Research This assignment can be 3-5 pages sample is attached

Three primary conceptual models have been posited to explain the observed relationship between alcohol use and interpartner violence: (a) the spurious model, (b) the indirect effects model, and (c) the proximal effects model (Leonard & Quigley, 1999). In the spurious model, the relationship between alcohol use and partner violence is the result of these behaviors being related to other factors that influence both drinking and aggression. For example, individuals who have extensive drinking problems may also have certain personality characteristics, such as sociopathy, that are associated with interpersonal violence. Such individuals may en-gage in both drinking and partner violence frequently, and in turn, these behaviors appear directly linked when they are not.

Proponents of the indirect effects model argue that, over the course of a marriage or intimate relationship, unhappy and con- flictual dyadic and family environments evolve from long-term alcohol misuse by one or both partners. In turn, it is the lower relationship satisfaction and poor dyadic adjustment observed in dyads with partners who misuse alcohol, rather than the intoxicat- ing effects of alcohol per se, that are associated with partner violence.

According to the proximal effects model, individuals who con- sume alcohol are more likely to engage in partner violence because alcohol intoxication facilitates violence, which may be mediated through the psychopharmacologic effects of ethanol on cognitive processing (e.g., Chermack & Taylor, 1995) or the expectancies associated with intoxication (Critchlow, 1983). In contrast to the chronic effects of long-term alcohol misuse on relationship quality assumed in the indirect effects model, it is the acute effects of alcohol that are associated with increased partner violence in the proximal model, thus suggesting violence is more likely shortly after consumption of alcohol. Although spurious variable relation- ships and indirect pathways most likely explain, to some degree, the relationship between alcohol use and interpartner aggression, investigations that have controlled for these factors have nonethe- less continued to observe a significant relationship between alco- hol consumption and partner violence (e.g., Leonard & Senchak, 1996). This suggests other models might better explain the ob- served link between alcohol use and partner physical aggression.

Findings from four different lines of research often have been cited as support for the proximal effects model of alcohol intoxi- cation and violence: (a) laboratory studies, (b) marital interaction experiments, (c) reports by victims of partner physical aggression, and (d) event-based studies. However, much of this support has been weakened when attempts were made to generalize findings from these lines of research to explain the role of drinking in episodes of partner violence.

For example, laboratory studies have consistently found alcohol consumption by study participants leads to increased aggression, as measured by levels of administered electric shock (e.g., Bush- This project was supported, in part, by Grants R01DA12189 and R01DA14402 from the National Institute on Drug Abuse and by the Alpha Foundation and Old Dominion University.

I thank John Schafer, Department of Psychology, University of Cincin- nati; Christopher M. Murphy, Department of Psychology, University of Maryland at Baltimore County; Timothy J. O’Farrell, Department of Psy- chiatry, Harvard Medical School; and Gary R. Birchler, Department of Psychiatry, University of California at San Diego Medical School, for their helpful comments throughout this investigation. Additionally, I wish to thank Katherine Masyn, Graduate School of Education and Information Studies, University of California, Los Angeles, for her assistance with data analysis and Kenneth E. Leonard, Research Institute on Addictions, State University of New York at Buffalo, for his insightful suggestions on earlier versions of this article.

Correspondence concerning this article should be addressed to William Fals-Stewart, Research Institute on Addictions, State University of New York at Buffalo, 1021 Main Street, Buffalo, New York 14203. E-mail:

[email protected] Journal of Consulting and Clinical PsychologyCopyright 2003 by the American Psychological Association, Inc.

2003, Vol. 71, No. 1, 41–520022-006X/03/$12.00 DOI: 10.1037/0022-006X.71.1.41 41 men, 1997; Richardson, 1981). Although such findings provide strong support for the proximal model of alcohol use and aggres- sion, these studies share several characteristics that are markedly different from those observed in circumstances of aggression be- tween intimate partners and thus make the extent to which the results of these investigations generalize to partner violence less clear. More specifically, these investigations usually involve male- to-male behavior (rather than male-to-female), contrived and un- natural aggressive responses (i.e., electric shock), and a very narrow, preprogrammed set of responses from confederates (vs.

the natural and wide array of dynamic responses between intimate partners in conflict).

Marital interaction experiments, in which partners are asked to discuss major conflict areas, also have been conducted to explore the effects of alcohol on negative verbal behavior. For example, Leonard and Roberts (1998) asked intimate partners to talk about an important conflict and then to discuss their most serious area of conflict after the male partner received either (a) no alcohol, (b) an active placebo, or (c) an intoxicating dose of alcohol. Alcohol consumption led to increased negativity over baseline sessions, whereas those in the no-alcohol and the placebo conditions dem- onstrated no increased negativity. Although some investigators have not found this effect (e.g., Frankenstein, Hay, & Nathan, 1982), most studies of this type have tended to support the hy- pothesis that administration of alcohol in the context of marital conflict appears to increase negative interactions.

Many of the limitations of the laboratory aggression studies noted earlier are not shared by the marital interaction experiments.

In particular, the conflicts in the marital interaction studies occur between intimate partners who engage in a variety of responses they are likely to use in other dyadic conflict situations. However, some limitations of these studies, at least with respect to general- izing to episodes of physical aggression between partners, should be highlighted. Although negative verbal behavior is a precursor to escalating conflict, it is not, in fact, physical aggression. Addition- ally, partners are examined in laboratory settings discussing se- lected topics for a relatively brief time period, typically while being observed and videotaped, which is in contrast to naturally occurring interactions between partners over the course of ex- tended periods in their natural environments (e.g., homes, neighborhoods).

Reports from victims of partner violence frequently indicate acute intoxication accompanies the episodes of physical aggres- sion (e.g., Gayford, 1975; Nisonoff & Bitman, 1979). In these surveys, intimate partners are typically asked a general question about whether alcohol was a factor in the abuse. As noted by Leonard and Jacob (1988), interpretation of the role of alcohol and violence based on these reports is difficult because alcohol use and violence may co-occur by chance. For example, if a male partner drinks alcohol daily, any episode of violence will be linked to drinking, regardless of whether or not alcohol plays a causal role.

In addition, the reports of partners may also be influenced, to a certain degree, by a perception bias; that is, if reporters have a preconceived notion that alcohol consumption plays a role in partner violence, they may have a tendency to recall these behav- iors as linked.

In event-based studies, victims of partner violence are inter- viewed about specific violent episodes (or“events”) to determine the role of alcohol in these circumstances. Fagan, Stewart, andHansen (1983) interviewed 270 women who had been involved in domestic violence intervention programs and asked them about the event that resulted in their entry to treatment. Although alcohol use was common in these events, it was not related to the extent of injuries that occurred during these episodes. Pernanen (1991) interviewed a community-based sample about the most recent episode of domestic violence they experienced and found 13% were victimized when no alcohol was involved, whereas 26% of victims of partners who were intoxicated were injured. Leonard and Quigley (1999) interviewed 366 newlywed couples after their first year of marriage about their most severe verbal conflict and their first and most severe episode of physical aggression. They found husbands’drinking was more likely during episodes of physical aggression than during episodes of verbal aggression.

Although event-based studies provide support for the proximal effects model, certain limitations of these investigations should also be noted. Retrospective reports may be influenced by the perception bias discussed previously; moreover, interviewees may not recall clearly the circumstances surrounding the violence, particularly when events may have happened several months or even years earlier. These studies also tend to focus on single events or a very small number of specific episodes; thus, it is not possible from these data to discern the relationship of multiple drinking occasions and different partner violence episodes over the course of an extended period of time.

Although these different lines of research provide converging evidence in support of the proximal effects model of alcohol intoxication and aggression in the context of partner violence, investigations examining the temporal relationship between epi- sodes of alcohol consumption and occurrences of physical aggres- sion between partners, in their natural environments, over an extended time interval, are needed. This information would allow investigators to explore the co-occurrence of drinking and partner violence across multiple episodes, as well as the temporal rela- tionship of alcohol use and partner physical aggression within a given episode. Investigations in other areas (e.g., alcohol con- sumption and sexual activity) have effectively used daily diaries, which combine event-based assessment with panel design meth- odology, to make this type of measurement possible (e.g., Leigh, 1993). Although psychometrically established daily measures of alcohol use have been available for some time, a reliable and valid daily measure of interpartner violence has only recently been developed; thus, examination of the temporal relationship of these variables is now possible.

Thus, the purpose of the present study was to analyze the day-to-day relationship between alcohol consumption and partner physical aggression. Consistent with the proximal model, it was hypothesized that, after controlling for dyadic adjustment and alcohol use severity, the likelihood of any and severe male-to- female physical aggression between partners would be higher on days of male partners’drinking than on days when male partners did not consume alcohol. Furthermore, if the day-to-day relation- ship between alcohol use and partner violence was significant in these models, the relationship between the amount of drinking on a given day (i.e., categorized as heavy drinking, nonheavy drink- ing, and no drinking) and the occurrence of any and severe violence would also be explored. Additionally tested was the hypothesis that on days when drinking and partner physical ag- gression co-occur, violence would be more likely during male 42 FALS-STEWART partners’drinking or shortly after his drinking ended than it would be as the elapsed time between drinking and partner violence increased. Method Participants Participants were male and female intimate partners who reported at least one act of male-to-female physical aggression in their relationship during the previous year. Male clients entering one of two treatment programs, both of which were located in the northeastern United States, were recruited for participation. One sample consisted of men entering a 12-week domestic violence outpatient treatment program and their female partners. A separate sample was recruited from domestically violent male patients entering a 12-week outpatient alcoholism treatment program and their female partners. Several investigations (e.g., Gondolf & Foster, 1991; O’Farrell & Murphy, 1995) of alcoholic clinical populations have found a high incidence of male-to-female physical aggression among married or cohabiting patients (i.e., 50%–60% of alcoholic couples report at least one act of physical aggression between partners during the year before treat- ment entry). This alcoholism treatment-seeking sample consisted of dyads in which at least one partner reported one or more acts of male-to-female physical aggression during the year before treatment entry.

For the sample recruited from the domestic violence program, consec- utive male admissions (n 216) were approached to participate in the study. Of these, 22 (10%) men refused to be in the investigation. The intimate partners of 17 men (8%) who initially agreed to be in the study refused to participate when the investigation was described more fully; data from these couples were not included. Men in 16 couples (7%) who agreed to participate left treatment before any assessment data were collected.

Twenty-four couples (11%) did not provide sufficient data during the posttreatment period (e.g., only one partner provided data, partners refused to participate in interviews in one or more of the posttreatment interviews, only female participants provided data and were unable to provide specific information about the male partners’drinking) leaving data from 137 couples for analysis. The sociodemographic and background characteristics of this final sample are located in Table 1. Statistical comparisons of the sociodemographic characteristics of the couples who participated andthe 79 dyads who were not included, using analysis of variance (ANOVA) and binomial tests, revealed that the partners who did not participate in the investigation were significantly (i.e.,ps .05) younger (M 26.4 years, SD 6.4) and had shorter lengths of relationship (M 3.8 years, SD 3.6).

Of the 339 married or cohabiting men consecutively admitted to the alcoholism treatment program, at least one of the partners in 210 (62%) of these dyads reported at least one act of male-to-female aggression during the year before the pretreatment assessment and were asked to participate.

Of these, 21 (10%) men refused to be in the investigation. The intimate female partners of 14 men (8%) who initially agreed to be in the study refused to participate when the investigation was described more fully; data from these couples were not included. Men in 11 couples (5%) who agreed to participate left treatment before any assessment data were collected; an additional 29 (14%) did not provide sufficient data during the posttreat- ment period for analysis. The sociodemographic and background charac- teristics of the final sample (n 135) of couples recruited from the alcoholism treatment program are located in Table 1. Statistical compari- sons of the sociodemographic characteristics of the couples who partici- pated and the 75 dyads who were not included, using ANOVA and binomial tests, revealed that the partners who did not participate were significantly (ps .05) younger (M 33.4 years,SD 6.7) and had shorter lengths of relationship (M 4.4 years,SD 5.0). Materials and Measures Diaries.At admission to the treatment programs, male and female partners participated in an initial didactic training conducted by a research assistant, during which they were each given a 15-month supply of log books and were instructed how to fill out the daily log sheets. Each log book contained a 3-month supply of diary sheets to record daily episodes of both male partners’drinking and male-to-female physical aggression.

The daily drinking log included space for three different drinking episodes per day. For each day of drinking by the male partners, both partners were asked to record each time a drinking episode began and ended and the quantity of alcohol consumed. For female partners’logs of their male partners’drinking behavior, along with recording the specific information about the drinking behavior if known, respondents could also use the following codes: (a) drinking, amount unknown, (b) drinking, behavior unknown, and (c) drinking, time unknown.

In the daily physical aggression log, the participant recorded the time each episode began and the type of partner violence that occurred. Violence type was classified by the categories used in the Conflict Tactics Scale (CTS; Straus, 1979): (a) threw something at partner; (b) pushed, grabbed, or shoved; (c) slapped; (d) kicked, bit, or hit; (e) hit, or tried to hit, with something; (f) beat up; (g) threatened with a knife or gun; (h) used a knife or gun; or (i) used other types of physical aggression not otherwise specified. It is possible more than one type of violent behavior could occur as part of any circumscribed episode; all types of partner physical aggres- sion occurring on a given day and during a given episode were recorded in the log. As with the CTS, Items d through h are types of severe violence (Straus, 1990). On each day, the log also allowed respondents to note if they had any face-to-face contact. Partners were considered to have face- to-face contact even in circumstances in which the contact was brief (e.g., picking up children for visitation). Any face-to-face contact represented an opportunity for male-to-female physical aggression to occur between partners.

Research assistants contacted participants by telephone each month to remind and encourage them to complete the daily logs. In addition, par- ticipants were contacted 2 to 3 days before each face-to-face interview to remind them to bring the logs for use as part of the quarterly face-to-face interviews.

Calendar assessments.Partners were interviewed separately with two semistructured calendar-based interviews: the Timeline Followback Spou- sal Violence Interview (TLFB-SV; Fals-Stewart, Birchler, & Kelley, in Table 1 Baseline Characteristics of Participants From the Domestic Violence and Alcoholism Treatment-Seeking (Alcoholic) Samples CharacteristicSample Domestic violence Alcoholic Male partners meeting criteria for aDSM–IValcohol use disorder, No. (%) Alcohol abuse 21 (15) 10 (7) Alcohol dependence 31 (23) 125 (93) CTS subscale score,M(SD) Male-to-female Verbal Aggression 68.3 (18.1) 60.2 (15.4) Overall Violence 17.3 (13.8) 5.3 (3.9) Severe Violence 4.9 (4.0) 1.2 (2.4) Female-to-male Verbal Aggression 40.2 (35.6) 34.8 (33.6) Overall Violence 12.6 (14.3) 7.6 (14.2) Severe Violence 2.2 (3.7) 0.9 (1.3) Note. DSM–IV Diagnostic and Statistical Manual of Mental Disorders (4th ed.);CTS Conflict Tactics Scale. 43 PARTNER PHYSICAL AGGRESSION AND ALCOHOL USE press) to assess episodes of male-to-female physical aggression and the Timeline Followback Interview (TLFB; Sobell & Sobell, 1996) to assess the male partners’episodes of alcohol use. These interviews were con- ducted quarterly for 15 months (i.e., at the posttreatment period, which was 3 months after admission to treatment, and every 3 months thereafter for 1 year).

Both interviews use a daily calendar to assess days on which the target behaviors occurred in the time interval under study. Using their daily logs as memory aids, interviewers solicited from respondents and noted on the calendars (a) days on which male partners drank alcohol, (b) times of day alcohol was consumed (i.e., when each episode began and ended), (c) how much alcohol was consumed in each drinking episode, (d) days on which acts of male-to-female physical aggression occurred, (e) times of day the violent act occurred, (f) type of violent acts that occurred in each episode, and (g) days on which partners had no face-to-face contact.

As reported by Fals-Stewart et al. (in press), the test–retest reliability of the TLFB-SV was excellent, with scales derived from the TLFB-SV (e.g., Proportion of Days of Male-to-Female Physical Aggression, Proportion of Days of Female-to-Male Physical Aggression) having intraclass correla- tions ranging from .91 to 1.0 over a 2-week period. The TLFB-SV subscales also have moderate to high correlations with other measures of partner violence and general marital distress but comparatively low (and nonsignificant) correlations with a measure of positive impression man- agement. In the original psychometric analysis of the TLFB-SV, conducted on data from a different sample than the one used in the present investi- gation, day-to-day agreement between partners about occurrence of vio- lence was low at pretreatment but was high after the diaries were intro- duced (i.e., Yule’sYvalues, a measure of interrater agreements, ranged from .61 to .76, with values greater than .50 considered acceptable; Hoffman & Ninoneuvo, 1994). In the present study, day-to-day agreements between partners on the occurrence of any male-to-female physical aggres- sion and severe male-to-female physical aggression were acceptable, with mean (standard deviation) Yule’sYvalues of .71 (.13) and .79 (.14), respectively.

The TLFB is the most widely used calendar method to assess alcohol use frequency. Studies of the psychometric properties of the TLFB for the measurement of alcohol consumption have shown high temporal stability, with most test–retest correlations exceeding .85. Similar correlations be- tween self- and collateral-reports and between self-reports and official records of verifiable events such as hospitalizations and jail stays have been found when assessing the TLFB’s criterion validity (for a detailed review, see O’Farrell & Langenbucher, 1988). In a recent study, Fals- Stewart, O’Farrell, Freitas, McFarlin, and Rutigliano (2000) found day-to- day agreement between patients and collateral informants at different assessment intervals to be acceptable, with Yule’sYvalues ranging from .61 to .71. In the present study, day-to-day agreements between partners for any drinking and for heavy drinking (i.e., 6 oz [170 ml] or more of alcohol) by the male partners were acceptable, with mean (standard deviation) Yule’sYvalues of .74 (.18) and .79 (.14), respectively.

Relationship violence.The CTS was used to assess the frequency of male- and female-perpetrated verbal aggression, overall violence, and severe violence. Respondents rated each CTS item on a 7-point scale (never,once,twice,3–5 times,6–10 times,11–20 times,more than 20 times) for the frequency that they and their intimate partners engaged in the behavior in the past 12 months. Straus (1990) recommended scoring the frequency of violent acts using the middle of yearly frequency range for each CTS response category as follows:never 0;once 1;twice 2; 3–5 times 4;6–10 times 8;11–20 times 15;more than 20 times 25. A combined self- and partner-report was used. For example, if either the male partner or his intimate partner reported the male partner had been violent in the last year, he would be coded as violent. This approach is frequently used to address concerns about underreporting and leads to higher prevalence rates than self-reports alone.The CTS Verbal Aggression subscale has six items (e.g., yelled and/or insulted, threatened to hit or throw something at the partner). The CTS Violence subscale has eight items: (a) threw something at the partner; (b) pushed, grabbed, or shoved; (c) slapped; (d) kicked, bit or hit; (e) hit, or tried to hit, with something; (f) beat up; (g) threatened with a knife or gun; and (h) used a knife or gun. Frequency scores based on all eight items are labeledoverall violence; scores based on items d–h provide a measure of severe violence(Straus, 1979, 1990).

Substance use.Along with the TLFB, the extent of the male partners’ alcohol problems was evaluated with the Michigan Alcoholism Screening Test (MAST; Selzer, 1971). The MAST is a widely used 25-item instru- ment designed to measure the extent and severity of alcohol misuse with well-established reliability and validity. Scores can range from 0 to 56, with higher scores indicating increased problems with alcohol.

Male partners were interviewed with the alcohol abuse and dependence modules of the Structured Clinical Interview forDSM–IV(SCID; First, Spitzer, Gibbon, & Williams, 1995), administered by one of two master’s- level research assistants (both of whom were trained by William Fals- Stewart, who has extensive experience administering the SCID). Interrater reliability was assessed using a paired-rater design. Videotaped interviews of 20 patients entering a drug abuse treatment center were independently observed by both primary interviewers and by William Fals-Stewart.

Kappas between the two primary interviewers for the alcohol use disorders ranged from .90 to 1.0. Kappas between William Fals-Stewart and the primary interviewers also ranged from .90 to 1.0. These kappas reflect good-to-excellent observer agreement (Landis & Koch, 1977).

Relationship disharmony and conflict.Partners completed the Dyadic Adjustment Scale (DAS; Spanier, 1976), a widely used inventory of general relationship adjustment and satisfaction. Higher scores indicate higher levels of general relationship satisfaction. Traditionally, scores less than 98 indicate significant relationship dissatisfaction. Procedure Assessment.Partners were interviewed by master’s-level research as- sistants at pretreatment, posttreatment, and quarterly thereafter for 1 year.

At pretreatment, partners completed the DAS, MAST, CTS, and SCID; as noted, they were also trained in the use of the daily log books. At the scheduled end of treatment (i.e., 3 months after program admission) and quarterly thereafter for 12 months, the TLFB-SV and the TLFB were administered. Although all interviews were scheduled to be conducted in a face-to-face format, either at the treatment program location from which the participants were recruited or at the participants’homes, a small proportion of interviews were conducted by telephone (i.e., less than 10%) when participants were either away from the area or refused to participate in face-to-face interviews. Partners were paid for participation in the baseline, the posttreatment, and all follow-up interviews.

Partners reported high compliance with completing the log books. For nearly all scheduled interviews, partners came to the interviews with completed logs. At the end of the study, partners were asked if they completed the diaries daily as they had been instructed; 93% of the partners from both samples reported they completed the diaries daily or nearly every day, with many noting that the reminder telephone calls prompted them to continue filling out the provided log books.

Treatment.Men in the domestic violence treatment program partici- pated in facilitated 90-min support groups that met weekly for 12 weeks.

These groups were designed to provide a supportive milieu to allow members to explore feelings and learn nonviolent methods of coping with conflict and improving communication skills. According to treatment pro- viders, alcohol use was an ancillary topic area covered in the groups; patients identified with alcohol problems were referred to alcoholism treatment programs after completion of batterers’treatment.

Men in the alcoholism treatment program participated in weekly indi- vidual and group counseling over a 12-week period. The intervention followed a traditional 12-step facilitation model of alcoholism treatment, 44 FALS-STEWART while also encouraging participants to engage in self-help meetings, such as Alcoholics Anonymous. According to treatment providers in this pro- gram, partner violence was only discussed if the issue was raised in the context of group or individual treatment. Patients who reported that they engaged in interpartner violence were referred to a domestic violence treatment program after completion of alcoholism treatment.

It is important to emphasize that this study was not a treatment outcome study; men were recruited from treatment programs because such settings allowed for convenient access to and recruitment of men who had engaged in partner violence along with their intimate partners, both of whom could then be trained in the use of the drinking and partner violence logs. Use of the diaries was a critical component of the investigation, increasing the reliability of daily reports of both drinking and violence.

Primary Statistical Analytic Method To examine the daily conditional relationship between alcohol use and partner physical aggression, I conducted two-level hierarchical generalized linear model analyses (HGLM; Snijders & Bosker, 1999) using the HLM 5 program (Raudenbush, Bryk, Cheong, & Congdon, 2001). In these models, the occurrence of male-to-female physical aggression on a given day was the dependent measure and male partners’drinking on the same day was the primary independent measure of interest. The focus of the analysis was on the joint dependency of these two behaviors (i.e., on the extent to which the likelihood of an event or behavior is influenced by the occurrence of another behavior). The change in the odds of male-to-female physical aggression occurring on days of face-to-face contact between partners when the male partner drank alcohol before the violent episode compared with days of no drinking or drinking only after the occurrence of violent episodes was examined. 1 Because the dependent measure can take values of either 0 (i.e., no partner violence) or 1 (i.e., occurrence of partner violence), the HGLMs used a Bernoulli sampling distribution and a logit link function. Drinking was treated as a Level 1 time-dependent covariate. Partner violence and drinking were nested within participants (Level 2). An unrestricted covari- ance structure was used and a Laplace approximation was implemented for parameter estimation (Raudenbush, Yang, & Yosef, 2000).

In these models, to control for levels of dyadic adjustment and alcohol use severity, I entered couple DAS scores and male partners’MAST scores in the models as Level-2 predictors. In addition to serving as control variables, their inclusion in the models allowed for the examination of these variables’main effects (i.e., the effect of MAST and DAS scores on the general likelihood of partner violence on a given day, regardless of alcohol consumption) and their cross-level interactions with day-to-day drinking (i.e., the effect of MAST and DAS scores on the conditional daily relationship between drinking and partner violence).

Consistent with the proximal model, partner physical aggression and alcohol use by the male partners were counted as linked on a given day only if the first episode of drinking preceded the first act of male-to-female violence. For the present study, if multiple types of violence occurred on a given day, the type of violence occurring on the day was considered “severe”if any of the violent acts were severe. If the occurrence of drinking was found to be significantly related to the occurrence of violence in the multilevel models, follow-up simple contrasts were conducted to examine the relationship between different amounts of drinking on a given day (i.e., heavy and nonheavy drinking) and violence.

It is also important to recognize that the conditional probability observed between two behaviors is a function not only of the joint dependencies between the behaviors but also of the serial dependency or autocorrelation of each behavior (Dumas, 1986; Gottman & Ringland, 1981). Conceptu- ally, the problem is that the occurrence of a given behavior of interest is often best predicted by the previous occurrence of that behavior versus some other behavior. More specific to the present study, the occurrence of previous episodes of violence may be a better predictor of subsequentepisodes of violence than the occurrence of male partners’alcohol con- sumption. Therefore, the extent to which a behavior of interest (i.e., male-to-female physical aggression) is dependent on the occurrence of another behavior (i.e., male partner’s drinking) can only be established by first controlling for the previous occurrence of the behavior of interest (i.e., an episode of male-to-female aggression on the previous day). 2Therefore, it is important to partial out this autodependence before concluding that another variable (in this investigation, drinking) is related to the occurrence of a target behavior (in the present study, male-to-female physical aggres- sion). Thus, the occurrence of partner violence was lagged 1 day and was included as a control Level-1 variable in the models, thereby controlling for autocorrelation by controlling for the occurrence of violence on the pre- vious day (Allison & Liker, 1982).

Although obtained from both partners, data provided by the female partners regarding episodes of male-to-female physical aggression and their reports of face-to-face contact were used in the analyses reported.

Female reports of male-to-female physical aggression were used because some studies suggest, perhaps due to issues related to social desirability, that men report fewer episodes of male-to-female physical aggression than their female counterparts (e.g., Archer, 1999). However, other investiga- tions have not found this pattern, although most have tended to find that female partners report higher levels of severe aggression than their male counterparts (e.g., Heyman & Schlee, 1997). Violence reports from male partners, which were very similar to the reports of female partners, were also analyzed and the findings are briefly discussed.

Additionally, drinking information provided by the male partners was used in the analytic models. Although men’s reports of their own drinking could also be influenced by positive impression management, men were actually the only individuals in the dyads who had full knowledge of when they drank and how much they consumed, particularly given much of the male partners’drinking could have occurred in settings where the female partner is not present. An important advantage of using different partners’ reports of violence and drinking, respectively, is that the effect of the perception bias held by any single member of the dyad that violence and drinking are related is largely reduced or eliminated if the reports of the behaviors of interest are given by different respondents. Results Male Partners’Drinking and Male-to-Female Physical Aggression During the Assessment Period For the couples from the domestic violence sample, the mean (standard deviation) number of days of face-to-face contact during the 15-month assessment period was 264.2 (206.6). Among these couples, 49 (36%) of the female partners reported at least one act 1For the two samples, there were nine episodes (i.e., six for the alco- holism treatment-seeking sample and three for the domestic violence sample) in which male partners began drinking, perpetrated an act of male-to-female physical aggression, and then continued drinking after the violent event. Before the event, the amount of alcohol consumption would have been considered nonheavy drinking, but for the day (i.e., both before and after the violent episode), the amount of drinking would have been considered heavy. In the analyses presented, these were counted as non- heavy drinking days; however, the data were also analyzed with these days considered as heavy drinking days, and the results did not change substantively.

2In theory, autocorrelation can exist between any two error terms (i.e., at adjacent and nonadjacent time periods in a repeated measures design). In practice, a frequently made assumption is that only immediately adjacent errors are correlated to a significant degree; this is referred to as first-order autocorrelation and was what was controlled in this study. 45 PARTNER PHYSICAL AGGRESSION AND ALCOHOL USE of male-to-female physical aggression and 82 (60%) of the male partners reported drinking alcohol on one or more of the days of face-to-face contact during the 15-month assessment period.

Pooled across couples, the number of days of male-to-female physical aggression between partners was 562; of these, 165 (29%) were days of severe violence. On 416 (74%) days of male-to- female physical aggression, the male partner had consumed alco- hol before the first episode; on 134 (81%) days of severe violence, the male partner drank before the episode.

For the alcoholism treatment-seeking sample, the mean (stan- dard deviation) number of days of face-to-face contact was 305.5 (264.6). Forty-one (31%) of the female partners in these couples reported at least one act of male-to-female physical aggression and 57 (42%) of the male partners reported drinking on one or more of the days of face-to-face contact. Pooled across couples, the number of days of male-to-female physical aggression was 432; of these, 135 (31%) were episodes of severe aggression. On 276 (64%) days of any physical aggression and on 109 (81%) days of severe violence, the male partner drank alcohol before the violent episode.

The Day-to-Day Relationship Between Alcohol Use and the Occurrence of Domestic Violence Domestic violence sample.The results of the HGLM examin- ing the day-to-day relationship between male partners’alcohol consumption and the occurrence of male-to-female physical ag- gression for the domestic violence sample are located in Table 2.

After controlling for autodependence, couple DAS scores, and MAST scores, the odds of any type of male-to-female aggression occurring were more than 8 times higher on days of any alcohol consumption by male partners than on days when male partnersdid not drink. Furthermore, the odds of severe male-to-female physical aggression were more than 11 times higher on days of male partners’drinking than on days of no drinking.

In addition, for both models, there were also main effects for scores on the DAS and the MAST, indicating that couples with lower satisfaction (i.e., lower DAS scores) and higher alcohol use severity (i.e., higher MAST scores) were more likely to engage in male-to-female physical aggression on any given day, regardless of daily alcohol consumption. However, the cross-level interac- tions between DAS and MAST scores and any drinking were not significant, indicating neither dyadic adjustment nor alcohol use severity had a significant effect on the joint relationship between drinking and male-to-female physical aggression.

Given the conditional probabilities of any and severe male-to- female physical aggression were significantly higher on male partners’days of alcohol use than on days of no drinking, follow-up analyses were conducted to examine the likelihood of any and severe violence on days of nonheavy and heavy drinking versus days of no drinking. As shown in Table 3, although the odds of any and severe male-to-female physical aggression were more than three times higher on days of nonheavy drinking, heavy drinking by male partners appears to be very strongly related to the likelihood of male-to-female physical aggression. In contrast to days of no drinking by male partners, on a day of heavy drinking by the male partner, the odds of an occurrence of any male-to- female physical aggression were over 18 times higher and the odds of severe aggression were more than 19 times higher. Main effects for DAS and MAST scores were significant in both of these models; however, none of the cross-level interactions were significant.

It should also be noted, although they are not presented, all other possible interaction terms were tested in all models (e.g., DAS Table 2 Parameter Estimates for Two-Level Hierarchical Generalized Linear Models Examining the Day-to-Day Relationship of Drinking and the Occurrence of Any and of Severe Male-to-Female Physical Aggression Among Couples in the Domestic Violence Sample Fixed effectsBSE z pOR Any physical aggression Intercept 5.62 0.09 58.44 .01 0.01 1-day lag any violence 0.53 0.21 2.50 .05 0.59 Any drinking 2.19 0.06 34.22 .01 8.94 DAS 0.04 0.02 2.31 .05 0.96 MAST 0.18 0.09 2.05 .05 1.19 DAS Any Drinking 0.02 0.03 0.55ns0.98 MAST Any Drinking 0.07 0.11 0.64ns1.06 Severe physical aggression Intercept 6.74 0.06 112.33 .01 0.00 1-day lag severe violence 0.65 0.15 4.30 .01 0.52 Any drinking 2.44 0.05 46.92 .01 11.47 DAS 0.06 0.02 3.76 .01 0.94 MAST 0.16 0.05 2.84 .01 1.17 DAS Any Drinking 0.01 0.04 0.30ns0.99 MAST Any Drinking 0.06 0.12 0.49ns1.06 Note.The 1-day lag any violence and 1-day lag severe violence were variables entered into the models to control first-order autodependence of the episodes of any and severe male-to-female physical aggression, respectively. OR odds ratio; DAS Couple Dyadic Adjustment Scale score; MAST Michigan Alcoholism Screening Test. 46 FALS-STEWART MAST, DAS MAST Any Drinking). Although they are not reported, none were found to be significant (i.e., allps .05).

Alcoholism treatment-seeking sample.As shown in Tables 4 and 5, a very similar pattern of results was obtained when dataprovided by the partners from the alcoholism treatment-seeking sample were analyzed. Compared with days of no drinking by the male partner, on days of male partners’drinking, the odds of the occurrence of any male-to-female physical aggression and severe Table 3 Parameter Estimates for Hierarchical Generalized Linear Models Examining the Day-to-Day Relationship of Nonheavy and Heavy Drinking and the Occurrence of Any and Severe Male-to- Female Physical Aggression Among Couples in the Domestic Violence Sample Fixed effectsBSE z pOR Any physical aggression Intercept 5.25 0.09 58.34 .01 0.01 1-day lag any violence 0.52 0.22 2.38 .05 0.59 Nonheavy drinking 1.34 0.15 8.92 .01 3.83 Heavy drinking 2.92 0.05 54.07 .01 18.54 DAS 0.07 0.03 2.31 .05 0.93 MAST 0.16 0.08 2.18 .05 1.17 Nonheavy Drinking DAS 0.07 0.05 1.39ns0.93 Nonheavy Drinking MAST 0.15 0.12 1.24ns1.24 Heavy Drinking DAS 0.06 0.06 1.03ns0.94 Heavy Drinking MAST 0.14 0.12 1.16ns1.15 Severe physical aggression Intercept 6.72 0.06 112.02 .01 0.00 1-day lag severe violence 0.64 0.14 4.67 .01 0.53 Nonheavy drinking 1.24 0.09 13.80 .01 3.46 Heavy drinking 2.99 0.06 42.97 .01 19.89 DAS 0.04 0.02 2.22 .05 0.96 MAST 0.16 0.07 2.29 .05 1.18 Nonheavy Drinking DAS 0.03 0.02 1.32ns0.97 Nonheavy Drinking MAST 0.16 0.12 1.31ns1.17 Heavy Drinking DAS 0.04 0.03 1.21ns0.96 Heavy Drinking MAST 0.15 0.11 1.51ns1.16 Note.The 1-day lag any violence and 1-day lag severe violence were variables entered into the models to control first-order autodependence of the episodes of any and severe male-to-female physical aggression, respectively. OR odds ratio; DAS Couple Dyadic Adjustment Scale score; MAST Michigan Alcoholism Screening Test. Table 4 Parameter Estimates for Two-Level Hierarchical Generalized Linear Models Examining the Day-to-Day Relationship of Drinking and the Occurrence of Any and of Severe Male-to-Female Physical Aggression Among Couples in the Alcoholism Treatment-Seeking Sample Fixed effectsBSEz pOR Any physical aggression Intercept 4.79 0.11 43.55 .001 0.01 1-day lag any violence 0.48 0.18 2.74 .05 0.62 Any drinking 2.41 0.07 35.97 .001 11.13 DAS 0.05 0.01 4.61 .001 0.95 MAST 0.10 0.08 1.21ns1.10 DAS Any Drinking 0.05 0.03 1.41ns0.96 MAST Any Drinking 0.09 0.09 0.98ns1.09 Severe physical aggression Intercept 7.05 0.18 39.17 .001 0.00 1-day lag severe violence 0.74 0.12 6.17 .001 0.48 Any drinking 2.40 0.08 28.93 .001 11.02 DAS 0.07 0.02 2.70 .05 0.92 MAST 0.09 0.07 1.22ns1.07 DAS Any Drinking 0.08 0.06 1.23ns0.92 MAST Any Drinking 0.08 0.08 1.01ns1.07 Note.The 1-day lag any violence and 1-day lag severe violence were variables entered into the models to control first-order autodependence of the episodes of any and severe male-to-female physical aggression, respectively. OR odds ratio; DAS Couple Dyadic Adjustment Scale score; MAST Michigan Alcoholism Screening Test. 47 PARTNER PHYSICAL AGGRESSION AND ALCOHOL USE male-to-female physical aggression were both roughly 11 times higher. As with the domestic violence sample, heavy drinking by the male partner had a strong relationship to the likelihood of male-to-female physical aggression. More specifically, compared with days of no drinking, on days of heavy drinking, the odds of any physical aggression and severe physical aggression were nearly 17 times and more than 19 times higher, respectively.

One difference between the model results with the domestic violence sample and the alcoholism treatment-seeking sample was that, for the alcoholism treatment-seeking sample, only the main effect for DAS scores, not MAST scores, was significant. Thus, for the alcoholism treatment-seeking sample, lower DAS scores (i.e., indicating reduced relationship satisfaction) were related to a re- duced likelihood of male-to-female physical aggression on any given day, regardless of drinking. However, as with the domestic violence sample, none of the cross-level interactions were signif- icant. Although they are not presented, all other possible interac- tions were tested in all models; none were significant.

Results of Analyses Using Other Report Combinations As noted earlier, data from female partners’reports of male-to- female physical aggression and male partners’reports of his own drinking are reported. However, because both partners provided data on the two behaviors of interest, the following different combinations of data sources were also used in the models: (a) male partners’reports of both behaviors, (b) female partners’ reports of both behaviors, and (c) combined highest report (i.e., if either partner reported a day of male-to-female violence or malepartners’drinking, the day was counted as a violence day or a drinking day, respectively). The analyses from data generated using these different approaches resulted in findings very similar to those reported here. Using male partners’reports of both the occurrence of male-to-female physical aggression and drinking resulted in odds ratios of partner violence on days of drinking that were nearly identical to those obtained when using female part- ners’reports of violence (i.e., all were within 3% of those presented). Analysis of female partners’reports of drinking and violence and combined reports of both behaviors tended to provide slightly higher odds ratios (5%–8% higher) than revealed in anal- yses presented, but these differences were not substantial. 3 An Examination of the Trend in the Temporal Relationship Between Drinking and Violence If it is assumed alcohol intoxication serves to facilitate acts of violence, it would follow that, on days when drinking and violence co-occur, acts of violence are more likely to occur close in time to episodes of drinking than to occur when drinking and violence are temporally distant. To explore this hypothesis, I isolated all days on which both male partners’drinking and male-to-female phys- ical aggression occurred for both samples. On each of these days, the time of day of the first episode of violence was identified from the female partners’reports. From the male partners’reports, all 3The complete results of these analyses are available from William Fals-Stewart on request. Table 5 Parameter Estimates for Hierarchical Generalized Linear Models Examining the Day-to-Day Relationship of Nonheavy and Heavy Drinking and the Occurrence of Any and of Severe Male- to-Female Physical Aggression Among Couples in the Alcoholism Treatment-Seeking Sample Fixed effectsBSEz pOR Any physical aggression Intercept 4.77 0.11 45.43 .001 0.01 1-day lag any violence 0.48 0.17 2.81 .05 0.62 Nonheavy drinking 1.29 0.09 14.34 .001 3.63 Heavy drinking 2.84 0.08 35.49 .001 17.12 DAS 0.04 0.02 2.44 .05 0.96 MAST 0.07 0.07 1.02ns1.07 Nonheavy Drinking DAS 0.03 0.03 0.98ns0.97 Nonheavy Drinking MAST 0.06 0.07 0.88ns1.06 Heavy Drinking DAS 0.05 0.04 1.28ns0.95 Heavy Drinking MAST 0.08 0.08 0.95ns1.08 Severe physical aggression Intercept 7.03 0.16 43.94 .001 0.00 1-day lag severe violence 0.65 0.14 4.51 .001 0.52 Nonheavy drinking 1.42 0.08 17.75 .001 4.14 Heavy drinking 2.96 0.09 32.89 .001 19.30 DAS 0.06 0.02 2.51 .05 0.94 MAST 0.08 0.06 1.31ns1.08 Nonheavy Drinking DAS 0.05 0.06 0.80ns0.95 Nonheavy Drinking MAST 0.08 0.09 0.84ns1.08 Heavy Drinking DAS 0.02 0.02 1.04ns0.98 Heavy Drinking MAST 0.08 0.10 0.79ns1.09 Note.The 1-day lag any violence and 1-day lag severe violence were variables entered into the models to control first-order autodependence of the episodes of any and severe male-to-female physical aggression, respectively. OR odds ratio; DAS Couple Dyadic Adjustment Scale score; MAST Michigan Alcoholism Screening Test. 48 FALS-STEWART drinking episodes on that day occurring before the episode of violence were marked; of these, the drinking episode occurring closest in time to the act of violence was identified. The drinking episode was considered to be closest if the male-to-female aggres- sion was bracketed in time by the start and completion of a drinking episode. In instances when the violence did not occur during a drinking period, drinking ending nearest in time to the act of violence was considered closest. The amount of time between the end of the closest episode of male partners’drinking and the act of partner violence was calculated; when active drinking was occurring when the violent episode took place, the amount of time between drinking and violence was assumed to be zero. The time between drinking and violence was transformed into an ordered multinominal time variable: (a) 0–1 hr 59 min, (b) 2–3 hr 59 min, (c) 4–5 hr 59 min, (d) 6–7 hr 59 min, and (e) 8 hr or more. 4 The relationships between the ordered multinominal time vari- able and both any and severe partner violence were examined. The percentage of acts of any male-to-female physical aggression hap- pening in each time period for the domestic violence and alcohol- ism treatment-seeking sample is shown in Figure 1A. To test the trend of the relationship between likelihood of any male-to-female aggression and time of drinking, I performed an HGLM on these data. Occurrence of violence in each time frame was the binary dependent variable; the multinominal time variable was a Level-1 predictor. Participants who engaged in violence represented Level 2. Conceptually, the occurrence of male-to-female physical ag- gression during the different time intervals (i.e., represented by the multinominal time variable) was nested within participants.

Linear, quadratic, and cubic trends were sequentially tested until the first nonsignificant result was found; the highest order orthog- onal polynomial that was found to be significant was considered to reflect the functional relationship between the variables. Consis- tent with visual presentation in Figure 1A, the highest order significant polynomial relationship between any male-to-female physical aggression and male partners’drinking was the quadratic trend, for both the domestic violence sample,B 1.80,SE 0.24,t(2078) 7.39,p .01, and the alcoholism treatment- seeking sample,B 1.67,SE 0.22,t(1378) 7.46,p .01.

The percentage of acts of severe male-to-female physical ag- gression occurring in each time period for the domestic violence and alcoholism treatment-seeking samples is shown in Figure 1B.

As with the relationship between drinking and any violence, the highest order significant polynomial relationship between severe male-to-female physical aggression and male partners’drinking was the quadratic trend, for both the domestic violence sample,B 2.02,SE 0.23,t(668) 46.21,p .01, and the alcoholism treatment-seeking sample,B 1.84,SE 0.24,t(543) 7.68, p .01. 5 Discussion The purpose of the present investigation was to examine the day-to-day relationship between male partners’drinking and the occurrence of male-to-female physical aggression. Consistent with the proximal model, it was hypothesized male-to-female physical aggression would be more likely on days of drinking by male partners than on days of no drinking or days when drinking occurred only after the violent episode. Across two separate sam- ples of domestically violent men, significant and comparatively similar conditional daily relationships were found between male partners’drinking and the occurrence of male-to-female physical aggression. Perhaps more important, the likelihood of male-to- 4This variable was transformed from an interval-level continuous vari- able to an ordered multinominal variable because of concerns about the precision of the time estimates of drinking provided by male partners.

However, the results using time as a continuous measure in the models were similar to those reported here, with a quadratic trend fitting the models best. The analyses of the time variable used as a continuous measure are available from William Fals-Stewart on request.

5The complete results of these analyses are available from William Fals-Stewart on request. Figure 1.The percentage of episodes of any male-to-female physical aggression (Panel A) and severe male-to-female physical aggression (Panel B) occurring at different time periods after male partner’s drinking ended for the domestic violence and alcoholism treatment-seeking (alcohol) samples. 49 PARTNER PHYSICAL AGGRESSION AND ALCOHOL USE female physical aggression was substantially higher on days of drinking by male partners compared with days of no drinking, even after controlling for levels of relationship disharmony and alcohol severity. For men entering the domestic violence treatment pro- gram (alcoholism treatment program odds are given in parenthe- ses), the odds of any male-to-female physical aggression on a given day were more than 8 times (11 times) higher on days when men drank alcohol than on days of no drinking. Additionally, the odds of severe male-to-female physical aggression were more than 11 times (11 times) higher on days of men’s drinking than on days of no drinking. Moreover, compared with days of no drink- ing, on days of heavy drinking by the male partners the odds of any male-to-female violence were more than 18 times (17 times) higher and the odds of severe violence were more than 19 times (19 times) higher.

On days when male-to-female physical aggression and drinking co-occur, a negatively decelerating relationship between the amount of time between the cessation of drinking and the occur- rence of male-to-female physical aggression (any and severe) was observed in both samples. As hypothesized, these findings indi- cated that violence was more likely during male partners’drinking or shortly after drinking episodes ceased compared with the like- lihood of violence occurring at a temporally distant time after drinking ceased.

The results of the present investigation lend further support to the proximal effects model of aggression and alcohol use as applied to partner violence. Examined across multiple couples and, in many instances, more than one violence episode within a given dyad over an extended period, use of alcohol by the male partner is a significant risk factor for the occurrence of physical aggression among couples with a history of interpartner violence. These effects remained not only significant, but also large, even after controlling for relationship satisfaction and drinking severity. Ef- fort was made to ensure the findings were not artifacts of system- atic reporter or measurement bias. Reports of drinking and reports of violence were provided by different partners, which served to reduce the influence of perception bias of either partner. Yet, even when reports of drinking and male-to-female physical aggression were taken from a single partner or when a combined report was used, the results were nonetheless very similar to those reported here.

The main effect of DAS scores in the models also provides some support for the indirect effects model, suggesting that lower rela- tionship satisfaction was associated with increased likelihood of male-to-female physical aggression. However, relationship satis- faction did not significantly influence the conditional daily rela- tionship between drinking and partner violence. The robust find- ings indicating partner violence was more likely on a day of drinking than on a day of no alcohol consumption, and was most likely close in time to the drinking episode, would not be predicted by the indirect effects model. Moreover, because it was controlled in the data analytic models, general levels of dyadic adjustment did not account for the significant conditional daily relationship be- tween drinking and partner violence obtained in the present study.

Alcohol use severity was associated with increased likelihood of male-to-female physical aggression in the domestic violence sam- ple only, although it did not significantly influence the conditional daily relationship between alcohol use and partner violence. How- ever, the main effect of alcohol use severity in the models indicatesindividuals seeking treatment for domestic violence who have more severe problems with alcohol misuse are generally more likely to engage in partner violence on any given day, regardless of drinking, than their counterparts who do not have problems with drinking. It is unclear from the data collected if elevated alcohol use severity was associated with another factor, such as sociopa- thy, that may, in turn, be related to a propensity for partner violence (as would be suggested by the spurious model). Interest- ingly, alcohol severity was not significantly related to the likeli- hood of partner violence in the alcoholism treatment-seeking sam- ple. Because all of these patients, by definition, had problems with alcohol misuse, there may not have been enough variability in alcohol use severity (i.e., a ceiling effect) for alcohol use severity to account for a significant amount of variance in the likelihood of partner violence in these models.

It is clear from these results that, for couples in which male partners have a fairly recent history of perpetrating partner vio- lence, drinking (and particularly heavy drinking) by male partners represents a highly significant risk factor for recurrence of physical aggression. For dyads in which partners abuse alcohol or other drugs, partners in these relationships commonly report they often engage in arguments and other conflict resolution discussions when male partners are intoxicated. A frequent admonition to female intimate partners of alcoholic and drug-abusing men in treatment is not to engage in highly conflictual discussions or arguments when their spouses are drunk or otherwise intoxicated (e.g., O’Farrell & Fals-Stewart, 2000). Although the present study did not test whether this advice would reduce violence, the find- ings suggest that male partners who have a history of violence may be substantially more dangerous when they have been drinking, and thus avoiding conflict during those times would be prudent advice and may, in fact, reduce the overall risk of violence in these dyads.

This investigation has a number of strengths that should be highlighted. This is the first study to obtain prospective, daily reports of both drinking and partner violence over an extended period of time, thus allowing for a comprehensive analysis of these behaviors’day-to-day conditional relationship. Moreover, analy- ses of data from two separate, and relatively large, samples of domestically violent men revealed very similar joint relationships between male partners’drinking and occurrence of male-to-female physical aggression. Partners were interviewed with psychometri- cally sound measures of both daily alcohol consumption and daily episodes of partner violence, the latter of which (i.e., the TLFB- SV) has only recently been available. Although the results were based on male partners’reports of drinking and female partners’ reports of violence, findings using other report combinations (e.g., male partners’reports of drinking and violence, female partners’ reports of drinking and violence) were consistent with and similar to the findings provided here.

This investigation also had several important limitations. The study examined two circumscribed sets of behaviors: male part- ners’alcohol consumption and male-to-female physical aggres- sion. Although both of these behaviors are of critical importance in understanding the role of alcohol in the occurrence of domestic violence, the investigation did not explore other aspects of these behaviors. More specifically, partner violence is best viewed as a multidimensional phenomenon, marked by any number of differ- ent behaviors, which broadly includes such behaviors as female- 50 FALS-STEWART to-male physical violence, sexual aggression, and verbal and emo- tional abuse by either partner. Whereas the present study focused on an important type of partner violence (i.e., male-to-female physical aggression), it did so to the exclusion of other behaviors that would be considered under the rubric of partner violence.

Moreover, only male partners’drinking was considered; informa- tion on female partners’drinking was not examined, nor was information collected from partners on use of other psychoactive substances. All of these behaviors could have influenced the like- lihood of domestic violence in these dyads.

To a certain degree, the behaviors of interest in the study were chosen not only because of their import in earlier partner violence studies but also because of the practical realities of the investiga- tion that only allowed for an examination of a fairly small number of behaviors. The project was labor-intensive, requiring substantial time and effort by the data collection teams, data entry personnel, and by the participants themselves. Additionally, even though partners generally reported they were completing the diaries reg- ularly and as instructed and almost always brought them to the quarterly interviews, there is no guarantee that these diaries were completed daily (e.g., a partner could have filled out the logs on the day of the scheduled interview but told the interviewer that he or she was completing them daily throughout the given assessment interval).

To address some of these practical issues, the availability of the comparatively new interactive voice response (IVR) technologies could effectively be applied to the interview and data collection methods used in the present investigation. For example, some investigators are using telephone-assisted data collection for such studies, whereby participants call a specific toll-free telephone number and, within a few minutes, complete a structured interview about the previous day’s behaviors, responding to forced-choice queries by using the telephone keypad. If individuals fail to call over a predefined time period, the investigators are automatically alerted and can then contact participants. This approach has been used in the study of alcohol consumption among male drinkers and has been shown to produce reliable and valid self-reports (e.g., Searles, Perrine, Mundt, & Helzer, 1995); its use has also been extended to assess binge eating (Bardone, Krahn, Goodman, & Searles, 2000). IVR is highly flexible and could also be pro- grammed to query respondents about male-to-female physical ag- gression and other forms of partner violence once the reliability and validity of this assessment approach were established.

Although data were collected from a relatively large sample, it is important to recognize that participants were domestically vio- lent men entering outpatient treatment either for battering or alco- holism, drawn from two clinical programs located in one geo- graphic location in the United States; how these results would generalize to other populations of domestically violent men from other types of treatment programs or other geographic regions remains unclear. Relatedly, participation in intervention programs for problem drinking and violent behavior could have influenced the conditional probabilities of these behaviors over time. Thus, it is unclear if the findings reported would generalize to couples who are not involved in treatment.

Data were collected regarding when the episodes of drinking and violence occurred, but the role of intoxication is not entirely clear. Although both partners noted anecdotally that, on days of violence and drinking, male partners were most often intoxicatedwhen they perpetrated the violent acts, it is simply not possible to obtain even a rough index of the level of intoxication, such as blood alcohol levels, on the basis of the data collected. Information on the timing of the violent episodes; the timing, duration, and amount of male partners’alcohol consumption; and other variables (e.g., male partners’weight) would have to be collected with far more precision than was done in the present study to examine the role of intoxication level and the occurrence of violence.

Despite these limitations, the results of the present study lend support to the proximal model of alcohol use and aggression as applied to domestic violence. In fact, all of the different types of studies that have produced results that tend to support the proximal model have certain limitations when applied to interpartner vio- lence. However, the results from these respective types of inves- tigations, including the present study, appear to converge on the notion that alcohol use plays some role in the facilitation of aggression in the context of certain marital relationships. In most instances, alcohol consumption, even heavy drinking, does not lead to domestic violence. Thus, alcohol use and intoxication are perhaps best viewed as only one of several dynamically interacting factors that help to create the context in which partner violence results. References Allison, P. D., & Liker, J. K. (1982). Analyzing sequential categorical data on dyadic interaction: A comment on Gottman.Psychological Bulle- tin, 91,393–403.

Archer, J. (1999). Assessment of the reliability of the Conflict Tactics Scales: A meta-analytic review.Journal of Interpersonal Violence, 14, 1263–1289.

Bardone, A. M., Krahn, D. D., Goodman, B. M., & Searles, J. S. (2000).

Using interactive voice response technology and timeline follow-back methodology in studying binge eating and drinking behavior: Different answers to different forms of the same question?Addictive Behav- iors, 25,1–11.

Bushman, B. J. (1997). Effects of alcohol on human aggression: Validity of proposed explanations. In M. Galanter (Ed.),Alcoholism(pp. 227–242).

New York: Plenum Press.

Chermack, S. T., & Taylor, S. P. (1995). Alcohol and human physical aggression: Pharmacological versus expectancy effects.Journal of Stud- ies on Alcohol, 56,449–456.

Critchlow, B. (1983). Blaming the booze: The attribution of responsibility for drunken behavior.Personality and Social Psychology Bulletin, 9, 451–473.

Dumas, J. E. (1986). Controlling for autocorrelation in social interaction analysis.Psychological Bulletin, 100,125–127.

Fagan, J. A., Stewart, D. K., & Hansen, K. V. (1983). Violent men or violent husbands: Background factors and situational correlates. In D.

Finkelhor, R. J. Gelles, G. T. Hotaling, & M. A. Straus (Eds.),The dark side of families: Current family violence research(pp. 49–67). Thou- sand Oaks, CA: Sage.

Fals-Stewart, W., Birchler, G. R., & Kelley, M. (in press). The Timeline Followback Spousal Violence Interview to assess physical aggression between partners: Reliability and validity.Journal of Family Violence.

Fals-Stewart, W., O’Farrell, T. J., Freitas, T. T., McFarlin, S. K., & Rutigliano, P. (2000). The Timeline Followback reports of psychoactive substance use by drug-abusing patients: Psychometric properties.Jour- nal of Consulting and Clinical Psychology, 68,134–144.

First, M., Spitzer, L., Gibbon, M., & Williams, J. (1995).Structural Clinical Interview for Axis I DSM–IV disorders (SCID).Washington, DC: American Psychiatric Association. 51 PARTNER PHYSICAL AGGRESSION AND ALCOHOL USE Frankenstein, W., Hay, W. M., & Nathan, P. E. (1982). Effects of intox- ication on alcoholics’marital communication and problem solving.

Journal of Studies on Alcohol, 46,1–6.

Gayford, J. J. (1975). Wife battering: A preliminary survey of 100 cases.

British Medical Journal, 1,194–197.

Gondolf, E. W., & Foster, R. A. (1991). Wife assault among VA alcohol rehabilitation patients.Hospital and Community Psychiatry, 42,74–79.

Gottman, J. M., & Ringland, J. T. (1981). The analysis of dominance and bidirectionality in social development.Child Development, 52,393– 412.

Heyman, R. E., & Schlee, K. A. (1997). Toward a better estimate of the prevalence of partner abuse: Adjusting rates based on the sensitivity of the Conflict Tactics Scale.Journal of Family Psychology, 11,331–338.

Hoffman, N. G., & Ninonuevo, F. G. (1994). Concurrent validation of substance abusers self-reports against collateral information: Percentage agreement vs. vs. Yule’sY.Alcoholism: Clinical and Experimental Research, 18,231–237.

Landis, J. R., & Koch, G. G. (1977). The measurement of observer agreement for categorical data.Biometrics, 33,159–174.

Leigh, B. C. (1993). Alcohol consumption and sexual activity as reported with a diary technique.Journal of Abnormal Psychology, 102,490–493.

Leonard, K. E., & Jacob, T. (1988). Alcohol, alcoholism, and family violence. In V. B. Van Hasselt, R. L. Morrison, A. S. Bellack, & M.

Hersen (Eds.),Handbook of family violence(pp. 383–406). New York:

Plenum Press.

Leonard, K. E., & Quigley, B. M. (1999). Drinking and marital aggression in newlyweds: An event-based analysis of drinking and the occurrence of husband marital aggression.Journal of Studies on Alcohol, 60, 537–545.

Leonard, K. E., & Roberts, L. J. (1998). The effects of alcohol on marital interactions of aggressive and nonaggressive husbands and their wives.

Journal of Abnormal Psychology, 107,602–615.

Leonard, K. E., & Senchak, M. (1996). Prospective prediction of husband marital aggression within newlywed couples.Journal of Abnormal Psy- chology, 105,369–380.

Nisonoff, L., & Bitman, I. (1979). Spouse abuse.Victimology, 4,131–140.

O’Farrell, T. J., & Fals-Stewart, W. (2000). Behavioral couples therapy for alcoholism and drug abuse.Journal of Substance Abuse Treatment, 14, 1–7.

O’Farrell, T. J., & Langenbucher, J. (1988). Time-line drinking behavior interview. In M. Hersen & A. Bellack (Eds.),Dictionary of behavioral assessment techniques(pp. 477–479). Elmsford, NY: Pergamon Press.

O’Farrell, T. J., & Murphy, C. M. (1995). Marital violence before and after alcoholism treatment.Journal of Consulting and Clinical Psychol- ogy, 63,256–262.Pernanen, K. (1991).Alcohol in human violence.New York: Guilford Press.

Quigley, B. M., & Leonard, K. E. (2000). Alcohol, drugs, and violence. In V. B. Van Hasselt & M. Hersen (Eds.),Aggression and violence: An introductory text(pp. 259–283). Boston: Allyn & Bacon.

Raudenbush, S., Bryk, A., Cheong, Y. F., & Congdon, R. (2001).HLM 5:

Hierarchical linear and nonlinear modeling.Chicago: Scientific Soft- ware International.

Raudenbush, S., Yang, M. L., & Yosef, M. (2000). Maximum likelihood for hierarchical linear models via higher-order, multivariate Laplace approximation.Journal of Computational and Graphical Statistics, 9, 141–157.

Richardson, D. (1981). The effect of alcohol on male aggression toward female targets.Motivation and Emotion, 5,333–334.

Searles, J. S., Perrine, M. W., Mundt, J. C., & Helzer, J. E. (1995).

Validation of daily self-reported alcohol consumption using interactive voice response (IVR) technology.Journal of Studies on Alcohol, 56, 487–490.

Selzer, M. L. (1971). The Michigan Alcoholism Screening Test: The quest for a new diagnostic instrument.American Journal of Psychiatry, 127, 1653–1658.

Snijders, T., & Bosker, R. (1999).Multilevel analysis: An introduction to basic and advanced multilevel modeling.Thousand Oaks, CA: Sage Publications.

Sobell, L. C., & Sobell, M. B. (1996).Timeline followback user’s guide: A calendar method for assessing alcohol and drug use.Toronto, Ontario, Canada: Addiction Research Foundation.

Spanier, G. (1976). Measuring dyadic adjustment: New scales for assessing the quality of marriage and similar dyads.Journal of Marriage and Family, 38,15–30.

Straus, M. A. (1979). Measuring intrafamily conflict and violence: The Conflict Tactics (CT) Scales.Journal of Marriage and Family, 41, 75–78.

Straus, M. A. (1990). The Conflict Tactics Scale and its critics: An evaluation and new data on validity and reliability. In M. A. Straus & R. J. Gelles (Eds.),Physical violence in American families(pp. 49–73).

New Brunswick, NJ: Transaction Publishers.

Straus, M. A., & Gelles, R. J. (1990).Physical violence in American families: Risk factors and adaptations to violence in 8,145 families.New Brunswick, NJ: Transaction Publishers. Received July 10, 2001 Revision received October 31, 2001 Accepted November 8, 2001 52 FALS-STEWART