Thinking When I’m Drinking: The Interaction of Mood and Rumination Across Drinking and Non-Drinking Days

Abstract

Background:

Prior research has explored the links between emotion regulation, rumination, and problematic use. However, the extant literature has yet to investigate daily fluctuations among these factors on drinking and non-drinking days comparatively.

Method:

College student drinkers (n= 85; 64% female; M age = 20.04 years) were enrolled in a 21-day ecological momentary assessment study (EMA) examining rumination, via the brooding subscale of the ruminative response scale, and emotion, from the mood circumplex model, on both drinking and non-drinking days. Participants completed up to 11 assessments each day measuring composite positive mood, sadness, anger, anxiety, rumination, and alcohol use. Multilevel regression examined the influence of rumination and mood trajectories across drinking and non-drinking days.

Results:

On drinking days, positive mood peaked at drinking initiation and declined after. On non-drinking days, positive mood was low across the day. Both sadness and anger trajectories were moderated by rumination on drinking days. On drinking days, at high rumination both sadness and anger peaked several hours before drinking initiation and then declined across the day leading up to a drinking episode. In the hours after drinking, sadness and anger changed directions and escalated rapidly. Anxiety was lower on drinking days and not moderated by rumination.

Discussion:

Findings support differences emotional functioning on drinking days and non-drinking days, particularly for individuals with rumination; highlighting the importance of rumination in the context of overall affect and alcohol use.

Introduction

Alcohol is one of the most widely used substances in the United States (Rehm et al., 2009). Although current recommendations for healthy alcohol consumption in the United States are one or fewer standard drinks per day for women and two or fewer standard drinks per day for men, (U.S. Department of Agriculture and U.S. Department of Health and Human Services, 2020), a significant number of adult drinkers engage in excessive or problematic use. In fact, approximately 7% of adult drinkers engaged in heavy drinking (i.e., 8+/15+ standard drinks per week for women/men) and approximately 17% of adult drinkers engaged in binge drinking in 2022 (i.e., 4+/5+ standard drinks per 2 hour period for women/men; Centers for Disease Control and Prevention, 2022).

Problematic alcohol use is associated with many adverse outcomes including a variety of physical and mental health problems (Barbería-Latasa et al., 2022; Castillo-Carniglia et al., 2019; Rehm et al., 2021) and an increased risk of mortality (Esser et al., 2022; Ricci et al., 2020; Tian et al., 2023). Understanding factors that may promote or maintain problematic use is of critical public health importance. Prominent theoretical models and decades of empirical research have highlighted the association between emotion and problematic alcohol use (Baker et al., 2004; W. Gorka et al., 2017; Koob & Le Moal, 2008) suggesting that the development and maintenance of problematic alcohol use may be driven, at least in part, by an individual’s emotional or affective states.

Emotion and Alcohol Use

Emotional experiences have both stable and variable aspects. Trait emotions are posited to be fairly stable constructs. In contrast, state emotions are momentary and fluctuate across situations (Nett et al., 2017). As such, emotional experiences in alcohol research are often dichotomized into distinct trait emotions (i.e., positive affect and negative affect), and different state emotions (i.e., anxiety, sadness, distress, anger; (Cho et al., 2019; De Leon et al., 2020; Dvorak et al., 2018; Emery et al., 2023; Sher & Grekin, 2007; Watson & Clark, 1994) and have been found to play a critical role in alcohol-related problems. For instance, individuals high in negative affect may use alcohol to cope with various negative emotions, and experience greater alcohol-related problems (Cooper et al., 1992; Cox & Klinger, 1988; Wray et al., 2012). Similarly, positive affect has been associated with greater alcohol consumption as well as alcohol-related problems (Cooper et al., 2000; McCreary & Sadava, 2000; Wray et al., 2012). Despite different sampling methods (i.e., cross-sectional, ecological momentary assessment, daily diary), findings regarding the impact of emotion on drinking, both at the state and trait level, remain inconsistent (Emery et al., 2023; Kassel, 2010). This inconsistency warrants further inquiry to understand fluctuations in emotions within the context of maladaptive emotion regulation strategies, such as rumination, as this may be uniquely linked to consumption.

Rumination and Alcohol Use

Rumination is perseveration on negative content and excessive consideration of the thought processes themselves, resulting in emotional distress (Sansone & Sansone, 2012). High rumination is associated with increased mental health vulnerabilities and exacerbation of existing psychopathology (Nolen-Hoeksema et al., 2008; Watkins & Roberts, 2020). Rumination has been consistently linked with depressive symptomology (Nolen-Hoeksema et al., 2008; Wolitzky-Taylor et al., 2021) and is also linked to various other mental health consequences including problematic alcohol use in both clinical (Wolitzky-Taylor et al., 2021) and community samples (Pollack et al., 2021; Watkins & Roberts, 2020).

The links between rumination and alcohol use are somewhat variable and tend to be specific to rumination content and/or mood. For example, rumination about a specific stressful life event can be predictive of same-day alcohol use, while rumination over an angering event has been associated with weekly alcohol use (Pollack et al., 2021). Thus, further understanding of the proximal factors and theoretical mechanisms implicated in the associations between rumination and alcohol use are required. With regard to alcohol, rumination may lead to attentional fixations on negative aspects of daily experiences and subsequently drive emotion-drinking associations which may be further exacerbated by alcohol use.

Alcohol Myopia Theory

Alcohol myopia theory (AMT) suggests potential links between rumination, affect, and alcohol use. Broadly, AMT posits that alcohol use is associated with narrowed processing of environmental stimuli, including decreased attention towards distal cues and increased appraisal of proximal cues (Fleming et al., 2013; Steele & Josephs, 1990). Inhibition conflict or attending to cues that promote behavior instead of cues that inhibit behavior, also plays a role in information processing under alcohol use (Steele & Southwick, 1985). High levels of alcohol consumption exacerbate these attentional biases (Field et al., 2004; Huntley, 1973; Steele & Southwick, 1985). Together, these attentional biases interact with stimuli and pressures from one’s environment to guide behavior, affect, cognition, and substance use within a particular setting (Field & Cox, 2008).

This valuation of “convenient” stimuli provides one explanation for varying emotional states that accompany alcohol use. Individuals may attune to different cues depending on their context and affect while drinking, hence producing disparate alcohol-related outcomes. Further, it remains unclear whether cognitive mechanisms, such as rumination, impact affectivity before and after alcohol consumption. This relationship appears plausible, given the impact of rumination on relevant, non-relevant, and emotion-based information processing (Bernblum & Mor, 2010; Whitmer & Gotlib, 2013) and may produce distinct emotional states post-drinking through myopia-related information processing.

Putting it All Together

Given these associations, a connection between rumination and alcohol use appears plausible. This may be, in part, attributable to the role of rumination in emotional processing. Rumination has been described as an “emotional magnifier” (Watkins & Roberts, 2020), and excessive rumination has been implicated in exacerbating and prolonging negative emotional states such as sadness and anger (Verduyn & Lavrijsen, 2015; Watkins & Roberts, 2020). Rumination also increases negative emotional reactivity while slowing emotional recovery (Verduyn & Lavrijsen, 2015), which may increase feelings of distress. Given the distress experienced by those high in rumination, it has been suggested that alcohol use may be an attempt to escape from or cope with emotional distress related to rumination (Sorid et al., 2021a). However, the interconnectivity among these relationships has yet to be fully explored.

Current Study

The present study aimed to examine changes in mood across the day as a function of rumination and drinking in a sample of moderating drinking college students. It was hypothesized that there would be a significant difference in emotional functioning on drinking days vs. non-drinking days and that this would vary by rumination. Previous research has examined possible associations between rumination, problematic drinking and depressive symptoms in the general population (Caselli et al., 2008), across countries (Bravo et al., 2018), and between genders (Nolen-Hoeksema & Harrell, 2002). Thus, while several studies have focused on negative affect in relation to drinking, there is a lack of research on the associations between drinking and positive affect and on the comparison of emotional regulation across the day on drinking vs. non-drinking days. To date, research that does focus on positive affect and problematic drinking has not included rumination as a factor in their analysis. Nor has any research examined the role of rumination on indices of negative affect and alcohol use and daily level. The current study aims to close the gap that is present in the literature by using EMA methodology to understand how emotional states and rumination might interact to impact alcohol use.

Methods

Participants

The sample (n = 85) was recruited from a Midwest University in the United States, and included slightly more women (n = 54, 63.53%) than men (n = 31, 36.47%). The average age was 20.04 years (SD = 1.66). The sample was predominantly White (94%), and Black (2.35%), Multiracial (2.35%), and Native American/Alaskan Native (1.18%). One person endorsed Hispanic/Latinx ethnicity. All participants were treated in accordance with APA ethical guidelines (Sales & Folkman, 2000).

Procedure

Data was collected from fall 2014 to spring 2016. Potential participants (n = 1,226) completed an online screening to assess alcohol use and aspects of emotion regulation including rumination. Those who endorsed alcohol consumption over the last two weeks were eligible to participate in the EMA phase of the study (n = 85). During the EMA phase, participants carried a personal data device (PDD) for up to 21 days. During this time, participants responded to three different assessments: (1) random: occurring up to eight times per day which asked about current mood and alcohol use, (2) self-initiated drinking: marked the time a person began drinking, (3) morning assessments: drinks consumed the prior night. The PDD could be set to ‘Do Not Disturb’ during times participants could not respond to surveys (e.g., class, bedtime, etc.). Participants were compensated at a rate of $0.25/random assessment and $1.00/morning assessment. All data and analysis code are available via an open data repository hosted at Open Science Framework: https://osf.io/j7xd8/

Between-subject Measures

Demographic variables included age, race, ethnicity, and biological sex. Both age and biological sex were added as covariates to control for potential differences in emotional functioning across these variables.

Alcohol Use at baseline was assessed using the Daily Drinking Questionnaire (DDQ) to assess quantity and drinking frequency item. The frequency item asked: “How often do you consume alcohol?” (0 – never, 1 – less than once a month, 2 – two to three times a month, 3 – once or twice a week, 4 – three to four times a week, 5 – every day). The DDQ asks participants to rate the number of drinks they typically consume for each day of the week over the last 3 months. Participants report a number for each day and this is summed to give typical weekly consumption (Dimeff et al., 1999).

Rumination was assessed by the Brooding subscale of the Ruminative Response Scale (RRS; Nolen-Hoeksema & Morrow, 1991). The RRS has three subscales (Brooding, Reflection, and Depression). The current analysis used only the Brooding subscale. This subscale focuses on the negative and comparative components of ruminative cognition. In contrast, the reflection subscale is more introspective, and solution focused. The Brooding subscale includes 5 items rated on a 1 (never) to 4 (always) Likert-type scale. (e.g., I think “What am I doing to deserve this?”). The five items had good internal consistency (α = .86).

Within-subject Measures

Emotional Functioning was assessed by 18 items from subscales of the PANAS-X (Watson & Clark, 1994) and the mood circumplex model (Russel, 1980). Items had the root: How [insert mood] have you been feeling for the last 30 minutes?” with a responses scale of 1 (not at all) to 7 (extremely). Four facets of mood were selected. Three negative mood states - anxiety (anxious, nervous, jittery; α = .74), anger (angry, hostile, irritated, α = .77), and sadness (down, blue, sad; α = .87). Five positive mood states (excited, enthusiastic, energetic, happy, joyful; α = .93) were used to form the positive mood indicator. This approach has been used in previous EMA studies of mood-alcohol associations (Dvorak et al., 2016, 2018).

Current Alcohol Use was assessed using two approaches. First, participants reported current drinking during random assessments. They were asked “have you drank since your last assessment?” The first time they answered ‘yes’ was time stamped and used to mark drinking initiation for the day (i.e., to separate mood assessments before and after drinking). Second, participants reported on morning assessments if they had consumed alcohol the night before. If they had consumed alcohol, but had not indicated this in prior random assessments the day prior the day was marked as a drinking day, and the “average” drinking initiation time from confirmed drinking days was used to separate days into pre- and post-drinking. This approach for identifying drinking days and times has been used in previous EMA studies (De Leon et al., 2020; Dvorak et al., 2014).

Data Preparation

The overall sample consisted of 85 participants who completed up to 21 days of monitoring. To identify the start time of drinking episodes, we used two approaches. First, we identified the first EMA assessment in which participants reported they started drinking. We then identified the last non-drinking assessment and split the time between these two assessments. On days in which a participant reported drinking in the morning assessment but did not report it in the EMA assessments the prior day, we used their average start drinking time from confirmed drinking times. We then used this time (average start drinking time) to split non-drinking days into typical drinking and non-drinking times to compare mood trajectories across the day on drinking and non-drinking days. There were 10,939 momentary assessments in the dataset. To maintain proximity to drinking events, we limited analyses to assessments that were up to 10 hours prior to drinking and 6 hours after drinking resulting in a total of 9,984 momentary assessments for analysis.

Analysis Plan

Multilevel modeling was conducted using the mixed command in Stata SE 16 (Stata Corp, 2019). A series of 3-level (moments nested in days nested in persons) mixed-effects models were specified for each emotion (i.e., positive mood, sadness, anxiety, and anger) to examine within-subject mood trajectories across drinking and non-drinking days. Each model examined linear, quadratic, and cubic effects of time across the day, with time centered at (a) the time the person reported drinking (e.g., the time split between the first drinking assessment and the last nondrinking assessment) or (b) the average drinking time if option ‘a’ was not available (see data preparation). We initially examined random variance components for the model intercept and each of the time slopes, however, only the intercept and linear time slopes (across models) were significant – these were allowed to vary randomly. We built each model in a stepwise approach. Model 1 examines linear slopes with interactions of day type and rumination. Model 2 adds a quadratic slope with interactions of day type and rumination. Model 3 adds the cubic slope with interactions of day type and rumination. To test the primary hypotheses, we examined trait-level rumination as a moderator of the within-subject mood trajectories across day types. Interactions were examined by calculating simple slopes of trajectories across the two-day types (drinking vs non-drinking) and high (+1SD), mean, and low (−1SD) levels of rumination. Age, biological sex, and six dummy-coded day of week indicators were added to the model as covariates.

Each mood model took the following form:

Level 1:

$${\text{Mood}}_{it}={\pi }_{0i}+{\pi }_{1i}\left({\text{Time}}_{it}\right)+{\pi }_{2i}\left({{\text{Time}}^2}_{it}\right)+{\pi }_{3i}\left({{\text{Time}}^3}_{it}\right)+{\pi }_{4i}\left({\text{Drinking Day}}_{it}\right)+{\pi }_{5i}\left({\text{Time}\times \text{Drinking Day}}_{it}\right)+{\pi }_{6i}\left({\text{Time}}^2\times {\text{Drinking Day}}_{it}\right)+{\pi }_{6i}\left({\text{Time}}^3\times {\text{Drinking Day}}_{it}\right)+e_{it}$$

Level 2 (note: x = parameter tied to each level 1 effect):

$${\pi }_{\text{x}i}=B_{\text{x}0}+B_{\text{x}1}(\text{Rumination})+u_{0i}$$

Results

Descriptive Statistics

At baseline 16.47% participants reported drinking 2–3 times/month, 62.35% reported drinking once or twice a week, and 21.18 reported drinking 3–4 times/week. Participants reported typically drinking an average of 20.17 (SD = 9.16) drinks per week over the past three months. Men (M = 25.77, SD = 11.28) endorsed higher typically weekly drinking quantity than women (M = 16.94, SD = 5.64), t(83) = 4.81, p < .001, Cohen’s d = 0.96. Bivariate correlations of between-subject data is in Table 1.

Table 1.

Descriptive and Bivariate statistics

	1.	2.	3.	4.	5.	6.
1. Days Complete	----
2. Positive Mood	.04	----
3. Angry Mood	−.14	−.29*	----
4. Sad Mood	−.12	−.28*	.74*	----
5. Anxious Mood	−.09	−.20	.60*	.64*	----
6. Rumination	−.15	−0.13	.25*	.15	.13	----
Mean	20.31	3.52	1.42	1.42	1.82	9.98
Standard Deviation	2.76	1.07	0.38	0.43	0.61	2.94
Upper Range	22	6.75	2.71	2.87	3.60	18
Lower Range	5	1.56	1.00	1.01	1.03	5
Skew	−4.50	0.94	1.28	1.58	0.95	0.42

Note.

p < .05*

Compliance Statistics

There were a possible 1,785 days (85 participants 21 day); however, there were missing days across participants resulting in a total of 1,726 days of data (96.70% of possible days). Participants reported drinking on 513 (29.72%) days. Participants reported drinking an average of 7.21 (SD = 6.29) standard drinks on drinking nights. There were a possible 13,808 momentary assessments (1,726 days × 8 assessments/day); however, there were only 10,939 momentary assessments available for analysis (compliance = 90.54%). As noted above, assessments occurring 10 hours prior to, or 6 hours following, alcohol initiation were excluded from analyses resulting in 9,984 observations.

Emotional Functioning and Rumination across Drinking and Non-drinking Days

Positive Mood.

For state positive mood there was a positive effect of day type indicating that positive mood was higher at typical drinking initiation on drinking days relative to non-drinking days. On non-drinking days there was a negative linear slope across the day which was qualified by negative quadratic and cubic effects. All three of these slopes (linear, quadratic, and cubic) differed across drinking days (see Table 2). The linear effect on drinking days was positive (B = 0.065, p < .001) and this effect was qualified by negative quadratic (B = −0.021, p < .001) and cubic (B = −0.002, p < .001) effects. None of these slopes varied by level of brooding (ps = .067 to .839). However, there was a significant linear slope x brooding x day interaction. On non-drinking days, the linear slope varied by brooding. At high levels of brooding, there was a robust negative linear slope (B = −0.062, p < .001) while at low levels of brooding this slope was attenuated, but still significant (B = −0.028, p = .001). On drinking days, none of the slopes varied by brooding (ps = .292 to .466). This resulted in slopes with a general negative trend on non-drinking days, that were steeper for those higher in brooding; and slopes with a positive trend that peaks around drinking initiation and then declines post-drinking on drinking days (see Figure 1a).

Table 2.

Multilevel Model of Rumination and Positive Mood Across the Day on Drinking and Non-Drinking Days

Predictor Variables	Model 1	Model 2	Model 3
	B (SE)	B (SE)	B (SE)
Fixed Effects
Intercept	3.292* (0.119)	3.392* (0.119)	3.417* (0.119)
Time	−0.019* (0.004)	−0.053* (0.005)	−0.045* (0.006)
Time2	-	−0.008* (0.001)	−0.010* (0.001)
Time3	-	-	−0.001* (0.000)
Drinking Day	0.567* (0.045)	0.545* (0.048)	0.640* (0.051)
Time×Drinking Day	0.084* (0.007)	0.081* (0.009)	0.110* (0.011)
Time2×Drinking Day	-	−0.000 (0.001)	−0.010* (0.002)
Time3×Drinking Day	-	-	−0.002* (0.000)
Rumination	−0.055 (0.038)	−0.053 (0.038)	−0.056 (0.038)
Rumination×Time	−0.003* (0.001)	−0.004* (0.002)	−0.006* (0.002)
Rumination×Time2	-	−0.000 (0.000)	0.000 (0.000)
Rumination×Time3	-	-	0.000 (0.000)
Rumination×Drinking Day	−0.000 (0.015)	−0.002 (0.016)	0.005 (0.017)
Rumination×Time×Drinking Day	0.006* (0.002)	0.007* (0.003)	0.009* (0.004)
Rumination×Time2×Drinking Day	-	0.000 (0.000)	−0.001 (0.001)
Rumination×Time3×Drinking Day	-	-	−0.000 (0.000)
Level 1 Random Effects
Intercept	0.368* (0.020)	0.372* (0.020)	0.373* (0.020)
Time	0.005* (0.001)	0.005* (0.001)	0.006* (0.001)
Level 2 Random Effects
Intercept	0.992* (0.157)	0.989* (0.156)	0.987 (0.156)
Model Fit
Wald χ2(df)	370.30 (15)	550.45 (19)	613.75 (23)
LRT with prior model	-	175.47*	60.05*

Note. Drinking Day (uncentered) was coded as: 0 = non-drinking day, 1 = drinking day; thus, time slopes in each model are the trajectory of emotion on non-drinking days (trajectories on drinking days are in text). Six dummy-coded day of week variables were included as covariates to control for day-level variation in emotion-drinking patterns.

p<.05*

Figure 1.

Emotion across the day on drinking and non-drinking days as a function of rumination.

Note. DD = Drinking Day, ND = Non-Drinking Day. Slopes with the same markers indicate significant differences at corresponding levels of rumination.

Angry Mood.

For anger (see Table 3) there was a negative effect of day type (B = −0.135, p <.001) indicating that anger was lower at the typical drinking time on drinking days relative to non-drinking days. On non-drinking days there was a positive linear slope across the day, with negative non-significant effects for both the quadratic and cubic slopes. The linear, quadratic, and cubic slopes all differed on drinking days vs. non-drinking days. The linear effect on drinking days was negative and non-significant (B = −0.006, p = .398), however positive effects were revealed for both the quadratic (B = 0.004, p= .005) and cubic (B = 0.000, p= .039) slopes. This resulted in slopes with a general negative trend on non-drinking days; and slopes with a general negative trend that plateaus directly prior to drinking initiation and then increases post-drinking on drinking days. On non-drinking days there were no significant variations in the linear, cubic, or quadratic slopes based on level of brooding (ps = .091 to .288), thus these were not probed. However, the analysis indicated significant variation in the linear, quadratic, and cubic slopes on drinking days by brooding (ps = < .001 to .012). Specifically, on drinking days at high levels of brooding the linear slope (B= −0.018, p= .053) was qualified by quadratic (B= 0.012, p= .015) and cubic (B= 0.001, p= .022) effects. In contrast, at low levels of brooding, on drinking days, there were no significant linear, quadratic, or cubic effects (ps = .060 to .320). These findings suggest that individuals with high levels of brooding experience a decline in anger across the day leading up to drinking initiation, a switch from a negative slope to a positive slope around drinking initiation, and then a rapid escalation in angry mood in the hours after drinking initiation (see Figure 1b).

Table 3.

Multilevel Model of Rumination and Angry Mood Across the Day on Drinking and Non-Drinking Days

Predictor Variables	Model 1	Model 2	Model 3
	B (SE)	B (SE)	B (SE)
Fixed Effects
Intercept	1.456* (0.050)	1.450* (0.049)	1.460* (0.050)
Time	0.005* (0.002)	0.007* (0.003)	0.010* (0.004)
Time2	-	0.000 (0.000)	−0.001 (0.001)
Time3	-	-	−0.000 (0.000)
Drinking Day	−0.091* (0.029)	−0.101* (0.031)	−0.135* (0.034)
Time×Drinking Day	−0.010* (0.005)	−0.005 (0.006)	−0.015* (0.008)
Time2×Drinking Day	-	0.001 (0.001)	0.005* (0.002)
Time3×Drinking Day	-	-	0.001* (0.000)
Rumination	0.039* (0.014)	0.040* (0.014)	0.043* (0.015)
Rumination×Time	0.002* (0.001)	0.002 (0.001)	0.002 (0.001)
Rumination×Time2	-	−0.000 (0.000)	−0.000 (0.000)
Rumination×Time3	-	-	−0.000 (0.000)
Rumination×Drinking Day	−0.004 (0.009)	−0.012 (0.010)	−0.031* (0.011)
Rumination×Time×Drinking Day	−0.004* (0.002)	−0.001 (0.002)	−0.007* (0.003)
Rumination×Time2×Drinking Day	-	0.001* (0.000)	0.003* (0.001)
Rumination×Time3×Drinking Day	-	-	0.001* (0.000)
Level 1 Random Effects
Intercept	0.121* (0.008)	0.122* (0.008)	0.122* (0.008)
Time	0.001* (0.000)	0.001* (0.000)	0.001* (0.000)
Level 2 Random Effects
Intercept	0.121* (0.021)	0.122* (0.021)	0.121* (0.021)
Model Fit
Wald χ2(df)	54.11 (15)	61.91 (19)	82.19 (23)
LRT with prior model	-	7.81	20.13*

Note. Drinking Day (uncentered) was coded as: 0 = non-drinking day, 1 = drinking day; thus, time slopes in each model are the trajectory of emotion on non-drinking days (trajectories on drinking days are in text). Six dummy-coded day of week variables were included as covariates to control for day-level variation in emotion-drinking patterns.

p<.05*

Sad Mood.

For sad mood (see Table 4) there was a negative effect of day type (B = −0.134, p< .001) indicating that sadness was lower at the typical drinking initiation time on drinking days relative to non-drinking days. On non-drinking days, there was a positive linear slope that was not qualified by quadratic or cubic effects. Only the linear slope differed across days. On drinking days there was a negative linear slope across the day (B = −0.016, p = .028). However, both the quadratic (B = 0.004, p = .057) and cubic effects (B = 0.001, p = .026) demonstrated positive effects. The quadratic slope across days was not significantly moderated by Brooding. Nevertheless, brooding moderated the day differences in the linear and cubic slopes. We examined the linear and cubic effects across the day on drinking and non-drinking days at high and low levels of brooding. On drinking days, at high brooding, there was a significant negative linear effect (B = −0.029, p = .013), while at low brooding the linear slope was not significant (B = −0.003, p = .657). In contrast on non-drinking days at high levels of brooding there is a robust positive linear effect (B = 0.023, p = .001), while at low levels of brooding the linear effect was no longer significant (B = 0.001, p = .808). Finally, on drinking days at high levels of brooding there was a positive cubic effect (B = 0.001, p = .041), while at low levels of brooding the cubic effect was no longer significant (B = 0.000, p = .675). In contrast on non-drinking days, at high levels of brooding the cubic effect was negative and non-significant (B = −0.000, p = .191), while at low levels the cubic effect was positive and approached statistical significance (B = 0.000, p = .053). This pattern of effects manifested as a decline in sad mood at approximately 7 hours prior to drinking on drinking days, then reversed at drinking initiation and escalated rapidly thereafter; this pattern was more robust both prior to and following drinking for those high in rumination (see Figure 1c). In contrast, on non-drinking days sad mood had a consistent positive trend across the day that did not vary by brooding.

Table 4.

Multilevel Model of Rumination and Sad Mood Across the Day on Drinking and Non-Drinking Days

Predictor Variables	Model 1	Model 2	Model 3
	B (SE)	B (SE)	B (SE)
Fixed Effects
Intercept	1.513* (0.055)	1.503* (0.055)	1.500* (0.056)
Time	0.010* (0.003)	0.013* (0.003)	0.012* (0.004)
Time2	-	0.001 (0.001)	0.001 (0.001)
Time3	-	-	0.000 (0.000)
Drinking Day	−0.111* (0.033)	−0.102* (0.034)	−0.134* (0.037)
Time×Drinking Day	−0.016* (0.005)	−0.019* (0.007)	−0.028* (0.008)
Time2×Drinking Day	-	−0.001 (0.001)	0.003 (0.002)
Time3×Drinking Day	-	-	0.001* (0.000)
Rumination	0.032* (0.016)	0.031* (0.016)	0.036* (0.016)
Rumination×Time	0.001 (0.001)	0.002 (0.001)	0.004* (0.001)
Rumination×Time2	-	0.000 (0.000)	−0.000 (0.000)
Rumination×Time3	-	-	−0.000* (0.000)
Rumination×Drinking Day	−0.002 (0.011)	−0.001 (0.011)	−0.015 (0.012)
Rumination×Time×Drinking Day	−0.003* (0.002)	−0.004 (0.002)	−0.008* (0.003)
Rumination×Time2×Drinking Day	-	−0.000 (0.000)	0.002* (0.001)
Rumination×Time3×Drinking Day	-	-	0.001* (0.000)
Level 1 Random Effects
Intercept	0.188* (0.010)	0.188* (0.010)	0.188* (0.010)
Time	0.002* (0.000)	0.002* (0.000)	0.002* (0.000)
Level 2 Random Effects
Intercept	0.149* (0.026)	0.149* (0.026)	0.149* (0.026)
Model Fit
Wald χ2(df)	57.89 (15)	60.61 (19)	76.46 (23)
LRT with prior model	-	2.68	15.80*

Note. Drinking Day (uncentered) was coded as: 0 = non-drinking day, 1 = drinking day; thus, time slopes in each model are the trajectory of emotion on non-drinking days (trajectories on drinking days are in text). Six dummy-coded day of week variables were included as covariates to control for day-level variation in emotion-drinking patterns.

p<.05*

p<.01**

Anxious Mood.

For anxious mood (see Table 5) there was a negative effect of day type (B = −0.096, p = .018) indicating that anxious mood was lower on drinking days relative to non-drinking days. The linear effect was different on drinking days versus non-drinking days (B = −0.032, p < .001). On drinking days, the linear effect was negative and significant (B = −0.039, p <.001). However, on non-drinking days the linear effect was not significant. None of the slopes across day types varied by brooding (ps = .052 to .721). Thus, anxious mood is both lower on drinking days and declines across the day leading up to drinking (see Figure 1d).

Table 5.

Multilevel Model of Rumination and Anxious Mood Across the Day on Drinking and Non-Drinking Days

Predictor Variables	Model 1	Model 2	Model 3
	B (SE)	B (SE)	B (SE)
Fixed Effects
Intercept	1.762* (0.074)	1.761* (0.075)	1.771* (0.075)
Time	−0.011* (0.003)	−0.010* (0.004)	−0.007 (0.005)
Time2	-	0.000 (0.001)	−0.001 (0.001)
Time3	-	-	−0.000 (0.000)
Drinking Day	−0.080* (0.036)	−0.050 (0.038)	−0.096* (0.041)
Time×Drinking Day	−0.004 (0.005)	−0.017* (0.007)	−0.032* (0.009)
Time2×Drinking Day	-	−0.003* (0.001)	−0.002 (0.002)
Time3×Drinking Day	-	-	0.001* (0.000)
Rumination	0.035 (0.023)	0.032 (0.023)	0.031 (0.023)
Rumination×Time	−0.001 (0.001)	0.000 (0.001)	0.000 (0.002)
Rumination×Time2	-	0.000 (0.000)	0.000 (0.000)
Rumination×Time3	-	-	0.000 (0.000)
Rumination×Drinking Day	−0.011 (0.012)	−0.006 (0.012)	−0.012 (0.013)
Rumination×Time×Drinking Day	−0.002 (0.002)	−0.004 (0.003)	−0.006 (0.003)
Rumination×Time2×Drinking Day	-	−0.000 (0.000)	0.000 (0.001)
Rumination×Time3×Drinking Day	-	-	0.000 (0.000)
Level 1 Random Effects
Intercept	0.217* (0.012)	0.217* (0.012)	0.217* (0.012)
Time	0.002* (0.000)	0.002* (0.000)	0.002* (0.000)
Level 2 Random Effects
Intercept	0.342* (0.056)	0.342* (0.056)	0.342* (0.055)
Model Fit
Wald χ2(df)	72.79 (15)	84.36 (19)	94.34 (23)
LRT with prior model	-	11.51*	9.91*

Note. Drinking Day (uncentered) was coded as: 0 = non-drinking day, 1 = drinking day; thus, time slopes in each model are the trajectory of emotion on non-drinking days (trajectories on drinking days are in text). Six dummy-coded day of week variables were included as covariates to control for day-level variation in emotion-drinking patterns.

p<.05*

p<.01**

Discussion

This study examined the influence of rumination on positive and negative affective states across drinking and non-drinking days. While previous research has revealed an association between rumination and problematic drinking, little is known regarding the impact of these constructs on positive and negative moods. Our results suggest that on drinking days, the trajectories of negative moods across the day vary considerably by rumination. The same was not true for positive mood. These trajectories are discussed by mood type below.

Positive mood was higher and increased throughout the day on drinking days and decreased throughout the day on non-drinking days. This is consistent with prior literature (Dvorak et al., 2018). Rumination did not influence these trajectories. While this finding may seem counterintuitive, as rumination should amplify negative emotion (i.e., low positive mood) (Verduyn & Lavrijsen, 2015; Watkins & Roberts, 2020), there is evidence to suggest positive and negative mood should be conceptualized as separable dimensions rather than opposite poles on one continuum (Larsen & Diener, 1987; Watson & Clark, 1994). Consequently, although rumination potentiated negative moods, positive mood may be more affected by perceptions around alcohol consumption and mood enhancement (e.g., expectancies that drinking leads to outcomes that amplify positive mood; Baker et al., 2004). Elevated positive mood around the time of consumption (i.e., evenings) and a dip in positive mood on non-drinking days around that same time, may be more related positive alcohol expectancies, which could lead to more alcohol consumption over time (Jones et al., 2001). This remains a question for future research.

Sadness was observed to be higher on non-drinking days, with increases in sadness throughout the day compared to drinking days. This is consistent with research showing elevated sadness is associated with non-drinking days (O’Hara et al., 2014). Consequently, those with higher levels of rumination experienced greater sadness that increased throughout the day on non-drinking days. There was a different story on drinking days, where sadness was consistently low across the day for those with low rumination. In contrast, among those with high rumination sadness spiked several hours out from drinking initiation, dropped precipitously as drinking initiation approached, and then spiked again after drinking had commenced. This pattern seems to involve two distinct processes. First, those high in rumination seem to experience elevated sadness well before drinking. However, it remains unclear whether this drives future drinking. Examining alcohol use motivation as a function of sadness in the hours before drinking would help to better elucidate this finding. Nonetheless, sadness decreased as the drinking event approached, which may be indicative of a tension-reduction expectancy effect. Next, sadness increased rapidly after drinking initiation among those high in rumination. This is consistent with Alcohol Myopia Theory (AMT) in which attentional focus shifts toward salient emotional cues as individuals begin to consume alcohol. Future research should fully explore these two patterns.

Anger was lower on drinking days relative to non-drinking days. Further, among individuals with high levels of rumination, anger appeared to increase throughout the day on non-drinking days. Similar to sadness, anger varied on drinking days as a function of rumination. While the declines leading up to drinking were not as robust as those observed for sadness, the spike observed post drinking was notable. This, too, is consistent with AMT, and suggests that the role of AMT may be even more prominent for this high valence negative mood. Recent research has suggested that distraction may be important for regulation of anger (Rutherford et al., 2023). However, individuals high in rumination may experience more difficulty with cognitive distraction (Nolen-Hoeksema & Morrow, 1993). This is complicated by alcohol, which limits attention to the most salient aspects of our environment, potentially limiting one’s ability to disengage from stimuli and/or distract themselves (Steele & Joesphs, 1990). Thus, individuals higher in rumination may experience spikes in anger after drinking, as their attentional resources become limited in scope and their ability to distract themselves from anger-provoking stimuli are diminished.

Anxiety was lower on drinking days relative to non-drinking days with no significant effect of rumination. The trajectory of anxiety was also consistent across both drinking and non-drinking days. These findings are consistent with prior studies demonstrating lower levels of negative affect on drinking days compared to non-drinking days (De Leon et al., 2020), including anxiety. Further, findings are consistent with research specifically identifying lower levels of anxiety on drinking days versus non-drinking days (Dvorak et al., 2018). Notably, rumination did not significantly impact levels of anxiety or trajectories of anxiety on drinking days compared to non-drinking days. This finding contributes to a more nuanced understanding of the role of rumination in the relationship between negative affect and alcohol use and suggests that rumination may not be a clinically significant treatment target among individuals who drink specifically to regulate anxiety.

Limitations

Several limitations should be considered. First, these analyses used a small sample (n = 85) of primarily White (e.g., 94%) college students. It is possible that the findings may not be generalizable to a more ethnically or racially diverse population. Though, it is worth noting that effects were across the day, with nearly 10,000 momentary assessments in the analyses. Additionally, this was not a clinical sample. Thus, it remains possible that the associations identified may not be observed in samples of drinkers who meet clinical thresholds. Future research should attempt to examine these associations in clinical samples. Along these lines, the sample was comprised of college students. In this population, drinking is mostly frequently characterized by socialization. The trajectories we observed, both for positive mood and for the negative moods, could be a function of (a) prosocial expectancies about drinking or (b) disappoint related to lack of social contact (which occurs through drinking); a point we have made in prior research (see Dvorak et al., 2018). Future research should seek to parse out emotion-driven vs socially facilitated drinking occasions as a way to understand the role of emotional trajectories. Further, data collection for the present study spanned an average of 21 days. Longitudinal approaches examining these constructs may be warranted to better understand how these associations may function over time. Similarly, a participant’s drinking episode was identified using a proxy, based on the average time between drinking and non-drinking assessments. Although care was taken to maintain proximity to drinking events, it is possible that this did not completely or accurately identify drinking vs. non-drinking episodes. Future research should use more precise drinking initiation time points to better elucidate associations between aspects of emotional functioning and alcohol consumption. Finally, it is important to note that rumination was solely operationalized within the present study using the brooding subscale of the RRS (Nolen-Hoeksema & Morrow, 1991). This was chosen given research suggesting that this aspect of rumination is most linked to negative emotional functioning (Ciesla & Roberts, 2007; Sorid et al., 2021b), and thus may be uniquely associated with problematic alcohol consumption, however, other aspects of rumination may also play a role.

Conclusions

The findings of the current study may support differences in reinforcement-related emotional functioning on drinking days compared to non-drinking days, particularly for individuals who experience high levels of rumination. Further, these findings highlight the importance of examination of constructs such as rumination in the context of overall affect and alcohol use. Specifically, the experience of high levels of rumination may elevate one’s risk for greater alcohol use to mitigate feelings of distress. However, this was also characterized by escalation in negative mood following consumption in a manner consistent with alcohol myopia. Overall, the current study provides valuable insights into the impacts of rumination on the associations between mood and consumption of alcohol across the day. Future research should seek to elucidate why negative mood drops in the hours prior to use and escalates shortly thereafter to better understand the mechanisms (expectancies, social facilitation, attentional narrowing) that produce these associations in those high in rumination.

Highlights.

• Rumination affects negative, but not positive, mood slopes on drinking days

• High rumination is associated with declines in negative mood prior to drinking

• High rumination is associated with spikes in negative mood after drinking