History of Abuse
One of the most consistent predictors of intimate partner violence (IPV) perpetration in prior research is a history of abuse, particularly within the family of origin. However, not all victims of child maltreatment go on to perpetrate violence within their intimate partnerships. A number of theoretical explanations help distinguish who may be most at risk for perpetuating this “cycle of violence.”
Historically, intimate partner violence (IPV) has often been viewed as a primarily male-perpetrated phenomenon. However, research in recent years has demonstrated that women also perpetrate violence against their loved ones. National surveys estimate that more than one in three women and more than one in four men have experienced rape, physical violence, or stalking by an intimate partner in their lifetimes (CDC 2010). In community surveys specifically, where less severe forms of violence, often referred to as “common couple violence,” are studied, it is often found that women perpetrate violence at rates equivalent to or even higher than men (e.g., Kaufman-Parks et al. 2017). In order to reduce IPV perpetration and its associated costs to victims, perpetrators, and society at large (e.g., CDC 2010), it is imperative to understand the potential factors that increase the risk for such violence.
Prior research has documented a number of negative sequelae of childhood abuse and neglect. One of the most frequently researched consequences has been coined the “cycle of violence” hypothesis, whereby it is presumed that individuals who are abused as children grow up to face an increased likelihood of perpetrating violence toward their loved ones. Supporting this hypothesis, past empirical research has documented associations between harsh physical discipline and physical abuse of children and subsequent IPV perpetration throughout both adolescence and adulthood (e.g., Capaldi et al. 2012). Yet, research has also demonstrated that exposure to violence during childhood does not lead in a deterministic fashion to violent behaviors, including IPV perpetration, in later life (e.g., Smith et al. 2011). Rather, it appears that accounting for potential mediating factors is essential if we are to understand why individuals with a history of child abuse are more likely to engage in IPV (e.g., Capaldi et al. 2012). A number of theoretical explanations lend insight into these potential mediating factors, including individuals’ attitudes and beliefs, the ability to develop and maintain healthy attachments with others, cognitive decision-making processes, personality traits and disorders, genetic or neurophysiological risks, and adoption of specific life history strategies.
Social Learning Mechanisms
Social learning theorists argue that attitudes and behaviors are acquired through processes of observation, learning, and reinforcement with key social agents (Bandura 1977). As part of the natural aging and socializing processes, children observe the behaviors of key figures in their lives and encode their observations. At a later time, they may copy or imitate these observed behaviors, which are then reacted to by others. If others reinforce the newly performed behavior through praise or reward, the child is likely to continue performing the behavior. If reactions by others are negative, continued performance of the behavior is likely to decline or cease entirely. Accordingly, because the family is one of the first and main socializing agents for most individuals, relationships between parents and between parents and children provide models for how individuals should behave in relationships with others.
In the context of IPV, social learning theorists propose that children exposed to violence in their families of origin learn that hitting is an appropriate response to conflict with others. Upon entering intimate relationships as adolescents and adults, these individuals then often have limited resources for dealing with relationship conflict in constructive ways and resort to violence as a strategy for conflict resolution. They may also develop an expectation for violence in their own relationships or feel that violence is necessary to maintain control and power in their lives (Foshee et al. 1999; Kaufman-Parks et al. 2016). For example, in a community sample of young adult men and women, it was found that harsh physical discipline by parents (i.e., physical abuse) in adolescence increased the risk for IPV perpetration in young adulthood among both men and women. Moreover, while the risk of IPV also increased among all individuals when they experienced verbal conflict with their intimate partners, this was especially the case among individuals who had previously been maltreated (Kaufman-Parks et al. 2016). As concluded by the researchers, such findings demonstrated that individuals exposed to childhood physical abuse may have a lower tolerance for verbal aggression in their romantic relationships and react violently as a result.
Despite the evidence linking child abuse and later IPV perpetration, empirical research also has found that the vast majority of individuals exposed to violence in the family of origin do not report IPV in adulthood (e.g., Smith et al. 2011). Similarly, global endorsements of IPV in the general population are often relatively low (Simon et al. 2001). In other words, most individuals with a history of abuse still do not perceive violence as an acceptable way of dealing with conflict in their adolescent and adult intimate relationships. What accounts for whether abuse leads to the adoption of attitudes accepting of IPV may thus rest in factors considered in combination with and external to childhood victimization experiences specifically. For example, analyzing the development of attitudes toward IPV, Copp et al. (2016) noted a positive, significant association between neighborhood poverty and attitudes accepting of IPV, even after accounting for exposure to violence in the family of origin via interparental violence and “coercive parenting” (i.e., physical abuse). Thus, IPV perpetration among those with a history of child abuse may be especially likely if they are also placed within a greater structural and cultural context (i.e., the neighborhood) that espouses violence as a normative, or at least justifiable, behavior.
It is also important to note that although most social learning explanations of IPV focus primarily on family of origin experiences, the process of learning continues throughout the life course. In other words, individuals’ attitudes toward and engagement in violence with intimate others may be influenced by those external to the family unit and in contexts beyond the childhood years. For example, prior research has indicated that association with peers who perpetrate violence increases risk for individuals’ own IPV perpetration (Capaldi et al. 2012). Relatedly, individuals who experienced controlling behaviors from partners in earlier intimate relationships are more likely to accept violence as a justified response, under certain circumstances, toward an intimate partner in subsequent relationships (Copp et al. 2016). Thus, IPV perpetration among those with a history of abuse may be especially likely if such individuals are exposed to peers and intimate partners who further reinforce violence as acceptable. Conversely, individuals who, despite a history of abuse, are able to surround themselves with others who reject the acceptability of violence may decrease their chances of later IPV perpetration. As noted by Giordano (2010), children are not only influenced by their environments but, because of human agency, are able to influence their environments. Although children are arguably much more constrained than adults in selecting their immediate environments, they may still have the power to attend more to some family members than others or to select role models outside the family, who do not model and reinforce violence as justified or socially normative.
Interestingly, however, some prior research has indicated that even after accounting for a history of family violence and a number of additional potential risk factors (i.e., sociodemographic characteristics, IPV experiences in previous relationships, current relationship status and duration), women may be more likely than men to endorse the use of violence under a number of different circumstances (Copp et al. 2016). What accounts for this difference is not entirely clear, but it is possible that the general societal belief that female-perpetrated IPV is less problematic or more acceptable than males may lead women to minimize the harms associated with their violent behavior.
Prior research has found that children who are abused are more likely to describe relationships with others as threatening or painful, attribute hostile intentions toward social partners, and are less likely to establish or maintain social interactions overall (e.g., Wolfe et al. 2001). This association between childhood abuse and later relationship perceptions may be best understood through an attachment perspective. Attachment theory (Bowlby 1982) rests on the premise that individuals begin to form early cognitive models of relationships with others based on the interactions they have with their parents and other adult caregivers. These cognitive models often entail such notions of others as being predictable and trustworthy, of the self as being lovable and competent, and of relationships in general as being rewarding and worthwhile.
Building from this foundation and illustrating the relative stability of attachment styles throughout the life course, Bartholomew and Horowitz (1991) developed and tested a four-category model of attachment styles based on adults’ perceptions of self and expectations of intimate others. Three of the attachment styles reflected insecure attachments, and two of the three have been found especially likely to lead to IPV perpetration (e.g., Dutton et al. 1994; Henderson et al. 2005). The first insecure attachment style linking childhood abuse to later IPV perpetration has been labeled “fearful” (Bartholomew and Horowitz 1991). Individuals with a fearful attachment orientation experience a sense of unworthiness and the expectation that others will be untrustworthy and rejecting. This insecure attachment style results when children place blame on both themselves and others for their caregivers’ physically abusive behaviors. However, despite this negative model of self and others, fearful individuals still desire social contact and intimacy. This seeming contradiction results in a chronic sense of frustration, as fearful individuals become enmeshed in, and simultaneously try to withdraw from, relationships with romantic partners due to continual anxieties over rejection and abandonment (Bartholomew 1990). The second style of attachment linking childhood abuse to IPV perpetration has been labeled “preoccupied” (Bartholomew and Horowitz 1991). Preoccupied individuals experience a sense of unworthiness but a positive evaluation of others. This attachment style results when children place blame solely on themselves for the abuse that they have incurred. Relationships with others are sought in order to increase preoccupied individuals’ sense of self-worth. Yet, entering these relationships with a negative view of self, preoccupied individuals are as equally anxious about potential rejection and abandonment as their fearful counterparts. They are only differentiated from fearful persons in their positive evaluation of others, making preoccupied individuals less likely to withdraw from partners or blame partners for their anxieties.
How insecure attachment styles may lead to IPV perpetration can largely be understood as a consequence of negative emotional states (Dutton et al. 1994; Goldenson et al. 2007). For example, in a study of 120 men referred to treatment for wife assault, Dutton et al. (1994) found that both preoccupied and fearful men scored higher on levels of jealousy, borderline personality organization, anger, and trauma symptomology. More specifically, due to insecurely attached individuals’ hypersensitivity to rejection and abandonment, fearful and preoccupied individuals are much more likely to experience romantic jealousy in their intimate relationships. Faced with the anticipated or perceived loss of their romantic partners, these individuals then attempt to regain partners’ attention and restore their relationships. However, their ability in doing so is limited as a result of dysfunctional parenting practices rooted in violence, which were unlikely to convey the social skills necessary to handle relationship problems in constructive ways. These limited social skills, combined with the trauma symptoms associated with a history of childhood abuse, then lead to increased anger and the potential for violence as an outlet for such anger. Illustrating that such processes are not confined to male perpetrators, a study of 33 female offenders receiving mandated treatment for IPV perpetration indicated that female offenders reported less attachment security, more trauma-related symptoms, and more personality psychopathology than did a clinical comparison group of women who had not engaged in IPV perpetration (Goldenson et al. 2007).
From an attachment theory perspective, these findings demonstrate that IPV perpetration may be a reaction to romantic partners analogous to the angry behavior of a child who has been separated from their caregiver (Dutton et al. 1994; Henderson et al. 2005). Although the consequences of IPV perpetration for romantic relationship functioning generally are not positive, they do often represent an active attempt on the part of perpetrators to maintain or improve relationships they perceive to be in danger. Thus, for those with a history of childhood abuse, IPV perpetration may be likely if individuals also suffer from attachment insecurities.
Social Psychological Processes and Personality Traits
Complementing both social learning and attachment theories, the social information-processing (SIP) model argues that childhood abuse may lead to IPV perpetration in adulthood through SIP deficiencies. Originally formulated by McFall (1982) to describe social skills among the general population, Holtzworth-Munroe (1992) applied this model to martially violent men specifically, given many therapy programs’ focus on addressing social skills deficits as a strategy to treat domestic violence. In later years, recognizing that many of the pathways leading to IPV perpetration may be similar among men and women, this model was later applied to domestically violent women as well (e.g., Fite et al. 2008).
According to McFall (1982), social skills are acquired and performed in a three-stage process. In the first stage, individuals receive and interpret incoming social stimuli, a process known as “decoding.” Once decoded, individuals must choose an appropriate response to the social situation. This requires individuals to think of all the available potential ways in which they can respond, determine the response type that best fits the situation under examination, and select the behavior from their behavioral repertoire that best reflects their desired response. This process is referred to as “decision-making.” In the final stage, labeled “enactment,” individuals must execute their chosen response and monitor the situation to determine whether their response had the desired impact.
For individuals with a history of abuse, there may be breakdowns in the SIP stages that increase the risk for violence in a number of given social situations. For example, in the first stage, and in line with attachment theories (Bowlby 1982), individuals may misinterpret social situations due to various cognitive deficits, including unrealistic expectations, faulty attributions, and irrational beliefs (Holtzworth-Munroe 1992). For abuse survivors, who are likely to view others as untrustworthy and relationships as unrewarding or threatening (e.g., Wolfe et al. 2001), such cognitions might include being hypervigilant toward hostile social cues, perceiving harm, or ill-intent in situations where none exists. Similarly, and in line with social learning theories (Bandura 1977), individuals with a history of abuse may be unable to produce nonviolent response options in the decision-making stage of the SIP model. They may also perceive violent options as superior in producing the desired situational outcome (Holtzworth-Munroe 1992). Intuitively, each of these breakdowns would be likely if individuals were exposed to violence and come to see violence as an acceptable, or at least justifiable, behavioral response.
Using data from the longitudinal Child Development Project, Fite et al. (2008) provided empirical support for using SIP models to understand IPV perpetration. Specifically, the researchers analyzed whether SIP models mediated the relationship between witnessing interparental relationship conflict during childhood and children’s own romantic relationship conflict in young adulthood. Among children who had witnessed interparental relationship conflict, there were SIP breakdowns in both response generation and response evaluation stages, and these breakdowns partially mediated the intergenerational transmission of relationship conflict. In other words, witnessing interparental conflict during childhood inhibited individuals’ abilities to generate constructive, non-conflictual social responses when interacting with romantic partners in young adulthood. They were also less able to effectively evaluate the potential outcome of their chosen (and often conflictual) responses. Importantly, no significant gender differences were found; both young adult men and women were equally affected by interparental conflict and SIP deficits in their own romantic relationships.
In addition to potential cognitive deficits, survivors of childhood abuse may suffer from personality disorders or exemplify certain personality traits that increase their risk for IPV perpetration. In regard to personality disorders, results from a community sample followed for over 20 years demonstrated that suffering from either Cluster A (i.e., paranoid, schizoid, and schizotypal) or Cluster B (i.e., borderline, narcissistic, antisocial, and histrionic) personality disorders mediated the effect of childhood family violence exposure and adult partner violence (Ehrensaft et al. 2006). The researchers concluded that individuals exposed to interparental violence were more likely to suffer from Cluster A personality disorders and that Cluster A disorders were suggestive of preexisting symptoms of mistrust, suspiciousness, and distortions in cognitions, all of which increased the risk for IPV perpetration. Similarly, individuals exposed to either interparental violence or childhood sexual abuse showed elevated Cluster B symptoms, which increased the risk of IPV perpetration due to individuals’ propensity to engage in aggressive and antisocial acts.
In assessing personality traits, a considerable body of research has documented associations between anger and IPV perpetration. In a meta-analytic review of 33 studies, Norlander and Eckhardt (2005) found that IPV perpetrators, compared to nonviolent men, consistently reported higher levels of anger and hostility, and anger levels also helped to distinguish men in low-moderate and moderate-high IPV relationships. Accordingly, a number of hypotheses have been put forward to understand the connection between anger and IPV, given the recognition that anger does not always lead to aggression. As one example, some scholars have argued that anger may reduce inhibitions against violence, either by aiding in the justification of violence as a viable response option or by disrupting cognitive processes that would defend against a violent response (Norlander and Eckhardt 2005). Each of these propositions, in turn, builds upon social learning and SIP theories, where individuals with a history of abuse already face greater likelihood of accepting violence as justified.
Importantly, while most of the research on anger and IPV to date has focused on male perpetrators, a comparably smaller body of research indicates that female perpetrators may also experience a strong disposition toward angry feelings, often referred to as “trait anger.” For example, in a study of 80 women arrested for domestic violence, Shorey et al. (2011) found that both traits anger and impulsivity were significantly associated with women’s physical and psychological aggression toward intimate partners and that trait anger mediated the relationship between impulsivity and aggression perpetration. Taken together, these findings indicate that individuals with abuse histories may be more likely to perpetrate IPV in adulthood if they also experience SIP deficiencies, suffer from certain personality disorders, or have higher state-trait anger.
Neurophysiological and Genetic Explanations
To date, sociological and psychological theories have dominated the literature in understanding intergenerational linkages of intimate violence. Yet, in recent years, some scholars have argued that attention be paid to potential neurophysiological and genetic explanations. Such explanations compliment more traditional IPV theories, as it is likely that individuals’ physiology interacts with such things as expectations about and attitudes toward violence, emotional regulation, and social skill deficiencies (Margolin et al. 2016). Accordingly, a number of physiological and genetic factors may help to explain why some, but not all, individuals who were abused as children grow up to perpetrate violence against romantic partners.
One potential neurophysiological link between a history of abuse and IPV perpetration is the hypothalamic-pituitary-adrenocortical (HPA) axis (Margolin et al. 2016), which regulates many bodily processes, including reactions to stress, and the regulation of moods and emotions. Exposure to early life stressors, such as childhood abuse, may lead to the dysregulation of the HPA axis. In turn, HPA reactivity, specifically in regard to cortisol activity, has been linked to aggression and hostility in adult intimate relationships. For instance, neuroimaging research conducted by George et al. (2004) demonstrated that abusive men had lower cortisol levels, resulting in hypersensitivity to social stimuli, including certain looks or critical statements from romantic partners.
In addition to the more indirect HPA dysregulation processes which may result from early exposure to violence, brain structure and functioning may be affected by childhood abuse more directly via traumatic brain injury. Severe childhood abuse may result in traumatic brain injury, later increasing the risk for IPV perpetration. Prior research has reported that abusive men have higher rates of head injury, with rates ranging between 40% and 62%, compared to an estimated 6% in the general population (Howard 2012). The orbitofrontal and anterior temporal lobes are the most common sites of brain contusion and laceration. In turn, the orbitofrontal cortex is the area of the brain that maps rewards and punishments through cortical activity. When damaged, this can lead to deficient social competencies in the arenas of judgment and behavior, the ability to experience certain moods and emotions, and the capacity for realizing others’ thoughts and feelings.
Finally, some research has noted that genetic susceptibility may help to differentiate between maltreated children who do and do not go on to become aggressive and violent in adulthood. Deficient monoamine oxidase A (MAOA) gene activity, which metabolizes neurotransmitters such as norepinephrine, serotonin, and dopamine, may predispose individuals to respond hyperreactively toward perceived threats in incoming social stimuli. Examining the effects of maltreatment status and MAOA activity on antisocial behavior, Caspi et al. (2002) conducted research among a sample of individuals assessed from early childhood (age 3) to young adulthood (age 26). When measured by itself, MAOA activity had no significant effect on antisocial behavior. However, among maltreated children, low MAOA activity predicted four antisocial behavioral measures: adolescent conduct disorder, adult violent criminal convictions, self-reported disposition toward violence, and informant-reports of antisocial personality disorder symptoms. Taken together, these findings lead to the conclusion that, among individuals with a history of abuse, certain neurophysiological deficiencies and genetic susceptibilities may further increase risk for perpetrating IPV in later life. Importantly, while the vast majority of neurophysiological and genetic research on IPV perpetration to date has been conducted among men, it is not implausible that such findings may also hold true for women.
Integrating social psychological and biological models, contemporary evolutionary scientists have proposed a life history strategy for understanding the connections between a history of childhood abuse and IPV perpetration in adulthood. Life history theory is based on the premise that all living organisms face trade-offs in dividing their energy and effort between important survival goals, given that resources for survival and reproduction are finite (Kaplan and Gangestad 2005). The trade-offs that organisms, or individuals, make determine what is referred to as their “life history strategy,” which exists on a fast to slow continuum. A fast life strategy entails making decisions based often on immediate gratification or short-term gain with little investment in long-term maintenance or growth. Fast life strategies are most often exhibited in environments that are considered dangerous or limited in resources. Because individuals in these environments experience shorter life expectancies, it pays more to invest in short-term gains instead of devoting time and energy into long-term planning and relationships. Sexual promiscuity and aggressive behaviors are commonly cited examples of a fast life strategy (e.g., Brumbach et al. 2009). Conversely, a slow life strategy entails making long-term investments that signify the preservation and progression of both physiological systems and “embodied capital” (e.g., knowledge and skills) (Kaplan and Gangestad 2005). Slow life strategies are commonly adopted in environments that are considered safe, stable, and plentiful in resources. Delaying reproduction until individuals are able to sufficiently provide for their offspring financially and emotionally is one example of a slow life strategy.
Importantly, a great degree of variability exists between individuals’ life history strategies; and some scholars have argued that such variation may be accounted for by the family. For example, building on Bowlby’s (1982) attachment theory, Belsky et al. (1991) suggested that the family environment is central to the social, emotional, and behavioral development of children. Accordingly, when family functioning is disrupted through environmental uncertainty and its consequent stressors, secure attachments between parents and children become damaged, leading children to adopt internal working models of relationship uncertainty. In turn, internal working models of significant others and the environment overall as uncertain and unpredictable increase the risk for adoption of a fast life history approach. Analyzing data from the National Longitudinal Study of Adolescent and Adult Health (Add Health), Brumbach et al. (2009) provided empirical support for the importance of family in predicting young adults’ life history strategies and a wide variety of life history traits (e.g., mental and physical health, sexual attitudes and behaviors, social deviance). Of most relevance to the study of IPV perpetration, the researchers found that environmental harshness during adolescence via exposure to violence was predictive of adolescent delinquent behavior. Similarly, environmental unpredictability, which measured frequent changes or ongoing inconsistency in individuals’ childhood environments (e.g., parents failing to provide for basic needs, being removed by social services), was predictive of delinquency during young adulthood. In turn, adolescent and young adult delinquency was associated with other negative life history traits (e.g., mental and physical health, sexual attitudes and behaviors) indicative of fast life history strategy. Considering that aggression is one potential form of delinquency, individuals who (unconsciously) adopt a fast life history approach as a result of childhood abuse may be more likely to perpetrate IPV in adolescence and young adulthood.
Being a victim of or witness to violence in early life is a significant risk factor for perpetrating violence throughout the life course. Yet, most individuals with a history of abuse still do not become violent toward romantic partners in adulthood. As demonstrated throughout, a full understanding of this variance and the complex processes by which intergenerational violence is transmitted requires a multifaceted theoretical framework. Specifically, those individuals exposed to violence in childhood via child maltreatment or interparental violence may be especially likely to perpetrate IPV in adulthood if they are also placed in multiple contexts that model violence as an appropriate way of dealing with conflict. These contexts may include associations with deviant peer groups and intimate partners, as well as more structural contexts such as neighborhoods (Capaldi et al. 2012; Copp et al. 2016). Relatedly, when combined with a history of abuse, attachment insecurities (Dutton et al. 1994), cognitive difficulties (Holtzworth-Munroe 1992), personality disorders or certain personality traits (Ehrensaft et al. 2006; Shorey et al. 2011), neurophysiological deficiencies (Margolin et al. 2016), genetic susceptibilities (Caspi et al. 2002), and adoption of a fast life history strategy (Brumbach et al. 2009) may each further increase the risk for violence toward a romantic partner in adulthood.
- Bandura, A. (1977). Social learning theory. Englewood Cliffs: Prentice-Hall.Google Scholar
- Bowlby, J. (1982). Attachment and loss: Vol. 1, Attachment. New York: Basic Books.Google Scholar
- Caspi, A., McClay, J., Moffitt, T.E., Mill, J., Martin, J., Craig, I.W., … & Poulton, R. (2002). Role of genotype in the cycle of violence in maltreated children. Science, 297, 851–853.Google Scholar
- Centers for Disease Control and Prevention, National Center for Injury Prevention and Control, Division of Violence Prevention. (2010). The national intimate partner and sexual violence survey, 2010 summary report. http://www.cdc.gov/ViolencePrevention/NISVS.
- Copp, J. E., Giordano, P. C., Longmore, M. A., & Manning, W. D. (2016). The development of attitudes toward intimate partner violence: An examination of key correlates among a sample of young adults. Journal of Interpersonal Violence, 1–31. https://doi.org/10.1177/0886260516651311.
- Fite, J. E., Bates, J. E., Holtzworth-Munroe, A., Dodge, K. A., Nay, S. Y., & Pettit, G. S. (2008). Social information processing mediates the intergenerational transmission of aggressiveness in romantic relationships. Journal of Family Psychology, 22, 367–376.CrossRefPubMedPubMedCentralGoogle Scholar
- George, D. T., Rawlings, R. R., Williams, W. A., Phillips, M. J., Fong, G., Kerich, M.…., & Hommer, D. (2004). A select group of perpetrators of domestic violence: Evidence of decreased metabolism in the right hypothalamus and reduced relationships between cortical/subcortical brain structures in position emission tomography. Psychiatry Research: Neuroimaging, 130, 11–25.CrossRefPubMedGoogle Scholar
- Kaplan, H. S., & Gangestad, S. W. (2005). Life history theory and evolutionary psychology. In D. M. Buss (Ed.), Handbook of evolutionary psychology (pp. 68–95). Hoboken: Wiley.Google Scholar
- Kaufman-Parks, A. M., DeMaris, A., Giordano, P. C., Manning, W. D., & Longmore, M. A. (2016). Parents and partners: Moderating and mediating influences on intimate partner violence across adolescence and young adulthood. Journal of Social and Personal Relationships, 1–29. https://doi.org/10.1177/0265407516676639.
- Kaufman-Parks, A. M., DeMaris, A., Giordano, P. C., Manning, W. D., & Longmore, M. A. (2017). Intimate partner violence perpetration from adolescence to young adulthood: Trajectories and the role of familial factors. Journal of Family Violence, 1–15. https://doi.org/10.1007/s10896-017-9924-5.