Sources and References

Primary texts — Akers, R. L. (2009). Social Learning and Social Structure: A General Theory of Crime and Deviance. New Brunswick, NJ: Transaction Publishers; McGloin, J. M. & Thomas, K. J. (2019). Peer influence and delinquency. Annual Review of Criminology, 2, 241–264; Sutherland, E. H. (1947). Principles of Criminology (4th ed.). Philadelphia: Lippincott; Gottfredson, M. R. & Hirschi, T. (1990). A General Theory of Crime. Stanford University Press.

Supplementary texts — McCarthy, B. (1996). The attitudes and actions of others: Tutelage and Sutherland’s theory of differential association. British Journal of Criminology, 36(1), 135–147; Gallupe, O., McLevey, J., & Brown, S. (2016). An experimental test of deviant modeling. Journal of Research in Crime and Delinquency, 53(4), 482–505; Weerman, F. M. & Smeenk, W. H. (2005). Peer similarity in delinquency for different types of friends. Criminology, 43(2), 499–524; Sampson, R. J. (1999). Techniques of research neutralization. Theoretical Criminology, 3(4), 438–451; Posick, C. & Rocque, M. (2019). Is crime natural or do we learn it? Journal of Criminal Justice Education, 30(1), 98–118; Giordano, P. C. (2020). Continuing education: Toward a life-course perspective on social learning. In D. P. Farrington, L. Kazemian, & A. R. Piquero (Eds.), The Oxford Handbook of Developmental and Life-Course Criminology; McCuddy, T. (2021). Peer delinquency among digital natives. Journal of Research in Crime and Delinquency, 58(4), 445–480; Hoeben, E. M. & Thomas, K. J. (2019). Peers and offender decision-making. Criminology & Public Policy, 18(4), 971–999; Haynie, D. L. (2001). Delinquent peers revisited: Does network structure matter? American Journal of Sociology, 106(4), 1013–1057; Boman, J. H. & Mowen, T. J. (2017). Building the ties that bind: How parenting styles and peer relationships shape reentry. Criminology & Public Policy, 16(3), 753–774; Taxman, F. S. (2017). Are you asking me to change my friends? Criminology & Public Policy, 16(3), 787–793; Papachristos, A. V. & Kirk, D. S. (2015). Changing the street dynamic: Evaluating Chicago’s Group Violence Reduction Strategy. Criminology & Public Policy, 14(3), 525–558; Warr, M. (2002). Companions in Crime. Cambridge University Press; Burgess, R. L. & Akers, R. L. (1966). A differential association-reinforcement theory of criminal behavior. Social Problems, 14(2), 128–147; Bandura, A. (1977). Social Learning Theory. Englewood Cliffs, NJ: Prentice Hall.

Online resources — Oxford Bibliographies in Criminology; Annual Review of Criminology archives.

Chapter 1: Introduction to Peers and Crime

The relationship between peers and criminal behaviour is one of the most robust and long-standing findings in all of criminology. Since at least the early twentieth century, researchers have observed that offending tends to cluster within social groups: individuals who associate with delinquent peers are far more likely to engage in delinquency themselves. This correlation holds across samples, across time periods, across cultures, and across virtually every type of offence that has been studied. Yet despite the strength and consistency of this association, its meaning remains deeply contested.

At the heart of the debate lies a deceptively simple question: does associating with delinquent peers cause delinquency, or does the correlation arise for other reasons? Some scholars argue that peers actively teach, model, and reinforce criminal behaviour – that crime is fundamentally a learned phenomenon. Others contend that the apparent peer effect is largely or entirely spurious, driven by selection effects (birds of a feather flock together) or by the projection of one’s own behaviour onto descriptions of one’s friends. Still others take intermediate positions, acknowledging both influence and selection while trying to disentangle their relative contributions.

This course examines the peer-crime relationship in depth. We begin with the major theoretical frameworks – differential association theory and social learning theory – that provide the intellectual foundation for the peer influence perspective. We then explore how peer effects are measured, what methodological challenges arise, and how critics from the self-control and life-course traditions have challenged the causal interpretation of peer influence. Later chapters examine how peer influence operates across the life course, in digital environments, within network structures, in situational decision-making contexts, and through the lens of intervention design.

The goal is not merely to survey these topics but to develop the capacity for critical engagement with a complex, multi-layered body of scholarship. Understanding the peer-crime nexus requires grappling with questions of theory, measurement, causality, and policy – and recognizing that these questions are deeply intertwined.

Chapter 2: Theoretical Foundations

Sutherland’s Differential Association Theory

Edwin Sutherland first articulated his theory of differential association in the 1939 edition of Principles of Criminology, revising it into its mature nine-proposition form in the 1947 edition. Differential association theory was revolutionary for its time because it offered a general, social-psychological explanation of crime that applied across social classes and demographic categories. Where earlier theories had focused on poverty, biological defect, or psychological pathology, Sutherland insisted that criminal behaviour was learned through the same processes by which all behaviour is learned.

The Nine Propositions

Sutherland’s theory can be summarized through its nine core propositions:

1. Criminal behaviour is learned.
2. Criminal behaviour is learned in interaction with other persons through a process of communication.
3. The principal part of the learning of criminal behaviour occurs within intimate personal groups.
4. The learning includes (a) techniques of committing crime and (b) the specific direction of motives, drives, rationalizations, and attitudes.
5. The specific direction of motives and drives is learned from definitions of the legal codes as favourable or unfavourable.
6. A person becomes delinquent because of an excess of definitions favourable to violation of law over definitions unfavourable to violation of law. This is the principle of differential association.
7. Differential associations may vary in frequency, duration, priority, and intensity.
8. The process of learning criminal behaviour by association with criminal and anti-criminal patterns involves all of the mechanisms that are involved in any other learning.
9. While criminal behaviour is an expression of general needs and values, it is not explained by those general needs and values, since non-criminal behaviour is an expression of the same needs and values.

Key Contributions and Limitations

Sutherland’s insistence that crime is learned – not inherited, not the product of poverty per se, and not reducible to individual pathology – was a landmark contribution. The theory directed attention to the content of what is learned (techniques, motives, definitions) and to the social context in which learning occurs (intimate personal groups rather than media or casual contacts).

However, the theory also had significant limitations. Sutherland never specified the precise learning mechanisms by which definitions are acquired and translated into behaviour. He spoke of “all the mechanisms that are involved in any other learning” without identifying what those mechanisms were. The concept of “definitions favourable to law violation” remained somewhat vague – are these verbal statements, cognitive schemas, normative orientations, or something else entirely? And the theory was notoriously difficult to test empirically, since the ratio of favourable to unfavourable definitions could not be easily measured across a person’s entire history of associations.

McCarthy’s Extension: Tutelage

Bill McCarthy (1996) extended Sutherland’s framework by emphasizing the concept of tutelage – the direct, concrete instruction in criminal techniques and opportunities that occurs between experienced and novice offenders. McCarthy argued that much of the criminological literature had focused on the attitudinal component of differential association (definitions favourable to crime) while neglecting the behavioural component (techniques of committing crime). Using data on homeless youth, McCarthy demonstrated that tutelage in specific criminal skills – how to shoplift, how to sell drugs, how to break into cars – was a significant predictor of subsequent offending, even after controlling for attitudes and prior behaviour. This work highlighted that peer influence is not only about changing how people think but also about equipping them with the practical knowledge to act.

Akers’ Social Learning Theory

Ronald Akers, initially in collaboration with Robert Burgess (Burgess & Akers, 1966), developed social learning theory as a more comprehensive and testable reformulation of Sutherland’s differential association. Where Sutherland had left the learning mechanisms unspecified, Akers drew on operant conditioning (from B. F. Skinner) and social learning theory (from Albert Bandura) to provide a detailed account of how criminal behaviour is acquired, maintained, and modified.

The Four Central Concepts

Akers’ social learning theory rests on four interrelated concepts:

1. Differential Association

This concept retains Sutherland’s core insight but refines it. Differential association refers to the process by which individuals are exposed to normative definitions, models of behaviour, and social reinforcements through their interactions with others. The most important associations are those that occur earliest in life (priority), most frequently (frequency), over the longest duration (duration), and with the greatest emotional closeness or prestige (intensity). Akers emphasized that differential association is primarily a concept about exposure – it describes the social context in which learning takes place.

2. Definitions

Definitions are the attitudes, beliefs, orientations, and rationalizations that an individual holds regarding whether particular behaviours are right or wrong, justified or unjustified, appropriate or inappropriate. Akers distinguished between general definitions (broad moral or religious beliefs about right and wrong) and specific definitions (attitudes toward particular acts in particular situations). He also distinguished between positive definitions (which actively approve of deviant conduct), negative definitions (which disapprove), and neutralizing definitions (which do not positively endorse crime but provide justifications, excuses, or rationalizations that free the individual to commit it). Neutralizing definitions are particularly important because they are more common than outright positive definitions – most offenders do not believe crime is morally good, but they develop ways of neutralizing their moral inhibitions.

3. Differential Reinforcement

Differential reinforcement refers to the balance of anticipated and actual rewards and punishments associated with a given behaviour. Behaviour is strengthened when it is rewarded (positive reinforcement), when it removes an aversive stimulus (negative reinforcement), and when alternative behaviours are punished. Behaviour is weakened when it is punished (positive punishment) or when rewards are withdrawn (negative punishment). Akers emphasized that reinforcement can be social (approval, status, acceptance from peers) or non-social (the intrinsic physiological effects of drug use, the material gains from theft). The concept of differential reinforcement introduced a dynamic, consequentialist dimension to the theory: behaviour is not only learned through prior associations but is maintained or extinguished through its ongoing consequences.

4. Imitation

Imitation (or modelling) refers to the process by which individuals learn behaviours by observing them in others. Drawing on Bandura’s work, Akers argued that individuals are more likely to imitate behaviours that they observe being rewarded, that are performed by models who are admired or similar to the observer, and that are salient or vivid. Imitation is particularly important in the initiation of new behaviours – a person who has never used drugs may begin after watching a friend use them – but becomes less important relative to differential reinforcement in the maintenance of behaviour over time.

The Social Structure–Social Learning Model (SSSL)

In Social Learning and Social Structure (2009), Akers extended his theory by specifying how macro-level social structural variables – including sociodemographic characteristics, community context, and social disorganization – affect individual behaviour. His argument was that social structure operates through social learning mechanisms. Structural conditions do not directly cause crime; rather, they shape the patterns of differential association, the content of definitions, the schedules of reinforcement, and the availability of models to which individuals are exposed. A neighbourhood with high poverty and low collective efficacy, for example, produces crime not because poverty is criminogenic per se but because it creates social environments in which pro-criminal definitions are more prevalent, criminal behaviour is more frequently modelled, and the social rewards for crime (status, material gain) outweigh the rewards for conformity.

Empirical Support

Akers’ social learning theory has received extensive empirical support across a wide range of studies. Meta-analyses consistently find that the social learning variables – particularly differential association and definitions – are among the strongest correlates of criminal and deviant behaviour. The theory has been applied to drug use, violent crime, property crime, white-collar crime, sexual offending, and numerous other forms of deviance. Critics acknowledge the empirical strength of the theory but raise questions about whether the observed correlations truly reflect causal learning processes, as we shall explore in subsequent chapters.

Chapter 3: Measuring Peer Effects

The Measurement Problem

One of the most consequential methodological debates in the peer influence literature concerns how peer delinquency is measured. The vast majority of studies in this area rely on survey self-reports, in which respondents are asked both about their own behaviour and about the behaviour of their friends. This creates a fundamental measurement challenge that has generated decades of controversy.

Perceptual vs. Behavioural Measures

The standard approach in criminological research is to ask respondents to estimate how many of their friends have engaged in various delinquent behaviours (e.g., “How many of your close friends have stolen something worth more than $50 in the past year?”). These are called perceptual measures because they capture the respondent’s perception of peer behaviour rather than measuring peer behaviour directly.

An alternative approach is to collect behavioural measures by surveying the peers themselves – asking each friend to report on their own behaviour and then using those self-reports as the measure of peer delinquency. This approach is far more labour-intensive and is only possible in designs where friendship networks can be identified and all members surveyed.

The distinction matters enormously because perceptual and behavioural measures produce very different results. Studies that use perceptual measures consistently find a very strong association between peer delinquency and personal delinquency – often the strongest predictor in multivariate models. Studies that use behavioural measures find a much more modest association, often reduced by half or more.

Projection Bias

The leading explanation for this discrepancy is projection bias (also called assumed similarity or false consensus). Projection bias occurs when respondents attribute their own behaviours, attitudes, and values to their friends. A respondent who shoplifts may assume – perhaps incorrectly – that their friends also shoplift. This tendency inflates the correlation between self-reported delinquency and perceived peer delinquency, making it appear that peer influence is stronger than it actually is.

Weerman and Smeenk (2005): Peer Similarity Revisited

Weerman and Smeenk (2005) directly addressed the perceptual-behavioural measurement gap using data from Dutch adolescents. Their study was notable for its ability to compare perceptual measures (respondents’ reports about their friends) with behavioural measures (friends’ own self-reports) within the same sample. They found that:

1. Perceptual measures substantially overstated the degree of peer similarity in delinquency.
2. When behavioural measures were used, the association between peer delinquency and personal delinquency was weaker but still statistically significant.
3. The degree of actual similarity varied across types of friendship – similarity was greater among reciprocated (mutual) friendships than among unreciprocated ones.

These findings suggest that while projection bias inflates the peer-crime correlation, it does not eliminate it entirely. There is a real, if more modest, association between friends’ actual behaviour and one’s own behaviour.

Gallupe et al. (2016): An Experimental Test

Gallupe, McLevey, and Brown (2016) took a different approach to the measurement problem by conducting an experimental study of deviant modelling. Rather than relying on survey correlations, they placed participants in controlled laboratory settings where they could observe the behaviour of confederates (research assistants posing as participants) who either did or did not engage in mildly deviant acts. This experimental design allowed the researchers to establish a causal relationship between observing deviant behaviour and engaging in it oneself, bypassing the projection bias problem entirely.

The results showed that participants who observed confederates engaging in deviant behaviour were significantly more likely to do so themselves, providing direct experimental evidence for imitation effects consistent with social learning theory. This study is important because it demonstrates that peer influence is not merely an artefact of measurement or cognitive bias – at least some component of the peer effect operates through genuine social learning processes.

Additional Measurement Challenges

Beyond projection bias, several other measurement issues complicate the study of peer effects:

Temporal ordering: Cross-sectional studies cannot determine whether peer associations precede and cause delinquency, or whether delinquency precedes and causes changes in peer associations (selection). Longitudinal designs improve on this but still face challenges in establishing the precise timing of influence versus selection processes.

Network boundary specification: Researchers must decide which relationships count as “peers.” Are they best friends? All friends? Classmates? Neighbourhood acquaintances? Online contacts? Different boundary specifications yield different results, and most studies rely on a small number of nominated friends rather than capturing the full scope of an individual’s social network.

Heterogeneity of influence: Peer influence may vary in strength depending on the type of offence, the nature of the relationship, the characteristics of the individual, and the social context. A single estimate of “the peer effect” may obscure important variation.

Chapter 4: Challenges to the Peer Influence Perspective

The Self-Control Challenge: Gottfredson and Hirschi

The most influential challenge to the peer influence perspective comes from Gottfredson and Hirschi’s (1990) general theory of crime, which locates the primary cause of crime in individual differences in self-control rather than in peer relationships. In A General Theory of Crime, Gottfredson and Hirschi argued that people with low self-control are impulsive, insensitive, risk-seeking, short-sighted, and oriented toward simple rather than complex tasks. These traits, which are established early in life through ineffective parenting, are the primary and sufficient cause of criminal behaviour.

From this perspective, the observed correlation between peer delinquency and personal delinquency is spurious – it arises because people with low self-control both tend to commit crimes and tend to associate with others who have low self-control. The correlation does not reflect a causal influence process but rather a selection effect: birds of a feather flock together.

Gottfredson and Hirschi went further, arguing that offenders do not actually have deep, stable friendships at all. Because people with low self-control are self-centred, insensitive, and volatile, their relationships tend to be superficial and transient. They may co-offend, but this co-offending reflects convergence in time and space (driven by shared low self-control) rather than genuine peer influence. In this view, “delinquent peers” are not so much friends who teach crime as fellow low-self-control individuals who happen to be present when opportunities for crime arise.

Evaluating the Self-Control Challenge

Empirical tests of the self-control versus peer influence debate have generally found that both variables matter. When self-control and peer delinquency are entered into the same regression model, both typically remain significant predictors of crime. This suggests that the peer-crime association is not entirely spurious – even after accounting for individual propensity, peer context adds explanatory power. However, the relative importance of the two factors varies across studies, samples, and types of offending.

Sampson’s “Techniques of Research Neutralization”

Robert Sampson (1999) offered a provocative methodological critique of the peer influence literature in his essay “Techniques of Research Neutralization.” The title deliberately echoed Sykes and Matza’s famous “techniques of neutralization” theory, suggesting that researchers themselves use rhetorical strategies to avoid confronting the weaknesses in their evidence.

Sampson’s central argument was that the peer influence literature systematically neutralizes the threat of alternative explanations through several “techniques”:

1. Denial of selection: Researchers acknowledge that selection effects exist but then proceed as though their statistical controls adequately address the problem, when in fact they typically do not.
2. Appeal to correlation: The strength and consistency of the peer-crime correlation is treated as evidence of causation, when strong correlations can arise from confounding.
3. Condemnation of the condemners: Critics who raise selection or measurement concerns are dismissed as ideologically motivated rather than substantively engaged.
4. Definitional tautology: In some formulations, “peer delinquency” and “personal delinquency” are measured so similarly that the correlation between them is partly definitional.

Sampson did not argue that peer influence is non-existent; rather, he argued that the evidence for it is weaker and more ambiguous than its proponents typically acknowledge. His critique pushed the field to take selection effects and measurement artefacts more seriously and contributed to the development of more rigorous research designs.

Posick and Rocque (2019): Nature vs. Nurture in Crime

Posick and Rocque (2019) raised a different kind of challenge by asking whether criminal behaviour might have a natural or innate component rather than being entirely learned. Drawing on evolutionary psychology and biosocial criminology, they questioned the assumption – central to both differential association and social learning theory – that crime is wholly a product of social learning. They did not argue that biology determines crime, but they suggested that certain behavioural tendencies (aggression, risk-taking, sensation-seeking) have evolutionary roots and may interact with social learning processes in ways that the purely social learning perspective fails to capture.

This challenge is important because it questions the foundational premise of the peer influence perspective: that criminal behaviour is learned through social interaction. If some component of criminal behaviour reflects innate tendencies, then the role of peers may be more limited than social learning theory suggests – peers may provide opportunities and contexts for the expression of pre-existing tendencies rather than creating those tendencies through learning.

Distinguishing Influence from Selection: Empirical Strategies

Given the persistent challenge of disentangling peer influence from peer selection, researchers have developed several empirical strategies:

Longitudinal panel designs measure peer associations and delinquency at multiple time points, allowing researchers to assess whether changes in peer associations predict subsequent changes in delinquency (influence) and whether changes in delinquency predict subsequent changes in peer associations (selection). Most studies find evidence for both processes, though their relative importance varies.

Fixed-effects models control for all time-invariant individual characteristics (including unmeasured self-control), thereby addressing the selection critique more rigorously than standard regression. When fixed-effects models still find peer effects, this provides stronger evidence for influence.

Experimental and quasi-experimental designs (like Gallupe et al., 2016) establish causality by manipulating exposure to deviant peers or models while holding other factors constant.

Social network analysis uses mathematical models of network formation and behaviour change (such as SIENA models) to simultaneously estimate selection and influence effects within a network over time.

Chapter 5: The Life-Course Perspective on Peer Influence

Beyond Adolescence

The peer influence literature has historically concentrated almost exclusively on adolescence, treating the teenage years as the critical – and implicitly the only – period in which peer influence on crime operates. This focus is understandable: adolescence is when peer relationships become most intense, when offending rates peak, and when the peer-crime correlation is strongest. But it has created a somewhat distorted picture by neglecting how peer influence operates across the full life course.

Giordano (2020): Continuing Education

Peggy Giordano (2020) argued for a life-course perspective on social learning that extends the analysis of peer influence beyond adolescence into adulthood and across major life transitions. Drawing on the broader life-course criminology tradition (particularly the work of Sampson and Laub), Giordano contended that social learning is a continuous, lifelong process – people continue to learn from their associates throughout adulthood, and changes in peer networks associated with life transitions (marriage, employment, parenthood, incarceration, military service) can promote either desistance from or persistence in crime.

Key Arguments

Adult peer influence is real but different from adolescent peer influence. Adult relationships tend to be more stable, more selective, and more embedded in institutional contexts (workplaces, families, community organizations). The mechanisms of influence may shift: where adolescent peer influence often operates through direct modelling and status seeking, adult peer influence may operate more through normative expectations, social support, and the reinforcement of conventional identities.

Life transitions restructure peer networks. Marriage, for example, tends to reduce contact with delinquent peers and increase contact with conventional associates – the partner, the partner’s family, couples’ friends. This network restructuring can promote desistance by altering the balance of pro-criminal and pro-social definitions and reinforcements. Conversely, imprisonment tends to increase exposure to criminal associates and criminal learning opportunities, potentially deepening criminal involvement.

Cognitive transformations interact with peer influence. Giordano emphasized that peer influence is not a purely mechanical process of exposure and conditioning. Individuals actively interpret and respond to their social environments, and cognitive changes – shifts in identity, in future orientation, in the meaning attributed to relationships – mediate the effects of peer influence. A person who develops a “replacement self” (a conventional identity that is incompatible with offending) may become resistant to pro-criminal peer influence even if their network has not changed.

Implications for Sampson and Laub’s Age-Graded Theory

Giordano’s life-course perspective on social learning creates an interesting dialogue with Sampson and Laub’s age-graded theory of informal social control. Sampson and Laub argued that desistance from crime is primarily driven by the acquisition of social bonds (marriage, employment, military service) rather than by changes in peer associations per se. From their perspective, the key mechanism is social control – the stake in conformity that adult social bonds create – not social learning.

Giordano’s framework suggests a more integrated view: life transitions affect crime partly through social control (by increasing the costs of offending) and partly through social learning (by restructuring peer networks and altering the patterns of definitions and reinforcements to which individuals are exposed). These mechanisms are not mutually exclusive; they likely operate simultaneously and reinforce one another.

Chapter 6: Digital Peers and Social Media

The Digital Transformation of Peer Influence

The rise of social media and digital communication technologies has fundamentally transformed the landscape of peer interaction, raising new questions about how peer influence on crime operates in the digital age. Traditional theories of peer influence were developed in an era when peer interaction was primarily face-to-face and geographically bounded. Today, adolescents and young adults maintain extensive peer networks that span online and offline contexts, and much of their interaction occurs through digital platforms.

McCuddy (2021): Peer Delinquency among Digital Natives

Timothy McCuddy (2021) examined how peer delinquency operates in digital contexts, introducing the concept of digital native peer influence. His work addressed several key questions:

Does Online Peer Exposure Matter?

McCuddy found that exposure to delinquent behaviour and pro-delinquent definitions through social media and digital platforms was associated with personal delinquency, even after controlling for offline peer associations. This suggests that the social learning processes described by Akers – differential association, definitions, modelling, and reinforcement – operate in online environments as well as offline ones.

How Does Digital Peer Influence Differ from Traditional Peer Influence?

Several features of digital communication may alter the dynamics of peer influence:

Expanded reach: Social media allows exposure to a much larger and more diverse set of peers than traditional face-to-face interaction. This can increase exposure to deviant models and pro-criminal definitions, but it can also increase exposure to pro-social influences.

Permanence and visibility: Posts, images, and videos on social media can be viewed by large audiences and persist over time, potentially amplifying the modelling effects that Akers described. A single video depicting violence or drug use can reach thousands of viewers, functioning as a kind of mass-mediated modelling.

Reduced social cues: Online communication often lacks the non-verbal cues (facial expressions, body language, tone of voice) present in face-to-face interaction. This may make it easier to transmit pro-criminal definitions without the tempering effects of social disapproval that might be expressed in person.

Anonymity and disinhibition: Some digital platforms allow anonymous or pseudonymous interaction, which may reduce the social costs of endorsing deviant behaviour and increase willingness to share deviant content.

Blurred boundaries between peers and non-peers: On social media, the distinction between intimate friends and distant acquaintances becomes blurred. Sutherland emphasized that the most important learning occurs within “intimate personal groups,” but social media may extend influence effects to more peripheral contacts – or even to strangers.

Cybercrime and Online Deviance

Beyond the digitization of traditional peer influence, the internet has created entirely new forms of deviance (hacking, online fraud, cyberbullying, distribution of illegal content) that are often learned through online peer networks. The forums, chat rooms, and encrypted messaging groups where cybercriminals share techniques and knowledge function as digital equivalents of the “intimate personal groups” described by Sutherland, and the learning processes they facilitate align closely with both differential association and social learning theory.

Chapter 7: Peer Networks and Social Structure

Beyond Dyadic Peer Relationships

Most research on peers and crime has focused on dyadic relationships – the connection between an individual and their friends considered one at a time. But individuals are embedded in complex social networks, and the structure of those networks may shape the dynamics of peer influence in important ways. Social network analysis provides tools for examining how the architecture of peer relationships – not just their content – affects criminal behaviour.

Haynie (2001): Does Network Structure Matter?

Dana Haynie’s (2001) landmark study used data from the National Longitudinal Study of Adolescent Health (Add Health) to examine whether features of adolescents’ friendship networks predicted delinquency beyond the simple proportion of delinquent friends. She measured several structural properties of peer networks:

Network Density

Network density refers to the degree to which an individual’s friends are also friends with one another. A dense network is one in which everyone knows everyone else (high interconnection); a sparse network is one in which the individual’s friends are largely strangers to each other. Haynie found that network density moderated the effect of peer delinquency: the influence of delinquent peers was stronger in dense networks. This makes theoretical sense because dense networks create more consistent exposure to pro-criminal definitions and reinforcements (everyone in the group is reinforcing the same norms) and make it harder for the individual to escape the influence of the group.

Network Centrality

Centrality measures how connected an individual is within the broader network. Central individuals have many connections and occupy structurally important positions; peripheral individuals have few connections. Haynie found that more central individuals were more susceptible to peer influence – suggesting that being deeply embedded in a network amplifies the learning processes described by social learning theory.

Popularity and Status

Haynie also examined the role of popularity (measured by the number of friendship nominations received). Popular adolescents were more delinquent on average, possibly because status-seeking and risk-taking behaviours that contribute to delinquency also contribute to popularity in adolescent social hierarchies.

Implications for Social Learning Theory

Haynie’s findings enriched social learning theory by demonstrating that the structure of social relationships matters, not just their content. Two individuals with the same proportion of delinquent friends may experience very different levels of peer influence depending on whether their networks are dense or sparse, whether they are central or peripheral, and whether their delinquent friends know each other. This insight points toward a more sociologically sophisticated version of social learning theory that takes network architecture seriously.

Peer Groups vs. Gangs

An important distinction in the peer influence literature is between informal peer groups and more formalized gangs. While much of the research discussed in this course focuses on informal peer relationships, gangs represent an intensified form of peer influence in which group identity, status hierarchies, initiation rituals, and territorial norms create particularly strong pressures toward criminal behaviour. Gang membership is associated with elevated rates of offending that exceed what would be predicted by the delinquency of one’s associates alone, suggesting that the organizational structure of the gang adds something beyond mere peer influence.

Mark Warr (2002), in Companions in Crime, emphasized that most delinquency is a group phenomenon – the majority of juvenile offences are committed in the company of others. But the group dynamics that produce co-offending are not always well captured by social learning theory, which focuses on the learning of attitudes and behaviours over time. Some co-offending may reflect situational dynamics – the excitement, competition, and mutual encouragement that occur in the moment of a criminal event – rather than prior learning.

Chapter 8: Decision-Making and Situational Peer Influence

Peers in the Moment

Much of the peer influence literature focuses on the developmental question: how do peer associations over time shape an individual’s propensity toward crime? But peers also influence crime in real time, at the moment when criminal opportunities arise and decisions are made. This situational dimension of peer influence has received increasing attention.

Hoeben and Thomas (2019): Peers and Offender Decision-Making

Hoeben and Thomas (2019) examined how the presence of peers shapes offender decision-making processes. Drawing on rational choice and situational action theories, they argued that peers affect not only an individual’s general propensity toward crime (through social learning) but also their assessment of specific criminal opportunities in the moment.

Mechanisms of Situational Peer Influence

Encouragement and dares: Peers may directly encourage or dare individuals to commit offences, creating social pressure to act that the individual might not experience alone.

Audience effects: The presence of an audience – particularly an audience of peers whose approval is valued – can alter the perceived rewards of criminal behaviour. Acts that might seem pointless when alone (vandalism, fighting, petty theft) may become rewarding when they earn admiration, laughter, or respect from peers.

Diffusion of responsibility: When multiple individuals are present, each may feel less personally responsible for the outcome, lowering the perceived moral cost of the act. This is related to the well-documented bystander effect in social psychology.

Information and opportunity provision: Peers may provide concrete information about opportunities (e.g., an unlocked car, an unguarded store) or about the likelihood of detection, thereby altering the individual’s assessment of the risks involved.

Emotional amplification: Group settings can heighten emotional arousal – excitement, anger, bravado – that pushes individuals toward action. The group process literature emphasizes that collective emotional dynamics can produce behaviours that none of the individual participants would have engaged in alone.

Integrating Developmental and Situational Perspectives

Hoeben and Thomas argued for an integration of developmental and situational approaches to peer influence. Social learning theory explains how individuals develop a general orientation toward crime through exposure to pro-criminal definitions, models, and reinforcements. But whether this general orientation is translated into action depends on situational factors, including the presence and behaviour of peers at the moment of decision. A complete account of peer influence must encompass both the distal, developmental processes emphasized by Akers and the proximate, situational processes emphasized by rational choice and routine activities perspectives.

Unstructured Socializing and Routine Activities

The connection between peers and crime is also illuminated by routine activities theory and the concept of unstructured socializing. Osgood and colleagues demonstrated that spending time with peers in unstructured, unsupervised settings (hanging out with friends without organized activities or adult supervision) is one of the strongest situational predictors of delinquency. This effect holds even after controlling for the delinquency of one’s peers, suggesting that the mere context of unstructured socializing creates criminogenic opportunities independent of social learning processes.

The convergence of motivated offenders, suitable targets, and the absence of capable guardians – the three elements of routine activities theory – is most likely when adolescents gather in unstructured settings. Peers do not need to teach or encourage crime in these settings; they simply need to be present in an environment where opportunities and temptations abound and social controls are weak.

Chapter 9: Interventions and Policy Implications

From Theory to Practice

If peers influence criminal behaviour, then interventions that alter peer associations, disrupt pro-criminal learning processes, or leverage peer networks for pro-social influence should be effective in reducing crime. The design and evaluation of such interventions represents one of the most practically important applications of the peer influence literature.

Boman and Mowen (2017): Building the Ties that Bind

Boman and Mowen (2017) examined the role of peer relationships in the reentry process for individuals returning to the community after incarceration. They argued that successful reentry depends critically on the development of pro-social peer networks and the severing (or reduction) of ties to anti-social associates. Drawing on social learning theory, they emphasized that reentry interventions should not merely focus on removing negative peer influences but should actively help returning citizens build positive social connections.

Key Findings and Recommendations

Boman and Mowen found that parenting styles during adolescence shaped the quality of peer relationships in adulthood, which in turn affected reentry outcomes. Individuals who had experienced authoritative parenting (warm, supportive, but with clear boundaries) were more likely to develop pro-social peer networks as adults, which in turn reduced recidivism. This finding highlights the intergenerational dimension of peer influence: the family context in which social skills and relationship patterns are first learned shapes the peer relationships that subsequently influence criminal behaviour.

Their policy recommendations included:

• Family-based interventions that strengthen parenting skills and family bonds to facilitate the development of pro-social peer relationships.
• Mentoring programs that connect returning citizens with pro-social role models who can serve as sources of conventional definitions and reinforcements.
• Structured reentry programs that provide opportunities for positive peer interaction (employment training, community service, support groups) while helping participants navigate the challenge of distancing themselves from anti-social associates.

Taxman (2017): “Are You Asking Me to Change My Friends?”

Faye Taxman (2017) offered a critical perspective on peer-based interventions by highlighting the practical and ethical challenges of asking individuals to change their peer networks. The title of her article captures a fundamental tension: criminal justice interventions routinely instruct offenders to “avoid anti-social associates,” but this directive ignores the social reality that peer relationships are deeply embedded in individuals’ lives and identities.

Practical Challenges

Peer networks are not easily changed. Individuals’ friendships are shaped by neighbourhood, school, work, family, and institutional contexts. Simply telling someone to “find new friends” without altering the structural conditions that constrain their social opportunities is often unrealistic.

Pro-social networks may not be accessible. For individuals with criminal records, limited education, and few resources, access to conventional peer networks (through employment, community organizations, or educational institutions) may be severely constrained. The very factors that contribute to criminal involvement also limit access to pro-social alternatives.

Severing ties has costs. Even anti-social peer relationships provide emotional support, companionship, and practical assistance. Asking individuals to abandon these relationships without providing adequate alternatives may leave them socially isolated, which is itself a risk factor for reoffending.

Papachristos and Kirk (2015): Changing the Street Dynamic

Papachristos and Kirk (2015) evaluated Chicago’s Group Violence Reduction Strategy (GVRS), a programme modelled on David Kennedy’s Cure Violence (formerly CeaseFire) approach. This intervention represents a fundamentally different approach to leveraging peer networks: rather than asking individuals to change their friends, it attempts to change the normative environment within existing peer networks.

The Cure Violence / Focused Deterrence Model

The Cure Violence model treats violence as an epidemic – a contagion that spreads through social networks much as an infectious disease spreads through physical contact. The intervention has several key components:

Violence interrupters: Community members with credibility and connections in high-risk networks (often former gang members) are employed as mediators who intervene in conflicts before they escalate to violence. These individuals operate as counter-models – they demonstrate that violence is not inevitable and that conflicts can be resolved without it.

Focused deterrence / “call-ins”: Law enforcement and community leaders deliver a direct message to members of high-risk groups (typically gangs or crews identified through network analysis as most likely to be involved in gun violence). The message combines a deterrence component (“If your group is involved in violence, the full resources of law enforcement will be directed at your group”) with a services component (“If you want to change, we will help you with employment, housing, and education”).

Community mobilization: The programme seeks to shift community norms against violence by engaging clergy, teachers, parents, and other community stakeholders in a coordinated campaign to change the “street code” that legitimizes violent responses to disrespect and conflict.

Evaluation Findings

Papachristos and Kirk’s evaluation of the Chicago GVRS found mixed but promising results. In some target areas, the intervention was associated with significant reductions in gun violence; in others, the effects were more modest. The variability in outcomes appeared to be related to the quality of implementation – the programme was most effective when the full complement of components (deterrence, services, community mobilization) was delivered consistently and when violence interrupters had genuine credibility within the target networks.

Several broader lessons emerge from the Cure Violence evaluation:

1. Network-based interventions can work, but they require sophisticated understanding of local network structures and dynamics.
2. Implementation quality matters enormously. Theoretical elegance does not guarantee practical effectiveness; the gap between programme design and programme delivery is often large.
3. Combining deterrence with services appears more effective than either approach alone. Pure enforcement may suppress violence temporarily but does not change the underlying norms and learning processes; pure service provision may not be sufficient to overcome the powerful reinforcements associated with the street code.
4. Sustainability is a challenge. Maintaining violence interruption programmes requires ongoing funding, community trust, and institutional support, all of which are vulnerable to political and economic fluctuations.

Broader Intervention Approaches

Beyond the specific programmes discussed above, the peer influence literature suggests several general principles for intervention design:

Cognitive-Behavioural Interventions

If criminal behaviour is sustained by pro-criminal definitions (attitudes, beliefs, rationalizations), then interventions that target cognition – helping individuals recognize and challenge their pro-criminal thinking patterns – should reduce offending. Cognitive-behavioural therapy (CBT) programmes have substantial empirical support in corrections and are broadly consistent with social learning theory’s emphasis on definitions as a key mediator of behaviour.

Prosocial Modelling

If imitation is a key learning mechanism, then providing access to pro-social models – individuals who demonstrate that success and satisfaction can be achieved through conventional means – should promote desistance. Mentoring programmes, restorative justice conferences (where offenders encounter the perspectives of victims and community members), and therapeutic communities all incorporate prosocial modelling as a core element.

Reinforcement-Based Approaches

If behaviour is maintained through differential reinforcement, then interventions should ensure that the rewards for conformity exceed the rewards for crime. Employment programmes, educational opportunities, housing assistance, and other material supports alter the reinforcement contingencies facing individuals by increasing the tangible benefits of conventional behaviour.

Group-Based Interventions: Cautions

An important caution from the peer influence literature is that group-based interventions can backfire. Programmes that aggregate high-risk youth (such as group therapy, boot camps, or juvenile detention) create precisely the conditions that social learning theory predicts will increase offending: concentrated exposure to deviant models, pro-criminal definitions, and social reinforcement for anti-social behaviour. The phenomenon known as deviancy training – in which group-based interventions inadvertently teach criminal techniques and attitudes – is well documented and represents a serious iatrogenic risk.

Chapter 10: Integration and Future Directions

Synthesizing the Evidence

The evidence reviewed across this course supports several broad conclusions about the relationship between peers and crime:

1. Peer influence on criminal behaviour is real, but its magnitude is often overstated. The strongest correlations between peer delinquency and personal delinquency arise in part from measurement artefacts, particularly projection bias. When behavioural measures are used, the association is weaker but still significant. Experimental studies confirm that at least some component of the peer effect operates through genuine social learning processes (imitation, modelling).

2. Selection effects are also real and substantial. Individuals choose friends who are similar to themselves in attitudes and behaviour, and this selection process accounts for a significant portion of the peer-crime correlation. Neither a pure influence model nor a pure selection model is adequate; both processes operate simultaneously.

3. The mechanisms of peer influence are multiple and varied. Peer influence operates through the learning of definitions (attitudes, beliefs, neutralizations), through imitation and modelling, through differential reinforcement (social rewards and punishments), through the provision of criminal opportunities and techniques (tutelage), and through situational dynamics (encouragement, dares, emotional amplification, diffusion of responsibility). No single mechanism is sufficient to explain the full range of peer effects.

4. Context matters. The strength and nature of peer influence depend on the structure of peer networks (density, centrality), the stage of the life course (adolescence vs. adulthood), the medium of interaction (face-to-face vs. digital), the type of offence, and the broader social and institutional context.

5. Interventions must be designed with care. Effective peer-based interventions leverage social learning processes to promote pro-social behaviour (through prosocial modelling, cognitive restructuring, reinforcement of conventional behaviour, and norm change). Poorly designed interventions that aggregate high-risk individuals can inadvertently amplify the very peer influence processes they aim to counteract.

Unresolved Questions

Several important questions remain unresolved and represent promising directions for future research:

How do online and offline peer influence interact? As digital communication becomes increasingly central to social life, understanding how online exposure to deviant content interacts with offline peer relationships will be essential. Do online and offline peer influences operate through the same mechanisms, or does the digital context create qualitatively different processes?

How does peer influence operate for different types of crime? Most research focuses on common street crime and substance use. Less is known about peer influence in white-collar crime, domestic violence, sexual offending, terrorism, and other domains. The mechanisms of peer influence may differ substantially across offence types.

What is the role of network dynamics in desistance? While we know that network changes are associated with desistance from crime, the causal mechanisms are poorly understood. Do individuals first decide to stop offending and then change their networks, or do network changes drive cognitive and behavioural change? How do structural constraints (neighbourhood, employment, incarceration) shape the possibility of network transformation?

How can biosocial factors be integrated with social learning theory? The challenge from biosocial criminology – that biological predispositions interact with social learning processes – remains largely unaddressed in the mainstream peer influence literature. Understanding gene-environment interactions in the context of peer influence could lead to more nuanced and effective interventions.

What role do weak ties and peripheral contacts play? Sutherland emphasized “intimate personal groups,” and most research focuses on close friends. But the sociological literature on the “strength of weak ties” suggests that more distant contacts can also be influential – particularly in spreading information, norms, and opportunities. The role of weak ties in criminal learning is understudied.

Conclusion

The study of peers and crime sits at the intersection of criminological theory, social psychology, network science, and public policy. The field has made substantial progress since Sutherland’s original formulation of differential association theory nearly a century ago. We now have a much more nuanced understanding of the mechanisms through which peers influence crime, the conditions under which this influence is stronger or weaker, the competing and complementary explanations for the peer-crime correlation, and the practical challenges of designing interventions that harness peer dynamics for public safety.

At the same time, the field’s persistent challenges – disentangling influence from selection, overcoming measurement limitations, integrating individual-level and structural perspectives, accounting for digital transformation – ensure that the study of peers and crime will remain a vibrant and evolving area of scholarship for years to come. The most productive path forward lies not in defending any single theoretical framework but in pursuing rigorous empirical research that tests competing explanations, develops better measurement strategies, and evaluates interventions with the same critical scrutiny that we apply to the theories on which they are based.