Alcohol and intimate partner violence: when can we say that heavy drinking is a contributing cause of violence?


Intimate partner violence (IPV) has been recognized as a prevalent and serious problem for more than 20 years. At a fundamental level violence against women, as well as many other forms of violence, occurs because physical aggression is accepted or tolerated, at least under some circumstances. However, many other factors contribute to the occurrence of violence, and ‘the field appears to be developing toward an integrative, metatheoretical model of violence that considers multiple variables operating at different times in a probabilistic fashion’ (Crowell & Burgess 1996, p. 69).

Of the multiple variables that have been proposed, excessive alcohol consumption is one of the most controversial. While there is broad agreement that partner-violent men are often heavy drinkers and heavy drinking often accompanies the violence, there is substantial disagreement regarding whether alcohol consumption plays any causal role in IPV. Some argue that there is no credible evidence that excessive drinking, either acutely or chronically, is a cause of marital aggression, that alcohol simply serves as an ‘excuse’ for the aggression, and that the aggression does not stop with successful treatment of the problem drinking (Zubretsky & Digirolamo 1996). There is, however, a different perspective, one that argues that acute alcohol consumption is a contributing cause of marital aggression, and that successful resolution of heavy drinking will often lead to a reduction, and in some cases a cessation, of IPV.

The evidence regarding alcohol and marital violence has been reviewed extensively. Briefly, case–control studies of violent men, alcoholic men, and abused women, and epidemiological studies of the general population and of women in health-care settings have consistently found an association between indices of heavy drinking and the occurrence of domestic violence. In a meta-analytical summary of this literature, Lipsey et al. (1997) reported an effect size of 0.22 and interpreted this as indicating that the upper half of drinkers had twice the risk of violence as the lower half. Many of these studies did not control for other factors that might be associated with both drinking and marital violence, but a growing number of studies have controlled for such variables in the context of multivariate analyses and continued to find that drinking behavior is associated with domestic violence, independent of other factors, such as age, socioeconomic or occupational status, and race/ethnicity (e.g. Leonard et al. 1985; Pan, Neidig & O’Leary 1994). At a more substantive level, the relationship has been observed after controlling for hostility and antisocial behavior (e.g. Leonard & Senchak 1993; Reider et al. 1988), normative views of aggression (Kaufman Kantor & Straus 1990) and drug problems (Pan et al. 1994). In addition, most of these studies also controlled for marital satisfaction, relationships discord or similar variables and continued to find a significant relationship between alcohol and marital violence (e.g. Leonard & Senchak 1993; McKenry,  Julian,  &  Gavazzi  1995;  see  Leonard  2001 for a review). Moreover, there have been several longitudinal studies that have reported that drinking patterns are longitudinally predictive of subsequent domestic violence (Leonard & Senchak 1996; Leonard & Quigley 1999). These studies, in fact, controlled for previous domestic violence. Clearly, there is the need for further examination of third variables. However, at present the research suggests that third variables do not account for the entire association between heavy drinking patterns and domestic violence, although they do account for some of the association.

Recently, there have a number of reports that suggest treatment for alcohol dependence is associated with reductions in intimate partner violence (O’Farrell & Choquette 1991; O’Farrell et al. 2003; Stuart et al. 2003), that this reduction is also apparent for verbal aggression (O’Farrell et al. 2000) and that this reduction is observable up to 2 years post-treatment (O’Farrell et al. 1999). In addition, this research has demonstrated that alcoholics who relapsed did not reduce their violence, whereas alcoholics in remission did reduce their violence (O’Farrell & Murphy 1995). In addition, O’Farrell et al. (2004) found that couple's involvement with behavioral couples treatment (BCT) was predictive of lower post-treatment partner violence in alcoholic men, and mediation analyses suggested that relationship functioning and drinking mediated this relationship. Similarly, Fals-Stewart et al. (2002) found that among drug abusing men, many with alcohol problems, behavioral couple's therapy reduced their violence from pre-treatment (43%) to post-treatment (17%). Moreover, men in the behavioral couples therapy engaged in less post-treatment violence than men in the individual treatment, an effect that was mediated by the better treatment outcome of BCT for both marital adjustment and drinking/drug use.

Although the evidence is quite strong that drinking patterns marked by current, very heavy consumption or by other alcohol problems are associated with domestic violence, these studies provide only suggestive evidence that intoxication, per se, has any association with domestic violence. A heavy drinker who has engaged in domestic violence may have done so only while sober or both while sober and while intoxicated. Laboratory studies of aggressive behavior indicate that subjects who receive alcohol are more aggressive than subjects who receive no alcohol or subjects who receive placebo beverages (Bushman & Cooper 1990; Ito et al. 1996; Lipsey et al. 1997). Giancola & Zeichner (1995) and Richardson (1981) found that the effect of alcohol on laboratory-studied aggression was also present when a woman was the target of the aggression. In addition, several studies of marital behavior have demonstrated that alcohol administration to men increases the extent of negative verbal behavior displayed by the men and their partners (e.g. Jacob & Krahn 1988; Leonard & Roberts 1998).

Critics of this literature have often called for research into naturally occurring violent events to examine whether acute drinking is associated with the occurrence or severity of these events. While early research failed to find drinking to be related to assault episodes (Bard & Zacker 1974), more recent research has been supportive. In a community sample of newlywed couples, Leonard & Quigley (1999) found that husbands’ alcohol consumption was more prevalent among physical violence events than among verbal aggression events. Testa, Quigley & Leonard (2003) reported that episodes of violence in which the husband was drinking involved more acts of violence and were more likely to involve severe violence compared to sober violence episodes. Similar findings have been reported in clinical samples. Murphy et al. (in press) asked alcoholics and their spouses to report on conflict episodes that either involved physical violence or did not. Alcoholic men were more likely to be drinking during violent events according to wives’ accounts, and more likely to have consumed 6 or more drinks prior to violent events, according to husbands’ accounts. According to the alcoholic husbands, they had approximately 14 standard drinks and an estimated blood alcohol level of 0.19 in violent events. In contrast, they reported having consumed an average of 8 standard drinks and an estimated blood alcohol level of 0.11 in conflict events that did not include violence. Fals-Stewart (2003) assessed men entering treatment for alcoholism or domestic violence. Using a time-line follow-back method to determine which days in the past year substance use occurred and, independently, on which days marital violence occurred, he found that severe violence was much higher on days of heavy drinking (six or more drinks) than on days of no substance use and that the violence was most likely to occur within 4 hours after drinking. In addition, several studies have reported that alcohol use is more common among serious physical assault events than among less serious events (Martin & Bachman 1997; Thompson, Saltzman & Bibel 1999).

As careful researchers, we usually caution that findings are not be taken to suggest causality. Nevertheless, that is the goal of much of our research. If we can establish causality, we have the basis for developing interventions, preventions and policies that ‘cause’ improvements in health and wellbeing. In the area of alcohol and violence, this means addressing the excessive drinking behavior of individuals who have behaved in a violent way while drinking. In contrast, if there is no evidence of causality, it is not possible to argue that reducing the excessive drinking is a critical aspect of reducing violence.

Thus, a critical question is, ‘What methodology is needed to demonstrate a causal relationship between excessive alcohol use and marital violence?’. It is well understood that cross-sectional associations between alcohol use and marital violence do not establish causality. This is true even for the event-based designs in which a variable such as stress, which differs from day to day, may cause both drinking and violence. Longitudinal associations serve basically to establish temporal precedence, but the possibility of spurious variables creating the association is not ameliorated by this approach. Sophisticated longitudinal modeling has the possibility of misspecification; that is, omitting the critical variables. Experimental studies can demonstrate causality, but only within the framework of the experiment. Hence, an experiment can demonstrate that the administration of alcohol causes increased shock settings in a competitive paradigm, but whether that generalizes outside the laboratory is a concern. Treatment studies may remove or reduce a putative causal factor, and observe that the hypothesized effect is similarly reduced. However, the cause  of  a  cure  is  not  necessarily  directly  indicative  of the cause of the disorder.

In short, there is no design that will definitively demonstrate causality. Causality, at least in this instance, is a scientific attribution based on the convergence of evidence drawn from varied, but by themselves flawed, sources. In the face of this realization, we must ask what more evidence is needed to tip the balance from ‘there is no evidence of causality’ to ‘there is evidence that heavy alcohol consumption is a contributing cause of violence’. My own view is that we have reached the point where we should conclude that heavy drinking is a contributing cause of violence.

In asserting that heavy drinking contributes to violence, we should not gloss over the several important caveats. This assertion does not mean that the presence of alcohol is the only or even the primary determinant of whether violence will occur. Alcohol is neither a necessary nor sufficient cause of violence. Of most critical importance, alcohol's influence on intimate partner violence is not uniform. Instead, it is clear that alcohol contributes to violence in some people under some circumstances. There remains much to be learned about the specific conditions under which alcohol contributes to violence. However, there is evidence to suggest that alcohol is unlikely to predict the occurrence of a violent episode among individuals who have very low hostile motivations. Instead, alcohol appears to act synergistically with hostile motivations to predict the occurrence of violence (Bailey & Taylor 1991; Jacob et al. 2001; Giancola 2002; Giancola et al. 2003). However, there is also evidence that among individuals with very high levels of hostile motivations, alcohol does not appear to contribute to the occurrence of violence (Blane, Miller & Leonard 1988; Rice & Harris 1995). Very recently, we have reported evidence that suggests that drinking on a specific day is associated with the occurrence of severe violence among antisocial men, but that it is not associated with the occurrence of non-severe violence among these antisocial men (Fals-Stewart, Leonard & Birchler, in press). Taken together, these findings suggest that the occurrence of a violent episode among the most violent, antisocial men is not related to alcohol consumption, but that alcohol consumption may increase the severity of the  violent episode. Given the limited research into moderators of the alcohol–intimate partner violence relationship, this conclusion should be regarded as tentative and in need of further research.

The assertion that alcohol contributes to the occurrence of violence has very important implications for alcoholism treatment. For men in alcoholism treatment who have engaged in violence, the successful reduction or elimination of heavy drinking should have a dramatic impact on the occurrence of violence. In addition, the treatment literature has found that behavioral couples treatment appears to have an enhanced effect on drinking outcomes, and through this and its impact on marital functioning is an effective treatment for violence. At present, we have little information regarding the value of adding other interventions, such as anger management, to alcoholism treatment, and this is an area of further development. The findings that alcohol is not associated with the occurrence of violence among very aggressive and antisocial men, but may be associated with the severity of violence provides an important potential limitation. While the alcoholism treatment literature provides some optimism that successful alcoholism treatment will lead to reductions in severe violence and overall violence, it is also the case it does not entirely eliminate the violence among all alcoholics, but only reduces it to the level observed in non-alcoholics. Among alcoholic men with histories of frequent, severe aggression, successful alcoholism treatment would probably not eliminate all IPV, but it might reduce the severity of violence. For these severely violent, alcoholic men, other interventions that are successful in reducing IPV would be of vital importance. While such interventions for men who have engaged in severe violence have not proven to be highly effective to date (Babcock, Green & Robie 2004), it will be important to utilize these tools when they become available and to evaluate their usefulness in the larger context of alcoholism treatments.