Fluent Essays

I uploaded both theories discussed in the prompt https://doi.org/10.1177/0886260517710486 Journal of Interpersonal Violence 1 –27 © The Author(s) 2017 Reprints and permissions: sagepub.com/journalsPermissions.nav DOI: 10.1177/0886260517710486 journals.sagepub.com/home/jiv Article Social Learning Theory, Gender, and Intimate Partner Violent Victimization: A Structural Equations Approach Ráchael A. Powers, PhD,1 John K. Cochran, PhD,1 Jon Maskaly, PhD,2 and Christine S. Sellers, PhD3 Abstract The purpose of this study is to examine the applicability of Akers’s Social Learning Theory (SLT) to explain intimate partner violence (IPV) victimization. In doing so, we draw on the Intergenerational Transmission of Violence Theory (IGT) to extend the scope of SLT to the explanation of victimization and for a consideration of uniquely gendered pathways in its causal structure. Using a structural equation modeling approach with self- report data from a sample of college students, the present study tests the extent to which SLT can effectively explain and predict IPV victimization and the degree, if any, to which the social learning model is gender invariant. Although our findings are largely supportive of SLT and, thus, affirm its extension to victimization as well as perpetration, the findings are also somewhat mixed. More significantly, in line with IGT literature, we find that the social learning process is not gender invariant. The implications of the latter are discussed. 1University of South Florida, Tampa, USA 2The University of Texas at Dallas, USA 3Texas State University, San Marcos, USA Corresponding Author: Ráchael A. Powers, Department of Criminology, University of South Florida, 4202 E. Fowler Ave., SOC 107, Tampa, FL 33620, USA. Email: [email protected] 710486 JIVXXX10.1177/0886260517710486Journal of Interpersonal ViolencePowers et al. research-article2017 https://us.sagepub.com/en-us/journals-permissions https://journals.sagepub.com/home/jiv mailto:[email protected] 2 Journal of Interpersonal Violence 00(0) Keywords social learning, intimate partner violence, victimization, gender, intergenerational transmission of violence Introduction A longstanding theoretical perspective on intimate partner violence (IPV) is the Intergenerational Transmission of Violence Theory (IGT; Straus, Gelles, & Steinmetz, 1980). This theory argues that experiencing household violence (direct victimization) and/or witnessing it (indirect/vicarious victimization), particularly during childhood, leads to subsequent IPV, either perpetration or victimization. The causal process involved in IGT is most often attributed to a learning process (e.g., Alexander, Moore, & Alexander, 1991). In this way, IGT shares much in common with Akers’s Social Learning Theory (SLT; see Fox, Nobles, & Akers, 2011; Sellers, Cochran, & Branch, 2005; Sellers, Cochran, & Winfree, 2003; Wareham, Boots, & Chavez, 2009a, 2009b). For instance, IGT and SLT stress exposure to influential role models (parents) who perpetrate or experience interpersonal violence within the household that children witness and later imitate. IGT and SLT also articulate the impor- tance of the transmission of beliefs, values, and norms conducive to IPV. However, SLT can also explicitly accommodate other common explanations for IPV including extrafamilial socialization, gender roles and violent mascu- linity, and the role of differential reinforcement. On the contrary, Akers’s SLT is explicitly a theory of perpetration, and makes no claim to account for victimization. Conversely, IGT can account for both IPV perpetration and victimization. Given the high level of concep- tual and propositional congruity between the theories, it is our contention that IGT can offer a theoretical basis for expanding the scope of SLT to include victimization as well as perpetration. In addition, some of the literature on IGT has suggested that the causal processes are highly gendered, whereas SLT is ostensibly gender invariant. Although the results are not conclusive, previous research has found that men and women react differently to violence in the family of origin, and therefore, a gendered application of IGT may be warranted (see Stith et al., 2000). Extending this rationale to SLT, an other- wise gender invariant theory, it is possible that its causal processes do not operate identically for men and women with regard to IPV victimization. To that end, the primary purpose of the present study, and one of its more significant contributions to the literature, is to examine the applicability of SLT to explain IPV victimization. In so doing, we are able to provide a further test of its theoretical scope, which, as a “general” theory of Powers et al. 3 behavior, it should be able to accommodate. Moreover, we also examine the degree to which SLT as an explanation for IPV victimization is gender invariant, another important contribution and purpose of the present study. To the extent that SLT cannot account for IPV victimization, or to the extent that it is not gender invariant, it may not be as “general” a theory as some may claim it to be. Finally, this studycontributes to the literature testing SLT by employing a full structural equation model of the theory with com- plete representation and strong measurement properties for all four of the key SLT constructs while also accounting for the feedback/reciprocal influ- ence of IPV victimization. Intergenerational Transmission of Violence Experiencing physical violence in childhood has been associated with a sub- stantial increase in the odds of IPV perpetration (e.g., Gómez, 2011) and victimization (e.g., Hamby, Finkelhor, & Turner, 2012). The most common explanation for this relationship focuses on learning processes. Children learn behavior from their experiences and observations of social interactions. These observations are particularly salient when the modelers are of high status, such as parents and caregivers (Bandura, 1973). Therefore, when chil- dren experience violence or hostile parenting practices, they learn that vio- lence is an acceptable means of conflict resolution and will later model that behavior in their relationships (Akers & Sellers, 2009). Although the direct link between childhood exposure to or experiences of interparental violence and later life involvement in IPV is presumably well established, it is far from conclusive. Recent studies using prospective meth- ods or advanced statistical procedures (e.g., propensity score matching) sug- gest that this relationship may not be causal; once other adverse childhood experiences or selection bias has been taken into account, the direct impact of childhood abuse and childhood exposure to domestic violence on later IPV perpetration and victimization disappears (Jennings et al., 2014; Widom, Czaja, & DuMont, 2015). In a meta-analysis of 39 studies of IGT conducted by Stith and colleagues (2000), they suggest that support for the theory is weak to moderate. With regard to victimization, they suggest that child abuse and witnessing interparental aggression have weak to moderate effects on later intimate partner victimization. They point to the need for more “com- plex studies” that are able to move beyond the examination of the direct rela- tionship between violence in the family of origin and later IPV. For example, Messing and colleagues (2012) suggested that posttraumatic stress disorder (PTSD) may partially mediate the relationship between some forms of child- hood trauma and later adult victimization. 4 Journal of Interpersonal Violence 00(0) Likewise, several studies that are in line with the processes outlined in Akers’s SLT model have been employed to further disentangle this relation- ship between childhood experiences and adult IPV. The majority of these studies focus on the role of attitudes and beliefs surrounding violence and how they shape the risk for later perpetration and victimization. Several stud- ies have found that experiences of child abuse are related to later acceptance or condoning of violence against women, which may increase the likelihood of perpetration or risk of entering into a violent relationship (e.g., Markowitz, 2001). Indeed, this is often considered a crucial link in the learning process between childhood abuse and adult IPV. Taken together, this research suggests that there is a complex relationship between childhood experiences of violence and adult IPV. However, the learning processes and mechanisms by which violence is transferred are not well understood. Akers’s SLT articulates some of these learning mechanisms and processes more explicitly. Akers’s SLT SLT (Akers, 1998) proposes that crime and conformity are learned through interactions with other people that expose the individual to definitions and behaviors, reinforcements, and role models that either favor or oppose crime. Depending on the unique configuration of associates with whom one inter- acts, as well as the weight of each one’s influence on the individual, one may be exposed to attitudes, behaviors, reinforcements, and models that, on bal- ance, favor or oppose crime. In brief, SLT predicts that criminal behavior is likely to increase as association with criminal individuals outweighs associa- tion with noncriminal individuals; when this occurs, rewards for crime out- weigh the costs of crime, the number of criminal role models outweighs the number of conforming role models, and one’s positive or neutralizing defini- tions of crime outweigh one’s own negative definitions of crime. SLT posits a processual model whereby differential association exerts both a direct effect on criminal behavior and a partially mediated or indirect effect via its influence on differential reinforcement, imitation, and defini- tions, which likewise exert direct effects on criminal behavior. Moreover, Akers (1998) argued that the model incorporates reciprocal and feedback effects, in which an increase in criminal behavior also amplifies association with others favorable to crime, which then continues one’s exposure to the other three elements of SLT. Akers advanced SLT in a series of statements beginning with the differen- tial association-reinforcement theory (Burgess & Akers, 1966) and culminat- ing with the social structure-social learning model (Akers, 1998). Few Powers et al. 5 empirical investigations of the theory were conducted until Akers himself published the first test of the full social learning model (Akers, Krohn, Lanza- Kaduce, & Radosevich, 1979), which demonstrated remarkable predictive accuracy of SLT in accounting for alcohol and drug use among adolescents. Tests of SLT flourished thereafter. Most of these studies were simple tests of the direct, linear, independent effects of one or more of the four social learn- ing variables on a dependent variable, the latter most frequently a form of substance use or common delinquency (for a review and meta-analysis, see Pratt et al., 2010). Far less common in the body of empirical research on SLT are tests of the causal sequencing of the full social learning model. Akers and Lee (1996) used structural equation modeling (SEM) to estimate both causal and reciprocal/feedback effects of social learning and teenage smoking (see also Cochran, Maskaly, Jones, & Sellers, 2017; Krohn, Skinner, Massey, & Akers, 1985), confirming (with the exception of the imitation variable) the hypothesized social learning effects. Lee, Akers, and Borg (2004) found sim- ilar results in their SEM analysis of the Social Structure Social Learning (SSSL) model of adolescent alcohol and marijuana use. Extending the social learning causal model to physical aggression rather than substance use, Cochran and colleagues (2017) demonstrated direct and indirect effects of all social learning variables on violence perpetrated against an intimate partner; moreover, IPV also exerted reciprocal/feedback effects on the four social learning variables. Criminological theories like SLT are advanced explicitly to account for offending behavior. However, these theories in some instances are also pos- ited as viable explanations of criminal victimization (see, for instance, Schreck, 1999, regarding low self-control; Smith & Jarjoura, 1988, regarding social disorganization theory; or Zavala & Spohn, 2013, regarding general strain theory), in part because of the undeniable overlap between criminal offending and victimization (Lauritsen, Sampson, & Laub, 1991). There is at least some evidence of similar overlap in the victimization and perpetration of IPV (Graham-Kevan & Archer, 2003), especially in instances of what Johnson (1995) refers to as “common couple violence.” Prior research pro- vides some support of SLT as an explanation of IPV perpetration and less often, victimization (Cochran et al., 2017; Fox et al., 2011; Sellers et al., 2005, 2003; Wareham et al., 2009a, 2009b). Within the context of SLT, the likelihood of IPV victimization increases not only as exposure to a violent partner increases, but also as exposure to other victims of violence increases (differential association), especially when other victims express attitudes (neutralizing definitions) that excuse or rationalize the violence perpetrated against them (e.g., I was asking for it; he was drunk; that’s how a man shows he loves me). SLT acknowledges that these socialization processes can occur 6 Journal of Interpersonal Violence 00(0) external to the family, including the cultural acceptance of violence as a means of conflict resolution (Krug, Mercy, Dahlberg, & Zwi, 2002) and the influence of traditional gender roles and violent masculinity. With regard to reinforcement, there is research to suggest that tenets of operant conditioning and reinforcement may explain the risk of IPV victimization or the stay– leave decision of victims. For example, Miller, Lund, and Weatherly (2012) applied operant learning principles to the examination of stay–leave deci- sions among women in violent relationships. The unpredictable pattern of offending behavior and subsequent reconciliation after abusive episodes pro- vides partial positive and negative reinforcements. Websdale (1998) sug- gested that the balance of reinforcements (e.g., the “honeymoon phase,” financial dependence) versus costs (e.g., physical injuries, emotional trauma, presence of children) of IPV victimization may at times tip toward repeated victimization. The response-cost of IPV may be quite high (Miller et al., 2012), and therefore, victimization, and repeated victimization, becomes more likely. Gendered Learning Processes in IPV IPV is a gendered phenomenon with regard to both perpetration and victim- ization (e.g., Johnson, 1995). As a general theory, SLT purports to account for the behavior of both men and women. In general, men are far more likely than women to be offenders and slightly more likely than women to be vic- tims of crime. SLT would explain that men are more likely than women to operate in learning environments that are more conducive to offending and victimization. However, some research on IPV finds that men and women are equally likely to be perpetrators as well as victims of aggression (Johnson & Ferraro, 2000). Furthermore, the social learning predictors of IPV differ by gender for both perpetration (Sellers et al., 2003) and victimization (Cochran, Sellers, Wiesbrock, & Palacios, 2011), but these tests were restricted to sim- ple regression-based approaches which do not allow for both direct and indi- rect/mediated social learning processes. What progress has been made to investigate the processual features of SLT has largely focused exclusively on offending or deviant behavior such as sub- stance use. However, SLT has begun to be found to be a viable explanation of victimization as well and less often, of gendered processes in IPV. Although research is scant, social learning variables have been associated with stalking (Fox et al., 2011) and IPV (Cochran et al., 2011; Sellers et al., 2003). Sellers and colleagues (2003) tested both the efficacy of SLT to explain IPV perpe- tration and the gender invariance of SLT. They found that SLT could account for IPV perpetration but that several of the SLT measures were not gender Powers et al. 7 invariant. Similarly, Cochran and colleagues (2011) tested the efficacy of Akers’s SLT against self-report data on repetitive intimate partner victimiza- tion; they found that for both male and female victims, repetitive IPV victim- ization was associated with both differential association and differential reinforcement. However, they did not test for gender invariance. Moreover, neither the Sellers and colleagues (2003) study nor the Cochran and col- leagues (2011) study employed an SEM approach that would have permitted them to examine both the direct and indirect effects of SLT variables on IPV and the reciprocal effects of IPV on the SLT process. The moderating effects of gender on IPV perpetration and victimization have been more fully articulated and explored in the IGT literature. Several studies have found that the conclusions regarding the influence of experienc- ing or witnessing violence in the home are contingent on the gender of the child or the parental aggressor, which suggests that there are gendered pro- cesses in the transmission of violence. Whereas many have found that both male and female children are adversely impacted by exposure to or experi- ences of violence in the home, some have found that this relationship holds only for men (Alexander et al., 1991) or women (Douglas & Straus, 2006). Marshall and Rose (1988) found that experiencing child abuse was correlated with both IPV perpetration and victimization for men, but only victimization for women. Others have attempted to disentangle this relationship by focusing on the gender dynamics of the parental relationship. Jankowski, Leitenberg, Henning, and Coffey (1999) suggested that the gender of the child and the parental aggressor interact such that vicarious victimization increased the likelihood of later IPV perpetration only when the child and parent were of the same sex. Likewise, Laporte, Jiang, Pepler, and Chamberland (2011) found that although both male and female teens who experienced child abuse were more likely to perpetrate IPV, these effects were stronger for men in general and strongest for men who experienced abuse by their fathers. In sum, SLT provides more causal mechanisms and perhaps more explic- itly articulated mechanisms for explaining how experiencing or witnessing violence in the home leads to later IPV perpetration. However, it has rarely been extended to examine IPV victimization. Furthermore, the theory assumes gender invariance. IGT, on the contrary, has been extended to vic- timization and provides compelling evidence that these learning processes may be gendered. The present study draws from the theoretical basis of IGT as well as the empirical literature to examine the utility of SLT; it tests the efficacy of SLT to explain IPV victimization; it tests for gender invariance in the SLT process; and it examines the direct, indirect, and reciprocal relation- ships between SLT constructs and IPV through an SEM analytic approach. 8 Journal of Interpersonal Violence 00(0) Method Data The data for this study were gathered through a self-administered survey of students attending a large urban university in Florida. The students were surveyed in graduate and undergraduate classes randomly selected from the course offerings of five colleges (Arts and Sciences, Business Administration, Education, Engineering, and Fine Arts) during the first 4 weeks of the spring 1995 semester. Courses were sampled from each col- lege in proportion to the enrollments each college contributed to the uni- versity’s total enrollment. This sampling strategy targeted a total of 2,500 students; however, absenteeism on the day of the survey and enrollments of students in more than one sampled course produced an overall response rate of 73\%. The current study is based on those students who completed the questionnaire, who report being currently involved in an intimate rela- tionship (i.e., married or dating), and who also report having had at least one previous serious relationship (n = 1,124). The sociodemographic pro- file of the sample was very similar to that of the total enrollment at the university. Importantly, these data, unlike most other self-reported data collections, were specifically designed to examine the efficacy of Akers’s SLT on IPV. Finally, while these self-report data are derived from a sample of college students, it is noteworthy that a substantial number of the respondents were married or cohabiting, and as we report below, the prev- alence and frequency of IPV among the students sampled was quite substantial. Measures Dependent variable: IPV. The dependent variables used in this study were latent constructs developed from a single set of measures of self-reported intimate partner violence victimization (IPV-V) by one’s current partner: a total scale composed of eight items and a subscale of the five more serious/ injurious items. All are drawn from the physical aggression items in Straus’s (1979) Conflict Tactics Scale (CTS)—The data were collected prior to the development of the CTS-II. Specifically, respondents were asked with regard to their current marital or dating relationship how many times their partner had done any of the following eight acts of IPV: (a) threw something; (b) pushed, grabbed, or shoved; (c) slapped; (d) kicked, bit, or hit with a fist; (e) hit with something; (f) beat up; (g) threatened with a knife or gun; and (h) used a knife or gun. Responses to these items were never, once or twice, 3 to 5 times, 6 to 10 times, 11 to 20 times, and 21 or more times, coded from 0 to Powers et al. 9 7. Identical constructs were also constructed for victimization by a previous romantic partner. These were employed as exogenous variables and provide a method for assessing the reciprocal/feedback effects of prior victimization experience on the social learning process. Independent variable: Social learning constructs. The independent variables in this study are first- or second-order latent constructs representing each of Akers’s four social learning concepts: differential association, imitation, defi- nitions, and differential reinforcement. We endeavored to measure the con- structs using items and scales derived near exactly as they were measured by Akers and colleagues (1979), though modified to reflect IPV rather than ado- lescent substance use. Differential association is a second-order latent construct comprised of a single-item measure of the respondents’ estimation of the proportion of their best friends who had been physically victimized by a romantic partner (1 = none or almost none, 2 = less than half, 3 = more than half, and 4 = all or almost all), and two first-order latent constructs. The first of these first-order latent constructs is comprised of four items measuring mother’s, father’s, partner’s, and best friend’s attitudes toward IPV. For these items, respondents were asked to indicate to what degree each of these significant others would approve/disapprove of the use of physical violence against a partner (1 = strongly disapprove to 4 = strongly approve). The second of these two first- order latent constructs used to constitute differential association is composed of five indicators of physical violence used against significant others. Specifically, respondents were asked to indicate how often their mother, father, siblings, other family members, and best friends were victims of IPV (1 = never, 2 = seldom, 3 = usually, and 4 = always). Imitation is measured by a first-order latent construct comprising seven different admired role models the respondent had actually seen being physi- cally victimized (i.e., hit, slapped, kicked, or punched) by an intimate partner during a disagreement. These admired models included actors on television or in movies, mother, father, siblings, other family members, friends, and others. Definitions is another second-order latent construct comprising a single- item measure of respondents’ own approval/disapproval of the use of physi- cal violence against a partner (1 = strongly disapprove to 4 = strongly approve), and three first-order latent constructs. The first of these three first- order latent constructs is a two-item measure of respondents’ attitudes favor- able toward the violation of the law in general and indicated by the extent to which respondents agreed/disagreed with the following Likert-type state- ments (1 = strongly agree to 5 = strongly disagree): “We all have a moral duty 10 Journal of Interpersonal Violence 00(0) to abide by the law” (reverse coded) and “It is okay to break the law if we do not agree with it.” The next of these three second-order latent constructs rep- resents definitions approving of IPV indicated by three Likert-type state- ments (e.g., “It is against the law for a man to use violence against a woman even if they are in an intimate relationship”). Finally, the third of these first- order latent constructs measures neutralizing definitions and is composed of responses to three Likert-type statements (e.g., “Physical violence is a part of a normal dating/marital relationship”). The last social learning construct, differential reinforcement, is a second- order latent construct comprised of two first-order constructs and two single- item measures. First, respondents reported the actual or anticipated reaction of three different sets of significant others (i.e., parents, other family mem- bers, and best friends) to the respondent’s physical victimization by their partner. Respondents indicated that these significant others would either 1 = disapprove and report to the authorities, 2 = disapprove and try to stop it, 3 = disapprove but do nothing, 4 = neither approve nor disapprove, and 5 = approve and encourage it. Second, a single 3-point, ordinal measure of the overall balance of reinforcement for partner violence was included. This item measured the respondent’s perception of the usual or anticipated net outcome from being victimized by their current partner (1 = mostly bad, 2 = about as much good as bad, and 3 = mostly good). Third, the net rewards-to-costs of being physically victimized by their partner was measured by asking respon- dents to indicate which, if any, of five social and nonsocial rewards and seven social and nonsocial costs they associated with IPV by their current partner. An example of a reward was “It showed me my partner really loved me,” and an example of a cost was “My friends criticized me.” To compute the net rewards-to-costs, the sum of the identified costs was subtracted from the sum of the identified rewards; this produced a measure with values ranging from −7 (all costs and no rewards) to 5 (no costs and all rewards). The results from the full confirmatory factor analysis (CFA) are included in the appendix. Analytic Strategy Considering that the purpose of the current study is to examine the direct and indirect effects of various components of the social learning process on IPV victimization, the most appropriate analytical technique is SEM. Following the two-step process (see Kline, 1998), we first develop and test a measure- ment model using CFA. Following the recommendations of Hoyle and Panter (1995), we report several fit indices (i.e., χ2, standardized root mean square residual [SRMSR], root mean square error of approximation [RMSEA], and comparative fit index [CFI]). We follow the recommended general criterion Powers et al. 11 values for fit statistics (e.g., Hu & Bentler, 1995). Following the measure- ment model, we test the structural model. Our structural models proceed in two phases, examining how SLT predicts (a) IPV victimization by one’s cur- rent partner and (b) the extent to which the social learning model of IPV victimization is gender invariant (measurement models available upon request). Although no variable had more than 15\% missing, using the standard list- wise deletion procedure would have resulted in approximately 30\% attrition. Markov Chain Monte Carlo (MCMC) simulation was used to impute 10 data- sets to fill in missing values. All models were estimated in MPlus 7.4 using the weighted least squared Muthen version (WLSMV) estimator to account for the limited nature of the indicators (Rhemtulla, Brosseau-Liard, & Savalei, 2012). Results The baseline social learning model is presented in Figure 1. It allows for an examination of both direct and indirect effects of the theoretically expected paths between the social learning constructs and IPV victimization. Importantly, it also controls for the anticipated feedback or reciprocal effects of IPV victimization on the social learning constructs as expressed by the effects of IPV victimization by one’s past partner on the social learning pro- cess that predicts IPV victimization by one’s current partner. Overall, the model fit …