A dyadic approach to attachment anxiety and avoidance, marital conflict, and drinking

Abstract

In this research, we integrate attachment theory and dyadic methodology to examine how attachment anxiety and avoidance might interact with marital conflict to influence alcohol consumption, drinking motives, and alcohol-related problems in a sample of 280 married and cohabiting couples over 3 years. Both husband and wife attachment anxiety were related to higher levels of own drinking to cope and alcohol-related problems. Additionally, both husband and wife reports of marital conflict were associated with own alcohol-related problems. For wives, significant interactions between anxiety and marital conflict suggested that anxiety was more strongly associated with alcohol consumption, coping, and problems at higher levels of conflict. For husbands, significant interactions between avoidance and conflict indicated that avoidance was more strongly associated with coping and problems at lower levels of conflict. This research suggests two main patterns of attachment and alcohol use, both exacerbated by marital conflict and different for husbands and wives.

Despite the many benefits of high-quality romantic relationships for mental and physical health (Kiecolt-Glaser, 2018), navigating daily life with another person requires managing divergent preferences, opinions, and values. Further, the way people approach and resolve conflict differs as a function of each partner’s internal working models of attachment. Attachment theory provides a theoretical lens through which both romantic relationship behaviors (e.g., conflict resolution) and various risky health behaviors (e.g., coping drinking motives, alcohol-related problems) can be understood (Bowlby, 1973; Hazan & Shaver, 1994). Previous work has identified attachment insecurity as a risk factor for alcohol use (Fletcher et al., 2014; Kassel et al., 2007), but few studies have examined dyadic effects. Moreover, it is not yet known under what circumstances attachment dimensions place individuals at greatest risk for developing alcohol problems. We explore marital conflict as one exacerbating circumstance in a sample of married couples followed over three years.

Relationship processes, alcohol use, and gender

Research supports the notion that the association between romantic relationship processes and heavy alcohol use in couples is bidirectional. That is, negative interactions with a partner are related to increases in drinking, which are related to further increases in conflict (Fairbairn et al., 2018; Marshal, 2003; Rodriguez, Neighbors, et al., 2014). Studies using cross-sectional, daily diary, and dyadic designs show that negative relationship interactions increase the likelihood of subsequent drinking behaviors (DeHart et al., 2008; Derrick & Testa, 2017; Levitt & Cooper, 2010; Mohr et al., 2001; Rodriguez, Knee, et al., 2014).

Some research also suggests that the association between relationship processes and alcohol use may be more robust for women than men. Women (but not men) are more likely to drink alcohol (and drink at heavier levels if they do drink) on days when they feel less emotionally connected to their partner (Levitt & Cooper, 2010). Similarly, wives’—but not husbands’—reports of alcohol use are related to greater alcohol problems in the presence of higher levels of conflict, indicating a stress-exacerbating effect of conflict on alcohol problems more evident among women (Windle & Windle, 2019).

Conversely, positive relationship processes mitigate the likelihood of alcohol problems. Relationship characteristics (e.g., partner support) provide a buffer for environmental stressors’ influence on the development of alcohol problems (Jarnecke & South, 2014). This buffering effect may also be stronger among women than men: Positive interactions with a romantic partner are associated with lower subsequent intoxication levels among women (but not men) with a past-year alcohol use disorder diagnosis (Cranford et al., 2015). In summary, some individuals use alcohol as a result of experiencing negative interactions with a romantic partner, and the association between romantic relationship processes and alcohol use may be stronger among women partners in the dyad.

Attachment theory and links to alcohol use

Attachment theory provides a theoretical framework for understanding how people differ in managing interpersonal stress, and why some couples are at greater risk of using alcohol in response to conflict (Bowlby, 1969; Shaver & Hazan, 1993). Although researchers have operationalized adult attachment in a variety of ways, a widespread approach is to assess attachment insecurity via dimensions of anxiety and avoidance (Brennan et al., 1998; Shaver & Mikulincer, 2002), with individuals low in both being considered more secure. Whereas those higher in attachment anxiety tend to be hyper-vigilant to relationship threat and partner negative affect (e.g., Feeney & Ryan, 1994; Simpson, 1990), those higher in attachment avoidance are characterized by a lower desire for closeness and a heightened focus on maintaining independence and emotional distance (Mikulincer & Shaver, 2007). Related to the current study design (i.e., measurements every year for three years), attachment anxiety and avoidance fluctuate over time through naturally occurring relationship processes (Arriaga et al., 2018; Kirkpatrick & Davis, 1994). These within-person fluctuations have been linked to characteristics of the individual’s relationship, suggesting that attachment orientations can be responsive to relationship dynamics (e.g., Park et al., 2019). Baldwin and Fehr (1995) estimated that around 30% of adults experience statistically significant changes in their attachment patterns across relatively short time spans (less than 1 year).

Links to drinking patterns.

Individuals with greater attachment anxiety and avoidance are more likely to engage in health-risk behaviors such as heavy drinking (Fairbairn et al., 2018; McNally et al., 2003; Molnar et al., 2010). Some researchers posit this is due to how insecurely attached individuals manage affective distress. Indeed, substances may be an attractive way to “self-medicate” in response to experiencing negative emotions, stress, or relationship distress or threat (Khantzian, 1997). Whereas securely attached individuals are better able to seek, utilize, and maintain social support and emotion regulation strategies in efforts to cope, insecurely attached individuals are more likely to utilize less adaptive coping strategies (Feeney & Ryan, 1994; Mikulincer & Florian, 1998; Mikulincer & Orbach, 1995; Ognibene & Collins, 1998; Shaver & Clark, 1994; Shaver & Hazan, 1993), including drinking to cope (Brennan & Shaver, 1995).

Substance use in response to certain emotions might differ based on one’s specific attachment orientation. Generally, existing literature suggests that attachment anxiety is more strongly related to coping motives and drinking-related problems than is avoidance (e.g., Cooper et al., 1998; Kassel et al., 2007; Levitt & Leonard, 2015; McNally et al., 2003; Molnar et al., 2010). Molnar et al. (2010) observed in a sample of college students that greater attachment anxiety was related to experiencing more alcohol-related consequences, as well as greater endorsement of all drinking motives. Results from a clinical sample (i.e., clients at an addiction treatment facility) indicated that attachment anxiety was associated with higher alcohol consumption and alcohol-related consequences. Although attachment avoidance was not associated with either drinking or alcohol-related consequences in either sample, avoidance was associated with higher levels of coping motives in the student sample. In another study of college students, higher attachment anxiety (but not avoidance) was linked with greater coping motives and negative alcohol-related consequences (McNally et al., 2003). However, a recent meta-analysis found that the relationship between attachment subtype (anxious vs. avoidant) and alcohol use was nonsignificant after controlling for type of measure used (Fairbairn et al., 2018). Although evidence is fairly consistent that attachment anxiety drives heavier drinking, somewhat mixed findings for avoidance highlight the need to examine both attachment dimensions in the context of marital strains and alcohol use and related problems.

Current research

The tenets of attachment theory suggest the anxiety—drinking association should be especially pronounced under relationship stress (Feeney, 1998; Rholes et al., 1998; Simpson & Rholes, 2017), a time when anxious individuals’ attachment systems are activated and they are most affected by negative interactions. Despite a widespread call by theoretical experts (Simpson & Howland, 2012), few studies examining attachment and substance use have adopted a dyadic approach in their design, resulting in an underdeveloped understanding of how both partners’ attachment relates to drinking and how relationship conflict might moderate associations between attachment and drinking. Moreover, although research now strongly supports the notion that relationship partners influence each other, there is little research investigating how one person’s attachment might be associated with their partner’s drinking patterns.

In this research, we utilized a dyadic, longitudinal approach to explore between- and within-person actor and partner effects of attachment anxiety and avoidance, marital conflict, and their interaction on different alcohol outcomes in a sample of married couples. We expected that greater attachment anxiety (henceforth termed anxiety) would be associated with greater alcohol use, coping motives, and alcohol-related problems (Hypothesis 1). We expected all effects to emerge at the between- and within-person levels. In other words, we expected that compared to less anxiously attached (henceforth termed less anxious) individuals, more anxiously attached (henceforth termed more anxious) individuals would report heavier drinking and coping patterns (i.e., a between-person effect), and that when individuals felt more anxious than usual, they would report heavier drinking and coping patterns (i.e., a within-person effect). Importantly, we expected the anxiety–drinking behavior association to be stronger in the presence of greater marital conflict (Hypothesis 2), and we expected these effects to be stronger for wives compared to husbands (Hypothesis 3). Given the lack of consistent results in previous research, we did not specify a priori hypotheses for attachment avoidance.

With regard to partner effects, very few studies have evaluated partner effects of attachment on drinking outcomes. Levitt and Leonard (2015) found that partner anxiety was related to relationship-specific drinking-to-cope motives. Other work showing that partner attachment insecurity—particularly anxiety—is related to lower satisfaction (Banse, 2016; Erol & Orth, 2013; Feeney, 1994; Mondor et al., 2011) and negative affectivity (Davila et al., 1998) suggests the possibility that partner attachment insecurity—particularly anxiety—might be related to drinking patterns. We explore this possibility in this study.

Method

Participants

Data for the current manuscript were drawn from a larger three-wave longitudinal study comprised of heterosexual married couples and unmarried cohabitating (>1 year) couples from the larger Buffalo, NY community (N = 280 dyads) (Derrick et al., 2010; Testa et al., 2012). To ensure that this sample would be large enough to test our hypotheses, we conducted power analyses using a Monte Carlo simulation approach described in Muthén and Muthén (2002). Results indicated that we would have power greater than .80 to detect Level 1 effects as small as d = .12 and Level 2 effects as small as d = .20.

Participants were recruited via mailed packet containing a letter describing the study and a screening survey sent to households in Erie County, NY that had a high likelihood of containing a married couple between the ages of 18 and 45. The response rate was 26% (5,463/21,000 responses were received, 226 returned due to incorrect address). Couples were excluded if either partner reported a current medical condition that would impair cognitive functioning or history of a serious head injury or loss of consciousness. To ensure sufficient sample size of heavy drinking women and men, couples in which one or both partners was a heavy episodic drinker (4/5 drinks per occasion, at least weekly) were oversampled (see Derrick et al., 2010; Testa et al., 2012, for details on sampling and recruitment). To avoid the bias associated with disproportionate sampling (Korn & Graubard, 1995; Pfeffermann, 1993), each couple’s data were weighted as a proportion of the strata in the population divided by the proportion of the strata in the sample. In this study, there were four strata: (1) husband-only heavy drinker (12.3% of the population); 2) wife-only heavy drinker (4.1%); (3) both heavy drinkers (8.5%); and (4) neither heavy drinker (75%). In the sample, there were 28.6% husband-only, 15.6% wife-only, 28.2% both, and 28.6% neither. Thus, the sample weights for the husband only, wife only, both, and neither strata were .430, .263, .301, and 2.62, respectively.

The average age at Wave 1 was 35.4 (SD = 5.9) for women and 36.9 (SD = 5.8) for men. The sample was largely White (91% each), well-educated (67% of women and 58% of men had completed college education), and employed at least part-time (80% of women and 91% of men). Median income for women was in the US$20,000–US$29,999 range, and median income for men was in the US$40,000–US$54,999 range. Most couples in the sample were married (87%) and had been together for an average of 9.84 (SD = 5.41) years. Nearly 80% of the sample had children: 15% had one child, 38% had two, 19% had three, and 7.5% had more than three. Of the 280 couples who participated at Wave 1, 259 (92.5%) completed Wave 2 and 243 couples (87% of original sample) completed Wave 3. Husbands and wives completed 2.8 Waves on average.

Procedure

Participants provided informed consent, completed a series of questionnaires through the mail, and attended a laboratory assessment. Mailed questionnaires were sent separately to wives and husbands and consisted of background information, attitudes, and personality measures. At the laboratory assessment, couples completed questionnaires assessing alcohol and drug use and relationship functioning. Couples also completed a series of executive cognitive functioning tasks and a conflict interview. Each participant received US$100 for Wave 1 and US$75 for each of the shorter Wave 2 and 3 assessments.

Measures

Attachment dimensions.

Adult attachment styles were assessed using a four-part categorical measure of attachment styles (Bartholomew & Horowitz, 1991). Participants read a description of each attachment style and rated the extent to which the style described them using a 7-point scale ranging from 1 (not at all like me) to 7 (very much like me). We calculated the two continuous dimensions of attachment anxiety and attachment avoidance following procedures outlined by Griffin and Bartholomew (1994).

Marital conflict.

The frequency of perceived negative interaction with one’s partner was assessed using an adapted version of the 18-item Test of Negative Social Exchange (Ruehlman & Karoly, 1991; Schumacher & Leonard, 2005). Participants reported how often during the past month their partner engaged in various types of negative interaction (e.g., “My partner lost his or her temper with me,” “My partner ignored my wishes or needs”). Participants responded on a 5-point scale ranging from 1 (not at all) to 5 (about every day). Responses were averaged to create a total score, with higher scores indicating more frequent negative interactions. Very few participants (less than 5%) indicated their partners had not engaged in any of the negative behaviors, and the scale was highly reliable (α = .92–.95 across the three waves).

Alcohol consumption.

Participants completed a standard quantity/frequency measure to assess alcohol consumption over the past year (Cahalan et al., 1969). The items were used to calculate an assessment of approximate drinks per week.

Coping drinking motives.

Drinking to cope with stress was assessed using two alcohol-related items on the Coping with Stress Scale (Rothbard, 1996; Schumacher et al., 2008); “I use alcohol to help me get through it” and “I drink alcohol in order to think about it less.” Participants were asked to indicate what they generally or usually do when under a lot of stress. Frequency of drinking to cope was measured on a 4-point scale ranging from 1 (I don’t do this at all) to 4 (I do this a lot). Responses on the two items were summed with higher scores reflecting greater drinking to cope motives.

Alcohol-related problems.

Excessive alcohol use severity was assessed using the 25-item Alcohol Dependence Scale (ADS; Skinner & Horn, 1984). Participants self-reported symptoms such as loss of behavioral control (e.g., blackouts, gulping drinks), withdrawal symptoms (e.g., hallucinations, hangovers, delirium tremens), and obsessive drinking style (e.g., sneaks drinks, always has a bottle at hand). Responses were summed to create a total score. The distribution of ADS scores was highly skewed, with 41.9% of men and 55.2% of women reporting scores of 0, and 4.7% of men and 1.9% of women scoring 10+. As such, scores were winsorized to reduce the impact of extremely high scores (Reifman & Keyton, 2010).

Analysis plan

The data for the current analyses were taken from both partners within couples assessed at three timepoints, and therefore, there were multiple sources of nonindependence in our data. Accordingly, we conducted multilevel path analyses with random intercepts and fixed slopes (Preacher et al., 2010) in Mplus (Muthén et al., 2017). Specifically, multiple time points (Level 1) were nested within couples (Level 2), and husbands and wives were treated as separate paths in simultaneous models. We included both Level 1 and 2 versions of each predictor in all analyses. Predictors at Level 2 were grand-mean centered, and predictors at Level 1 were person-mean centered. Therefore, significant effects at Level 2 indicate that an effect emerged for participants who were relatively lower or higher than the mean of the sample on that variable. Significant effects at Level 1 indicate that an effect emerged in years when participants were relatively lower or higher than “typical,” or their own mean, on that variable. Given the sampling frame, we included sampling weights in all analyses. Across all individuals, timepoints, and variables, 8.06% of data were missing. Missing data were handled using full information maximum likelihood estimation.

We used an actor–partner interdependence model (APIM) framework (Kenny et al., 2006) by including both a person’s own reports (i.e., actor effects) and his/her partner’s reports (i.e., partner effects) in each model (see Figure 1). We first examined both husbands’ and wives’ actor and partner main effects of attachment anxiety, avoidance, and conflict (H1). In a second step, we included all of the actor two-way anxiety × conflict and avoidance × conflict interactions and all of the partner two-way anxiety × conflict and avoidance × conflict interactions (H2). Significant interactions were followed up by decomposing into the simple slopes of anxiety or avoidance for those relatively low, medium, and high (i.e., 1 SD below the mean, at the mean, and 1 SD above the mean) on conflict (Aiken & West, 1991). As the models were just-identified, we tested for gender differences on each path using a simple 1 df χ² test (H3).

Figure 1.

Hypothesized model evaluating actor and partner effects of attachment anxiety, attachment avoidance, and marital conflict, as well as interactions between attachment dimensions and conflict, on alcohol outcomes.

Results

Descriptive statistics and preliminary associations

Descriptive statistics at all waves and intraclass correlation coefficients are provided in Table 1. Correlations among all variables at the within-person level are presented in Table 2. The significant correlations between spouses for consumption, coping motives, alcohol problems, and marital conflict indicate that the scores of dyad members were nonindependent. As expected, consumption, coping motives, and alcohol problems were positively correlated for both husbands and wives.

Table 1.

Descriptive statistics for study sample at all waves.

	Wave 1		Wave 2		Wave 3
Variable	Husbands M (SD)	Wives M (SD)	Husbands M (SD)	Wives M (SD)	Husbands M (SD)	Wives M (SD)
Consumption	2.31 (2.71)	1.16 (1.90)	1.96 (2.44)	.81 (1.38)	2.01 (2.62)	.84 (1.26)
Coping motives	1.57 (.71)	1.45 (.67)	1.60 (.76)	1.37 (.62)	1.60 (.77)	1.38 (.67)
Alcohol problems	3.92 (4.26)	2.80 (3.62)	3.88 (4.43)	2.52 (3.46)	3.72 (4.32)	2.25 (3.30)
Anxiety	2.37 (.97)	2.57 (1.12)	2.36 (.98)	2.92 (1.04)	2.43 (.94)	2.59 (1.20)
Avoidance	2.91 (1.16)	2.45 (1.06)	2.92 (1.04)	2.56 (1.11)	2.97 (1.08)	2.44 (1.05)
Marital conflict	1.87 (.62)	1.76 (.63)	1.95 (.67)	1.83 (.64)	1.95 (.65)	1.85 (.70)

Table 2.

Bivariate correlations at within-person level.

Variable	1	2	3	4	5	6
1. Consumption	.08*	.47***	.49***	.03	.02	.03
2. Coping motives	.47***	.12**	.52***	.07	−.01	.05
3. Alcohol problems	.51***	.56***	.16***	.02	.05	.02
4. Anxiety	.01	.03	.01	−.01	.20***	−.01
5. Avoidance	−.03	−.02	−.02	.02	−.06	.02*
6. Marital conflict	.02	.04	.03	.02	.01	.24***
ICC (husbands)	.44	.56	.73	.72	.69	.82
ICC (wives)	.54	.59	.74	.61	.61	.69

Note. Correlations for wives are presented above the diagonal. Correlations for husbands are presented below the diagonal. The correlation between husbands and wives for each subscale is presented along the diagonal. ICC = intraclass correlations.

^*p <.05;

^**p < .01;

^***p <.001.

Multilevel path analysis

Alcohol consumption.

Results evaluating consumption are presented for husbands and wives in Table 3. Results at Step 1 showed no within-person associations of attachment or conflict on consumption. However, there was a significant between-person actor effect of anxiety on consumption for husbands, suggesting that more anxious husbands reported greater alcohol consumption on average compared with less anxious husbands. This actor effect of anxiety on consumption was significantly different for husbands and wives, χ²(1) = 6.71, p = .010.

Table 3.

Actor–partner interdependence models evaluating husband and wife attachment dimensions and marital conflict on alcohol consumption.

	Outcome: Alcohol consumption
	Husband		Wife
	Within b (SE)	Between b (SE)	Within b (SE)	Between b (SE)
Time	−.96 (.47)*	—	−1.03 (.30)**	—
Husband anxiety	.44 (.55)	4.02 (1.21)***	−.47 (.40)	.96 (.76)
Husband avoidance	−.81 (.75)	−.34 (1.09)	.34 (.56)	−.40 (.56)
Husband marital conflict	.58 (1.13)	−1.48 (3.06)	−.01 (1.01)	−.58 (1.76)
Wife anxiety	.08 (.64)	−.80 (.98)	.29 (.32)	.36 (.54)
Wife avoidance	−.09 (.82)	1.18 (1.20)	.23 (.66)	1.15 (.82)
Wife marital conflict	2.15 (2.35)	1.92 (2.35)	1.43 (.88)	2.10 (1.54)
Husband anxiety × Husband marital conflict	.46 (2.64)	−1.29 (2.28)	5.42 (5.22)	1.03 (.86)
Husband avoidance × Husband marital conflict	−4.03 (2.81)	−.76 (1.79)	−5.73 (2.88)*	−.28 (1.13)
Wife anxiety × Wife marital conflict	3.33 (1.99)†	2.92 (1.36)*	6.74 (2.88)*	2.15 (.98)*
Wife avoidance × Wife marital conflict	−2.63 (3.42)	−1.20 (1.99)	−9.30 (5.59)†	.44 (1.45)

Note. Main effects come from a main effect only model. Interaction effects come from a model including main effects and interactions. Within = within-person level; Between = between-person level. Covariance between husband and wife consumption: between-person level b (SE) = 29.15 (10.20), p = .004; within-person level b (SE) = .46 (4.38), p = .916.

^†p < .10;

^*p < .05;

^**p < .01;

^***p < .001.

At Step 2, four significant interactions emerged. First, we observed two actor attachment × actor conflict interactions for wives. Specifically, her own anxiety interacted with her reports of marital conflict to predict consumption at both the between- and within-person levels. We illustrate this pattern for the between-person interaction in Figure 2. At higher levels of conflict, wife anxiety was marginally positively related to her consumption, an association that was not significant at mean or at lower levels of conflict. The pattern of results was generally consistent with the interaction at the within-person level (Figure 3)—wife anxiety was positively related to her consumption in years when she experienced relatively high levels of conflict, unrelated to consumption when she experienced typical levels of conflict, and negatively related to consumption when she experienced relatively low levels of conflict. This interaction differed significantly by gender at the within-person level, χ²(1) = 5.35, p = .021.

Figure 2.

Wife attachment anxiety × Wife conflict → Wife consumption (between-person). The effect of wife anxiety on her own alcohol consumption was not significant at low conflict, b = −1.127, SE(b) = .814, p = .166, nor at average conflict, b = .173, SE(b) = .534, p = .746, but it was significant and positive at high conflict, b = 1.473, SE(b) = .779, p = .059. ^†p < .10.

Figure 3.

Wife attachment anxiety × Wife conflict → Wife consumption (within-person). The effect of wife anxiety on her own alcohol consumption was negative at relatively low conflict, b = −3.675, SE(b) = 1.688, p = .029, unrelated to alcohol consumption at average conflict, b = .413, SE(b) = .343, p = .229, and was significant and positively related to consumption at relatively high conflict, b = 4.509, SE(b) = 1.868, p = .016. *p < .05.

Second, we observed one significant partner attachment × partner conflict interaction for husbands. Specifically, wife anxiety interacted with wife conflict to predict husband consumption at the between-person level. As illustrated in Figure 4, wife anxiety was unrelated to husband consumption at higher and at mean levels of conflict but was marginally related to lower husband consumption at lower levels of conflict.

Figure 4.

Wife attachment anxiety × Wife conflict → Husband consumption (between-person). Wife anxiety was associated with marginally lower husband alcohol consumption at lower levels of conflict, b = −2.724, SE(b) = 1.435, p = .058. Her anxiety was unrelated to husband consumption at average levels of conflict, b = −.955, SE(b) = 1.103, p = .346, and at higher levels of conflict, b = .813, SE(b) = 1.164, p = .485. ^†p < .10.

Third, we observed a partner avoidance × partner conflict effect for wives at the within-person level. Specifically, husband avoidance interacted with husband conflict to predict wife consumption. As presented in Figure 5, husbands’ avoidance was marginally negatively related to her drinking at high levels of conflict, was not related at typical levels of conflict, and was marginally positively related at low levels of conflict.

Figure 5.

Husband attachment avoidance × Husband conflict → Wife consumption (within-person). Husband avoidance was marginally positively related to his wife’s alcohol consumption when experiencing relatively low conflict, b = 3.848, SE(b) = 1.964, p = .050, unrelated to his wife’s consumption when experiencing average levels of conflict, b = .394, SE(b) = .443, p = .374, and was marginally negatively related to his wife’s consumption when experiencing relatively high conflict, b = −3.056, SE(b) = 1.595, p = .055. ^†p < .10.

Drinking to cope.

As shown in Table 4, Step 1 results at the between-person level indicated that both spouses’ anxiety was significantly associated with their own coping motives. Results at the within-person level suggested that wives’ anxiety and perception of conflict were independently associated with higher own coping motives.

Table 4.

Actor–partner interdependence models evaluating husband and wife attachment dimensions and marital conflict on drinking to cope.

	Outcome: Coping drinking motives
	Husband		Wife
	Within b (SE)	Between b (SE)	Within b (SE)	Between b (SE)
Time	.01 (.02)	—	−.03 (.02)	—
Husband anxiety	.03 (.03)	.19 (.05)***	−.05 (.03)	.06 (.05)
Husband avoidance	−.02 (.03)	−.02 (.04)	−.02 (.03)	−.01 (.04)
Husband marital conflict	.08 (.06)	.02 (.12)	−.01 (.06)	.09 (.08)
Wife anxiety	.03 (.05)	−.03 (.04)	.07 (.03)*	.10 (.04)*
Wife avoidance	.04 (.03)	.03 (.04)	−.02 (.04)	.03 (.04)
Wife marital conflict	.03 (.05)	.04 (.10)	.12 (.06)*	.03 (.07)
Husband anxiety Husband marital conflict	.20 (.14)	−.11 (.08)	.14 (.12)	.05 (.08)
Husband avoidance × Husband marital conflict	−.33 (.12)**	−.01 (.07)	.01 (.16)	.08 (.06)
Wife anxiety × Wife marital conflict	−.04 (.10)	.04 (.07)	.11 (.12)	.11 (.06)*
Wife avoidance × Wife marital conflict	.11 (.10)	.08 (.06)	−.04 (.14)	.02 (.07)

Note. Main effects come from a main effect only model. Interaction effects come from a model including main effects and interactions. Within = within-person level; Between = between-person level. Covariance between husband and wife consumption: Between-person level b (SE) = 02 (.02), p = .335; Within-person level b (SE) = .02 (.01), p = .044.

^*p < .05;

^**p < .01;

^***p < .001.

Step 2 (i.e., interaction) model results revealed two significant interactions. First, consistent with the consumption interaction, a between-person wife anxiety × conflict on wife coping interaction indicated her anxiety was positively related to her coping motives among wives experiencing higher levels and mean levels of conflict, but not among wives experiencing lower levels of conflict (Figure 6). This interaction was significantly different for husbands and wives, χ²(1) = 7.47, p = .006.

Figure 6.

Wife attachment anxiety × Wife conflict → Wife coping motives (between-person). Wife anxiety was positively associated with higher wife coping motives at higher conflict, b = .156, SE(b) = .046, p = .001, and at average conflict, b = .088, SE(b) = .037, p = .016, but unrelated to coping motives at lower conflict, b = .020, SE(b) = .053, p = .702. *p < .05; **p < .01.

Second, the husband avoidance × conflict on husband coping interaction at the within-person level suggested that for husbands, in years when they experienced more avoidance than average, they reported lower coping motives when experiencing relatively high relationship conflict and higher coping motives when experiencing relatively low relationship conflict. This interaction is presented in Figure 7 and was significantly different for husbands and wives, χ²(1) = 4.14, p = .042.

Figure 7.

Husband attachment avoidance × Husband conflict → husband coping motives (within-person). Husband avoidance was significantly associated with lower coping motives at relatively high levels of conflict, b = −.213, SE(b) = .073, p = .003, unrelated to coping at average levels of conflict, b = −.013, SE(b) = .025, p = .605, and was significantly associated with higher coping motives at relatively low levels of conflict, b = .186, SE(b) = .078, p = .017. *p < .05; **p < .01.

Alcohol problems.

As presented in Table 5, results from models at Step 1 revealed that husband anxiety was positively related to his own alcohol problems (p < .001) at the between-person level. Although this actor effect was also observed among wives (p = .020), it was significantly stronger among husbands, χ²(1) = 7.13, p = .008. Perceptions of greater conflict were associated with higher alcohol problems for both husbands (within-person level) and wives (between-person level).

Table 5.

Actor–partner interdependence models evaluating husband and wife attachment dimensions and marital conflict on alcohol-related problems.

	Outcome: Alcohol-related problems
	Husband		Wife
	Within b (SE)	Between b (SE)	Within b (SE)	Between b (SE)
Time	−.07 (.09)	—	−.26 (.10)*	—
Husband anxiety	.04 (.11)	1.56 (.31)***	−.09 (.14)	.03 (.28)
Husband avoidance	−.15 (.12)	−.29 (.25)	−.04 (.11)	−.10 (.23)
Husband marital conflict	.54 (.25)*	.68 (.77)	.02 (.24)	1.9 (.83)
Wife anxiety	.12 (.12)	−.13 (.30)	.01 (.15)	.51 (.22)*
Wife avoidance	−.17 (.15)	.36 (.28)	.37 (.28)	.39 (.23)†
Wife marital conflict	.10 (.28)	.21 (.59)	.26 (.38)	.93 (.38)*
Husband anxiety × Husband marital conflict	.68 (.61)	.54 (.66)	.41 (.96)	−.01 (.38)
Husband avoidance × Husband marital conflict	−1.00 (.51)*	−.68 (.40)	.61 (.60)	−.45 (.61)
Wife anxiety × Wife marital conflict	.15 (.62)	.48 (.43)	.68 (.61)	.60 (.28)*
Wife avoidance × Wife marital conflict	−.47 (.76)	.17 (.47)	.69 (.76)	.71 (.41)

^*Note. Main effects come from a main effect only model. Interaction effects come from a model including main effects and interactions. Within = within-person level. Between = between-person level. Covariance between husband and wife consumption: Between-person level b (SE) = 1.67 (.73), p = .023; Within-person level b (SE) = .09 (.202), p = .650.

^†p < .10;

^*p < .05;

^**p < .01;

^***p < .001.

Interaction results at Step 2 revealed a significant wife anxiety × conflict actor interaction for wives at the between-person level. As shown in Figure 8, similar to coping and consumption, her anxiety was significantly related to her alcohol problems at mean and higher levels of conflict but not at lower levels of conflict. Interaction results also revealed a marginally significant (p = .050) husband avoidance × conflict interaction on his alcohol problems at the within-person level. Similar to the results with coping and illustrated in Figure 9, when husbands reported relatively high levels of conflict, his avoidance was negatively related to his alcohol problems, an association not present at his typical and at relatively low levels of conflict.

Figure 8.

Wife attachment anxiety × Wife conflict → Wife alcohol problems (between-person). The effect of wife anxiety on alcohol consumption was not significant at low conflict, b = .106, SE(b) = .279, p = .705, but was significant and positive at average conflict, b = .466, SE(b) = .212, p = .028, and at high conflict, b = .826, SE(b) = .265, p = .002. *p < .05; **p < .01.

Figure 9.

Husband attachment avoidance × Husband conflict → Husband alcohol problems (within-person). Husband avoidance was related to lower levels of alcohol problems at relatively high levels of conflict, b = −.726, SE(b) = .338, p = .032. His avoidance was unrelated to alcohol problems at average conflict, b = −.128, SE(b) = .122, p = .293, and at relatively low levels of conflict, b = .469, SE(b) = .321, p = .144. *p < .05.

Discussion

The current research used a dyadic approach to examine how husbands’ and wives’ attachment insecurity was associated with alcohol consumption, coping motives, and alcohol-related problems, and if these associations were exacerbated under greater marital conflict. We hypothesized that greater attachment anxiety would be associated with heavier drinking patterns, we expected this to be stronger in the context of greater relationship turbulence, and we expected these effects to be stronger among wives compared to husbands.

Results were relatively consistent across outcomes, particularly coping motives and alcohol-related problems. Supporting Hypothesis 1, results revealed that more anxious wives endorsed greater drinking to cope motives and alcohol problems compared to less anxious wives. Similarly, more anxious husbands consumed more on average, endorsed greater drinking to cope motives, and reported more alcohol problems compared to less anxious husbands. These results are consistent with previous research demonstrating that more anxious individuals report heavier drinking patterns (e.g., Molnar et al., 2010). Moreover, our research suggests that when wives experienced greater anxiety than typical, they also reported greater coping motives. Although previous work has demonstrated that attachment does fluctuate within person over time (e.g., Arriaga et al., 2018), this is the first work of which we are aware showing that within-person deviations in attachment can influence drinking motives.

We were most interested in the intersection of attachment and marital conflict in drinking patterns. In support of Hypotheses 2 and 3, more anxious wives drank at greater quantity and frequency when conflict was high (compared to low), and wives drank more when they were more anxious than typical and their conflict was higher than typical. More anxious wives also reported greater coping motives and more alcohol problems than less anxious wives when their conflict was high. This is consistent with previous work demonstrating associations between attachment anxiety and drinking (e.g., Molnar et al., 2010) and with previous work showing links between relationship processes and drinking are stronger among women than men (Levitt & Cooper, 2010; Windle & Windle, 2019).

Further, partner effects showed that this spilled over to husband consumption: Wife anxiety was unrelated to husband consumption at higher conflict but was marginally related to lower consumption at lower conflict. For husbands, however, his own anxiety was associated with more drinking, drinking to cope, and alcohol-related problems regardless of level of conflict. This is consistent with work demonstrating partner effects of attachment anxiety on greater relationship-specific drinking-to-cope (Levitt & Leonard, 2015), lower relationship satisfaction (Feeney, 1994), and heightened negative affectivity (Davila et al., 1998). Future studies may wish to better understand whether, to what extent, and how individuals manage their partner’s attachment anxiety. In summary, anxiety was generally associated with increased drinking patterns for men, and anxiety was associated with increased drinking patterns when paired with marital tension for women. Findings are consistent with prior research suggesting negative interactions may affect women more than men (Kiecolt-Glaser & Newton, 2001; Smith et al., 2011), especially in regard to intimate or emotional aspects of the relationship (Levitt & Cooper, 2010).

A second, unexpected pattern of results emerged around avoidance, conflict, and drinking patterns, generally for husbands. In years when conflict was relatively high, relative increases in husbands’ avoidance were related to his reporting lower drinking to cope and fewer alcohol-related problems. Conversely, when conflict was relatively low, increases in his avoidance were associated with increases in drinking to cope motives. As such, it may be that the combination of avoidance and conflict is more representative of emotional suppression for men. When considering the alternative interpretation of the interaction by examining how relationship conflict is associated with drinking among husbands when reporting higher and lower avoidance, results indicate that among husbands lower in avoidance, greater relationship conflict is associated with more drinking and alcohol-related problems. However, this association is diminished among husbands relatively high in avoidance, findings consistent with attachment theory and which add to existing evidence that avoidant individuals outwardly indicate they are less afflicted or distressed by relationship tension compared to less avoidant individuals (perhaps as a way to keep their attachment system deactivated) (Mikulincer & Shaver, 2007) as well as research indicating avoidant men are less impacted emotionally by romantic conflict and separation than women (Simpson, 1990). Avoidant men may be better able to suppress negative cognitions and emotions associated with relationship conflict than avoidant women (Fraley et al., 2000; Fraley & Shaver, 1997).

Clinical and research implications

Based on our findings, clinicians may wish to incorporate measures of attachment anxiety and avoidance when working with couples who struggle with alcohol, especially if paired with relationship tension. Our results suggested rather consistently that attachment anxiety in particular was associated with all three indices of drinking for husbands and with greater coping motives and related problems for wives. These findings are consistent with other work suggesting that having a negative self-view is linked with experiencing more alcohol-related consequences (McNally et al., 2003), as well as that when more anxious people encounter relationship stressors, they perceive their partners and relationships more negatively and behave in more relationship-damaging ways (Collins et al., 2006; Gallo & Smith, 2001; Simpson & Rholes, 2017). Given that interpersonal conflict is among the most common factors contributing to relapse (Marlatt & Witkiewitz, 2005) and that using alcohol to cope with stress, negative affect, and/or interpersonal conflict is among the strongest risk factors for experiencing alcohol-related problems (Martens et al., 2008), clinicians may wish to target the combination of interpersonal turbulence and the unhealthy management of subsequent negative affect by providing ways to reappraise conflict, modify relationship-distressing attributions, and provide constructive suggestions for responding to negative affect (e.g., taking a walk together, yoga).

Highly avoidant individuals, in contrast, disengage behaviorally, emotionally, and cognitively when exposed to stressors (Simpson & Rholes, 2017). This pattern was observed with regard to all three alcohol indicators. Husbands reported lower drinking to cope as their avoidance increased when under higher marital conflict. However, they reported higher drinking to cope when under lower marital conflict. In the light of these findings, researchers and clinicians may also wish to take into account gender differences with regard to drinking to cope and relationship conflict. Previous work has found that whereas attachment avoidance was unrelated to all drinking indices for women, avoidance was indirectly related to both high-risk drinking and alcohol-related consequences via enhancement motives for men (Molnar et al., 2010). Similarly, we found more interactions between attachment avoidance and conflict on drinking outcomes for husbands than wives. Clinicians may wish to attend to how avoidantly attached individuals respond to conflict with their partner and ensure that any underlying distress is addressed.

A final interesting implication of these results is that reports of couples’ conflict were only modestly correlated (r = .24). It may seem surprising that this correlation was not stronger, but it suggests that what was captured here is more an operationalization of each partner’s perception of conflict, as opposed to an objective measure of how much conflict is actually taking place in the relationship. As is often found in social psychology (Ross & Ward, 1996) and more specifically with research on alcohol and relationships (e.g., Rodriguez & Neighbors, 2015), perception seems to be as much or more important in driving outcomes than objective reality. Clinicians doing couples work may wish to address subjective perceptions of conflict with couples to show that relationship strain can vary from individual to individual, even in the same relationship.

Limitations and future directions

This research should be considered in the light of its limitations, which provide suggestions for future investigation. First, while our sample is older than the college age or newlywed samples usually examined in alcohol and attachment research, the sample was predominantly White, highly educated, and mostly employed, and it is possible that our findings would not generalize to a more representative sample of the U.S. population. Similarly, couples who self-select to participate in dyadic research may have higher levels of commitment than those who do not, further limiting the generalizability of findings (Barton et al., 2020). Second, although the data were collected at three time points and were analyzed using an APIM framework, the correlational nature of this research does not allow us to determine the causal nature of relationships among attachment dimensions, conflict, and drinking behaviors. More refined research focusing on daily- or event-level fluctuations in attachment, conflict, and drinking would pair nicely to enhance our understanding of how these constructs interact on a more proximal level. Third is related to the specific measures used in this study. The attachment measure was selected for comparison with an earlier study that had included the Bartholomew and Horowitz (1991) questionnaire, and thus, is not consistent with modern industry standards. The measure of coping motives was part of a larger coping measure and did not assess other drinking motives (e.g., social, enhancement). Future research would benefit from assessing a wider range of drinking motives, using well-validated drinking motives measures, and a more nuanced focus on drinking to cope with depressed versus anxious feelings (Grant et al., 2007) to further isolate the mechanisms of interest. Future research might also investigate the significant interactions found here to understand what mechanisms (e.g., suppression, distraction, self-presentation for the husband avoidance interaction; negative affect reduction, relationship-specific alcohol expectancies for the wife anxiety interaction) underlie these associations. Finally, future research studies would benefit from exploring these associations with substances other than alcohol. Caspers et al. (2005) found a high prevalence of illicit substance use among insecurely attached individuals relative to secures. Other research has found that attachment anxiety is positively related to both drug use frequency and stress-motivated drug use (Kassel et al., 2007). Associations among attachment dimensions, presence of relationship tension, motives for use, and choice of substance may have important implications for research and clinical practice.

Conclusion

The current study provides further evidence that in community samples of couples, husband and wife attachment dimensions and marital conflict—independently and together—are related to the couple’s drinking behavior. Results provide relatively consistent support that attachment dimensions may operate differently for husbands and wives in their motivation for alcohol use. Marital conflict tends to exacerbate associations between attachment anxiety and drinking—especially for wives, whereas it mitigates associations between attachment avoidance and drinking—especially for husbands.

Open research statement.

As part of IARR’s encouragement of open research practices, the author(s) have provided the following information: This research was not pre-registered.

The data and materials used in the research cannot be publicly shared but are available upon request by emailing Lrodriguez12@usf.edu.