teach her the job. Almost every conversation about this film is about whether the relationship is defensible. The criminological question is narrower and stranger: she is right about how learning works.
// The theoryNine propositions, and the one everybody quotes wrong
Edwin Sutherland's theory of differential association (1947) is nine propositions long, and its reputation rests on a misreading. It does not say that bad company causes crime. It says that criminal behavior is learned in interaction with other persons in a process of communication, chiefly within intimate personal groups, and that what is learned includes both the techniques of committing the offense and the direction of motives, drives, rationalisations and attitudes. The eighth proposition is the one that gives the theory teeth: the process of learning criminal behavior involves all the mechanisms involved in any other learning. And the ninth closes the door on the easy explanation: criminal and non-criminal behavior are both expressions of the same general needs and values, so needs and values explain neither. A person does not steal because they want money; everyone wants money.
Akers's four mechanisms
Sutherland described a process he could not measure. Ronald Akers and Robert Burgess rebuilt it in 1966 on operant conditioning, and the result is the version that gets tested. It has four parts. Differential association: who you spend time with, weighted by priority, duration, intensity and frequency. Definitions: the attitudes and rationalizations that make an act seem permissible, necessary or neutral. Differential reinforcement: the anticipated and actual balance of rewards and punishments for the behavior. And imitation: modeling observed behavior, strongest when the model is admired and the behavior appears to work. This case supplies all four with unusual clarity — one intimate association of enormous intensity, a set of definitions delivered as professional ethics, reinforcement in the form of approval from the only adult who treats her as competent, and a model whose technique is displayed in detail. The theory's strength is that these are separable and each can be measured; its cost is that it is now four hypotheses wearing one name.
// The subjectThe apprentice who writes her own syllabus
What makes Mathilda useful is that she supplies the motivation herself and the training arrives afterwards. That ordering matters, because it separates two things social learning usually receives pre-mixed. Her intent—revenge for her four-year-old brother—is not learned from Léon; it predates him and he actively discourages it. What she learns from him is competence and a vocabulary: how to hold a rifle, how to case a building, and, more consequentially, the words “cleaner” and “no women, no kids”, which convert killing into a trade with standards. Sutherland would recognize that last item immediately. It is the transmission of definitions, and it is doing exactly what he said it does: not creating a desire, but making an act available as a category of legitimate work.
Albert Bandura's experiments supply the mechanism behind the imitation component, and one detail of them is usually lost. In the 1961 studies, children who watched an adult behave aggressively toward an inflatable doll reproduced the behavior — but reproduction depended heavily on whether the model was rewarded, punished or ignored, and on whether the child identified with the model. Mathilda identifies with Léon completely, and the environment rewards him with money, autonomy and a kind of dignity. The film even stages the vicarious-punishment condition: the one time she operates alone she is caught immediately, and the lesson lands.
Mathilda
Get the map of theories for free
Subscribe to the criminal archive —one case a week, no spam— and I'll send you The map of theories: the great currents of criminology on a single sheet.
// The crackWhich came first, the friend or the crime?
The oldest objection to differential association is also the hardest to answer: the direction of causality. Sheldon and Eleanor Glueck raised it in the 1950s and Travis Hirschi turned it into a slogan — birds of a feather flock together. If delinquent youths are found among delinquent friends, the association may be selection rather than socialization: people with a prior disposition seek out others like them. Decades of longitudinal work have produced a mixed answer with a clear shape — both processes operate, peer influence survives controls for prior delinquency, but the effect sizes shrink considerably once selection is modeled properly. A second, sharper criticism is that the theory struggles with offenses that require no tuition and no company: impulsive violence, a great deal of intimate-partner abuse, and the entirely solitary offender. A third is that “definitions favourable to violation of law” is measured with attitude scales so close to the outcome that some of the correlation is tautology. And there is a limit specific to this case that has nothing to do with criminology: the film frames an adult's instruction of a twelve-year-old as tenderness, and the framing has aged badly. The learning it depicts is real; the relationship that delivers it is grooming, and naming it is not a digression from the theory but an application of it — the intensity of the association is precisely what gives the definitions their force.
The responsibility lies with the adults
Mathilda is a child and the responsibility in this file sits with adults: the men who killed her family, and the man who handed a minor a rifle rather than a phone number for social services. Understanding her as an apprentice rather than a monster is not a softening — it is the more demanding reading, because it locates the failure in a chain of adults who each had an easier option. Social learning theory has one practical consequence worth holding onto: if criminal competence is taught, then it can be crowded out by teaching something else, and every functioning intervention for juveniles, from mentoring to vocational programs, is an attempt to win the same four levers.
Three key ideas
- Sutherland (1947): criminal behavior is learned in communication within intimate groups, and what is learned includes techniques and rationalizations. Needs and values explain nothing, because lawful behavior expresses the same ones.
- Akers rebuilt it as four measurable mechanisms — differential association, definitions, differential reinforcement and imitation. This case supplies all four, and the definitions arrive disguised as professional standards.
- The standing objection is selection versus socialization: delinquents choose delinquent friends. Both processes operate, but peer effects shrink once prior disposition is modeled. And the relationship delivering the training here is grooming, which is the theory working, not an exception to it.
Sources · to keep reading
- Sutherland, E. H. (1947). Principles of Criminology, 4th ed. Lippincott.
- Burgess, R. L. y Akers, R. L. (1966). “A Differential Association-Reinforcement Theory of Criminal Behavior”, Social Problems, 14(2).
- Akers, R. L. (1998). Social Learning and Social Structure. Northeastern University Press.
- Bandura, A. (1977). Social Learning Theory. Prentice Hall.
- Warr, M. (2002). Companions in Crime. Cambridge University Press.
- Pratt, T. C. et al. (2010). “The Empirical Status of Social Learning Theory: A Meta-Analysis”, Justice Quarterly, 27(6).
Case-file questionsFrequently asked questions
Who is Mathilda?
Mathilda is a villain from Léon: The Professional, analysed on KRIMINIS through the lens of criminology. Nobody recruits her.
Why does Mathilda become a villain?
Mathilda and social learning theory: the child who does not fall into crime but applies for it. The file on the four things an apprentice needs — which Sutherland listed before anyone else.
Which criminological theory explains Mathilda?
Mathilda's case is analysed through Social learning. Social learning theory solved a problem criminology had dodged for a century: nobody is born knowing how to offend. Edwin Sutherland set it out in 1947 in nine propositions, and the first is still the uncomfortable one: criminal behavior is learned, and it is learned the way anything else is — with techniques, with motives and with justifications. Ronald Akers later turned it into four measurable mechanisms. Mathilda is the textbook case because she reverses the usual direction: she is not a minor dragged along by bad company, she is a minor who identifies a trade, locates a practitioner and negotiates access to training. This file is not about a corrupted child. It is about what it takes for a person to learn anything, and why the list is identical for a violin and for a gun.
