Emotion differentiation among individuals in a randomized clinical trial for alcohol use disorder: Within- and between-person associations with affect, craving, and alcohol use in daily life

Abstract

Emotion differentiation refers to cognitively distinguishing among discrete, same-valenced emotions. Negative emotion differentiation (NED) is a transdiagnostic indicator of emotional functioning. The role of positive emotion differentiation (PED) in clinical disorders, including alcohol use disorder (AUD), is less understood. Further, despite consensus that emotions are highly variable, little is known about within-person fluctuations in NED/PED. The current study leveraged 84 consecutive daily smartphone surveys from participants (N=181) in a clinical trial of cognitive behavioral therapy for AUD to investigate whether between-person differences in overall NED/PED, or within-person variability in daily NED/PED, were associated with affect intensity, craving, drinking, and heavy drinking in daily life. Subsequent analyses explored whether associations were moderated by baseline alexithymia. At the between-persons level, greater average PED, but not NED, was associated with lower heavy drinking odds. At the within-persons level, higher-than-usual PED was associated with lower negative affect and odds of any drinking. Individuals with baseline alexithymia had stronger negative within-person associations between daily NED and both any and heavy drinking. PED is a skill linked to less alcohol use between- and within-persons irrespective of baseline alexithymia, whereas greater daily NED appears especially important for reduced alcohol use among individuals with co-morbid AUD and alexithymia.

Introduction

In 2021, 29.5 million people in the U.S. met criteria for an alcohol use disorder (AUD; SAMHSA, 2022). Affective states are strongly implicated in the development and maintenance of AUD (Baker et al., 2004). Many psychosocial AUD treatments help patients regulate affective states with the goals of reducing substance use (Berking et al., 2011) and increasing skills for tolerating emotional arousal (Stasiewicz et al., 2013). Yet, the mechanisms underlying one’s ability to cope with dysregulated affect remain largely unknown, in part because most research uses single-timepoint measures that have considerable limitations for capturing the dynamics of these processes (Fisher, 2015). Because affective states are variable and situational, researchers and clinicians have started capitalizing on experience sampling methods to understand the role of dynamic indices of affect in treatment processes and outcomes (Fisher et al., 2019). One indicator of within-person affect dynamics with potential as a mechanism of change is emotion differentiation (ED; Kashdan et al., 2015).

The current study used daily diary data to examine negative emotion differentiation (NED) and positive emotion differentiation (PED) among individuals in outpatient AUD treatment. We investigated whether between-person differences in overall NED/PED, and within-person variability in daily NED/PED, were associated with affect intensity, craving, alcohol use, and heavy drinking in daily life. We subsequently explored whether between-person differences in baseline alexithymia moderated within-person associations among NED/PED and outcomes during treatment.

Emotion differentiation (ED)

Emotion differentiation, also referred to as “emotional granularity,” is conceptualized as the degree to which individuals distinguish among discrete, same-valenced emotional states (Barrett et al., 2001). For example, an individual who distinguishes to a greater extent on average among negative emotional states such as “angry,” “frustrated,” and “irritable” (i.e., rates these states as relatively less similar to one another over time) has higher overall negative emotion differentiation, compared to an individual who makes less fine-grained distinctions among these experiences.

Low ED, especially for negative emotions, is a potential transdiagnostic risk factor. A recent meta-analysis found that lower NED was associated with substance use, impulsive aggression, binge eating, and non-suicidal self-injury (Seah & Coifman, 2022). Because feelings are theorized to provide information about how to act in a given situation (Schwarz, 2012), greater NED may provide more nuanced information about the emotional context and a clear target for emotion regulation. Lower NED, by contrast, provides less specific information about the emotional situation, and therefore less guidance for selecting regulation strategies. Supporting this conceptualization, “low differentiators” tend to exhibit lower self-regulation (Walters & Simons, 2022) and an overreliance on avoidance and disengagement regulation strategies (Brown et al., 2021; Pugach et al., 2023). “High differentiators” tend to use more emotion regulation strategies (Barrett et al., 2001), engage in more effective down-regulation of negative emotions (Kalokerinos et al., 2019), report more self-acceptance and less neuroticism (Grühn et al., 2013), and exhibit greater calmness (Lischetzke et al., 2021) and fewer internalizing symptoms (Nook et al., 2021) following exposure to stressful events. Neuroscience evidence also links NED with activity in areas of the brain involved in self-regulation and executive control (e.g., Kashdan et al., 2014). Thus, the empirical evidence supports theories purporting that NED facilitates effective emotion regulation.

Although extant research has primarily linked NED to adaptive outcomes, there is reason to hypothesize that PED may also be beneficial. The “broaden and build” theory (Fredrickson, 2001) suggests that positive emotions expand one’s momentary cognitive-behavioral repertoire to foster adaptive self-regulation. Accordingly, PED may expand cognitive capacities, such as savoring (Thompson et al., 2021), which in turn could promote adaptive behaviors (e.g., positive coping). Indeed, PED has been associated with personal growth and self-complexity (Grühn et al., 2013) and more effective coping, including less self-distraction under stress and less automatic responding (Tugade et al., 2004). Other findings indicate PED is at least as important as NED and perhaps more important in certain populations. PED, but not NED, was associated with behavioral adaptation among individuals high in borderline personality pathology (Dixon-Gordon et al., 2014) and within clinical populations more broadly; the reverse was observed for nonclinical populations (O’Toole et al., 2020). More deliberate attention to specific positive emotions may motivate adaptive behaviors associated with that specific emotion, especially among clinical populations who may have fewer positive experiences.

Emotion differentiation and substance use in daily life

Although ED is associated with various clinical disorders, fewer studies have examined ED in the context of substance use generally or AUD specifically. Important exceptions have shown that individuals with lower NED report a stronger association between affect and craving during a nicotine quit attempt (Walters et al., 2023) and are more likely to be heavy (compared to light) smokers (Sheets et al., 2015). Individuals with lower NED also exhibit stronger associations between alcohol use and aggression (Maloney et al., 2023) and stress (Kashdan et al., 2010), increased likelihood of lapse following SUD treatment (Anand et al., 2017), and greater alcohol use during early AUD recovery (Emery et al., 2022). Yet, there remains little research on the relationships of NED, as well as PED, to AUD treatment processes and outcomes.

Moreover, regardless of population, most studies have operationalized ED as a trait-like characteristic despite strong theoretical consensus that emotions are highly variable and situational (Frijda, 1986). Indeed, ED fluctuates over time within individuals (Erbas et al., 2018), and is purportedly adaptive because it can provide unique contextual information that points to appropriate regulation strategies in the moment (Kashdan et al., 2015). Yet, ED is typically computed from dynamic data as a person-level index. Resulting between-person findings cannot be generalized to within-person processes (Molenaar, 2004).

Researchers have increasingly acknowledged that ED varies within-persons, but also that patterns of associations may differ between- versus within-persons. Recent evidence indicates that NED and PED are related to affect intensity, stress, and impulsivity within, but not between, persons (Erbas et al., 2022; Tomko et al., 2015). Individuals were more likely to drink alcohol and less likely to smoke on days of lower-than-usual NED, but only marijuana use correlated with NED between-persons (this study did not examine PED; Lane & Trull, 2022). These findings indicate that investigations of ED should consider both between-person differences in average ED across repeated measurements as well as within-person fluctuations around this average.

Additionally, within-person fluctuations in ED could be more consequential for some individuals than others. Alexithymia, conceptualized as difficulty with understanding and describing emotions (Parker et al., 2001), is a potential moderator of these within-person processes. Related yet distinct from ED (Hoemann, Nielson, et al., 2021), alexithymia has been linked to smoking (Linn et al., 2020) and risky alcohol use via deficits in emotion regulation (Linn et al., 2021; Lyvers & Thorberg, 2023), including drinking to cope with negative moods (Lyvers et al., 2018). Therefore, the current study disaggregated the between- and within-person variance in ED, and examined whether alexithymia modified associations between ED and outcomes during treatment.

The current study

The current study addressed three aims. The first aim was to examine whether between-person differences in overall levels of NED or PED were associated with positive affect (PA) intensity, negative affect (NA) intensity, craving, and/or alcohol use. We also explored whether associations differed by treatment condition or after controlling for person-mean affect intensity. We hypothesized individuals with greater overall NED and PED would have higher PA, lower NA and craving, and lower odds of drinking and drinking heavily on average, controlling for person-mean PA/NA intensity (Dejonckheere et al., 2019). We also hypothesized that individuals receiving cognitive behavioral therapy (CBT) for AUD plus an emotion regulation treatment supplement (ERT) would demonstrate stronger associations between ED and outcomes compared to individuals receiving CBT plus a healthy lifestyles control condition (HLS).

The second aim was to examine whether within-person variability in daily NED or PED was associated with daily levels of PA/NA intensity, craving, and/or alcohol use. We also explored whether associations differed by treatment condition or after controlling for daily affect intensity and variability. We hypothesized that, on average, higher-than-usual ED would be associated with higher daily levels of PA, lower daily levels of NA and craving, and lower odds of drinking and drinking heavily that day. We hypothesized associations would be stronger for the ERT than HLS conditions.

The third aim was to examine whether between-person differences in baseline alexithymia moderated within-person associations among daily ED and outcomes, after controlling for daily affect intensity and variability. We hypothesized that individuals with (versus without) baseline alexithymia would have stronger negative associations between ED and NA, craving, and alcohol use.

Method

Participants

The current study is a secondary analysis of data drawn from an NIH-funded randomized clinical trial (R01AA024628) comparing the efficacy of an emotion regulation treatment supplement added to standard CBT for outpatient AUD treatment. Previous reports have analyzed aspects of the data collected as part of this larger project, including investigations of COVID-19’s impact on drinking (LaBarre et al., 2022) and associations between mean levels of affect and drinking (Linn et al., 2023; Zhao et al., 2023). Individuals calling in response to multi-media advertising were eligible if they: (1) met DSM-5 diagnostic criteria for moderate or severe AUD, (2) reported drinking heavily in response to negative affect situations (see Measures), (3) consumed any alcohol in the past 3 months, and (4) lived within commuting distance of the clinic. Participants were ineligible if they: (1) met DSM-5 criteria for severe mental illness (i.e., schizophrenia or bipolar disorders), (2) reported changes to prescription medication affecting mood in the past 3 months, (3) had a DSM-5 substance use disorder diagnosis other than nicotine or mild cannabis dependence, or (4) were mandated to attend treatment.

A comprehensive in-person screening/baseline assessment was completed within a week following initial eligibility screening. Of the 530 individuals initially screened, 354 (66.8%) were eligible for the in-person screening/baseline assessment. Of the 248 individuals who completed the in-person assessment, 194 (78.2%) were eligible to participate and were enrolled in the study. Twelve of the 194 individuals never attended the initial treatment session and one additional individual attended one treatment session but never initiated daily diary reporting and subsequently withdrew from the study, leaving a final sample size of N=181.

Procedure

Study procedures were approved by the University at Buffalo Institutional Review Board. Prior to participation, study participants completed the informed consent process and signed consent forms. Participants received up to 12 sessions consisting of 45 minutes of standard CBT for AUD adapted from Project MATCH (Kadden et al., 1995). In addition, participants were randomized to receive an additional 45 minutes of either an emotion regulation treatment (CBT + ERT) supplement or a health and lifestyle (CBT + HLS) treatment supplement intended as a time, attention, and expectation of benefit control (Stasiewicz et al., 2013). The ERT treatment supplement included adaptive emotion regulation strategies designed to help individuals develop the capacity to regulate negative affect in adaptive ways. The HLS treatment supplement was an active control that provided education about various health-related topics. Treatment assignment was based on the urn randomization procedure (Wei, 1978), which balanced participant gender, AUD severity, presence of a comorbid mood or anxiety disorder, and trait mindfulness. Treatment sessions were conducted by licensed, trained and supervised therapists. Prior to the COVID-19 pandemic, all assessments and treatment sessions were conducted in-person at a university-funded outpatient clinic. From April 2020 through the end of recruitment and follow-up, when the target sample size was reached, all sessions took place over video conferencing to comply with COVID-19 social distancing measures and public safety precautions.

After completing their first treatment session, individuals received training on the daily diary protocol. Each morning at 6:00 a.m. for 84 consecutive days during the 12-week treatment period, participants received a secure link via text or email to an online daily monitoring form. Participants without access to a smartphone or computer were provided with a restricted smartphone. Daily assessments asked participants to provide information about their prior day’s PA and NA, alcohol craving, and number of standard drinks consumed. All participants received a reminder text at noon to complete their daily reports. Mean survey completion time for the analytic sample was 10:03 a.m. (Mdn. = 9:40 a.m.) and ranged from 6:00 a.m. – 9:16 p.m. Approximately 32% of surveys were completed after 12:00 p.m. Patterns of results did not change when controlling for whether surveys were completed before or after noon. Participants earned $1.00 for each daily report completed, with bonuses for completing 6+ reports in a week ($10) and 6+ reports per week for the entire 12-week diary period ($50). Maximum compensation was $254.

Measures

Baseline assessment

Demographics.

Demographic information was collected at the phone screen and baseline assessment. The sample was 51.4% female and 93.5% Caucasian with an age range of 26-68 years (M_age = 50.76, SD_age = 10.68). More than half of the sample were married or living with a partner (51.6%), had a university or post-graduate degree (55.7%), and had personal income above $50,000 (62.1%).

Inventory of Drug-Taking Situations-Alcohol version (IDTS-A).

The IDTS-A is a 50-item measure assessing situations in which individuals report drinking heavily over the past year (Annis & Graham, 1995) from 1 (never) to 4 (almost always). Participants who scored above a 2 on either the “unpleasant emotions” or “conflict with others” subscales of the IDTS-A met study inclusion criteria for negative affect drinking.

Toronto Alexithymia Scale (TAS-20).

The TAS-20 (Bagby et al., 1994) is a 20-item measure of alexithymia consisting of three subscales: Difficulty Identifying Feelings, Difficulty Describing Feelings, and Externally Oriented Thinking. Items are rated from 1 (strongly disagree) to 5 (strongly agree) and summed to form a total score. A dichotomous variable was created indicating that baseline alexithymia was either absent (scores of 60 or lower) or present (scores of 61 or greater).

Daily diary

Positive and negative emotion.

Emotions were measured using the Positive and Negative Affect Scale (Watson et al., 1988), which consists of 10 negative and 10 positive emotion items assessed from 1 (very slightly or not at all) to 5 (extremely). The stem for each item was, “Thinking about yesterday, to what extent did you feel…”.

Alcohol craving.

Daily average alcohol craving was assessed with the item, “What was your AVERAGE level of craving or desire to drink over the course of the day?” Daily maximum alcohol craving was assessed with the item, “What was your HIGHEST level of craving or desire to drink?” The stem for both items was, “Thinking about yesterday…”. Responses to both questions were rated on a scale from 1 (none/very low craving) to 10 (extreme craving).

Alcohol use.

Daily alcohol use was assessed with two items: (1) “Did you consume any alcoholic beverages yesterday?” and (2) “Using the chart below, how many standard alcoholic beverages did you consume yesterday?” The first item had a dichotomous yes (1) / no (0) response format. The second item showed participants a standard drink chart, allowed them to enter the number of drinks consumed (from 0-75), and was treated as a count variable representing drinks per drinking day (non-drinking days were coded as missing). A dichotomized variable of heavy drinking was coded as yes (1) or no (0) (heavy drinking = > 3 standard drinks per day for women and > 4 standard drinks per day for men; National Institute on Alcohol Abuse and Alcoholism, n.d.).

Data preprocessing

Separate daily PA and NA intensity scores were calculated for each participant for each day as the mean of the 10 positive and negative emotion responses, respectively. Person-average PA/NA intensity was calculated as the mean of daily PA/NA intensity across all days for each participant. Separate within-day PA and NA variability scores were calculated for each participant for each day as the standard deviation (SD) of the 10 positive and negative emotion responses, respectively. Person-average PA/NA variability was calculated as the mean of within-day PA/NA variability across all days for each participant.

The function “calculate_ed” from the R package “emodiff” (Erbas et al., 2022) was applied to calculate NED and PED scores for each day and each person. These scores are derived from the intraclass correlation coefficient (ICC), which is the classical ED index used in most between-persons research (e.g., Kashdan et al., 2010). The function’s calculation of the daily index, separately for positive and negative emotions, involved taking the mean of the (centered) emotion items for each day, multiplied by the number of emotions and squared, and divided by the sum of the variances of all (centered) emotions (see Erbas et al., 2022). Strong deviations in the same direction from the mean for most emotions on a given day—e.g., all negative emotions rated as more negative than usual—resulted in higher scores, indicating greater undifferentiation. When emotions were experienced at their mean levels or when deviations were in opposite directions from the mean on a given day—e.g., lower-than-usual level of distress, usual level of upset, and higher-than-usual level of irritable—the resulting scores were lower, indicating lesser undifferentiation. As a final step, scores were Fisher r-to-z transformed and multiplied by −1. Daily NED and PED were more negative when differentiation was low, and approached zero when differentiation was high.

Person-mean variables were grand-mean centered, and within-person daily variables were person-mean centered. All variables were standardized for analysis. Individuals without any within-person variability in emotion across the daily diary period (N=1) or with negative ICCs for either positive or negative emotion (N=2) were excluded from further analysis (e.g., Kalokerinos et al., 2019), leaving a final person-level analytic sample of N=178.

Data analysis

There were six outcome variables for each set of mixed-effects models: affect intensity, average craving, maximum craving, any drinking (AD), heavy drinking (HD), and drinks per drinking day (DDD). All mixed-effects models were fit using the lme4 package in R (Bates et al., 2015) and p-values were calculated using lmerTest (Kuznetsova et al., 2017).

Aim 1: Examine between-person associations among overall NED/PED and affect intensity, craving, and alcohol use

Twelve separate mixed-effects models were estimated, six with person’s NED (PNED) as the focal predictor of each outcome and six with person’s PED (PPED). For example, the model regressing PA on PNED was:

$${DPA}_{di}={\gamma }_{00}+{\gamma }_{01}({PNED}_i)+u_{0i}+{\epsilon }_{di}$$ (1)

${DPA}_{di}$ (day’s PA) was the PA reported on day $d$ by individual $i$, ${\gamma }_{00}$ was the intercept representing the PA expected when PNED equaled the sample-average value ($u_{0i}$ captured person-to-person variability in this term), ${\gamma }_{01}$ indicated the expected difference in PA for each unit difference in PNED from the sample mean, and ${\epsilon }_{di}$ were errors.

Aim 2: Examine within-person associations among daily NED/PED and daily levels of affect intensity, craving, and alcohol use

Twelve separate mixed-effects models were estimated, six with day’s NED (DNED) as the focal predictor of each outcome and six with day’s PED (DPED). For example, the model regressing PA on DNED was:

$${DPA}_{di}={\gamma }_{00}+{\gamma }_{10}({DNED}_{di})+u_{0i}+u_{1i}({DNED}_{di})+{\epsilon }_{di}$$ (2)

${DPA}_{di}$ was the PA reported on day $d$ by individual $i$, ${\gamma }_{00}$ was the intercept representing the PA predicted when DNED equaled the person’s mean ($u_{0i}$ captured person-to-person variability in this term), ${\gamma }_{10}$ indicated the expected difference in PA for each unit deviation in DNED from the person’s mean ($u_{1i}$ was the random slope capturing the person-to-person variability in this term), and ${\epsilon }_{di}$ were errors.

Aim 3: Examine between-person differences in baseline alexithymia as a moderator of within-person associations among daily ED and affect intensity, craving, and alcohol use

Twelve separate mixed-effects models were estimated, six with interactions between DNED and alexithymia for each outcome and six with interactions between DPED and alexithymia. For example, the model regressing PA on DNED, alexithymia, and their interaction was:

$${DPA}_{di}={\gamma }_{00}+{\gamma }_{01}({ALE}_i)+{\gamma }_{10}({DNED}_{di})+{\gamma }_{11}{({ALE}_i)}^{\ast }({DNED}_{di})+u_{0i}+u_{1i}({DNED}_{di})+{\epsilon }_{di}$$ (3)

This model is the same as in Equation 2, except that (1) ${\gamma }_{00}$ represented the PA predicted for individuals without alexithymia when DNED equaled the person’s mean; (2) ${\gamma }_{01}$ indicated the expected difference in PA for individuals with alexithymia (ALE), relative to individuals without alexithymia and assuming DNED equaled the person’s mean; (3) ${\gamma }_{10}$ indicated the expected difference in PA for each unit deviation in DNED from the person’s mean for individuals without alexithymia, and (4) ${\gamma }_{11}$ indicated the expected shift in PA for each unit deviation in DNED from the person’s mean for individuals with ALE, relative to individuals without ALE.

Across all three sets of analyses, mixed-effects models with AD and HD as outcomes were fit with random intercepts and binomial distributions, and models with DDD as the outcome were fit with random intercepts and truncated Poisson distributions. We adjusted for factors that might be systematically related to the dependent variables in all models, including day in the study, treatment condition (ERT vs. HLS), and the number of completed surveys per person. Interaction terms between treatment condition and ED variables were tested to examine whether condition moderated any associations. Finally, affect intensity and variability were added in subsequent models to examine the unique effects of individuals’ discrimination among discrete negative and, separately, positive emotions, independent of the intensity and variability with which emotions were experienced. Doing so acknowledged that although intensity and variability of NA and PA may be associated with craving and drinking, the granularity with which individuals identify the particular negative or positive emotion(s) that is/are activated may confer additional information about craving or drinking risk.

Sensitivity analyses

First, models were rerun on a subset of participants with non-zero within-person variability in the outcome variable across the daily diary period. Craving analyses removed N=3 participants without within-person craving variability, either because they reported experiencing none/very low craving (N=2) or extreme craving (N=1) every day. Alcohol use analyses removed N=32 participants without within-person variability in any drinking, N=36 participants without within-person variability in number of drinks per drinking day, and N=51 participants without within-person variability in heavy drinking.

Second, we examined change in daily ED across the daily diary period (Hoemann, Barrett, et al., 2021; Widdershoven et al., 2019). Person-specific regression analyses estimated the relationship between study day and day’s NED and, separately, day’s PED, for each participant. One-sample t-tests compared PED/NED slopes to zero, and Cohen’s d estimated effect sizes.

Third, we considered the implications of handling days with identical ratings across negative and/or positive emotion items in different ways (resulting in zero variance; Lane & Trull, 2022). First, we only excluded days (N=57) where participants gave a score of “1” across both negative and positive emotion items. In a second set of analyses, we excluded days with zero variability (regardless of whether the scores were all 1’s, all 2’s, etc.; N=435) across both negative and positive emotion items. Third, we excluded days with no item variability separately for negative (N=3,000) and positive (N=1,024) emotion items (see Supplementary Material).

Results

Descriptive statistics

Means, within- and between-person SDs, and ICCs for daily variables are presented in Table 1. The 178 participants included in analyses completed M=67.04 (SD=20.39, Range=3 – 84) of the 84 possible days on average, for a total of 11,934 days. One hundred and nineteen participants (67%) completed ≥ 80% of daily reports (i.e., 67 or more days) and 18 completed all 84 days. Undifferentiation among negative emotions (ICC=0.79) and positive emotions (ICC= 0.83) was relatively high on average at the between-persons level. Within- and between-person correlations among daily variables are presented in Table 2. Of note, the correlations of NA intensity (rs=0.16 – 0.56) and within-day variability (rs=0.10 – 0.50) with craving and alcohol use outcomes were larger in magnitude than those of PA intensity (rs=0.15 – 0.30) and within-day variability (rs=0.03 – 0.16) with outcomes, both between- and within-persons.

Table 1.

Descriptive statistics for study variables.

Variable	M	Within-person SD	Between-person SD	ICC
Affect variables
Negative affect	1.67	0.45	0.54	0.55
Positive affect	2.81	0.50	0.65	0.60
Daily negative emotion non-differentiation	4.00	7.43	1.37	0.01
Daily positive emotion non-differentiation	4.56	6.46	1.57	0.03
Overall negative emotion non-differentiation	0.79	NA	0.16	NA
Overall positive emotion non-differentiation	0.83	NA	0.13	NA
Craving variables
Average daily craving	3.51	1.26	1.92	0.65
Maximum daily craving	4.77	1.58	2.34	0.64
Alcohol variables
Percent drinking days	37.54	30.77	33.25	0.45
Percent heavy drinking days	23.23	24.67	28.83	0.46
Drinks per drinking day	5.24	2.01	2.95	0.58

Notes. To aid in the interpretability of means, we include the raw ICC (i.e., prior to Fisher’s z transformation) for overall negative and positive emotion (non-) differentiation. M = Mean. SD = Standard deviation. ICC = Intraclass correlation coefficient. N_persons = 178; N_days = 11934.

Table 2.

Within- and between-person Spearman correlations among study variables.

Between-person
	NA	PA	NA SD	PA SD	NEND	PEND	Avg Craving	Max Craving	ADD	HDD
PA	−0.22
NA SD	0.89	−0.18
PA SD	0.09	−0.07	0.20
NEND	0.55	−0.16	0.40	0.09
PEND	0.15	0.02	0.08	−0.17	0.40
Avg craving	0.54	−0.20	0.47	−0.15	0.21	0.06
Max craving	0.56	−0.18	0.50	−0.10	0.22	0.04	0.89
ADD	0.34	−0.26	0.31	−0.16	0.08	0.12	0.56	0.60
HDD	0.43	−0.30	0.37	−0.16	0.19	0.21	0.59	0.61	0.85
DDD	0.31	−0.25	0.22	−0.12	0.25	0.26	0.36	0.34	0.27	0.71
Within-person
PA	−0.32
NA SD	0.75	−0.26
PA SD	0.12	−0.05	0.14
NEND	0.12	−0.07	0.01	0.02
PEND	0.07	−0.01	0.06	−0.12	0.15
Avg craving	0.31	−0.17	0.26	0.08	0.07	0.03
Max craving	0.33	−0.18	0.30	0.09	0.07	0.02	0.76
ADD	0.17	−0.17	0.15	0.05	0.05	0.02	0.36	0.40
HDD	0.16	−0.15	0.10	0.03	0.05	0.04	0.30	0.31	0.56
DDD	0.18	−0.15	0.13	0.05	0.08	0.08	0.28	0.27	−0.01	0.52

Notes. NA = Negative affect. PA = Positive affect. SD = Standard deviation. NEND = Negative emotion non-differentiation. PEND = Positive emotion non-differentiation. ADD = Any drinking day. HDD = Heavy drinking day. DDD = Drinks per drinking day.

Reliability of affect and ED estimates was examined in two ways. First, a generalizability theory approach (Shrout & Lane, 2012) assessed whether between-person differences and within-person change in PA and NA were measured reliably according to the dimensions of persons, days, and items. Between-person differences (NA R_KF = 0.99; PA R_KF = 0.99) and within-person change (NA R_C = 0.87; PA R_C = 0.89) were assessed with high reliability. Second, the I² statistic was applied to estimate the reliability of overall ED (Schneider & Junghaenel, 2022). Overall NED (I² = 0.89, 95% CI = 0.87, 0.91) and PED (I² = 0.90, 95% CI = 0.88, 0.92) were assessed reliably.

We also computed correlations between NED/PED and several emotion-related baseline variables to examine if our conceptualization of ED was empirically supported via patterns of relationships in the expected directions (i.e., convergent validity; see Supplementary Material). As hypothesized, PED and (especially) NED were negatively correlated with indicators of emotion dysregulation (e.g., DERS, BDI, ASI) and positively correlated with indicators of positive emotional functioning (e.g., MAAS). Therefore, both the reliability and convergent validity of our ED indices were supported.

Aim 1: Examine between-person associations among overall NED/PED and affect intensity, craving, and alcohol use

Results from analyses addressing the first aim are presented in Tables 3 and 4. Individuals with greater NED on average reported higher PA, lower maximum craving levels, and fewer DDD on average; NED was not associated with average craving levels or odds of drinking or drinking heavily. After controlling for person-mean NA intensity, only the association between NED and DDD remained.

Table 3.

Concurrent association between positive and negative person-level emotion differentiation and daily outcomes.

	PA			Avg Craving			Max Craving			Any Drinking			DDD			Heavy Drinking
NED	Est (SE)	95% CI	p	Est (SE)	95% CI	p	Est (SE)	95% CI	p	OR (SE)	95% CI	p	IRR (SE)	95% CI	p	OR (SE)	95% CI	p
Int	0.07 (0.08)	−0.10 – 0.23	0.429	0.02 (0.09)	−0.16 – 0.19	0.834	0.04 (0.09)	−0.13 – 0.22	0.611	0.42 (0.13)	0.23 – 0.77	0.005	4.35 (0.29)	3.82 – 4.96	<0.001	0.09 (0.03)	0.05 – 0.19	<0.001
Surveys	0.02 (0.04)	−0.06 – 0.10	0.605	−0.03 (0.04)	−0.11 – 0.06	0.502	−0.02 (0.04)	−0.11 – 0.06	0.567	0.84 (0.13)	0.62 – 1.14	0.269	0.96 (0.03)	0.90 – 1.02	0.225	0.81 (0.13)	0.58 – 1.12	0.197
Day	0.05 (0.01)	0.04 – 0.06	<0.001	−0.16 (0.01)	−0.17 – −0.15	<0.001	−0.14 (0.01)	−0.15 – −0.13	<0.001	0.68 (0.02)	0.64 – 0.71	<0.001	0.97 (0.01)	0.96 – 0.99	<0.001	0.68 (0.02)	0.64 – 0.72	<0.001
Cond	−0.13 (0.11)	−0.35 – 0.10	0.264	−0.04 (0.12)	−0.28 – 0.20	0.770	−0.09 (0.12)	−0.32 – 0.15	0.473	0.60 (0.26)	0.26 – 1.39	0.231	0.95 (0.09)	0.79 – 1.13	0.552	0.55 (0.26)	0.22 – 1.39	0.207
PNED	0.15 (0.06)	0.04 – 0.27	0.010	−0.08 (0.06)	−0.21 – 0.04	0.182	−0.12 (0.06)	−0.24 – −0.00	0.048	1.03 (0.23)	0.67 – 1.58	0.894	0.85 (0.04)	0.77 – 0.93	0.001	0.66 (0.16)	0.41 – 1.08	0.096
N	178 ID			178 ID			178 ID			178 ID			155 ID			178 ID
Obs	11934			11933			11931			11934			4332			11928
R2	0.032 / 0.602			0.036 / 0.678			0.041 / 0.664			0.024 / 0.704			0.077 / 0.904			0.042 / 0.739
	NA			Avg Craving			Max Craving			Any Drinking			DDD			Heavy Drinking
PED	Est (SE)	95% CI	p	Est (SE)	95% CI	p	Est (SE)	95% CI	p	OR (SE)	95% CI	p	IRR (SE)	95% CI	p	OR (SE)	95% CI	p
Int	−0.01 (0.08)	−0.17 – 0.15	0.900	0.02 (0.09)	−0.15 – 0.20	0.801	0.05 (0.09)	−0.12 – 0.22	0.575	0.42 (0.13)	0.23 – 0.77	0.005	4.41 (0.29)	3.87 – 5.02	<0.001	0.10 (0.03)	0.05 – 0.19	<0.001
Surveys	−0.12 (0.04)	−0.20 – −0.05	0.002	−0.04 (0.04)	−0.12 – 0.04	0.337	−0.04 (0.04)	−0.13 – 0.04	0.290	0.86 (0.13)	0.65 – 1.16	0.331	0.95 (0.03)	0.89 – 1.01	0.097	0.79 (0.13)	0.57 – 1.08	0.138
Day	−0.08 (0.01)	−0.09 – −0.07	<0.001	−0.16 (0.01)	−0.17 – −0.15	<0.001	−0.14 (0.01)	−0.15 – −0.13	<0.001	0.68 (0.02)	0.64 – 0.71	<0.001	0.97 (0.01)	0.96 – 0.99	<0.001	0.68 (0.02)	0.64 – 0.72	<0.001
Cond	0.02 (0.11)	−0.20 – 0.24	0.859	−0.04 (0.12)	−0.28 – 0.20	0.724	−0.10 (0.12)	−0.33 – 0.14	0.429	0.60 (0.25)	0.26 – 1.38	0.225	0.93 (0.08)	0.77 – 1.11	0.407	0.52 (0.24)	0.21 – 1.29	0.158
PPED	−0.08 (0.06)	−0.19 – 0.03	0.138	−0.04 (0.06)	−0.16 – 0.08	0.523	−0.01 (0.06)	−0.13 – 0.10	0.822	0.78 (0.17)	0.51 – 1.18	0.244	0.86 (0.04)	0.79 – 0.94	0.001	0.53 (0.12)	0.33 – 0.84	0.007
N	178 ID			178 ID			178 ID			178 ID			155 ID			178 ID
Obs	11934			11933			11931			11934			4332			11928
R2	0.033 / 0.561			0.031 / 0.679			0.027 / 0.664			0.030 / 0.704			0.079 / 0.905			0.063 / 0.739

Notes. NED = Negative emotion differentiation; PED = Positive emotion differentiation; Int = Intercept; Surveys = Number of surveys completed over the course of the study; Day = Day of study; Cond = Condition; PNED = Person’s negative emotion differentiation; PPED = Person’s positive emotion differentiation; PA = Positive affect; NA = Negative affect; DDD = Drinks per drinking day; OR = Odds ratio; IRR = Incidence rate ratio; CI = confidence interval; SE= standard error; R² = Marginal/conditional R². Two models were run for each outcome, one with DNED as main predictor and one with DPED.

Table 4.

Concurrent association between positive and negative person-level emotion differentiation and daily outcomes, controlling for the grand-mean-centered person-mean of negative or positive affect.

	PA			Avg Craving			Max Craving			Any Drinking			DDD			Heavy Drinking
NED	Est (SE)	95% CI	p	Est (SE)	95% CI	p	Est (SE)	95% CI	p	OR (SE)	95% CI	p	IRR (SE)	95% CI	p	OR (SE)	95% CI	p
Int	0.06 (0.08)	−0.10 – 0.23	0.452	0.03 (0.08)	−0.12 – 0.19	0.671	0.06 (0.08)	−0.10 – 0.21	0.458	0.43 (0.13)	0.24 – 0.78	0.005	4.33 (0.28)	3.81 – 4.92	<0.001	0.10 (0.03)	0.05 – 0.19	<0.001
Surveys	0.01 (0.04)	−0.07 – 0.09	0.821	0.02 (0.04)	−0.05 – 0.10	0.582	0.02 (0.04)	−0.05 – 0.10	0.557	0.94 (0.14)	0.70 – 1.26	0.688	0.98 (0.03)	0.92 – 1.04	0.453	0.93 (0.15)	0.67 – 1.27	0.628
Day	0.05 (0.01)	0.04 – 0.06	<0.001	−0.16 (0.01)	−0.17 – −0.15	<0.001	−0.14 (0.01)	−0.15 – −0.13	<0.001	0.68 (0.02)	0.64 – 0.71	<0.001	0.97 (0.01)	0.96 – 0.99	<0.001	0.68 (0.02)	0.64 – 0.72	<0.001
Cond	−0.12 (0.11)	−0.35 – 0.10	0.288	−0.06 (0.11)	−0.28 – 0.15	0.552	−0.11 (0.11)	−0.32 – 0.10	0.295	0.56 (0.23)	0.25 – 1.24	0.152	0.94 (0.08)	0.79 – 1.12	0.485	0.51 (0.23)	0.21 – 1.23	0.134
PNED	0.11 (0.07)	−0.02 – 0.24	0.097	0.10 (0.06)	−0.02 – 0.22	0.093	0.05 (0.06)	−0.07 – 0.17	0.387	1.55 (0.36)	0.99 – 2.44	0.058	0.90 (0.05)	0.81 – 0.99	0.031	1.09 (0.28)	0.66 – 1.81	0.735
NA mean	−0.10 (0.06)	−0.22 – 0.03	0.127	0.42 (0.06)	0.30 – 0.54	<0.001	0.40 (0.06)	0.28 – 0.51	<0.001	2.54 (0.58)	1.62 – 3.96	<0.001	1.14 (0.06)	1.03 – 1.25	0.009	3.06 (0.76)	1.88 – 4.98	<0.001
N	178 ID			178 ID			178 ID			178 ID			155 ID			178 ID
Obs	11934			11933			11931			11934			4332			11928
R2	0.039 / 0.603			0.173 / 0.677			0.161 / 0.663			0.085 / 0.701			0.129 / 0.906			0.118 / 0.737
	NA			Avg Craving			Max Craving			Any Drinking			DDD			Heavy Drinking
PED	Est (SE)	95% CI	p	Est (SE)	95% CI	p	Est (SE)	95% CI	p	OR (SE)	95% CI	p	IRR (SE)	95% CI	p	OR (SE)	95% CI	p
Int	0.00 (0.08)	−0.16 – 0.16	0.990	0.03 (0.09)	−0.14 – 0.21	0.696	0.06 (0.09)	−0.11 – 0.24	0.468	0.45 (0.13)	0.25 – 0.81	0.007	4.40 (0.28)	3.88 – 4.98	<0.001	0.11 (0.03)	0.06 – 0.20	<0.001
Surveys	−0.11 (0.04)	−0.18 – −0.03	0.004	−0.03 (0.04)	−0.11 – 0.05	0.493	−0.03 (0.04)	−0.11 – 0.05	0.453	0.93 (0.13)	0.70 – 1.23	0.599	0.96 (0.03)	0.90 – 1.02	0.203	0.86 (0.13)	0.64 – 1.15	0.310
Day	−0.08 (0.01)	−0.09 – −0.07	<0.001	−0.16 (0.01)	−0.17 – −0.15	<0.001	−0.14 (0.01)	−0.15 – −0.13	<0.001	0.68 (0.02)	0.64 – 0.71	<0.001	0.97 (0.01)	0.96 – 0.99	<0.001	0.68 (0.02)	0.64 – 0.72	<0.001
Cond	−0.00 (0.11)	−0.22 – 0.21	0.986	−0.07 (0.12)	−0.30 – 0.17	0.583	−0.12 (0.12)	−0.36 – 0.11	0.306	0.53 (0.22)	0.23 – 1.18	0.119	0.90 (0.08)	0.76 – 1.07	0.253	0.44 (0.19)	0.18 – 1.04	0.061
PPED	−0.09 (0.05)	−0.19 – 0.02	0.109	−0.04 (0.06)	−0.16 – 0.07	0.460	−0.02 (0.06)	−0.14 – 0.10	0.738	0.76 (0.16)	0.51 – 1.13	0.178	0.85 (0.04)	0.78 – 0.93	<0.001	0.51 (0.11)	0.33 – 0.79	0.002
PA mean	−0.14 (0.06)	−0.25 – −0.03	0.010	−0.15 (0.06)	−0.27 – −0.03	0.013	−0.17 (0.06)	−0.29 – −0.05	0.004	0.47 (0.10)	0.31 – 0.70	<0.001	0.85 (0.04)	0.78 – 0.93	<0.001	0.38 (0.08)	0.24 – 0.59	<0.001
N	178 ID			178 ID			178 ID			178 ID			155 ID			178 ID
Obs	11934			11933			11931			11934			4332			11928
R2	0.054 / 0.563			0.054 / 0.680			0.056 / 0.666			0.082 / 0.704			0.150 / 0.904			0.140 / 0.736

Notes. NED = Negative emotion differentiation; PED = Positive emotion differentiation; Int = Intercept; Surveys = Number of surveys completed over the course of the study; Day = Day of study; Cond = Condition; PNED = Person’s negative emotion differentiation; PPED = Person’s positive emotion differentiation; PA = Positive affect; NA = Negative affect; DDD = Drinks per drinking day; OR = Odds ratio; IRR = Incidence rate ratio; CI = confidence interval; SE= standard error; R² = Marginal/conditional R². Two models were run for each outcome, one with DNED as main predictor and one with DPED.

Individuals with greater PED on average had fewer DDD and a lower likelihood of HD on average; PED was not associated with NA, craving levels, or odds of AD. Both associations remained after controlling for person-mean PA intensity. Treatment condition initially appeared to moderate the association between PED and maximum craving, but results were not robust to subtle alterations of the model. Treatment condition did not moderate any other associations.

Aim 2: Examine within-person associations among daily NED/PED and daily levels of affect intensity, craving, and alcohol use

Results from analyses addressing the second aim are presented in Tables 5 and 6. On days when individuals reported higher-than-usual NED, they had higher PA, lower average and maximum craving, lower odds of drinking and drinking heavily, and fewer DDD. After controlling for daily NA mean and variability, only the association between NED and PA remained (and was now in the opposite direction).

Table 5.

Concurrent association between positive and negative daily emotion differentiation and daily outcomes.

	PA			Avg Craving			Max Craving			Any Drinking			DDD			Heavy Drinking
NED	Est (SE)	95% CI	p	Est (SE)	95% CI	p	Est (SE)	95% CI	p	OR (SE)	95% CI	p	IRR (SE)	95% CI	p	OR (SE)	95% CI	p
Int	0.06 (0.08)	−0.10 – 0.23	0.456	0.01 (0.09)	−0.16 – 0.19	0.908	0.03 (0.09)	−0.15 – 0.20	0.761	0.41 (0.13)	0.22 – 0.76	0.005	4.34 (0.30)	3.79 – 4.98	<0.001	0.09 (0.03)	0.05 – 0.19	<0.001
Surveys	0.06 (0.04)	−0.02 – 0.14	0.117	−0.05 (0.04)	−0.13 – 0.03	0.225	−0.06 (0.04)	−0.14 – 0.02	0.167	0.84 (0.13)	0.62 – 1.13	0.246	0.94 (0.03)	0.88 – 1.00	0.049	0.74 (0.12)	0.53 – 1.03	0.073
Day	0.04 (0.01)	0.03 – 0.05	<0.001	−0.15 (0.01)	−0.16 – −0.14	<0.001	−0.13 (0.01)	−0.14 – −0.12	<0.001	0.69 (0.02)	0.66 – 0.73	<0.001	0.98 (0.01)	0.96 – 0.99	0.003	0.70 (0.02)	0.65 – 0.74	<0.001
Cond	−0.11 (0.11)	−0.33 – 0.11	0.326	−0.02 (0.12)	−0.26 – 0.21	0.842	−0.06 (0.12)	−0.29 – 0.17	0.613	0.59 (0.26)	0.25 – 1.39	0.227	0.93 (0.09)	0.77 – 1.12	0.442	0.52 (0.25)	0.20 – 1.36	0.183
DNED	0.10 (0.01)	0.08 – 0.13	<0.001	−0.10 (0.01)	−0.13 – −0.07	<0.001	−0.11 (0.01)	−0.13 – −0.08	<0.001	0.75 (0.02)	0.71 – 0.79	<0.001	0.96 (0.01)	0.95 – 0.98	<0.001	0.76 (0.02)	0.71 – 0.81	<0.001
N	178 ID			178 ID			178 ID			178 ID			155 ID			178 ID
Obs	11934			11933			11931			11934			4332			11928
R2	0.021 / 0.628			0.040 / 0.706			0.039 / 0.691			0.031 / 0.714			0.024 / 0.907			0.035 / 0.748
	NA			Avg Craving			Max Craving			Any Drinking			DDD			Heavy Drinking
PED	Est (SE)	95% CI	p	Est (SE)	95% CI	p	Est (SE)	95% CI	p	OR (SE)	95% CI	p	IRR (SE)	95% CI	p	OR (SE)	95% CI	p
Int	−0.01 (0.08)	−0.17 – 0.15	0.908	0.02 (0.09)	−0.15 – 0.20	0.810	0.05 (0.09)	−0.13 – 0.22	0.596	0.42 (0.13)	0.22 – 0.77	0.005	4.38 (0.30)	3.83 – 5.02	<0.001	0.10 (0.03)	0.05 – 0.19	<0.001
Surveys	−0.13 (0.04)	−0.20 – −0.05	0.001	−0.05 (0.04)	−0.13 – 0.03	0.244	−0.05 (0.04)	−0.13 – 0.03	0.227	0.84 (0.13)	0.63 – 1.13	0.259	0.94 (0.03)	0.88 – 1.00	0.046	0.75 (0.12)	0.54 – 1.03	0.078
Day	−0.08 (0.01)	−0.09 – −0.06	<0.001	−0.16 (0.01)	−0.17 – −0.15	<0.001	−0.14 (0.01)	−0.15 – −0.13	<0.001	0.68 (0.02)	0.65 – 0.72	<0.001	0.97 (0.01)	0.96 – 0.99	<0.001	0.69 (0.02)	0.64 – 0.73	<0.001
Cond	0.02 (0.11)	−0.20 – 0.24	0.871	−0.05 (0.12)	−0.28 – 0.19	0.711	−0.09 (0.12)	−0.33 – 0.14	0.436	0.60 (0.26)	0.26 – 1.39	0.230	0.93 (0.09)	0.77 – 1.11	0.419	0.53 (0.25)	0.21 – 1.35	0.182
DPED	−0.08 (0.01)	−0.11 – −0.05	<0.001	−0.03 (0.01)	−0.05 – −0.00	0.018	−0.02 (0.01)	−0.04 – −0.00	0.040	0.88 (0.02)	0.84 – 0.93	<0.001	0.98 (0.01)	0.97 – 0.99	0.001	0.88 (0.02)	0.83 – 0.93	<0.001
N	178 ID			178 ID			178 ID			178 ID			155 ID			178 ID
Obs	11934			11933			11931			11934			4332			11928
R2	0.033 / 0.588			0.031 / 0.690			0.028 / 0.676			0.025 / 0.705			0.022 / 0.905			0.030 / 0.740

Notes. NED = Negative emotion differentiation; PED = Positive emotion differentiation; Int = Intercept; Surveys = Number of surveys completed over the course of the study; Day = Day of study; Cond = Condition; PNED = Person’s negative emotion differentiation; PPED = Person’s positive emotion differentiation; PA = Positive affect; NA = Negative affect; DDD = Drinks per drinking day; OR = Odds ratio; IRR = Incidence rate ratio; CI = confidence interval; SE= standard error; R² = Marginal/conditional R². Two models were run for each outcome, one with DNED as main predictor and one with DPED. Results did not change when controlling for person-level ED.

Table 6.

Concurrent association between positive and negative daily emotion differentiation and daily outcomes, controlling for the daily person-centered mean and standard deviation of negative or positive affect.

	PA			Avg Craving			Max Craving			Any Drinking			DDD			Heavy Drinking
NED	Est (SE)	95% CI	p	Est (SE)	95% CI	p	Est (SE)	95% CI	p	OR (SE)	95% CI	p	IRR (SE)	95% CI	p	OR (SE)	95% CI	p
Int	0.05 (0.08)	−0.11 – 0.22	0.532	0.02 (0.09)	−0.16 – 0.19	0.855	0.03 (0.09)	−0.14 – 0.21	0.711	0.40 (0.13)	0.21 – 0.76	0.005	4.28 (0.29)	3.74 – 4.90	<0.001	0.09 (0.03)	0.04 – 0.18	<0.001
Surveys	0.06 (0.04)	−0.02 – 0.14	0.127	−0.05 (0.04)	−0.13 – 0.03	0.226	−0.06 (0.04)	−0.14 – 0.02	0.171	0.84 (0.13)	0.62 – 1.13	0.249	0.94 (0.03)	0.88 – 1.00	0.063	0.74 (0.13)	0.53 – 1.03	0.077
Day	0.02 (0.01)	0.01 – 0.03	<0.001	−0.13 (0.01)	−0.14 – −0.12	<0.001	−0.12 (0.01)	−0.13 – −0.11	<0.001	0.71 (0.02)	0.67 – 0.75	<0.001	0.98 (0.01)	0.97 – 0.99	0.004	0.71 (0.02)	0.66 – 0.76	<0.001
Cond	−0.10 (0.11)	−0.32 – 0.12	0.393	−0.03 (0.12)	−0.27 – 0.21	0.788	−0.07 (0.12)	−0.30 – 0.17	0.570	0.58 (0.26)	0.24 – 1.39	0.221	0.92 (0.09)	0.77 – 1.11	0.386	0.52 (0.25)	0.20 – 1.36	0.180
DNED	−0.02 (0.01)	−0.05 – −0.00	0.037	−0.00 (0.01)	−0.03 – 0.02	0.750	0.00 (0.01)	−0.02 – 0.02	0.954	0.95 (0.03)	0.89 – 1.01	0.103	1.01 (0.01)	1.00 – 1.03	0.159	0.98 (0.04)	0.91 – 1.05	0.526
NA mean	−0.21 (0.01)	−0.23 – −0.19	<0.001	0.15 (0.01)	0.13 – 0.16	<0.001	0.16 (0.01)	0.14 – 0.18	<0.001	1.36 (0.05)	1.26 – 1.47	<0.001	1.08 (0.01)	1.06 – 1.10	<0.001	1.45 (0.06)	1.33 – 1.57	<0.001
NA SD	−0.03 (0.01)	−0.04 – −0.01	0.001	0.04 (0.01)	0.03 – 0.05	<0.001	0.05 (0.01)	0.04 – 0.07	<0.001	1.24 (0.05)	1.16 – 1.34	<0.001	1.01 (0.01)	0.99 – 1.02	0.504	1.14 (0.05)	1.05 – 1.24	0.003
N	178 ID			178 ID			178 ID			178 ID			155 ID			178 ID
Obs	11934			11933			11931			11934			4332			11928
R2	0.055 / 0.660			0.061 / 0.724			0.066 / 0.716			0.046 / 0.727			0.038 / 0.906			0.046 / 0.754
	NA			Avg Craving			Max Craving			Any Drinking			DDD			Heavy Drinking
PED	Est (SE)	95% CI	p	Est (SE)	95% CI	p	Est (SE)	95% CI	p	OR (SE)	95% CI	p	IRR (SE)	95% CI	p	OR (SE)	95% CI	p
Int	−0.01 (0.08)	−0.17 – 0.15	0.884	0.02 (0.09)	−0.16 – 0.20	0.823	0.05 (0.09)	−0.13 – 0.22	0.606	0.40 (0.13)	0.21 – 0.76	0.005	4.26 (0.29)	3.72 – 4.87	<0.001	0.09 (0.03)	0.04 – 0.18	<0.001
Surveys	−0.13 (0.04)	−0.21 – −0.06	0.001	−0.05 (0.04)	−0.13 – 0.03	0.244	−0.05 (0.04)	−0.13 – 0.03	0.218	0.84 (0.13)	0.62 – 1.13	0.252	0.94 (0.03)	0.88 – 1.00	0.066	0.74 (0.13)	0.53 – 1.03	0.078
Day	−0.05 (0.01)	−0.07 – −0.04	<0.001	−0.15 (0.01)	−0.16 – −0.14	<0.001	−0.13 (0.01)	−0.14 – −0.12	<0.001	0.71 (0.02)	0.67 – 0.75	<0.001	0.98 (0.01)	0.96 – 0.99	0.001	0.70 (0.02)	0.66 – 0.75	<0.001
Cond	0.02 (0.11)	−0.20 – 0.24	0.832	−0.04 (0.12)	−0.28 – 0.20	0.740	−0.09 (0.12)	−0.33 – 0.15	0.454	0.59 (0.26)	0.24 – 1.41	0.232	0.93 (0.09)	0.77 – 1.12	0.431	0.52 (0.26)	0.20 – 1.38	0.187
DPED	−0.06 (0.01)	−0.08 – −0.04	<0.001	−0.02 (0.01)	−0.04 – 0.00	0.068	−0.02 (0.01)	−0.04 – 0.00	0.122	0.93 (0.02)	0.89 – 0.98	0.009	0.99 (0.01)	0.98 – 1.00	0.176	0.95 (0.03)	0.90 – 1.01	0.075
PA mean	−0.22 (0.01)	−0.23 – −0.21	<0.001	−0.09 (0.01)	−0.11 – −0.08	<0.001	−0.10 (0.01)	−0.11 – −0.09	<0.001	0.62 (0.02)	0.58 – 0.65	<0.001	0.93 (0.01)	0.91 – 0.94	<0.001	0.59 (0.02)	0.55 – 0.62	<0.001
PA SD	0.07 (0.01)	0.06 – 0.08	<0.001	0.04 (0.01)	0.03 – 0.05	<0.001	0.05 (0.01)	0.04 – 0.06	<0.001	1.15 (0.03)	1.09 – 1.21	<0.001	1.02 (0.01)	1.00 – 1.03	0.019	1.12 (0.04)	1.05 – 1.19	<0.001
N	178 ID			178 ID			178 ID			178 ID			155 ID			178 ID
Obs	11934			11933			11931			11934			4332			11928
R2	0.086 / 0.632			0.041 / 0.698			0.039 / 0.686			0.045 / 0.727			0.038 / 0.905			0.051 / 0.759

Notes. NED = Negative emotion differentiation; PED = Positive emotion differentiation; Int = Intercept; Surveys = Number of surveys completed over the course of the study; Day = Day of study; Cond = Condition; PNED = Person’s negative emotion differentiation; PPED = Person’s positive emotion differentiation; PA = Positive affect; NA = Negative affect; DDD = Drinks per drinking day; OR = Odds ratio; IRR = Incidence rate ratio; CI = confidence interval; SE= standard error; R² = Marginal/conditional R². Two models were run for each outcome, one with DNED as main predictor and one with DPED. Results did not change when controlling for person-level ED.

On days when individuals reported higher-than-usual PED, they had lower NA, lower average and maximum craving, lower odds of drinking and drinking heavily, and fewer DDD. The associations between PED and NA and PED and odds of AD remained after controlling for daily PA mean and variability. Treatment condition did not moderate any associations.

Aim 3: Examine between-person differences in baseline alexithymia as a moderator of within-person associations among daily ED and affect intensity, craving, and alcohol use

Results from analyses addressing the third aim are presented in Table 7. After controlling for affect intensity and variability, alexithymia moderated daily within-person associations between NED and both AD and HD. Figure 1, Panel A shows the average change in the predicted probability of a drinking day across the range of NED, separately for individuals with (N=28) and without (N=150) alexithymia. Relative to individuals without alexithymia, individuals with alexithymia exhibited stronger negative within-person associations between NED and the probability of a drinking day on average, such that the probability of drinking was lower on days when NED was higher-than-usual. Figure 1, Panel B illustrates that relative to individuals without alexithymia, individuals with alexithymia also exhibited stronger negative within-person associations between NED and the probability of a heavy drinking day on average, such that the probability of drinking heavily was lower on days when NED was higher-than-usual. Alexithymia did not moderate any associations between PED and outcomes. Treatment condition did not moderate any associations.

Table 7.

Concurrent association between positive and negative daily emotion differentiation and daily outcomes, moderated by baseline alexithymia.

	PA			Avg Craving			Max Craving			Any Drinking			DDD			Heavy Drinking
NED	Est (SE)	95% CI	p	Est (SE)	95% CI	p	Est (SE)	95% CI	p	OR (SE)	95% CI	p	IRR (SE)	95% CI	p	OR (SE)	95% CI	p
Int	0.09 (0.09)	−0.08 – 0.27	0.290	0.03 (0.10)	−0.15 – 0.22	0.737	0.06 (0.09)	−0.12 – 0.25	0.498	0.51 (0.17)	0.26 – 0.99	0.048	4.41 (0.32)	3.83 – 5.08	<0.001	0.12 (0.04)	0.06 – 0.24	<0.001
Surveys	0.06 (0.04)	−0.02 – 0.13	0.149	−0.05 (0.04)	−0.13 – 0.03	0.222	−0.06 (0.04)	−0.14 – 0.02	0.164	0.83 (0.13)	0.61 – 1.12	0.212	0.94 (0.03)	0.88 – 1.00	0.058	0.74 (0.12)	0.53 – 1.02	0.065
Day	0.02 (0.01)	0.01 – 0.03	<0.001	−0.13 (0.01)	−0.14 – −0.12	<0.001	−0.12 (0.01)	−0.13 – −0.11	<0.001	0.71 (0.02)	0.67 – 0.75	<0.001	0.98 (0.01)	0.96 – 0.99	0.004	0.71 (0.02)	0.66 – 0.75	<0.001
Cond	−0.11 (0.11)	−0.33 – 0.11	0.315	−0.04 (0.12)	−0.28 – 0.20	0.759	−0.08 (0.12)	−0.31 – 0.16	0.520	0.53 (0.24)	0.22 – 1.27	0.154	0.91 (0.09)	0.76 – 1.10	0.339	0.47 (0.23)	0.18 – 1.24	0.127
ALE	−0.22 (0.16)	−0.53 – 0.10	0.176	−0.08 (0.17)	−0.41 – 0.24	0.614	−0.17 (0.17)	−0.49 – 0.16	0.316	0.26 (0.16)	0.08 – 0.87	0.028	0.84 (0.11)	0.65 – 1.09	0.186	0.24 (0.16)	0.06 – 0.90	0.035
DNED	−0.03 (0.01)	−0.05– −0.00	0.031	0.00 (0.01)	−0.02 – 0.03	0.777	0.01 (0.01)	−002 – 0.03	0.609	0.99 (0.03)	0.93 – 1.06	0.851	1.01 (0.01)	1.00 – 1.03	0.139	1.01 (0.04)	0.93 – 1.08	0.887
NA mean	−0.21 (0.01)	−0.23– −0.19	<0.001	0.15 (0.01)	0.13 – 0.16	<0.001	0.16 (0.01)	0.14 – 0.18	<0.001	1.35 (0.05)	1.25 – 1.46	<0.001	1.08 (0.01)	1.06 – 1.10	<0.001	1.44 (0.06)	1.33 – 1.57	<0.001
NA SD	−0.03 (0.01)	−0.04– −0.01	<0.001	0.04 (0.01)	0.03 – 0.05	<0.001	0.06 (0.01)	0.04 – 0.07	<0.001	1.25 (0.05)	1.16 – 1.35	<0.001	1.01 (0.01)	0.99 – 1.02	0.479	1.14 (0.05)	1.05 – 1.24	0.002
DNEDxALE	0.02 (0.03)	−0.04 – 0.07	0.563	−0.05 (0.03)	−0.11 – 0.01	0.138	−0.04 (0.03)	−0.09 – 0.02	0.232	0.79 (0.05)	0.69 – 0.91	0.001	0.99 (0.02)	0.96 – 1.02	0.530	0.84 (0.07)	0.72 – 0.98	0.029
N	178ID			178 ID			178 ID			178 ID			155 ID			178 ID
Obs	11934			11933			11931			11934			4332			11928
R2	0.062 / 0.661			0.061 / 0.725			0.069 / 0.717			0.064 / 0.727			0.046 / 0.906			0.064 / 0.754
	NA			Avg Craving			Max Craving			Any Drinking			DDD			Heavy Drinking
PED	Est (SE)	95% CI	p	Est (SE)	95% CI	p	Est (SE)	95% CI	p	OR (SE)	95% CI	p	IRR (SE)	95% CI	p	OR (SE)	95% CI	p
Int	−0.04 (0.09)	−0.21 – 0.13	0.671	0.04 (0.10)	−0.15 – 0.22	0.703	0.08 (0.09)	−0.11 – 0.26	0.409	0.50 (0.17)	0.26 – 0.99	0.045	4.39 (0.32)	3.81 – 5.06	<0.001	0.11 (0.04)	0.05 – 0.24	<0.001
Surveys	−0.13 (0.04)	−0.21 – −0.06	0.001	−0.05 (0.04)	−0.13 – 0.03	0.236	−0.05 (0.04)	−0.13 – 0.03	0.199	0.83 (0.13)	0.61 – 1.11	0.210	0.94 (0.03)	0.88 – 1.00	0.061	0.73 (0.12)	0.53 – 1.02	0.064
Day	−0.05 (0.01)	−0.07 – −0.04	<0.001	−0.15 (0.01)	−0.16 – −0.14	<0.001	−0.13 (0.01)	−0.14 – −0.12	<0.001	0.71 (0.02)	0.67 – 0.75	<0.001	0.98 (0.01)	0.96 – 0.99	0.001	0.70 (0.02)	0.66 – 0.75	<0.001
Cond	0.03 (0.11)	−0.19 – 0.25	0.772	−0.05 (0.12)	−0.29 – 0.19	0.706	−0.10 (0.12)	−0.34 – 0.14	0.400	0.54 (0.24)	0.23 – 1.29	0.165	0.92 (0.09)	0.77 – 1.11	0.381	0.48 (0.24)	0.18 – 1.25	0.133
ALE	0.13 (0.15)	−0.17 – 0.43	0.387	−0.09 (0.17)	−0.42 – 0.24	0.610	−0.17 (0.17)	−0.50 – 0.16	0.309	0.30 (0.18)	0.09 – 0.97	0.044	0.85 (0.11)	0.65 – 1.10	0.212	0.26 (0.18)	0.07 – 1.00	0.050
DPED	−0.05 (0.01)	−0.08 – −0.03	<0.001	−0.02 (0.01)	−0.04 – 0.00	0.104	−0.02 (0.01)	−0.04 – 0.00	0.091	0.93 (0.03)	0.88 – 0.99	0.015	0.99 (0.01)	0.98 – 1.00	0.157	0.95 (0.03)	0.89 – 1.01	0.118
PA mean	−0.22 (0.01)	−0.23 – −0.21	<0.001	−0.09 (0.01)	−0.11 – −0.08	<0.001	−0.10 (0.01)	−0.11 – −0.09	<0.001	0.62 (0.02)	0.58 – 0.65	<0.001	0.93 (0.01)	0.91 – 0.94	<0.001	0.59 (0.02)	0.55 – 0.63	<0.001
PA SD	0.07 (0.01)	0.06 – 0.08	<0.001	0.04 (0.01)	0.03 – 0.05	<0.001	0.05 (0.01)	0.04 – 0.06	<0.001	1.15 (0.03)	1.09 – 1.21	<0.001	1.02 (0.01)	1.00 – 1.03	0.018	1.12 (0.04)	1.05 – 1.19	<0.001
DPEDxALE	−0.04 (0.03)	−0.10 – 0.02	0.150	−0.00 (0.03)	−0.06 – 0.05	0.916	0.02 (0.03)	−0.04 – 0.07	0.493	1.01 (0.07)	0.89 – 1.15	0.877	1.01 (0.02)	0.97 – 1.04	0.785	0.99 (0.08)	0.85 – 1.15	0.853
N	178ID			178 ID			178 ID			178 ID			155 ID			178 ID
Obs	11934			11933			11931			11934			4332			11928
R2	0.089 / 0.634			0.042 / 0.699			0.043 / 0.687			0.060 / 0.725			0.045 / 0.905			0.066 / 0.758

Notes. NED = Negative emotion differentiation; PED = Positive emotion differentiation; Int = Intercept; Surveys = Number of surveys completed over the course of the study; Day = Day of study; Cond = Condition; ALE = Alexithymia; PNED = Person’s negative emotion differentiation; PPED = Person’s positive emotion differentiation; PA = Positive affect; NA = Negative affect; DDD = Drinks per drinking day; OR = Odds ratio; IRR = Incidence rate ratio; CI = confidence interval; SE= standard error; R² = Marginal/conditional R². Two models were run for each outcome, one with DNED as main predictor and one with DPED. Results did not change when controlling for person-level ED.

Figure 1.

The change in probability of (A) drinking day and (B) heavy drinking day across the range of day’s negative emotion differentiation, separately for individuals with and without baseline alexithymia.

Sensitivity analyses

First, the pattern of results for all adjusted person- and day-level NED and PED associations did not change when participants without within-person variability in craving or alcohol use were excluded.

Second, person-specific slope estimates indicated that change across the 84-day period in day’s NED (M=0.047, t=1.98, df=177, p=0.049), but not day’s PED (M=0.033 t=1.87, df=177, p=0.063), was significantly greater than zero. Yet, the effect sizes for these changes were negligible (d=0.15 and 0.14, respectively; see Supplementary Material). Further, individual differences in daily NED/PED change during treatment were not associated with treatment condition, the number of survey days completed, or person-mean NA/PA intensity (all ps > 0.2).

Third, results were robust to different ways of handling days with identical ratings across negative and/or positive emotion items. Analyses excluding days with (a) a score of “1” across both negative and positive emotion items (0.5% of days), (b) zero variability (regardless of whether the scores were all 1’s, all 2’s, etc.) across both negative and positive emotion items (3.7% of days), and (c) no item variability separately for negative (25.1% of days) and positive (8.6% of days) emotion items, all revealed no changes in the pattern of results for NED and PED.

Discussion

The current study investigated whether between-person differences in overall negative emotion differentiation (NED) or positive emotion differentiation (PED) across time in treatment and within-person variability in daily NED/PED were associated with affect intensity, craving, and alcohol use outcomes in daily life among individuals in outpatient treatment for alcohol use disorder (AUD). Subsequent analyses examined whether associations differed by treatment condition or baseline alexithymia, or after controlling for intensity and variability of affect. Results revealed that independent of the intensity with which emotions were experienced, NED was negatively associated with drinks per drinking day (DDD) between-persons and positive affect intensity (PA) within-persons. PED was negatively associated with DDD and heavy drinking (HD) between-persons and with negative affect (NA) and any drinking (AD) within-persons. Alexithymia moderated within-person associations between NED and alcohol use outcomes (AD and HD).

Implications for clinical practice

Findings from this study are unique in suggesting that PED is consistently linked to lower odds of alcohol use. At the between-persons level, individuals who demonstrated greater NED and PED on average drank fewer drinks per drinking day, but lower odds of heavy drinking were uniquely linked to greater overall differentiation of positive emotions. These findings are consistent with prior between-persons work establishing that PED, but not NED, was associated with adaptive behavior within clinical populations (Dixon-Gordon et al., 2014; O’Toole et al., 2020). Furthermore, the pattern of findings was similar at the daily, within-persons level: higher-than-usual daily NED was linked only to PA in adjusted main effects models, but higher-than-usual daily PED was associated with both lower NA and lower odds of drinking. Coinciding with the broaden-and-build framework for positive emotions (Fredrickson, 2001), more fine-grained attention to specific positive emotions, both overall and more than usual on a given day, may encourage adaptive cognition and behavior. For example, greater awareness of specific positive emotions such as “determined” (i.e., high differentiation) may motivate behaviors associated with that specific emotion (e.g., following through with a goal to avoid alcohol), more so than identifying positive emotions with less specificity such as “good” (i.e., low differentiation).

The few statistically significant main effects of NED between- and within-persons (after accounting for NA intensity and variability) contradict previous evidence linking overall NED to substance use generally (Anand et al., 2017; Walters et al., 2023) and alcohol use specifically (Emery et al., 2022). However, baseline alexithymia moderated within-person associations between daily NED and odds of both any and heavy drinking. These findings suggest that it is the within-person fluctuations in NED from day-to-day (which are indistinguishable from between-person variance in conventional person-level research) that may be most important, particularly for individuals with co-morbid AUD and alexithymia. In line with the “feelings as information” perspective (Schwarz, 2012), on days when such individuals are able to differentiate among their negative emotions more than usual, they may gain more nuanced insights into their emotional context than is typically available to them. This information can be used to apply more effective coping strategies (e.g., psychological distancing; Kashdan et al., 2015) that prevent distressing emotions from dictating behavior and lower the probability of alcohol use.

Together with previous research suggesting that ED is a modifiable skill (Vedernikova et al., 2021), findings from this study suggest that ED could be an important focus of treatment. Indeed, several effective substance use treatments teach individuals to accurately label their emotions (e.g., Axelrod et al., 2011) and increase awareness of their current emotional state (e.g., Stasiewicz et al., 2018; Witkiewitz & Bowen, 2010). It has been suggested that expanding individuals’ emotional vocabulary is at the heart of these interventions (Kashdan et al., 2015) and that inviting clients to sort emotions into categories could help develop expertise in ED (Jacobson et al., 2023). Mindfulness skills (Van Der Gucht et al., 2019) and self-monitoring (Hoemann, Barrett, et al., 2021; Widdershoven et al., 2019) have also been shown to improve ED. Applying such strategies to increase AUD patients’ PED skills and expertise might be particularly helpful in facilitating lower odds of heavy drinking overall, as well as greater emotional well-being (i.e., lower NA) and odds of abstinence on days when patients differentiate positive emotions more than usual. In the case of NED, therapists may want to use a more targeted treatment approach that involves identifying patients who struggle to sufficiently differentiate among negative emotions (i.e., those with alexithymia) and helping them construct their negative emotional experiences with more granularity in the context of their everyday lives. Such efforts early in treatment may correspond with enhanced PA and lower odds of drinking and heavy drinking.

Implications for future research

Our results demonstrate the importance of examining the unique effects of within-person emotion dynamics indicators, independent of the intensity and variability with which emotions are experienced (Dejonckheere et al., 2019). In unadjusted models, there were several associations among NED/PED and outcomes, especially at the within-person level. In adjusted models, NA/PA intensity and variability were strongly associated with both craving and alcohol use outcomes, whereas NED and PED were associated (either directly or through involvement in interactions) with alcohol use and affect but not with craving. These findings suggest that experiencing higher levels of affect across discrete categories is more robustly associated with alcohol craving, and furthermore, that the granularity with which the activated emotion(s) is/are identified may provide additional unique information about the propensity to consume alcohol. Future research should report on ED’s associations after accounting for level and variability in affect to continue parsing its unique associations with outcomes.

Additionally, our analyses revealed different patterns of associations within- vs. between-persons. Both NED and PED were consistently related to outcomes within-persons in unadjusted models and (to a lesser extent) in adjusted models, while between-person associations were more inconsistent. This pattern is generally in line with prior research operationalizing ED at the daily or momentary level, which has demonstrated consistent relationships between well-being and both NED and PED at the momentary level alongside inconsistent and weak associations at the person-level (Erbas et al., 2022; Tomko et al., 2015). Substance use research has also reported that smoking and alcohol use were associated with NED within-persons but not between-persons (Lane & Trull, 2022). That many associations appear at the day- or momentary-level and not at the person-level underscores the importance of disaggregating ED variance into its component stable and variable parts. It may also be one reason that findings are inconsistent with previous studies carried out exclusively at the person-level, which have typically found NED, and not PED, to be implicated in psychological well-being (Kashdan et al., 2015). Future research should continue exploring the within-person variability in, and correlates of, ED that may be hidden in conventional between-person indices. Future research should also seek to understand within-person change in ED over time during substance use recovery, and whether its relationship to outcomes differs during initial (up to 3 months; e.g., Linn et al., 2023) or early (3 months to 1 year; Emery et al., 2022) recovery stages as compared to sustained (1 to 5 years) and stable (greater than 5 years) recovery stages.

Strengths and limitations

One notable strength of the study’s design was that it assessed PA and NA with 10 items each, permitting an equal opportunity for differentiation. This approach overcomes limitations of previous studies (e.g., Lazarus & Fisher, 2021) which have used fewer items to assess positive compared to negative emotions, with unknown impacts on the PED index and subsequent findings.

Limitations to the current study should also be noted. First, assessing emotions once per day via the daily diary design may not have captured participants’ true range of emotions on a given day. Second, because ED and outcome variables were assessed on the same day, caution should be exercised when making inferences regarding the directionality of associations. Third, NED and PED were examined in separate models to minimize model complexity, limiting our ability to make direct comparisons. Fourth, our sample consisted of majority White negative affect drinkers with moderate to severe AUD. Generalizing to other populations should be done with caution.

Conclusion

The study’s findings suggest that PED may be an important and understudied skill linked to alcohol use behavior both between- and within-persons, irrespective of baseline alexithymia. In contrast, within-person variability in daily NED might be important specifically for individuals with co-morbid AUD and alexithymia. More broadly, the present findings demonstrate the importance of looking beyond the level of affect intensity at the within-person relationships among different emotion states (both negative and positive), the day-to-day fluctuations in these relationships, and their unique associations with alcohol use in daily life.