Abstract
Prior research indicates that criminal thought content and criminal thought process mediate the peer influence effect (peer delinquency → participant delinquency). This study sought to model the temporal order of these two categories of antisocial cognition in mediating peer influence. Responses provided by 1,795 (847 male, 948 female) members of the Gang Resistance Education and Training study (mean age = 12.11 years) on measures of criminal thought content (negative attitude toward the police) and criminal thought process (proactive criminal thinking in the form of neutralizing techniques) were analyzed. Only the “content before process” model achieved significance, however. Thus, while peer associations may stimulate criminal thought content and criminal thought process, the effect on criminal thought content may be more proximal and the effect on criminal thought process may be mediated by criminal thought content.
Similar content being viewed by others
Dividing cognitive functions into content and process is a common practice in psychiatry, psychology, and philosophy. In psychiatry, thought content and process are important features of the standard mental status exam (Trzepacz and Baker 1993). In psychology, the complex interplay of thought content and process is viewed as essential to normal human cognitive and social development (Vygotsky 1994). In philosophy, there is a branch of inquiry that examines the nature and meaning of mental events, functions, properties, and mind-body duality. This branch of philosophy, commonly referred to as the philosophy of mind, has as one of its principal goals, understanding the relationship between thought content and process (Gelfert 2015). The purpose of the current investigation is much less ambitious but no less important, namely, to determine whether there is a temporal relationship between criminal thought content (the substance or object of criminal thought or what a criminal thinks, e.g., negative attitudes toward the police and positive attitudes toward deviance) and criminal thought process (the pattern or form of criminal thought or how a criminal thinks, e.g., reactive/impulsive criminal thinking versus proactive/instrumental criminal thinking) in mediating the peer influence effect that begins with delinquent peer associations and ends with an increase in a person’s own involvement in delinquent and criminal behavior.
Although criminal thinking has not been as extensively researched in criminology as it has in forensic psychology, it does assume a central position in several popular theories of crime. Sykes and Matza’s (1957) techniques of neutralization, for instance, are designed to excuse criminal behavior and eliminate feelings of guilt following performance of a criminal act. These instruments of cognitive persuasion can be considered, along with moral disengagement (Bandura 1999), facets of proactive (planned, calculated, morally dismissive) criminal thought process or how an offender thinks. The alibis and excuses offenders use to justify and rationalize their involvement in crime are the essence of proactive criminal thinking. Sutherland’s (1947) definitions favorable to violation of the law, on the other hand, are more reflective of criminal thought content or what an offender thinks. Hence, the predominance of antisocial thought content over prosocial thought content learned over the course of one’s interactions with those already involved in crime is what is responsible for delinquency and criminal offending, at least according to Sutherland and other proponents of social learning theories of crime. Walters (2017) has taken these two forms of antisocial cognition—criminal thought process and criminal thought content—to create a cognitive theory of crime.
Borrowing the terms proactive and reactive from the developmental psychological literature on proactive and reactive childhood aggression (Dodge and Coie 1987), Walters (2017) contends that proactive and reactive criminal thinking represent distinct but overlapping dimensions of criminal thought process. Whereas proactive criminal thinking personifies the cold, calculated, and planned features of criminal thought process, reactive criminal thinking embodies the emotional, impulsive, and irresponsible aspects of criminal thought process. Both, however, represent the process or pattern of thought rather than its content. Thought process is the term used to describe the mental operations that give rise to decision-making, problem solving, critical analysis, and moral reasoning. In the criminal context, it is characterized by short-term decision-making, faulty problem solving, and weak critical analytic skills in the case of reactive criminal thinking, and neutralization of guilt, power-oriented motives, and lack of moral reasoning in the case of proactive criminal thinking. The proactive processes assist the individual in achieving various criminal goals by justifying continued involvement in antisocial behavior and making it easier for the individual to disregard the rights of others and the laws of society.
Criminal thought process, while important, is only half the antisocial cognition equation. The other half is criminal thought content. In developing and validating a self-report measure of criminal thought content, Walters and Morgan (2018) divided criminal thought content into three dimensions: negative attitude toward authority, positive attitude toward deviance, and criminal identity. The negative attitude toward authority dimension of criminal thought content delineates an individual’s lack of respect for and adherence to direction from authority figures: parents, teachers, and police officers, among others. By contrast, the positive attitude toward deviance dimension portrays a decidedly positive attitude toward non-criminal (e.g., lying to parents or spouse) and criminal (e.g., lying in court) forms of deviant behavior. The third and final dimension of criminal thought content—criminal identity—is marked by a sense of identification with all things criminal. In validating this three-dimensional model of criminal thought content, Walters and Morgan (2018) obtained moderately strong correlations between all three dimensions and self-reported criminal offending in two large samples of college students.
As social cognitive variables, criminal thought process and criminal thought content may well serve as mediators of important criminological relationships (Bandura 1986; Wu and Zumbo 2008). The prospect of using criminal thought process and content to mediate major criminological relationships was tested in a series of studies by Walters (2015, 2016b), the results of which showed that proactive but not reactive criminal thinking mediated the peer influence effect, which runs from peer delinquency to participant delinquency. In light of research showing that criminal thought content correlates significantly better with the proactive dimension of criminal thought process than with the reactive dimension (Morgan et al. 2010), it was reasoned that criminal thought content might also be capable of mediating the peer influence effect. A subsequent study showed that both criminal thought process, as represented by proactive criminal thinking, and criminal thought content, in the form of positive attitudes toward deviance, effectively mediated the relationship between prior peer delinquency and subsequent participant delinquency (Walters 2016a). In line with one of the research hypotheses for Walters (2016a), the effect revealed itself to be additive in that both mediators achieved significance when entered simultaneously in a parallel-mediator multiple mediation path analysis of the peer influence effect.
The Present Study
There were two features of the Walters (2016a) investigation that the current study sought to improve upon. First, the criminal thought content measure in Walters (2016a) was restricted to positive attitudes toward deviance items that asked whether it was wrong for someone to engage in various criminal and non-criminal antisocial acts. The proactive criminal thinking measure likewise asked whether it was okay to hit someone if they were provoked or lie to one’s parents or school officials to protect one’s friends. The similarity in item content between the two sets of items suggested that they may have been measuring the same construct to some extent and that greater conceptual distinction was required to provide a fair test of the research hypotheses. In the current study, negative attitude toward the police, one of many possible expressions of a negative attitude toward authority, was paired with proactive criminal thinking. This pairing was viewed as more conceptually distinct than the one between the proactive criminal thinking and positive attitude toward deviance measures in Walters (2016a). Second, the focus of the Walters (2016a) investigation was on moderately proximal mediation by multiple parallel mediators. The current study, by contrast, examined more distal mediation by cross-lagging mediators across two waves of data. Thus, while Walters (2016a) studied parallel relationships between mediators over a shorter period, the current study explored parallel and serial relationships between mediators over a longer period. In effect, the current study can be considered a replication and extension of the earlier Walters (2016a) investigation.
The purpose of the current investigation was to assess whether criminal thought process and criminal thought content are capable of mediating the peer influence effect (peer delinquency → participant delinquency). Control variables for this study included demographic measures (age, sex, race, parent education), treatment status (whether or not participants received gang-resistance training), and three social indicators (feel safe in the school neighborhood, parental knowledge, and unsupervised routine activities). The three social indicators were selected as control variables based on the fact each has been found to be moderately associated with peer and/or participant delinquency (Augustyn and McGloin 2013; Johnson et al. 2012; Lahey et al. 2008; McGloin and Shermer 2009). The first or replication hypothesis held that both criminal thought content and criminal thought process would mediate the relationship between peer delinquency and participant delinquency. The second or extension hypothesis held that by extending the mediator effect out another wave and cross-lagging the two mediators, significant serial mediator pathways would surface. One could argue on the basis of the second hypothesis that the mediating effect should be stronger going from criminal thought content (negative attitude toward the police) to criminal thought process (proactive criminal thinking) because content is easier to assimilate than process. Conversely, criminal thought process might precede criminal thought content in situations where criminal thought content evolves from a faulty or irrational foundation of criminal thought process. The second hypothesis, predicated on reciprocity (Thornberry 1987), predicted that criminal thought content should precede criminal thought process (“content before process”) and vice versa (“process before content”), thereby producing a bidirectional effect.
Method
Participants
Participants for this study were 1,795 (847 male, 948 female) youth from the Gang Resistance Education and Training (GREAT: Esbensen 2002) longitudinal study (mean age = 12.11 years, SD = .60). The current sample consisted of all GREAT participants who had complete data on at least three of the six main (independent, mediator, and dependent) variables for this study (50.3% of the 3,568 total participants). The 6-wave GREAT longitudinal study was conducted in six US cities (Philadelphia, PA; Portland, OR; Phoenix, AZ; Omaha, Ne; Lincoln, Ne; and Las Cruces, N.Mex) between 1995 and 1999. The first five waves of the GREAT longitudinal study were included in the present investigation. At wave 1, when the study began, participants were enrolled in the sixth or seventh grades (mean age = 12.15 years). At wave 5, when the study ended, most participants were in the ninth or tenth grades (mean age = 15.16 years). The racial/ethnic distribution of the sample was 53.8% White, 18.2% Hispanic, 15.8% African-American, 3.8% Asian/Pacific Islander, 2.5% Native American, and 6.1% other or mixed.
Measures
Independent Variable
The independent variable for this study was perceived peer delinquency measured at wave 2 of the GREAT longitudinal study. In completing this measure, participants were asked to estimate the proportion of friends (1 = none of them, 2 = a few of them, 3 = about half of them, 4 = most of them, 5 = all of them) who were involved in 10 different delinquent acts (i.e., “destroyed property,” “stole < $50,” “stole > $50,” “entered building to steal,” “stole a motor vehicle,” “hit someone,” “attacked someone with a weapon,” “committed armed robbery,” “sold marijuana,” and “sold other illegal drugs”) over the past year. The ratings were summed to produce a score that could range from 10 to 50. The internal consistency of this scale in the current sample of participants was excellent (α = .91).
Mediator Variables
The two mediator variables were cross-lagged at waves 3 and 4 of the GREAT study. One of these variables was a measure of criminal thought content (i.e., negative attitude toward the police: NAP) and consisted of 6 items (“police are honest,” “police are rude,” “police are hardworking,” “police are friendly,” “police are courteous,” “police are respectful to people like me”). Each item was rated on a five-point Likert-type scale (1 = strongly agree, 2 = agree, 3 = neither agree nor disagree, 4 = disagree, 5 = strongly disagree), with the “police are rude” item reverse-coded (1 = strongly disagree, 5 = strongly agree). The results were then summed to produce a score that could range from 6 to 30. The negative attitude toward the police scale achieved good internal consistency in the current sample of participants (α = .89).
The other mediator variable for this study was a measure of criminal thought process (i.e., proactive criminal thinking; PCT). This variable was assessed with 9 “neutralization” items (“small lies okay if no one hurt,” “okay to lie to keep friends out of trouble,” “okay to lie to keep you out of trouble,” “okay to steal from rich who can replace item,” “okay taking little things from stores,” “okay to steal if only way to get it,” “okay to physically fight if hit first,” “okay to physically fight to protect rights,” “okay to fight if threaten friends/family”), neutralization being a facet of proactive criminal thinking (Walters 2017). Although the item “okay to physically fight if hit first” may seem reactive at first blush, it actually reflects justification for physical aggression, which is a proactive mechanism. The 9 items were rated on a five-point Likert-type scale (1 = strongly disagree, 2 = disagree, 3 = neither agree nor disagree, 4 = agree, 5 = strongly agree) and the results summed to produce a score that could range from 9 to 45. The internal consistency of this 9-item scale in the current sample of participants was strong (α = .88).
Dependent Variable
The single dependent variable for this study was self-reported delinquent involvement at wave 4 (replication analysis) or wave 5 (extension analysis). Participants were asked to indicate how often they engaged in 14 different delinquent acts over the past six months (destroyed property, carried a weapon, spray painted a building, stole < $50, stole > $50, went into building to steal, stole a motor vehicle, hit someone, attacked someone with a weapon, committed armed robbery, involved in a gang fight, shot someone, sold marijuana, and sold other drugs). Frequency counts were organized into a seven-point scale (0 = no times, 1 = one or two times, 2 = three to five times, 3 = six to ten times, 4 = eleven to fifteen times, 5 = sixteen to twenty times, 6 = more than twenty times) and summed. Scores on this measure ranged from 0 to 84 and the one-year test-retest reliability between waves 2 and 5 ranged from .38 to .54. The lower correlation may reflect developmental factors in that the prevalence of offending rose moderately (26%) from wave 2 (mean age = 12.33 years) to wave 3 (mean age = 13.18), where it peaked and then gradually declined during waves 4 and 5.
Control Variables
The present study included eight control variables. Four of the control variables were demographic in nature—age (in years), sex (1 = male, 2 = female), race (1 = white, 2 = non-white), and parental education as a proxy for family socioeconomic status (1 = grade school or less, 2 = some high school, 3 = high school graduate, 4 = some college, 5 = college graduate, 6 = more than college). A fifth control variable was whether or not the child’s classroom received the GREAT anti-gang curriculum at wave 1 (1 = non-GREAT classroom, 2 = GREAT classroom). The last three control variables were whether the child felt safe in the school neighborhood (1= strongly disagree, 2 = disagree, 3 = neither agree nor disagree, 4 = agree, 5 = strongly agree), parental knowledge (“parents know where I am” and “parents know who I am with,” each rated on a 5-point Likert-type scale: 1 = strongly disagree, 5 = strongly agree), and child involvement in unsupervised routine activities (0 = no hours per week, 1 = one or two hours per week, 2 = three to five hours per week, 3 = six to ten hours per week, 4 = eleven to fifteen hours per week, 5 = sixteen to twenty hours per week, 6 = more than twenty hours per week).
Cole and Maxwell (2003) maintain that prior levels of a predicted variable should be controlled when performing mediation analysis. Because there were five predicted variables in this study (wave 3 negative attitude toward the police, wave 3 proactive criminal thinking, wave 4 negative attitude toward the police, wave 4 proactive criminal thinking, and wave 5 delinquency), there were five precursor measures. Specifically, the wave 2 negative attitude toward the police measure was included as a precursor in the equations predicting wave 3 and wave 4 negative attitude toward the police, the wave 2 proactive criminal thinking measure was included as a precursor in the equations predicting waves 3 and 4 proactive criminal thinking, and wave 2 delinquency was included as a precursor in the equation predicting wave 5 delinquency.
Research Design and Procedure
A four-wave longitudinal fixed panel design and five-wave longitudinal fixed panel design were used to conduct the replication and extension portions of this study, respectively. The first or replication part of the current study employed the first four waves of the GREAT longitudinal study. The second or extension part of the current study utilized the first five waves of the GREAT study. Eleven weeks separated waves 1 and 2 of the GREAT study and one year separated each subsequent pair of waves (2 and 3, 3 and 4, 4 and 5). There was no overlap between waves, denoting that the study employed a prospective design. In the replication portion of the study, the 8 control variables were assessed at wave 1, the independent variable (peer delinquency) was assessed at wave 2, the mediator variables (negative attitudes toward the police and proactive criminal thinking) were assessed at wave 3, and the dependent variable (offending) was assessed at wave 4. In the extension portion of the study, the 8 control variables were assessed at wave 1, the independent variable was assessed at wave 2, the first-stage mediators were assessed at wave 3, the second-stage mediators were assessed at wave 4, and the dependent variable was assessed at wave 5. Two different pathways were compared in the extension portion of this study: a “content before process” pathway (Peer-2 → NAP-3 → PCT-4 → Delinquency-5) and a “process before content” pathway (Peer-2 → PCT-3 → NAP-4 → Delinquency-5).
Data Analytic Plan
A path analysis was performed (Hayes 2013; Preacher 2015) using MPlus 8.1 (Muthén and Muthén 2017). Cases were clustered within classrooms and analyses were computed using a maximum likelihood with robust parameters and standard errors (MLR) estimator. The MLR estimator was used in order to accommodate the data, which were skewed and clustered within classrooms. Because MLR does not support bootstrapping, all indirect (mediated) effects were evaluated by means of 95% confidence intervals constructed using Preacher and Selig’s (2012) Monte Carlo Method for Assessing Mediation (MCMAM). A significant 95% confidence interval is one that does not include zero. The MCMAM procedure was performed with 20,000 repetitions.
Kenny’s (2013) “failsafe ef” procedure, (rmy.x) × (sdm.x) × (sdy.x)/(sdm) × (sdy), was used to test for omitted variable bias (Imai et al. 2010). The coefficient produced by the “failsafe ef” procedure indicates how strongly an unobserved covariate confounder would need to correlate with the mediating variable and partially correlate with the dependent variable, controlling for the mediator and independent variables, to completely eliminate the b path of a significant indirect effect. Because conditioning on the precursor to an outcome variable can create endogenous selection bias and inflate path coefficients (Elwert and Winship 2014), a second sensitivity analysis was performed whereby the precursor measures for the five regression equations in the extension portion of the study were removed from the analysis.
Missing Data
A little more than a third of the sample had complete data on all 18 variables (35.0%) in the extension portion of the study and close to a quarter of the sample had complete data on 17 out of 18 variables (22.7%). Another 10.8% were missing data on two variables, 10.3% were missing data on three variables, 13.9% were missing data on four or five variables, and 7.3% were missing data on six to eleven variables. Individual variables with more than 10% missing data included parental educational level (31.9%), feel safe in school neighborhood (10.2%), unsupervised routine activities (19.2%), Peer-2 (13.2%), NAP-2 (11.7%), NAP-4 (14.8%), PCT-2 (11.8%), PCT-4 (15.2), Delinquency-2 (10.9%), and Delinquency-5 (25.2%). Missing data were handled in both the replication and extension analyzing with full information maximum likelihood (FIML). The FIML procedure calculates model parameters and standard errors based on observed relationships between non-missing data.
FIML is robust to violations of its basic assumptions (Collins et al. 2001; Young and Johnson 2013) and has been found to yield significantly less biased results than traditional missing data procedures like listwise deletion, pairwise deletion, and simple regression-based imputation (Allison 2012). To further enhance the precision of FIML (Collins et al. 2001), 15 auxiliary variables (Peer-1, Peer-3, Peer-4, Peer-5, Peer-6, NAP-1, NAP-5, NAP-6, PCT-1, PCT-5, PCT-6, Delinquency-1, Delinquency-3, Delinquency-4, and Delinquency-6) were added to the procedure. Auxiliary variables are used to calculate parameters and standard errors for FIML but are not included in the actual analyses.
Results
Preliminary Analyses
Table 1 provides descriptive statistics and correlations for the 18 variables included in the replication and extension portions of the current investigation. Approximately two-thirds of the correlations between the 18 variables in Table 1 were significant using a Bonferroni-corrected alpha level (p = .00033). Testing the various regression equations for collinearity failed to reveal signs of multicollinearity: Tolerance = .433–.990; variance inflation factor = 1.010–2.308.
Nearly half of the sample (44.5%) reported engaging in at least delinquent act and about a third of the sample reported committing at least one of seven serious offenses (stole > $50, went into building to steal, stole a motor vehicle, hit someone, attacked someone with a weapon, committed armed robbery, shot someone) at wave 5.
Main Analyses
The replication portion of this study was conducted using a three-equation path analysis, the results of which are summarized in Fig. 1. As the results of this analysis indicate, three of the four path coefficients (a and b in the content-mediated pathway and b in the process-mediated pathway) achieved significance. In addition, the MCMAM procedure revealed a significant total indirect effect for the content-mediated pathway (Peer-2 → NAP-3 → Delinquency-4; 95% CI = .00141, .01903) but not for the process-mediated pathway (Peer-2 → PCT-3 → Delinquency-4; 95% CI = − .00103, .02527).
The results of the extension portion of this study are summarized in Table 2 (see also, Fig. 2). According to the results of a five-equation path analysis, the a (from independent variable to first mediator), d (from first-stage mediator to second-stage mediator), and b (from second-stage mediator to dependent variable) paths of the “content before process” pathway were significant, whereas only the d path of the “process before content” pathway was significant. The MCMAM results further revealed that only the total indirect effect for the “content before process” pathway (Peer-2 → NAP-3 → PCT-4 → Delinquency-5) was significant (see Table 3).
When the dependent variable was confined to the seven most serious offenses (stole > $50, went into building to steal, stole a motor vehicle, hit someone, attacked someone with a weapon, committed armed robbery, shot someone), the results did not change: the a, d, and b paths of the “content before process” pathway and the d path of the “process before content” pathway were significant, while the a and b paths of the “process before content” pathway were not. Likewise, the MCMAM confidence interval for the total indirect effect of the “content before process” pathway was significant (.00018, .00174) and the total indirect effect of the “process before content” pathway was non-significant (− .00018, .00063).
Sensitivity Testing
Because the “failsafe ef” can only account for one mediator at a time, two separate “failsafe ef” coefficients were calculated. The results revealed that an unobserved covariate confounder would need to correlate .31 with NAP-3 and .31 with PCT-4, controlling for Peer-2 and NAP-3 in the case of PCT-4, to eliminate the significant d path running from NAP-3 to PCT-4. Similarly, an unobserved covariate confounder would need to correlate .24 with PCT-4 and .24 with Delinquency-5, controlling for NAP-3 and PCT-4, to eliminate the significant b path running from PCT-4 to Delinquency-5.
In addition to testing for omitted variable bias, it is also recommended that researchers rule out endogenous selection bias or a collider effect as an alternate explanation for the results of a mediation analysis when precursor measures are used (Elwert and Winship 2014). Endogenous selection bias was assessed by removing the precursor variables from each of the five regression equations included in the extension path analysis. When precursor measures were removed from the path analysis, all four significant a, d, and b path coefficients from the original analysis improved and the a path from the “process before content” pathway achieved significance, findings inconsistent with a collider effect.
Discussion
Two hypotheses were tested in this study. The first hypothesis held that criminal thought process (proactive criminal thinking) and criminal thought content (negative attitude toward the police) would mediate the peer influence effect. In partial support of this hypothesis, criminal thought content but not criminal thought process successfully mediated the peer delinquency → participant delinquency relationship. Hence, the ability of both criminal thought content and criminal thought process to mediate the peer influence effect observed in the previous Walters (2016a) investigation was only partially replicated in the current investigation. One difference between the current study and prior Walters (2016a) investigation was that rather than assessing criminal thought content with a measure of positive attitudes toward deviance, the current study assessed criminal thought content with a measure of negative attitudes toward authority that did not overlap as much with proactive criminal thinking as the criminal thought content measure used in Walters (2016a). The goal of the second hypothesis was to extend the effect observed in Walters (2016a) by adding a serial mediator to each model and cross-lagging the two mediators. Congruent with the second research hypothesis, the pathway running from peer delinquency, to criminal thought content, to criminal thought process, to self-reported delinquency was significant. Incongruent with the second hypothesis, the pathway running from peer delinquency, to criminal thought process, to criminal thought content, to self-reported delinquency failed to achieve significance. Partial support for the second research hypothesis indicates that criminal thought content and criminal thought process mediate the peer influence effect in one direction but not the other, contradicting the reciprocity hypothesis (Thornberry 1987).
Implications
Peer delinquency is a robust and consistent correlate of offending behavior. Whether data are examined qualitatively, as in a literature review (Brechwald and Prinstein 2011; Hoeben et al. 2016), or quantitatively, as in a meta-analysis (Hubbard and Pratt 2002; Pratt et al. 2010), friends often display highly similar rates of crime and delinquency. Questions have been raised, however, as to the direction and nature of this relationship. Some have argued that peer delinquency leads to participant delinquency in a process known as peer influence (Akers 1998; Warr 2002), whereas others contend that participant delinquency leads to peer delinquency in a process known as peer selection (Gottfredson and Hirschi 1990; Kiesner et al. 2003). Most agree, however, that both processes play a role in peer effects (Knecht et al. 2010). Where there is less agreement is on the mechanisms responsible for each effect. Differential association and social learning theories of crime (Akers 1998; Sutherland 1947) maintain that reinforcement and modeling are at the heart of the peer influence effect, while others have argued that desire for higher status among peers (Berger 2008) and peer conformity as a means to a favorable self-identity (Gibbons et al. 2003) are the mechanisms that link peer delinquency to participant delinquency. Results from the current study suggest that additional social learning factors in the form of antisocial cognition—what Sutherland (1947) referred to as definitions favorable to violations of the law—should be considered potentially important mechanisms of peer influence and that the one significant two-mediator pathway (“content before process”) was moderately to highly robust to omitted variable bias.
From a social learning perspective, reinforcement and modeling relate more to the peer influence effect than they do to the peer selection effect. Cognitive factors, on the other hand, apply equally well to peer selection and influence. Prior research, in fact, has shown that the peer influence effect is specifically mediated by criminal thought content and the proactive dimension of criminal thought process (Walters 2015, 2016a, b), whereas the peer selection effect is mediated by the reactive dimension of criminal thought process (Walters 2016b). The current results suggest that not only are negative attitudes toward authority capable of mediating the peer influence effect but that they may do so in a particular order relative to proactive criminal thinking. In the current study and prior Walters (2016a) investigation, peer delinquency predicted criminal thought content and criminal thought process, although in the current study its impact on proactive criminal thought process only approached statistical significance. In both studies, however, the a path coefficient was 30–50% higher for the path mediated by criminal thought content. More importantly, wave 2 peer delinquency failed to predict either criminal thought content or criminal thought process at wave 4. The effect had to travel through the alternate mediator at wave 3 to produce a pathway capable of predicting delinquency at wave 5. This illustrates the principal contribution of the current study, namely, providing a long view of the peer influence effect. Certainly, there are short-term factors that contribute to the high correlation that exists for delinquency between friends, but the peer influence effect evolves over time and may depend on several different inter-related cognitive factors. The effect of criminal thought content on delinquency may be more immediate and less powerful than the effect of criminal thought process, but it may be indispensable in stimulating the slower and more powerful criminal thought processes that are more reliably connected to future delinquency.
The long view adopted in the current study suggests that several different cognitive factors facilitate the peer influence effect. Youth exposed to criminal others, will, over the course of several months, cultivate a way of thinking that supports criminal behavior. It is these cognitive patterns, learned in association with those already involved in crime, to which clinicians should turn their attention. Associating with prosocial peers, being raised by concerned and loving parents, and possessing good peer resistance skills can protect a child against negative peer influence, but in situations where these protective factors are absent or weak, additional intervention options are required (Warr 2002). One such option is cognitive-behavioral therapy (Landenberger and Lipsey 2005; Mpofu et al. 2018). What makes cognitive-behavioral therapy particularly attractive for use with juvenile offenders is that it is capable of addressing both criminal thought content and criminal thought process (Kroner and Morgan 2014). According to the present findings, the focus of attention during the early stages of an intervention should be on criminal thought content because it is more overt, concrete, and amenable to change than criminal thought process. Adopting a long view of the peer influence effect, it seems that the impact of criminal thought content on delinquency is largely indirect, via its effect on criminal thought process. As such, proactive thinking patterns should be targeted during the middle and later stages of an intervention, particularly when working with high risk delinquents.
Limitations
Generalizability could be considered a limitation of this study. The current investigation was conducted on early adolescent male and female youth living in the USA. It is uncertain exactly how well these results apply to older adolescents, adults, and youth from other countries.
A second limitation of this study is that nearly all of the variables were based on youth self-report. In general, using a single data source can potentially inflate path coefficients through shared method variance and mono-operational bias (Shadish et al. 2002). With respect to peer effects, it has been argued that perceptual measures of peer delinquency, of which the self-report peer delinquency measure used in the current study is an example, may confound peer and participant delinquency to the extent that the individual projects his or her own delinquency onto his or her peers (Haynie and Osgood 2005; Meldrum et al. 2009; Young et al. 2013). Projection was controlled in the current study by conditioning Delinquency-5 on a precursor measure of pre-existing delinquency (i.e., Delinquency-2). This said, future researchers should attempt to measure peer delinquency and participant delinquency in ways other than participant self-report to determine whether the results observed in the present study hold up when peer reports of delinquency and official measures of participant delinquency are used in place of self-report measures.
A third potential limitation of this study was the use of a measure of negative attitude toward the police as a proxy for criminal thought content. As previously noted, criminal thought content is presumed by Walters and Morgan (2018) to be composed of three dimensions (negative attitudes toward authority, positive attitudes toward deviance, and criminal identity). The current study employed a single instance of a single dimension (i.e., negative attitude toward the police as a proxy measure of negative attitudes toward authority) that covered only a fraction of the conceptual space used to define criminal thought content.
Future Research: Finding a Context
An important consideration in planning future research on the mechanisms responsible for the peer influence and selection effects is placing the research in its proper context or contexts. One potentially relevant context is supported by prior research and another is supported by the results of the current study. The context buoyed by prior research construes peer effects as existing within a developmental context. In other words, peer effects are not uniform across the lifespan but change over the course of development. The peer influence effect, for instance, rises to significance during early adolescence, gains momentum as the adolescent begins spending more time with friends than with parents (Brown and Bakken 2011), and then subsides as the individual enters early adulthood (Monahan et al. 2009). It would be a mistake, however, to conclude that researchers in the area of peer effects have ignored or underestimated the developmental context of the peer influence and selection effects. Developmental context has played a critical role in research on peer effects and should continue to guide this area of research (Brechwald and Prinstein 2011).
The context highlighted by the current results is one that views peer effects within a temporal context. Peer influence, as observed in two separate studies (Walters 2016a and current investigation), is mediated by the proactive dimension of criminal thought process and at least two of the dimensions of criminal thought content, with a two-year gap between peer and participant delinquency. The pathway becomes more complex when the gap is expanded to three years and the two cognitive mediators are measured twice and cross-lagged. In a study employing a two-week lag between waves, there was no peer influence effect, only situational effects based on changes in opportunity (Weerman et al. 2018). These results suggest that peer effects respond to different factors, depending, in part, on the time interval between waves. Two weeks is too short a time frame for thinking patterns to be observed, symbolized, and internalized and three years opens up a great many opportunities for variable cross-influence. Keeping the developmental and temporal contexts of peer effects in mind when studying the peer influence and selection effects would appear to be important in advancing research in this area.
References
Akers RL (1998) Social learning and social structure: a general theory of crime and deviance. Northeastern University Press, Boston
Allison PD (2012) Handling missing data by maximum likelihood. SAS Global Forum 2012 (Paper 312-2012). SAS Institute, Cary
Augustyn MB, McGloin JM (2013) The risk of informal socializing with peers: considering gender differences across predatory delinquency and substance use. Justice Quart 30:117–143
Bandura A (1986) Social foundations of thought and action. Prentice Hall, Englewood Cliffs
Bandura A (1999) Moral disengagement in the perpetration of inhumanities. Pers Soc Psychol Rev 3:193–209
Berger J (2008) Identity signaling, social influence, and social contagion. In: Prinstein MJ, Dodge KA (eds) Understanding peer influence in children and adolescents. Guilford, New York, pp 181–199
Brechwald WA, Prinstein MJ (2011) Beyond homophily: a decade of advances in understanding peer influence processes. J Res Adolescence 21:166–179
Brown BB, Bakken JP (2011) Parenting and peer relationships: reinvigorating research on family–peer linkages in adolescence. J Res Adolescence 21:153–165
Cole DA, Maxwell SE (2003) Testing mediational models with longitudinal data: questions and tips in the use of structural equation modeling. J Abnorm Psychol 112:558–577
Collins LM, Schafer JL, Kam CM (2001) A comparison of inclusive and restrictive strategies in modern missing data procedures. Psychol Methods 6:330–351
Dodge KA, Coie JD (1987) Social-information processing factors in reactive and proactive aggression in children’s peer groups. J Pers Soc Psychol 53:1146–1158
Elwert F, Winship C (2014) Endogenous selection bias: the problem of conditioning on a collider variable. Ann Rev Soc 40:31–53
Esbensen F-A (2002) National evaluation of the Gang Resistance Education and Training (G.R.E.A.T.) program: final report (94-IJ-CX-0058). U.S. Department of Justice, Washington
Gelfert A (2015) Inner speech, natural language, and the modularity of the mind. J Philos Sci 14:7–29
Gibbons FX, Gerrard M, Lane DJ (2003) A social reaction model of adolescent health risk. In: Suls JM, Wallson K (eds) Social psychological foundations of health and illness. Blackwell, Oxford, pp 107–136
Gottfredson MR, Hirschi T (1990) A general theory of crime. Stanford University Press, Stanford
Hayes AF (2013) Introduction to mediation, moderation, and conditional process analysis: a regression-based approach. Guilford, New York
Haynie DL, Osgood DW (2005) Reconsidering peers and delinquency: how do peers matter? Soc Forces 84:1109–1130
Hoeben EM, Meldrum RC, Walker D, Young JTN (2016) The role of peer delinquency and unstructured socializing in explaining delinquency and substance use: a state-of-the-art review. J Crim Justice 47:108–122
Hubbard DJ, Pratt TC (2002) A meta-analysis of the predictors of delinquency among girls. J Offender Rehabil 34(3):1–13
Imai K, Keele L, Tingley D (2010) A general approach to causal mediation analysis. Psychol Methods 15:309–334
Johnson SL, Burke JG, Gielen AC (2012) Urban students’ perceptions of the school environment’s influence on school violence. Child Sch 34:92–102
Kenny, D. A. (2013). Mediation: sensitivity analysis [webinar]. Retrieved from http://davidakenny.net/webinars/Mediation/Sensitivity/Sensitivity.html.
Kiesner J, Poulin F, Nicotra E (2003) Peer relations across contexts: individual-network homophily and network inclusion in and after school. Child Dev 74:1328–1343
Knecht A, Snijders TAB, Baerveldt C, Steglich CEG, Raub W (2010) Friendship and delinquency: selection and influence processes in early adolescence. Soc Dev 19:494–514
Kroner DG, Morgan RD (2014) An overview of strategies for the assessment and treatment of criminal thinking. In: Tafrate RC, Mitchell D (eds) Forensic CBT: a handbook for clinical practice. Wiley, Chichester, pp 87–103
Lahey BB, Van Hulle CA, D’Onofrio BM, Rodgers JL, Waldman ID (2008) Is parental knowledge of their adolescent offspring’s whereabouts and peer associations spuriously associated with offspring delinquency? J Abnorm Child Psychol 36:807–823
Landenberger NA, Lipsey MW (2005) The positive effects of cognitive-behavioral programs for offenders: a meta-analysis of factors associated with effective treatment. J Exp Criminol 1:451–476
McGloin JM, Shermer LO (2009) Self-control and deviant peer network structure. J Res Crime Delinq 46:35–72
Meldrum RC, Young JTN, Weerman FM (2009) Reconsidering the effect of self-control and delinquent peers: implications of measurement for theoretical significance. J Res Crime Delinq 46:353–376
Monahan KC, Steinberg L, Cauffman E (2009) Affliliation with antisocial peers, susceptibility to peer influence, and antisocial behavior during the transition to adulthood. Dev Psychol 45:1520–1530
Morgan RD, Fisher WH, Duan N, Mandracchia JT, Murray D (2010) Prevalence of criminal thinking among state prison inmates with serious mental illness. Law Hum Behav 34:324–336
Mpofu E, Athanasou JA, Rafe C, Belshaw SH (2018) Cognitive-behavioral therapy efficacy for reducing recidivism rates of moderate- and high-risk sexual offenders: a scoping systematic literature review. Int J Offender Ther Comp Criminol 62:170–186
Muthén B, Muthén L (2017) Mplus user’s guide, 8th edn. Muthén and Muthén, Los Angeles
Pratt TC, Cullen FT, Sellers CS, Winfree T, Madensen TD et al (2010) The empirical status of social learning theory: a meta-analysis. Justice Q 27:765–802
Preacher KJ (2015) Advances in mediation analysis: a survey and synthesis of new developments. Ann Rev Psychol 66:825–852
Preacher KJ, Selig JP (2012) Advantages of Monte Carlo confidence intervals for indirect effects. Commun Methods Meas 6:77–98
Shadish WR, Cook TD, Campbell DT (2002) Experimental and quasi-experimental designs for generalized causal inference. Mifflin, Boston
Sutherland EH (1947) Principles of criminology, 4th edn. J.B. Lippincott, Philadelphia
Sykes GM, Matza D (1957) Techniques of neutralization: a theory of delinquency. Am Sociol Rev 22:664–670
Thornberry TP (1987) Toward an interactional theory of delinquency. Criminology 25:863–891
Trzepacz PT, Baker RW (1993) The Psychiatric Mental Status Examination. Oxford University Press, Oxford
Vygotsky L (1994) The development of thinking and concept formation in adolescence. In: van der Veer R, Valsiner J (eds) The Vygotsky reader. Blackwell, Oxford, pp 185–265
Walters GD (2015) Proactive criminal thinking and the transmission of differential association: a cross-lagged multi-wave path analysis. Crim Justice Behav 42:1128–1144
Walters GD (2016a) Criminal thought content and criminal thought process as mediators of peer influence: combined versus individual effects. Crim Justice Rev 41:318–334
Walters GD (2016b) Friends, cognition, and delinquency: proactive and reactive criminal thinking as mediators of the peer influence and peer selection effects. Justice Q 33:1055–1079
Walters GD (2017) Modelling the criminal lifestyle: theorizing at the edge of chaos. Palgrave Macmillan, London
Walters GD, Morgan RD (2018) Assessing criminal thought content: preliminary validation of the Criminal Thought Content Inventory (CTCI). Psychol Crime Law 24:603–620
Warr M (2002) Companions in crime: the social aspects of criminal conduct. Cambridge University Press, Cambridge
Weerman FM, Wilcox P, Sullivan CJ (2018) The short-term dynamics of peers and delinquent behavior: an analysis of bi-weekly changes within a high school student network. J Quant Criminol 34:431–463
Wu AD, Zumbo BD (2008) Understanding and using mediators and moderators. Soc Indic Res 87:367–392
Young R, Johnson DR (2013) Methods for handling missing secondary respondent data. J Marriage Fam 75:221–234
Young JTN, Rebellon CJ, Barnes JC, Weerman FM (2013) Unpacking the black box of peer similarity in deviance: understanding the mechanisms linking personal behavior, peer behavior, and perceptions. Criminology 52:60–86
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of Interest
The author declares that there is no conflicts of interest.
Ethical Approval
Study approved by the Kutztown University Institutional Review Board.
Informed Consent
This is a secondary data analysis, but the authors of the original GREAT study did obtain informed consent from parents and informed assent from participating children.
Additional information
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
About this article
Cite this article
Walters, G.D. Mediating the Peer Influence Effect with Facets of Criminal Thought Process and Content in a Group of Early Adolescents: Replication and Extension. J Police Crim Psych 35, 207–218 (2020). https://doi.org/10.1007/s11896-019-09360-3
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11896-019-09360-3