Criminological Theory in a Dissertation (2026): Nine Theories, What Each One Predicts, and How to Test It

Most criminology dissertations that lose marks on theory lose them the same way. The theory is described in Chapter 3, correctly and at length, and then never appears again: the findings are reported, the discussion summarizes them, and the theory sits inert. A theoretical framework earns its place only when it produces a statement about your data that could turn out to be false. This guide takes the nine theories that dominate US criminology, and for each one gives the mechanism, the construct you would have to measure, the prediction the theory licenses, and the way that prediction usually fails. It is a companion to our pillar on writing a criminology dissertation, which covers structure, access and datasets.

What a working framework has to do

• Name a mechanism, not a topic. “Strain theory” is not a framework; “blocked opportunity produces negative affect, which produces offending among those without conventional coping resources” is.
• Identify the construct you can measure. Every theory below has a standard operationalization in the literature. If you cannot measure the construct in your data, choose another theory.
• State a prediction with a direction. A hypothesis that says two things are “related” cannot be wrong and therefore cannot be right.
• Name the rival explanation. A good criminology chapter says what a competing theory would predict instead, and what result would favor which.
• Return to the theory in the discussion. The best sentence in a criminology discussion chapter is the one explaining why the theory did not hold everywhere.

The nine frameworks, and what each one commits you to

Strain: measure the emotion, not just the disadvantage

Merton’s 1938 statement is macro: a culture that universalizes success goals while distributing legitimate means unequally produces adaptations, of which innovation is the criminal one. It is testable only with aggregate data, and a student who frames an individual-level survey in classic anomie has made the level error. Agnew’s general strain theory, published in Criminology in 1992, moved the argument to the individual and gave it a mediator: strains produce negative emotions, principally anger, and offending is one way of coping when legitimate coping is unavailable. That mediator is what makes the theory testable in a dissertation. The minimum credible test measures a strain, measures negative affect, measures offending, and models the indirect path — not a direct correlation between hardship and crime, which every theory in the table would also predict.

The standard failure is a study that measures strains and offending, finds an association, and declares support for general strain theory. Without the emotional mediator and a moderator for coping, the finding is equally consistent with social learning and with low self-control. Say what would have distinguished them.

Control theories: the prediction is generality

Hirschi’s social bond theory asks why people conform rather than why they offend, and locates the answer in four bonds: attachment, commitment, involvement and belief. Gottfredson and Hirschi’s later general theory replaces the bonds with a single stable trait, low self-control, established early and expressed in impulsivity, risk-seeking, short temper and preference for physical over mental tasks. The measurement standard is the Grasmick and colleagues (1993) self-control scale published in the Journal of Research in Crime and Delinquency, and if you use it you should report its reliability in your sample and say whether you treat it as unidimensional.

The distinctive commitment of control theory is generality: low self-control should predict every kind of misbehavior, including analogous non-criminal acts such as smoking, accidents and unstable employment. That is a falsifiable prediction, and it is the one your study should test. If low self-control predicts your outcome but not the analogous behaviors, that is a finding worth reporting, not an inconvenience.

Social learning: the peer problem you must address

Social learning theory, formalized by Burgess and Akers in Social Problems in 1966 and tested by Akers and colleagues in the American Sociological Review in 1979, holds that offending is learned through differential association, favorable definitions, imitation and differential reinforcement. It is the most consistently supported framework in the field and, for that reason, the easiest to test badly. The problem is selection: people who offend choose friends who offend, so a cross-sectional correlation between peer delinquency and one’s own delinquency is uninformative about causal direction. A defensible dissertation test needs longitudinal data with prior offending controlled, or a perceptual measure of peers’ definitions rather than a self-reported count of delinquent friends, and it needs to state plainly which of the four mechanisms it can and cannot separate.

Situational and neighborhood theories: match the theory to the geography

Routine activity theory (Cohen and Felson, 1979) is a theory of criminal events, not of criminal people: a crime requires a motivated offender, a suitable target and the absence of a capable guardian converging in space and time. It is the natural frame for studies of victimization risk, crime at places, repeat victimization, and the displacement effects of situational interventions. Its testable core is that changes in the routine activities of a population change crime rates without any change in criminal motivation, which is why it explains crime trends that offender-focused theories cannot.

Social disorganization moves the unit up to the neighborhood, and the 1997 Science paper by Sampson, Raudenbush and Earls gave the tradition its modern construct: collective efficacy, defined as social cohesion among neighbors combined with willingness to intervene for the common good, which they found mediated much of the association between concentrated disadvantage and violence. Testing it means multilevel data — individuals or incidents nested in neighborhoods — and the analysis follows accordingly; our guide to multilevel models for nested data covers the intraclass correlation and the level at which each predictor belongs. A common student error is to measure collective efficacy with a handful of items from one respondent per area, which reproduces the construct in name only.

Reaction theories: labeling, procedural justice and desistance

Labeling theory predicts that official intervention makes later offending more likely, by altering self-concept and by closing off conventional opportunities. Paternoster and Iovanni’s 1989 elaboration in Justice Quarterly is the source that made the theory testable, distinguishing the exclusion mechanism from the identity mechanism. Its prediction is directly opposed to deterrence, which is what makes it valuable in a dissertation: one framework says sanction reduces offending, the other says it increases it, and your data can adjudicate.

Procedural justice theory (Tyler) predicts that people comply with the law and cooperate with police because they regard the authority as legitimate, and that judgments of legitimacy rest on the fairness of the process rather than on the favorability of the outcome or the perceived probability of punishment. It is the standard frame for surveys of public attitudes to policing and is easy to operationalize with established scales for fairness, trust, obligation to obey, and willingness to cooperate.

Life-course criminology asks about change within people over time. Sampson and Laub’s 1990 paper in the American Sociological Review and their 2001 review in Crime and Justice frame desistance through turning points — marriage, stable employment, military service — that build informal social control. The theory’s specific claim is within-person: it is not that married people offend less than unmarried people, but that the same person offends less while married. That distinction dictates the analysis, which needs repeated measures and a within-person estimator, not a cross-sectional comparison of groups.

Turning the theory into your Chapter 3

1. Write the mechanism in one sentence, with a subject, a verb and an outcome. If you cannot, you do not yet have a framework.
2. Name the construct and the measure in the same paragraph. “Low self-control, measured with the 24-item Grasmick scale” is a framework doing work; “this study draws on self-control theory” is not.
3. Write the hypotheses directly under the theory, each one numbered and directional, and each traceable to a proposition you have just stated. Our guide to writing research questions covers the move from question to hypothesis.
4. Name one rival framework and its competing prediction. Label the result that would favor each. This single paragraph is what separates a first-class theory chapter from a competent one.
5. Reserve a discussion paragraph for the theory in advance. Draft the sentence that begins “If the association holds for property offenses but not for violence, that would suggest…” before you run the analysis.

The mechanics of the chapter itself — construct definitions, propositions, the model diagram — are the same in every field, and our step-by-step on writing a theoretical framework covers them. What differs is the catalogue: compare the criminology list above with the instruments in our guide to validated instruments for an education dissertation, or with the fourteen theories set out in our comparison of theoretical frameworks for a business dissertation, and the shape of each discipline’s theory chapter becomes clear.

Frequently asked questions

How many criminological theories should a dissertation use?

One as the primary framework. A second is defensible only when it makes a competing prediction you intend to adjudicate, which is a strength rather than a hedge. Describing four theories and testing none is the most common weakness in undergraduate criminology dissertations.

Can I use a macro theory with survey data?

Not as the primary framework. Classic strain and anomie are statements about societies and groups, and individual survey responses cannot test them. Use general strain theory instead, which was formulated at the individual level and specifies a mediator you can measure.

What is the difference between social learning and differential association?

Differential association, from Sutherland, holds that offending is learned through an excess of definitions favorable to law violation. Social learning theory keeps that idea and adds operant mechanisms: imitation and differential reinforcement. In practice social learning is the testable version, and it is the one most contemporary studies use.

How do I test collective efficacy without the Chicago data?

Use an existing community survey that includes cohesion and informal social control items aggregated to the area level, or an area-level proxy that you defend explicitly as a proxy. What you should not do is compute collective efficacy from a handful of respondents in each neighborhood and treat it as a neighborhood-level measure; report how many respondents per area you have and what that does to reliability.

Does labeling theory contradict deterrence theory?

On the effect of formal sanction, yes, and that is what makes the pair useful. Deterrence predicts that sanction lowers later offending through perceived cost; labeling predicts it raises later offending through exclusion and identity change. Set out both predictions, then say which pattern your data show.

Do qualitative criminology dissertations need a theoretical framework?

Yes, but it functions differently. In an interview or documentary study the framework supplies sensitizing concepts that shape the topic guide and the initial coding frame, and the analysis may extend or challenge it. Say at the outset whether you are testing the theory or building on it, because the two are written up differently.