CHAPTER SIX (Part 2)

(Discussion:  ... Continued)

Regardless of the modelling agent, it has been suggested that the exposure to violence in the family of origin teaches children the acceptability of violence (Kalmuss, 1984).  It also teaches them to resolve conflicts through the use of violent tactics.  While researchers agree on the strong modelling potential that exposure to violence in the family has on the perpetration of partner abuse in future relationships, the transfer of dysfunctional modes of conflict resolution to these relationships has not been formally addressed.

These findings provide strong support for the influence of observing one's same sex parent hitting the other on the perpetration of current partner abuse. However, it should also be noted that exposure to violence in the family of origin was not reported by 78.57 percent of males and 81.82 percent of females who perpetrated current partner abuse.  The perpetration of partner abuse by these individuals can be accounted for in part by measurement error as well as by a number of other factors that may include the modelling of violence by peers and the media.
 

Past perpetrated partner abuse.

Past perpetrated partner abuse emerged as a significant main effect as well as an interaction effect with stress for predicting current perpetrated partner abuse by males.  (A detailed discussion of this interaction effect will follow in the next section.)  By comparison, the effect of this variable was not as salient in the prediction of current perpetrated partner abuse by females.

The above finding is supported by the results of the correlations conducted on past and current perpetrated partner abuse for males and females.  The strength of the association between these measures were found to be greater for males (r=.46, p < .001) than they were for females (r=.29, p < .001).  The magnitude of these correlations are consistent with other research that also surveyed respondents over a two to 2.5 year period.  For example, O'Leary et al.'s (1989) study of stability in marital aggression within a community sample of men and women found that the correlations between past and current perpetrated partner abuse over a 30 month period were r=.31 for both males and females.  Marshall and Rose's (1990) study of premarital violence among college students provided similar results (r=.33, p < .001 for males and for females, r=.25, p < .001 for females).  These correlations lend support to this study's finding that past perpetrated partner abuse plays a role in predicting current perpetrated partner abuse by both males and females.

The linkage between past and current perpetrated partner abuse is also supported by personality theorists who argue that people's behaviour is consistent across time and situations (Davison & Neale, 1990).  This line of reasoning aids in our understanding of how partner abuse is sustained and transferred from one relationship to another (Kalmuss & Seltzer, 1986).  However, at the same time, the relationship between these variables has been shown to be anything but perfect.  For many, current perpetrated partner abuse is not always predicted by past perpetrated partner abuse and is likely to be influenced by a number of other factors.
 

Stress

Although the relationship between stress and the perpetration of partner abuse has not previously been clearly delineated, the results of this research appear to shed some light on the how these variables may be linked.  While this study's findings are consistent with other research demonstrating that batterers can be differentiated by measures of stress (Barnett et al., 1991; Neidig et al., 1985; Seltzer & Kalmuss, 1988), its results extend previous findings by examining the influence gender has on the relationship between these variables.

Analyses conducted on the individual stress items revealed that there were only two stress events common to males and females that significantly differentiated abusers from nonabusers (i.e., financial problems and stopping school).  While no other stress events differentiated female abusers from female nonabusers, male abusers on the other hand, were found to differ significantly from male nonabusers on three other stress events (i.e., lost job, changed job and spouse started work).  These findings suggest that males' experiences of a wider variety of stressors is more greatly reflected in the perpetration of current abuse than females'.

Analyses conducted on the full stress scale provide further insight into gender's influence on stress and the perpetration of current partner abuse. First, the significant interaction effects between gender and partner abuse based on both weighted and unweighted scales demonstrated that males who abused their partners during the past year had higher stress scores than females who did the same.  This finding suggests that the influence of stress on the perpetration of partner abuse is dependent upon one's gender.  Second, the results of Pearson correlations also lend support to a stronger linkage between stress and perpetrated partner abuse among males.  Results indicated that while stress was found to be significantly related to the perpetration of current partner abuse by females, the magnitude of the correlations were half those found for males based on both versions of the stress scale.

The most definitive characterization of sex differences with respect to the relationship between stress and the current perpetration of partner abuse emerged from the results of the logistic regression analyses.  Results clearly indicated that while stress is an important risk factor in current partner abuse perpetrated by males, it is not one for females.

As a significant main effect, increments in stress levels increased the likelihood for current perpetrated partner abuse by males by a factor of 1.696. Analyses conducted on the male data also revealed two significant interaction effects (i.e., stress by past perpetrated partner abuse and stress by age).  In each of these interactions, under high stress conditions, past perpetrated partner abuse, and being less than 50 years of age were associated with an increased likelihood of perpetrating current partner abuse.  Finally, a comparison of the log likelihood ratio of the main effects model, and the stress and alcohol models showed that the stress interaction model provided the best explanation of current perpetrated partner abuse by males.

The influence of past perpetrated partner abuse on partner abuse in future relationships likens itself to that of exposure to violence in the family of origin, in that it is demonstrates a reliance upon pre-existing and well established patterns of conflict resolution. When accompanied by high levels of stress, a partner abuse incident represents a form of negative reinforcement (Bandura, 1973) in which the immediate source of stress is reduced following the perpetration of abuse.

If the pattern of current perpetrated partner abuse is a reflection of what has transpired in previous relationships, then it is likely that the contingencies described above may have been operative in those earlier relationships.  When these contingencies are repeated over time, the result is a well established pattern of behaviour that is extremely difficult to extinguish.   Albeit dysfunctional, the combined effects of high stress and past partner abuse increases the likelihood of current perpetrated partner abuse among males.

Results of an experiment examining changes in the systolic blood pressures of college students suggest that aggression could be adaptive (Hokanson, 1970). Findings indicated that when subjects were given an opportunity to counteragress following a planned harassment-insult procedure, there was a dramatic decrease in blood pressure readings to prefrustrative levels compared to those subjects with no opportunity to aggress.  Thus, perpetrating partner abuse may be for some males an adaptive response to frustration and stress. Once the partner abuse incident is ended, so is the stress and frustration that proceed it.

The interaction between stress and age provides some interesting insight into the dynamics underlying the perpetration of current partner abuse by males. An examination of this interaction effect indicates that for males reporting no stress, low rates of current perpetrated partner abuse can be found across all age groups.  However, for males reporting stress, those in the age group 35-49 followed by those in the age group 18-34 appear to be the most vulnerable to perpetrate current partner abuse.  Finally, males who experienced stress and who were 50 years old or more, reported the lowest rate of current perpetrated partner abuse.

The above findings are better understood when the items included in the stress scale are considered.  For the most part, the stressors included in this study reflect the concerns of younger people (i.e., being fired, starting a new job, school, moving, having a baby, etc.).  Post hoc analyses also indicated that males between the ages of 18 to 34 years who experience high levels of stress were found to have low scores on ego strength (r=-.28,p < .001).  In other words, the stressors included in this study appear to be ego threatening to young males. It should therefore not be surprising that the rates of current perpetrated partner abuse are highest among those who are under 50 years of age and who report high levels of stress.
 

Alcohol Risk Factors

A comparison of Wave 1 and Wave 2 results demonstrate changes in the relationship between perpetrated partner abuse and alcohol consumption and dependence for males and females.  In Wave 1, multiple regression analyses demonstrated that alcohol dependence and the interaction between alcohol consumption and the neuroticism were significant risk factors in male perpetrated partner abuse.  Analyses conducted on Wave 2 male data found that these relationships lost their salience.

For females, the results of Wave 1 multiple regression analyses demonstrated that the perpetration of partner abuse was predicted by the interaction between alcohol consumption and EPQP scores.  The findings of Wave 2 also indicated that alcohol's effect on the perpetration of current partner abuse by females was dependent on the influence of other variables.  However, this time  current perpetrated partner abuse by females was predicted by the interaction between alcohol consumption and observing one's mother hitting one's father, past perpetrated partner abuse and neuroticism.

Comparisons of Wave 1 and Wave 2 correlation coefficients for the perpetration of partner abuse and alcohol measures also demonstrated a diminishment in effect in Wave 2 for both males and females. Whereas alcohol consumption, MAST, SADD and the alcohol dependence index were all found to be significantly correlated with the prevalence of perpetrated partner abuse in Wave 1 for males, only MAST and SADD attained levels of significance in Wave 2 for current perpetrated partner abuse.  For females, only SADD retained statistical significance in Wave 2 whereas previously both the MAST and SADD were found to be significantly correlated with the perpetration of partner abuse.

The diminished effect of the alcohol variables on perpetrated partner abuse can be explained in a number of ways.  First, alcohol's diminished effect for males is better understood when the effects of attrition with respect to alcohol consumption are considered.  Given that male dropouts consumed almost 30 percent more alcohol than male respondents who completed Wave 2, it is not surprising that lower levels of alcohol consumption are reflected in the relationship between alcohol and the perpetration of current partner abuse.  It also reasonable to suggest that high alcohol consumers are likely among those who score high on alcohol dependence.  If so, the loss of salience in the relationship between that alcohol dependence and partner abuse can be more easily discerned.

Although the instability of alcohol measures is less pronounced for females, it may also be affected somewhat by attrition.  For example, Wave 2 female completers consumed more alcohol than female dropouts suggesting that low alcohol consuming dropouts could have attenuated the correlations.

Aside from the issue of attrition, the relative instability of some measures compared to others can be more easily explained for both males and females if one considers what it is that each alcohol instrument measures.  For example, whereas the MAST assesses the social consequences associated with drinking (i.e., marital, family and work problems), the SADD and ALC3R assess its physiological and psychological effects (i.e., hangovers, blackouts and inability to control drinking).  The lower MAST scores reported in Wave 2 suggest that male partner abusers in Wave 2 experienced fewer social consequences associated with drinking compared to either physiological and psychological effects.

Finally, the lack of association between Wave 2 EPQL scores and alcohol measures for both males and females raises the possibility of yet another explanation for the diminished relationship between the perpetration of partner abuse and alcohol.  The results of the correlations between Wave 2 EPQL scores and various alcohol measures suggests that alcohol's reduced effect on perpetrated partner abuse was not related to pressures associated with social conformity. Instead, differences in the relationship between perpetrated partner abuse and alcohol for both males and females may instead reflect actual changes in drinking behaviour and/or be reflective of differences found in the reporting of partner abuse.

An examination of the interactions between alcohol consumption and observing mother hitting father, past perpetrated partner abuse, and neuroticism index scores reveals a number of complex sets of associations.  In all instances, females who did not observe their mothers hitting their fathers, did not perpetrate partner abuse in the past, and had low or moderate scores on the neuroticism index reported lower rates of current perpetrated partner abuse, independent of the amount of alcohol consumed.  On the other hand, the highest rates of current perpetrated partner abuse were found among females who scored high on neuroticism and who consumed the highest amounts of alcohol.  This pattern however, was not repeated for females who either observed their mothers hitting their fathers or perpetrated partner abuse in the past.

In both these latter two instances, highest rates of current perpetrated partner abuse were reported by females who either abstained from alcohol or consumed high amounts of alcohol.  The lack of research on the relationship between alcohol and female perpetrated partner abuse makes comparisons with other studies difficult.  Nevertheless, the interpretations of these results will again borrow upon the explanations advanced by Frieze and Schafer (1984).

According to these authors, considerable variability exists in the effects of alcohol on women.  Differences in women with respect to amounts of body fat, phase of the menstrual cycle, and use of oral contraceptives make it difficult to predict the effect that drinking may have on them.  Because of this, the rate and level of intoxication may vary from one woman to another given the same amount of alcohol.  Frieze and Schafer (1984) also suggested that the effects of alcohol in women may also depend on their cognitive interpretation of the physical sensations experienced when drinking. If the sensation of warmth associated with vasodilation is interpreted as power (as is often the case for men), then increased aggression may result.  On the other hand, if the interpretation is one of emotional warmth, the likelihood for violence is reduced.  The unpredictability of alcohol's effect on women together with the cognitive interpretation of its effect help to explain why the influence of alcohol consumption on current perpetrated partner abuse by females with respect to past abuse and observing abuse by mothers is nonlinear.

It is also possible that for high alcohol consuming females the exposure to violence in the family of origin may also include a pattern of heavy drinking. Given individuals' genetic predisposition toward alcoholism (Brook & Brook, 1992), this hypothesis seems to have theoretical merit.  While Brook and Brook (1992) noted that little is known about alcohol's genetic influences on family
interactions, post hoc analyses examining the relationship between "observing mother hitting father" and "mother's MAST scores" (assessed in Wave 1) revealed a correlation coefficient of .18 (p < .001).  It appears that mothers who perpetrated partner abuse against their husbands also had drinking problems. This finding suggests that in addition to a possible genetic predisposition toward alcoholism, there may be the modelling of problem drinking as well as partner abuse by the daughter.  It is possible that the intergenerational transmission of partner abuse is also somehow linked to the intergenerational transmission of problematic drinking.

Personality Risk Factors

The correlational and logistic regression analyses conducted on the personality measures with respect to the perpetration of partner abuse, have provided some of this study's most interesting findings. Logistic regression analyses conducted on the female data revealed that two of the personality measures found to be significant predictors of partner abuse by females in Wave 1 (i.e., EPQP and Neuroticism Index) were also found to be significant predictors of current perpetrated partner abuse in Wave 2.  This pattern, was not repeated for males as no single personality measure was found to predict current perpetrated partner abuse by them in Wave 2.

Measures assessing personality were found to be more stable across Wave 1 and Wave 2 for both males and females than was the case for the alcohol measures.  The stability found among personality measures is consistent with the argument that personality is biologically determined, and therefore remains relatively fixed across time (Buss & Plomin, 1984; Eysenck, 1965).  Given this premise, the following two questions are then raised: 1) Why did the relationship between the prevalence of perpetrated partner abuse and the EPQL (for males), ego-strength (for females), self-esteem (for females) and the MacAndrew Scale (for males and females) diminish in Wave 2?  2) Why did the Neuroticism Index (for males) and the MacAndrew Scale (for females) lose their salience as predictors of current perpetrated partner abuse in Wave 2?

With respect to the diminished effect of the relationship between perpetrated partner abuse and the EPQL, it is possible that for males who perpetrated partner abuse, their tendency to dissimulate diminished in Wave 2 because they felt more at ease with the interview process and questions being asked.  The lessening of the effect for the ego-strength and self-esteem scales among females may have more to with changes in the reporting pattern of partner abuse in Wave 2, and therefore be reflected in a weaker correlation. Finally, the loss of salience in the relationship between the prevalence of perpetrated partner abuse and MacAndrew Scale scores for both males and females may be associated with changes in the drinking patterns among males and females.  Given that the MacAndrew Scale gauges an individual's predisposition toward alcohol and drug dependency (MacAndrew, 1965), the weakening of the alcohol measures reported earlier, may in part have influenced the strength of the association between the MacAndrew Scale and the perpetration of current partner abuse.  Being that this is the first study to examine these relationships longitudinally, comparisons with other research cannot be made and as a result, limits one's ability to move beyond speculation.

The question regarding the loss of salience for the Neuroticism Index' for males and the MacAndrew Scale for females is a more difficult one to answer.  As suggested previously, it is possible that changes in the response patterns for Wave 2 reports of partner abuse may have influenced the relationship between the above described personality variables.  Equally likely, is the possibility that changes made to the partner abuse models tested in Wave 2 have also influenced the strength of these personality measures.  When Wave 1 and Wave 2 partner abuse models are compared, the following important differences are noted: 1) the Wave 2 model was based on longitudinal data whereas the Wave 1 model was not, 2) the Wave 2 model included all the variables tested in the Wave 1 model, as well as additional variables such as life stress events and exposure to violence in the family of origin, 3) the Wave 2 model included reported partner abuse from Wave 1 as one of its independent variables, and 4) Wave 1 and Wave 2 employed different statistical approaches to test the partner abuse models.  When all these differences are considered, changes in the relative importance of some variables from Wave 1 to Wave 2 are better understood.

Finally, the inclusion of past perpetrated partner abuse in the Wave 2 partner abuse model may have cancelled out the effects of individual differences previously reported for males in Wave 1.  This is supported by the significant correlation found between the neuroticism index and current perpetrated partner abuse (r=.16, p < .01) derived from the Wave 2 data.  Also recall that Wave 1 regression analyses found a link between the prevalence of perpetrated partner abuse and Neuroticism.  It was also this variable (i.e., Wave 1 prevalence of partner abuse) that constituted "past perpetrated partner abuse" and was found to be a significant predictor of current perpetrated partner abuse by males in the Wave 2 analyses.

In light of this linkage, one might expect that the effects of neuroticism on the perpetration of current partner abuse by males may have been suppressed and that the removal of past perpetrated partner abuse from the logistic regression model would improve the explanatory power of personality.  However, when this was done, the effects of personality still remained nonsignificant suggesting that its salience is likely to have been influenced by the factors previously noted or others not yet considered (i.e., stress).

The relationship between personality and current perpetrated partner abuse in this study is more clearly delineated by female data.  Results of the logistic regression analyses indicate that high scores on the EPQP and the Neuroticism Index were found to predict current perpetrated partner abuse.  Furthermore, the odds of perpetrating current partner abuse increased by a factor of 1.333 for those with high EPQP scores and by a factor of 1.355 for those with high Neuroticism scores.  As was described for males with high scores along this dimension (Eysenck, 1965), females who perpetrate current partner abuse also have a tendency to over-react and experience high levels of anxiety.  But in addition to these characteristics, they also possess the ability to be toughminded, uncaring and antisocial (Eysenck, 1965).

An interesting feature of the EPQP, is that high scores along this personality dimension are typically found among males (Eysenck, 1965).  Thus, the finding that high scores on the EPQP predict female perpetrated current partner abuse suggests that these women are characteristically more masculine.  This premise is supported by Kalichman (1988) who reported that females found guilty of domestic murders scored low on the MMPI MF scale which indicates higher masculinity.  In light of this study's findings of equivalent rates of current perpetrated partner abuse by males and females, investigations into possible sex differences in partner abuse might prove to be more beneficial if they were directed toward examining differences in hormone levels (i.e., testosterone), rather than differences in gender.

The personality characteristics just described are part of the multidimensional profile of the partner abuser described in family violence research.  Clinical research conducted by Gondolf (1985) found that male partner abusers experienced difficulties with impulse control, tolerance for stress and low self- esteem.  General population (Bland and Orn, 1986) and clinically based research (Kalichman, 1988; Hale et al., 1988; Scheurger and Reigle, 1988) found that male and female partner abusers exhibited antisocial behaviours similar to those characterized by the EPQP (i.e., social nonconformity, chronic lying and trouble with the law).  The challenge facing researchers is to determine the factors that predispose individuals to different types of abuser profiles.

The association between personality and the perpetration of current partner abuse by females, and what is possibly an indirect link to current partner abuse perpetrated by males supports the inclusion of personality measures in general population based research on partner abuse.  The diminished ability of personality measures to directly predict current perpetrated partner abuse among males also suggests that various characteristics may be subsumed within each other.  In order to overcome this problem, developing strategies that can isolate the effects of each remains an ongoing test for future research.

Next: Chapter 6 Part 3