Contextual Influences on Subjective Alcohol Response

Abstract

Prior research demonstrates contextual influences on drug responses in both animals and humans, though studies in humans typically focus on only one aspect of context (e.g., social), and examine a limited range of subjective experiences. The current study sought to address these limitations by examining the impact of both social and physical context on the full range of subjective alcohol effects. The sample included 448 young adult social drinkers (57% male, 66.5% White) randomly assigned to consume alcohol (target blood alcohol concentration (BAC) of .08 g%) or placebo in one of four contexts (solitary lab, group lab, solitary bar, group bar). Results indicated that high arousal positive (HAP) effects of alcohol (e.g., talkative, lively) were stronger in non-bar relative to bar contexts, and that low arousal positive effects (e.g., relaxed, calm) were only present in the group lab context. There were also main effects of social context such that high arousal effects (both positive and negative) were stronger in group contexts, regardless of beverage condition. These findings highlight the importance of considering context when examining alcohol effects. Studies designed to isolate pharmacological HAP effects may benefit from a non-bar setting, and studies of LAP effects might be most effective in a simulated living room or home environment, though future studies are needed to directly address this possibility. Further, studies with an explicit focus on expectancies or that need strong control for expectancies might benefit from a group context, particularly when studying high arousal effects.

Alcohol use disorder (AUD) is a pervasive mental health concern. In the United States, it was recently estimated that 13.9% and 29.1% of the population met criteria for past-year and lifetime AUD, respectively (Grant et al., 2015). Although costly, much of the public health burden associated with alcohol use is not related to individuals with AUD. Heavy drinking has been estimated to contribute to 10% of injuries in young adults and shortens overall life span by roughly 10% in the general population (Hingson et al., 2009; Lundin et al., 2015). Given these trends among young adults and the high prevalence of AUD, identifying risk factors for heavy drinking and alcohol-related problems is critical to improving public health. Individual differences in acute alcohol response (AR) are a well-established risk factor (Trim, Schuckit & Smith, 2009; King et al., 2011). However, efforts to apply the expanding knowledge base regarding AR to the prevention or treatment of alcohol-related problems are just beginning to develop (e.g., Schuckit et al., 2015; Schuckit et al., 2016). The limited development of empirically supported prevention and intervention programs targeting AR may be a result of our incomplete understanding of the many factors that contribute to these individual differences. The current study aims to advance our understanding of AR by building upon prior research demonstrating the important role of drinking context (Beck, Thombs, & Summons, 1993; Fairbairn & Sayette, 2014).

Classic behavioral theories of substance use suggest that contextual influences, including both the social and physical context, play a powerful role in initiation and later use of alcohol. Classical conditioning, for example, reinvents meaning of a stimulus by associating it with an alternative stimulus. Environmental and social cues may become paired with substance use through repeated exposures, resulting in novel responses to these environmental cues (Otto, Cleirigh, & Pollack, 2007). As a classic example of the impact of context, heroin users have been found to overdose, even when using a typical dose of the drug, if it is administered in a novel context (Seigel, 2016). This work argues that contextual cues at the time of administration become paired with the psychopharmacological response to the drug and, over time, such environmental or social cues come to elicit anticipatory physiological responses that temper the effect of the drug. When the drug is administered in the absence of such cues, this anticipatory physiological response is not initiated, which may result in an overdose. Of course, physical and social contexts may also impact drug responses in more subtle ways.

Social Context

Relative to other drugs, alcohol effects may be particularly susceptible to contextual influences given their diversity and lack of specificity (e.g., both sedative and stimulant effects). Prior studies have demonstrated the influence of social context (solitary versus group drinking) on AR (Fairbairn, 2017), suggesting that individuals may interpret the non-specific pharmacological effects of alcohol relative to the particular contexts in which alcohol is consumed. For example, early studies demonstrated that participants drinking in groups reported more subjective pleasure, enhanced sensations of warmth-glow (Sher, 1985), and greater reinforcement from drinking (Pliner & Cappell, 1974). In contrast, solitary drinking was shown to promote physical symptoms of sedation at comparable levels of intoxication (Pliner & Cappell, 1974).

More recent alcohol administration research has used small group paradigms to better understand how alcohol may facilitate social bonding and prosocial behavior in a social context. For example, Sayette et al. (2012a) found that moderate doses of alcohol within small groups of strangers increased time spent speaking to one another, decreased moments of silence, and increased self-reported bonding. In addition, strangers in alcohol-consuming groups were found to smile longer and more frequently, experience more mutual smiles, and show a decreased response to negative-affect facial expressions relative to groups who consumed a placebo or control beverage (Fairbairn et al., 2015, Sayette et al., 2012a). Catching another’s smile while drinking was also associated with increases in positive mood and social bonding, and decreases in negative mood (Fairbairn et al., 2015). In another study, social drinkers evidenced reduced sensitivity to others’ negative emotional expressions of anger, sadness, and fear (Dolder et al., 2017). Results of a recent meta-analysis of 21 studies and over 2,000 participants further demonstrate that alcohol facilitates prosocial behavior and mood enhancement more strongly for groups of strangers relative to acquaintances (Fairbairn, 2017). Findings of the Fairbairn meta-analysis suggest that alcohol significantly enhances social-emotional experiences as measured through behavioral and subjective reports of mood and in studies using strangers relative to familiar group members. Taken together, these findings provide strong evidence for examining the critical role of social context in AR and support a study design that focuses on groups of unacquainted participants.

More recent advances in technology have led to investigation of social drinking in real-world settings. For example, a recent study of heavy social drinkers captured photos of the social context while drinking, and participants rated level of acquaintance with all individuals captured in the photos. Alcohol consumption was assessed via transdermal alcohol concentration through ankle bracelets (Fairbairn et al., 2018). Positive mood was higher and negative mood was lower when photos depicted social contexts compared to when photos depicted no other individuals. Moreover, negative mood increased with the number of strangers present and decreased with average amount of time participants spent among depicted strangers. These findings illuminate the subtle ways in which social context may impact subjective experiences following alcohol consumption.

Physical Context

Relative to social context, research on physical contextual factors is sparse yet potentially critical in understanding alcohol effects. Drinking in bars, clubs, and parties, for instance, is associated with serious negative consequences such as alcohol-induced violence, binge drinking, and intoxicated driving (Casswell, Zhang, & Wyllie, 1993; Stockwell, Somerford, & Lang, 1991; Wells, Graham, Speechley, & Koval, 2005). At the same time, drinking in solitary settings is associated with drinking to cope, which is a robust risk factor for alcohol-related problems, including AUD (Corbin, Ladensack, Waddell, & Scott, In press). Although there is little direct research examining context effects on AR, prior studies have demonstrated that alcohol outcome expectancies, an overlapping but distinct construct from AR, vary across physical contexts. For example, Wall et al. (2001) found that participants in an on-campus bar expected more pleasurable disinhibition and stimulation than participants in a laboratory setting. An ecological momentary assessment (EMA) study also found that drinking events in settings like bars or restaurants were associated with more self-reported vigor than drinking in other contexts (Ray et al., 2010). Unfortunately, neither of these studies differentiated expected effects from actual pharmacological effects. This is critical, as placebo effects may be much stronger in naturalistic settings (e.g. simulated bar) than in typical lab settings. Both studies also focused on a very narrow range of subjective effects (i.e., impairment and intoxication).

We are aware of only one study, to date, that investigated the effects of physical context on AR using a broad measure of subjective effects (Corbin, Scott, Boyd, Menary, & Enders. 2015). This study found that low arousal positive effects of alcohol (e.g., relaxed, calm), relative to placebo, were stronger in the context of a traditional lab. In contrast, high arousal positive effects were more pronounced in the simulated bar under both alcohol and placebo. These findings suggest that individuals may be better able to differentiate reinforcing, pharmacological effects of alcohol from those of the drinking context in less stimulating environments. Interestingly, no significant beverage by context effects were found for negative effects, suggesting that physical context may uniquely impact positive responses to alcohol. Such preliminary findings are critical to enhancing our limited understanding of how arousal and valence of subjective experiences may differ across drinking contexts and how such differences may ultimately contribute to risk for alcohol abuse.

In summary, prior research demonstrates contextual (both social and physical) influences on both alcohol expectancies and subjective response to alcohol. To our knowledge, no prior studies have examined both social and physical context, and only one prior study of physical context effects on AR examined the full range of subjective experiences. The current study was designed to address these gaps in the literature by examining the effects of both social and physical context on the full range of subjective alcohol effects using a placebo-controlled alcohol administration paradigm.

Participants were randomly assigned to consume their beverages in one of four drinking contexts, crossing social (alone, group), and physical (lab, bar) contexts. Within each context, participants were randomly assigned to consume either alcohol or placebo. AR was assessed using a well-validated measure (SEAS; Morean et al., 2013) that captures the full valence by arousal affective space (i.e., low arousal positive (LAP), low arousal negative (LAN), high arousal positive (HAP), high arousal negative (HAN)). Although prior studies have typically captured high arousal positive (e.g., stimulation) and/or low arousal negative (e.g., sedation) effects, they have generally provided little coverage of high arousal negative (e.g. reckless, aggressive) or low arousal positive effects (e.g. relaxed, calm). The former may play an important role in alcohol-related risk taking, whereas the latter may play a more central role in the negative reinforcement value of alcohol. Individuals who experience strong low arousal positive effects may learn to use alcohol as a way to manage stress or cope with negative emotional states, and drinking to cope is a robust predictor of alcohol-related problems (Cooper, 1994). Moreover, the one prior study of physical context effects on AR found that physical context uniquely effected low arousal positive responses to alcohol (Corbin et al., 2015).

Based on the findings of prior studies, we hypothesized main effects of both beverage condition and context, qualified by significant beverage condition by context interactions. With respect to HAP effects, we anticipated main effects of beverage condition and physical context, with greater HAP effects under alcohol and in the bar setting. We also expected a social context by beverage condition interaction with stronger HAP effects under alcohol in group relative to solitary contexts. Hypotheses for HAN effects were less clear as prior studies have either focused on HAP effects or combined HAP and HAN effects. Thus, in addition to a main effect of beverage condition, we tentatively hypothesized a social context by beverage condition interaction with stronger HAN effects under alcohol in group relative to solitary contexts. For LAP effects, we expected a beverage condition by physical context interaction such that LAP effects would be stronger under alcohol in lab relative to bar contexts. Effects for social context were less clear as prior studies have either focused on LAN effects or combined LAN and LAP effects, but we tentatively hypothesized a beverage condition by social context interaction, with weaker LAP effects under alcohol in group relative to solitary settings. For LAN effects, we predicted a main effect of beverage condition and a beverage condition by social context interaction, such that LAN effects would be weaker under alcohol in group relative to solitary contexts. We did not have specific hypotheses regarding social by physical context interactions, or three-way interactions between beverage group, social context, and physical context given the lack of prior work addressing such effects. However, we evaluated all two and three-way interactions to provide a comprehensive understanding of the impact of context on AR.

Method

Participants

Participants were primarily recruited using online advertisements (e.g., Facebook, craigslist), and flyers posted at a large, public university and in the surrounding community. A total of 2452 individuals responded to recruitment efforts via telephone/email, and 1777 were successfully screened by telephone. Qualified participants must have reported consuming [4/5 drinks (women/men) in a single occasion] on at least one occasion in the past month. Other exclusion criteria included (1) contraindications to consuming alcohol (e.g. a flushing response to alcohol), (2) current use of psychotropic or prescription pain medications, (3) past month use of illicit drugs other than marijuana or daily or near daily use of marijuana, (4) current (past month) alcohol dependence, anxiety, or mood disorder, (5) current or past participation in an abstinence-oriented treatment program, and for women, (6) pregnancy or nursing.

Of the 716 who passed the initial screening criteria, 545 completed the initial survey/interview session. A total of 67 of the 545 session 1 completers were excluded for Alcohol Dependence or another current DSM-IV-TR Axis 1 diagnosis (e.g. Mood or Anxiety Disorder). Of the participants eligible for the alcohol challenge session, 27 were not successfully scheduled for the second session, and 3 declined further participation in the study. Thus, the resulting sample for analysis included 448 participants who completed both lab sessions. Of these participants, 254 were male (56.7%). The majority identified as White (66.5%), followed by those who identified as Asian (9.7%), African American (7.6%), and American Indian/Alaska Native (1.6%). An additional 14.6% selected “other” for race. With respect to ethnicity, 26% of the sample identified as Hispanic/Latinx.

Procedure

All procedures were approved by the Institutional Review Board (IRB) at Arizona State University (Protocol # 1210008481). Prior to the alcohol-challenge session, participants completed a timeline follow back interview of alcohol use over the past 30 days and a comprehensive battery of self-report measures. Surveys assessed drinking behavior, alcohol-related problems, and a range of other measures of relevance to AR. The full assessment took approximately 90 minutes. Structured clinical interviews for alcohol dependence, anxiety, and mood disorders were administered. Those who were screened out due to current diagnoses were paid for their time, provided with referral information for clinical services, and excused from further participation. Included participants were paid for the surveys upon completion of the second session (alcohol challenge).

It is beyond the scope of this paper to report on all of the behavioral outcomes that were assessed in the current study but, in addition to measures of subjective response, participants completed behavioral tasks of impulsivity (i.e., Balloon Analogue Risk Task, BART; Lejuez et al., 2002; MRBURNS, MacPherson, et al., 2012), and we assessed physiological responses (body sway, heart rate, cortisol and alpha amylase), and attentional bias toward alcohol cues (i.e., Implicit Association Test; Greenwald, McGhee, & Schwartz, 1998). Following beverage consumption, participants completed the BART on the ascending limb and the MRBURNS on the descending limb. All other measures were completed on both the ascending and descending limbs. Although analyses identified some main effects of beverage condition on these outcomes, no interactions between beverage condition and drinking context were observed.

Alcohol Challenge.

Participants were randomly assigned to consume alcohol or placebo in one of four contexts: solitary lab, group lab, solitary bar, and group bar. Each session was randomized to one of these contexts, with sessions further randomly assigned to alcohol (60%) or placebo (40%) conditions. We chose this unbalanced design with respect to beverage condition to fully power tests of associations between alcohol response and drinking outcomes. Figure 1 presents sample sizes by context and beverage condition. Given research suggesting that prosocial behavior and mood are stronger in the presence of strangers compared to acquaintances (i.e., Fairbairn, 2017), efforts were made to ensure that participants in group contexts were not familiar with one another (e.g., participants were not allowed to schedule together for the same night). Participants who indicated knowing one another outside the study context upon arrival to the lab were rescheduled for a different night.

Figure 1.

Flowchart of Assignment to Contexts and Beverage Conditions

Lab Contexts.

On evenings designated as solitary lab sessions, participants were escorted to individual rooms where they remained for the duration of the session following the consent process. The three individual rooms were approximately equivalent in size (100–110 square feet) and décor, with each including a desk and chair, computer, filing cabinets, and an absence of decoration (e.g. pictures on the walls) or alcohol-related stimuli. Each room was equipped with a camera feed to a computer in an adjacent room. This allowed participants to remain in a solitary setting while being monitored by project staff. Participants in the group lab condition completed study procedures in mixed gender groups of 2 to 3, in a laboratory space adjacent to the simulated bar lab. This room is larger than the individual lab rooms (approximately 200 square feet) but comparable in all other ways (e.g. computers, filing cabinets, absence of decoration and alcohol-related stimuli).

Bar Contexts.

Participants randomized to the group bar context completed all study procedures in mixed gender groups of 2 or 3 in a simulated bar. The simulated bar was replete with alcohol cues, including a backlit wall behind the bar with stemware and alcohol bottles. Opposite the bar is a seating area with lounge seating, and an entertainment system consisting of a stereo and flat screen televisions. Participants randomized to the individual bar setting completed all study procedures in the same simulated bar setting. They were monitored from a separate room via a one-way mirror.

Beverage Administration.

One to two weeks after the interview/survey session, participants returned to the lab to complete the beverage administration protocol. Upon arrival in the lab, participants’ legal drinking age was verified by trained research assistants who administered breathalyzer tests to ensure a .00 g% Blood Alcohol Concentration (BAC). Following consent, female participants completed a pregnancy test and verified a negative result. Participants then provided measures of baseline current subjective states (with revised instructions for the self-report measures of AR). Trained RAs calculated individual doses based on participants’ gender, height, and weight using a computer program developed by Curtin and Fairchild (2003). Drinks were poured from vodka bottles in full view of participants. In the alcohol condition, beverages contained a 1:3 mixture of 80 proof vodka to mixer (diet 7-up, cranberry juice, and lime juice) to achieve a target BAC of .08 g%. Similar procedures were used in the placebo condition. However, the vodka bottle in this condition contained decarbonated tonic. The rims of the glasses were moistened with vodka, and a squirt of vodka from a plastic lime was added to the top of each drink before serving.

Regardless of beverage condition, each participant was served a total of 3 drinks and given 6 minutes to consume each one with a one-minute break between drinks. The 20-minute dosing protocol was followed by an 8-minute absorption period. Breathalyzer tests then were conducted until the participant reached an ascending limb BAC of .06%, at which time the ascending limb measures of AR were taken. If the participant did not reach a BAC of .06%, ascending limb assessments began 60 minutes following the first BAC measurement. AR measures were repeated at peak BAC and on the descending limb when BACs returned to levels that matched ascending limb assessments as closely as possible. Following the descending limb assessment, BAC readings were taken every 30 minutes until participants were below .03%, at which time they were provided with transportation home. The RAs who tested BAC were aware of the participant’s beverage condition. However, RAs who administered beverages and AR measures were naive to beverage condition. Because the timing of assessments in the placebo condition could not be based on BACs, each placebo participant was yoked to an alcohol participant such that the timing of assessments was matched to a corresponding alcohol participant.

Measures

Interviews

Alcohol Use Disorder and Associated Disabilities Interview Schedule-IV (AUDADIS-IV; Grant & Dawson, 2000).

The AUDADIS-IV assessed current (past month) and past-year diagnoses of alcohol abuse and dependence as well as mood and anxiety disorders, as defined in the DSM-IV (American Psychiatric Association, 1994). The AUDADIS-IV has undergone rigorous evaluation and been found to perform well relative to similar measures including the CIDI (WHO, 1990) and NIMH-DIS (Robins et al., 1981). The AUDADIS-IV was designed for use by lay interviewers rather than trained clinicians and to be appropriate for normative as well as clinical samples. The reliability and validity of the AUDADIS-IV in normative populations has been demonstrated in prior studies (Grant et al., 2003).

Timeline Follow-Back Interview (TLFB; Sobell & Sobell, 1992).

Using standardized procedures, a graduate RA presented participants with a 30-day calendar and asked for daily drinking estimates. The TLFB provides a measure of drinking frequency (drinking days), as well as drinking quantity for each episode. Previous research finds the TLFB has adequate test-retest reliability (r = .92) and is positively associated with other indices of drinking frequency/quantity.

Lab-based Measures

Perceived Intoxication.

Following beverage consumption, participants were asked to estimate the number of standard drinks they had consumed and to estimate their current blood alcohol concentration (BAC). No specific information was provided about standard drinks though participants were previously familiarized with standardized drinks when they completed the TLFB. No information was provided regarding BAC cutoffs (e.g., .08 as the legal limit for intoxication).

Explicit Measures of AR.

A two-item manipulation check was used to assess perceived number of drinks consumed and perceived BAC. The following measure was completed at baseline, on the ascending limb, at peak BAC, and again on the descending limb. The 14-item Subjective Effects of Alcohol Scale (SEAS: Morean, Corbin, & Treat, 2013), was validated in a sample of 244 participants in an alcohol challenge study. The SEAS covers the full valence by arousal affective space, capturing unique effects not covered by other widely used measures: Low Arousal Positive (LAP; relaxed, calm, mellow, secure); Low Arousal Negative (LAN; dizzy, woozy, wobbly) High Arousal Positive (HAP; fun, lively, talkative, funny), and High Arousal Negative (HAN; rude, aggressive, demanding). Response options for the SEAS are on an 11-point scale ranging from 0 (Not at all) to 10 (Extremely). Reliability for the four subscales is good (alphas from .71 to .95) and the SEAS has demonstrated incremental validity in the prediction of alcohol-related outcomes relative to widely used measures including the BAES (Martin et al., 1993) and SHAS (Schuckit et al., 2000).

Data Analytic Plan

First, we examined distributions of all variables to identify any outliers or violations of assumptions of normal distributions. Next, we evaluated the effectiveness of the placebo manipulation by examining perceived BAC and perceived number of drinks consumed within the placebo condition. For participants who received alcohol, we examined the extent to which we achieved the target peak BAC (.08 g%) and matched BACs on the ascending and descending limbs of the BAC curve.

In two of the four drinking contexts, participants consumed their beverages and completed other study tasks in groups. Given the nested structure of the data, we evaluated the need for multilevel analyses by examining Intra Class Correlations (ICCs) and the extent to which there was significant level 2 (group) variability in the outcomes. These analyses (see results section) indicated minimal group level variability in outcomes. Combined with the large number of clusters (326) and small size of the clusters (1 to 3), multilevel models would likely yield comparable results to simpler models (McNeish, 2014). Thus, we conducted standard regression models but accounted for the nested structure of the data using the Cluster command and Type = Complex in Mplus.

Analyses used robust maximum likelihood (MLR) estimation Less than 4% of the data was missing for all measures and Little’s MCAR test for all variables included in any of the models was not statistically significant (χ² = 35.63, df = 39, p = .624). Thus, we used full information maximum likelihood (FIML) estimation of missing data. Covariates included gender, number of alcoholic drinks in the past 30 days from the TLFB, race (White or racial minority), ethnicity (non-Latinx or Latinx), and the baseline score for each outcome. Continuous covariates were grand mean centered. Because high arousal alcohol effects predominate on the ascending limb of the blood alcohol curve, and low arousal effects predominate on the descending limb, analyses used ascending limb scores for HAP and HAN effects, and descending limb scores for LAP and LAN effects (Martin et al., 1993). We have used this approach in our previous studies of AR (e.g., Corbin et al., 2015). For each outcome we initially tested a model that included main effects of beverage condition, social context, and physical context, all two-way interactions, and the three-way interaction. In the absence of a three-way interaction, models were retested with only main effects and two-way interactions. In the absence of two-way interactions, we tested a model with only main effects. Three and two-way interactions were decomposed by examining simple main effects of one variable at different levels of the other variable (e.g., effects of beverage condition in solitary vs. group contexts for a significant beverage condition by context interaction). Lower level main effects and interactions are only presented when they are not qualified by a significant higher order interaction (e.g., a main effect is not discussed if there was an interaction between that variable and another variable in the model).

Given prior work demonstrating greater cohesion in dyads relative to triads (e.g. Solano & Dunnam, 1985), and arguments by some that dyads do not represent true groups (Moreland, 2010), sensitivity analyses repeated each model using only groups of two in the group conditions as most groups were two-person groups (82 two-person groups; 20 three-person groups). If results were not replicated in the two-person groups (e.g., a non-significant effect became significant or a significant effect became non-significant), results are reported separately for two- and three-person groups.

Results

Preliminary Analyses

Examination of variable distributions indicated some departure from normality. There was modest positive skew for drinking in the past 30 days and both low and high arousal negative AR (all skewness values <= 3.02), which robust maximum likelihood estimation is well equipped to handle. At the first assessment of alcohol effects following beverage consumption, all participants reported on the number of standard alcoholic drinks they thought they consumed, and what they thought their current blood alcohol concentration (BAC) was. Only 4 participants in the placebo condition indicated that they thought they had consumed no alcoholic drinks, and only 2 reported that they thought their BAC was 0.00. A comparison of these six participants to the other participants in the placebo group identified minimal differences in reports of subjective effects following beverage consumption (partial eta squared values ranging from .001 to .006 with .01 representing a small effect) so all placebo participants were retained in the analyses. Among participants in the placebo condition, the average estimated number of alcohol drinks consumed was 2.680 (SD = 1.086), and the average estimated BAC was .044 (SD = .024). By comparison, among participants in the alcohol condition, the estimated number of drinks consumed was 3.350 (SD = 1.070), and the estimated BAC was .067 (SD = .024). Thus, the placebo response, relative to alcohol, was 80% for estimated drinks, and 66% for estimated BAC. Within the alcohol condition, the average ascending BAC was .068 g% (SD = .011), the average peak BAC was .084 g% (SD = .015), and the average descending BAC was .066 g% (SD = .011). Thus, BACs were well matched at the ascending and descending limb assessments of subjective response to alcohol. Descriptive statistics are reported by beverage condition and context in tables 1 and 2, respectively. Examination of intraclass correlations (ICCs) among participants assigned to a group context generally showed small ICCs (HAP = .03; HAN = .03; LAP = .14; LAN = .29). In addition, there were 35 clusters for HAN effects and 39 clusters for LAN effect with no group variability due to scores of 0 in all group members. Given the relatively small ICCs, large number of clusters with no variability for some outcomes, and small number of participants per cluster, we chose to use the Cluster command and Type = Complex within Mplus rather than testing multilevel models.

Table 1.

Descriptive Statistics for Subjective Response across Alcohol Conditions

	Alcohol Condition n = 270	Placebo Condition n = 178
	M (SD)	M (SD)
High Arousal Positive (HAP)	6.28 (2.18)	4.64 (2.23)
High Arousal Negative (HAN)	0.65 (1.24)	0.26 (0.64)
Low Arousal Positive (LAP)	6.37 (2.05)	6.23 (2.17)
Low Arousal Negative (LAN)	.96 (1.48)	0.22 (.60)

Note. M and SD are used to represent mean and standard deviation, respectively. Sample sizes for the groups are based on the number of participants assigned to the conditions. Actual means and standard deviations are based on participants with valid data. Two participant in each group were missing data on one or more of the AR measures.

Table 2.

Descriptive Statistics of Subjective Response across Physical and Social Contexts

	Bar Condition		Lab Condition
	Group n = 115	Solitary n = 109	Group n = 109	Solitary n = 115
	M (SD)	M (SD)	M (SD)	M (SD)
HAP	6.06 (1.81)	5.07 (2.42)	6.44 (2.21)	4.97 (2.55)
HAN	0.58 (1.14)	0.27 (0.63)	0.63 (1.37)	0.49 (0.93)
LAP	6.17 (1.80)	6.07 (2.34)	6.58 (2.09)	6.44 (2.13)
LAN	.71 (1.21)	0.48 (1.16)	0.83 (1.54)	0.63 (1.11)

Note. HAP = High Arousal Positive, HAN = High Arousal Negative, LAP = Low Arousal Positive, LAN = Low Arousal Negative. M and SD are used to represent mean and standard deviation, respectively. Sample sizes are based on all participants assigned to each condition but means and standard deviations are based on those with valid data. One participant in each condition did not have valid data on one or more of the AR measures.

Primary Analyses

High Arousal Positive (HAP) Effects.

In the full model with all two- and three-way interactions, covariates of gender, b = .48, standard error [SE] = .15, p = .002, and baseline HAP effects, b = .66, SE = .05, p < .001, were significant, with stronger HAP effects after beverage administration among men and those with higher baseline HAP. The three-way interaction was not significant, p = .81, so the model was re-estimated with only the two-way interactions. In this model, there was a significant beverage condition by physical context interaction, b = −.58, SE = .28, p = .04. Simple main effects of beverage condition were significant within both the bar, b = 1.36, SE = .27, p < .001, and non-bar settings, b = 2.031, SE = .28, p < .001, but the effect was stronger in the non-bar contexts (See Figure 2). There was also a significant main effect of social context, b = .78, SE = .15, p < .001, such that participants experienced stronger HAP effects in a social relative to a solitary context, regardless of beverage condition.

Figure 2.

Beverage Condition by Physical Context Interaction for High Arousal Positive Effects

High Arousal Negative (HAN) Effects.

In the full model with all two- and three-way interactions, covariates of gender, b = .26, SE = .08, p = .001, and baseline HAN effects, b = .53, SE = .10, p < .001, were significant, with stronger HAN effects after beverage administration among men and those with higher baseline HAN. The three-way interaction was not significant, p = .34, so the model was re-estimated with only the two-way interactions. In this model, there were no significant interactions. When the two-way interactions were removed, there were significant main effects of beverage condition, b = .35, SE = .07, p < .001, and social context, b = .16, SE = .07, p = .03, such that participants experienced stronger HAN effects under alcohol and in social contexts, relative to placebo and solitary contexts. When the model was tested separately including only two-person groups, the main effect of social context was not statistically significant, b = .10, SE = .08, p = .18. When the model was tested for only three-person groups within the group contexts, the main effect of social context was significant, b = .31, SE = .13, p = .02.

Low Arousal Positive (LAP) Effects.

In the full model with all two- and three-way interactions, the only significant covariate was baseline LAP effects, b = .74, SE = .05, p < .001, with stronger LAP effects after beverage administration among those with higher baseline LAP effects. The three-way interaction was not significant but there was a trend toward significance, b = −1.16, SE = .66, p = .08. The effect was similar when restricting the analysis to groups of 2 within the group contexts though this effect reached statistical significance (p = .04). When the two-way interactions between beverage condition and physical context were examined separately within the solitary and group contexts, neither was statistically significant, but the effects were in opposite directions (b = .69, SE = .50, p = .16 in the solitary context; b = −.59, SE = .44, p = .18 in the social context). In the solitary context, effects tended to be stronger in the bar than the non-bar context. In the group context, effects tended to be stronger in the non-bar than the bar context (See Figure 3). The group non-bar context was the only context in which there was a significant alcohol effect, relative to placebo, b = .74, SE = .34, p = .03. In the model without the three-way interaction, none of the two-way interactions were statistically significant (all p values > .47). In the main effects model, there was only a trend toward a main effect of physical context, with marginally stronger LAP effects in the non-bar relative to the bar context (b = −.27, SE = .16, p = .08).

Figure 3.

Marginal Three-Way Interaction between Beverage Condition, Physical Context, and Social Context for Low Arousal Positive Effects

Low Arousal Negative (LAN) Effects.

In the full model with all two- and three-way interactions, ethnicity, b = .29, SE = .13, p = .03, alcohol use in the past 30 days, b = −.007, SE = .002, p < .001, and baseline LAN effects, b = .50, SE = .15, p = .001, were significant predictors of LAN effects following beverage consumption. Latinx participants, lighter drinkers, and those who reported stronger baseline LAN effects reported stronger LAN effects following beverage consumption. The three-way interaction was not statistically significant, p = .75, and the model without the three-way interaction identified no significant two-way interactions (all p values > .15). In the main effects model, there was a significant effect of beverage condition, b = .71, SE = .10, p < .001, and a marginally significant effect of social context, b = .20, SE = .12, p = .096. Participants who consumed alcohol reported stronger LAN effects than those who consumed placebo, and participants who drank in social contexts reported marginally stronger LAN effects than those who drank in solitary contexts.

Discussion

The current study sought to examine the effects of both physical and social context on the full range of subjective alcohol effects using a placebo-controlled alcohol administration paradigm. To our knowledge, while prior studies have shown both social and physical contextual effects on subjective response to alcohol and alcohol expectancies, none have assessed effects of both the physical and social context, and only one has examined the full range of subjective response. Therefore, this work fills important gaps in the literature.

Interestingly, context seemed to have a greater impact on positive alcohol effects relative to negative alcohol effects, a finding that is consistent with some prior work (Corbin et al., 2015). More specifically, results showed that there was a significant main effect of beverage condition for high arousal positive (HAP) effects. However, this main effect was qualified by a significant condition by physical context interaction, such that those who consumed alcohol in the non-bar setting experienced significantly more HAP than those that consumed alcohol in the bar setting. This was unexpected, as it was hypothesized that those who drank in the bar would experience more positive stimulating effects of alcohol. However, examination of Figure 2 suggests that the smaller effect of alcohol in the bar context was driven, at least in part, by the stronger placebo response in this context. In other words, there was less room for increases in HAP effects among those who drank in the bar context relative to those who drank in the non-bar context. Taking the findings for both physical and social context into account, the group bar context provided the most robust placebo response with respect to high arousal positive effects.

There was also a marginally significant main effect of physical context for low arousal positive effects (LAP) that was qualified by a marginal three-way condition by social context by physical context interaction. While the two-way interactions between beverage condition and physical context were not significant in either the solitary or group context, the nature of the interactions were in the opposite directions. More specifically stronger LAP effects were reported in the bar relative to the lab setting for participants who drank in a solitary context, whereas stronger LAP effects were reported in the lab relative to the bar for those who drank in a group context. The only cell in which beverage condition significantly predicted low arousal positive effects was the group non-bar context. This is consistent with our previous finding that LAP effects were stronger in a non-bar group context than in a group bar context (Corbin et al., 2015). The significant impact of alcohol on LAP effects in group non-bar contexts suggests this might be the optimal setting for studying such effects. Studies that involve contexts that simulate a living room environment for example, that do not have cues of the bar but are still ecologically valid, may be even better suited to detecting these effects. The lack of LAP effects in other contexts may also be related to the developmental period of participants in the current study. Young adults may not be drinking for tension reduction motives that would likely map on to low arousal positive effects, particularly in a bar context where more stimulating responses to alcohol may predominate. Finally, alcohol effects may have been limited due to strong expectancies for tension reduction which would be activated by placebo consumption. Future studies that include both a no-alcohol and placebo condition are needed to examine this possibility.

Context effects for negative alcohol effects were relatively minimal and there were no interactions between beverage condition and context. With respect to HAN effects, there were only main effects of beverage condition and social context. Alcohol facilitated HAN effects relative to placebo and those who drank in the group context experienced stronger HAN effects than those who drank alone, regardless of beverage condition. Interestingly, sensitivity analyses showed that the latter effect was more pronounced in 3-person groups relative to 2-person groups. Thus, HAN effects may be even more prominent in larger groups, though additional studies are needed to directly test this hypotheses. Overall, the findings for HAP and HAN effects are consistent with prior research demonstrating stronger alcohol expectancies in group settings. An important implication of these findings is that studies with an explicit interest in expectancy effects, or that wish to have strong control for expectancy effects, might benefit from the use of a group context, particularly if the researchers are interested in outcomes associated with high arousal alcohol effects (either positive or negative).

Low arousal negative effects (LAN) showed the weakest impact of context. The only significant effect was for beverage condition with stronger LAN effects under alcohol, relative to placebo. The only sign of a context effect was a marginally significant main effect of social context, such that those in the solitary context experienced a trend toward more low arousal negative effects than those in the group contexts. There were no signs of interactions between beverage condition and context. The fact that LAN effects are relatively unaffected by context may help explain consistent links between these effects and later risk for alcohol problems despite most prior studies of these effects occurring in contexts with limited ecological validity (e.g., solitary lab settings). Thus, this aspect of AR may demonstrate more stable effects across contexts, which has important implications for research designed to capture AR in real-world drinking contexts (i.e., EMA studies).

In addition to lab-based studies, it is important to contextualize our findings relative to prior work using EMA. Our results largely support prior EMA studies that highlight the importance of contextual effects on drinking behavior (Freisthler, Lipperman-Kreda, Bersamin, & Gruenewald, 2014; Ray et al., 2010). For example, drinking in groups and/or in contexts with drinking cues has been shown to increase craving for alcohol and alcohol consumption (Kuerbis et al., 2020; O’Donnell et la., 2019; Trela, 2018). Overall, our results suggest that stimulating responses to alcohol may depend on the context, particularly the social context, in which one drinks. In real-world drinking contexts where expectancies are not separated from pharmacological effects, drinking in groups is likely to facilitate both positive and negative high-arousal effects across contexts. Thus, these effects may be particularly interesting to study in bar and party settings that are often characterized by heavy consumption and alcohol-related risk behavior. Future EMA studies that are able to capture these real-world drinking contexts would serve to further inform our understanding of contextual effects on stimulant alcohol responses.

In contrast to the stimulating effects of alcohol, low arousal positive effects may be sensitive to both the physical and social context with the strongest response occurring in low arousal group settings like a home or living room environment. Limited lab-based research has addressed such ecologically valid but low stimulation contexts, but this seems like a fruitful area for further inquiry. Finally, studies that use lab contexts with limited ecological validity may have the most success in accurately capturing low arousal negative responses to alcohol. This is important as it provides confidence in the vast literature demonstrating relations between LAN effects and risk for later alcohol problems.

In addition to informing future research on AR, broadening our knowledge of contextual effects has potentially important public health implications. If we can recognize high risk environments and have a better understanding of how these environments contribute to alcohol-related risk, we will be better able to intervene. For example, prevention efforts for those who drink in highly stimulating group contexts may be most effective if they focus on potential risks associated with high arousal, including over-consumption and engagement in high-risk behavior (e.g., risky sex, physical aggression). In contrast, those who drink in low stimulation group contexts may benefit most from efforts to prevent drinking for tension reduction or self-medication of other negative mood states. Targeting those who experience minimal LAN effects may be fruitful regardless of drinking context, and there is already some evidence to support the value of providing feedback about subjective alcohol effects as an aspect of prevention efforts (Schuckit et al., 2015; Schuckit et al., 2016). With greater understanding of the full range of effects and the contexts in which they predominate, such efforts may have even greater utility. Future prevention and intervention approaches might also train and encourage the use of protective behavioral strategies in settings that pose particular risk for specific individuals. Promoting use of protective strategies has shown benefits as a more global strategy (Araas & Adams, 2008; Barry & Merianos, 2018; Borden, et al., 2011), but tailoring such approaches to the contexts of greatest relevance for each individual may further enhance their efficacy.

Although the current study has potentially important implications for both lab-based studies of AR and prevention efforts targeting AR, the findings should be considered in light of several methodological limitations. First, as stated previously, the current study did not include a no-alcohol condition, so we cannot speak to the magnitude of expectancy effects. Although at least one prior study of affective responses that included both a placebo and no-alcohol control found similar results across these two groups (Sayette et al., 2012), the lack of a no-alcohol control complicates our interpretation of placebo efficacy/expectancy effects. For example, we suggest that placebo response for HAP effects is facilitated by the group context, but this presumes that HAP effects in a social context would differ between a placebo group and a no alcohol control. If these two groups were to look similar with respect to HAP effects, we might conclude that a group context simply facilitates positive mood rather than the efficacy of a placebo manipulation. Thus, future studies are needed to differentiate expectancy effects from alcohol effects, and to make stronger inferences about the extent to which context uniquely effects placebo response versus mood more generally. In addition to the lack of a no alcohol control, we used a single dose of alcohol, so it is not possible to generalize the findings to lower or higher alcohol doses. Future dose response studies are needed to determine the BACs at which contextual influences are most apparent. Our study was also limited in its ability to rule out effects of other drugs, as we did not conduct a urine drug screen prior to beverage administration. However, we did take precautions including instructing participants to not use any other substances 24 hours prior to the study start time and excluding participants who self-reported daily or almost daily use of marijuana, or use of other illicit drugs.

It is also important to note that participants completed the alcohol challenge in groups that varied in number and gender composition. While it is possible that group dynamics could have impacted feelings of subjective response, multilevel models did not identify substantial effects of the group in which participants completed the lab session. Thus, there was not substantial group variability that might be explained by differences in group size or gender composition. Our group conditions also included groups of individuals who were unfamiliar with one another. Although such group settings are common in the real world, they are probably less common than drinking that takes place in intact groups. Given evidence that alcohol has stronger effects on social bonding among unfamiliar versus familiar groups (Fairbairn, 2017), future studies directly comparing alcohol effects in novel vs. pre-existing groups may further our understanding of group processes involved in alcohol use and related behavior.

Finally, the study sample was limited in both age range and the range of drinking behavior. We only included participants who were between the ages of 21–25 to ensure they were above the legal drinking age and still in the peak period of risk for alcohol-related problems. Thus, we do not know whether we would find a similar pattern of results in an older population. For instance, older adults may be more likely to drink for tension reduction rather the stimulating/social effects of alcohol, which may lead to different contextual influences on AR. The range of drinking behaviors was also limited as we excluded very light drinkers and those with a past month AUD diagnosis. Although the sample had a truncated range of drinking behavior, it was similar to nationally representative samples with respect to mean levels of alcohol use (Grant, Stinson, & Hartford, 2001), and there are ethical constraints regarding administration of alcohol to light or dependent drinkers. Nonetheless, it is possible that the exclusion of both the lightest and heaviest drinkers may have impacted the findings. Although there are some important limitations of the study sample, the sample included both college and non-college students, had relatively equal samples of men and women, and was relatively diverse with respect to race and ethnicity. This provides some confidence that the results are likely to generalize beyond the particular sample used in the current study.

In summary, the current study extends upon previous literature by being the first, to our knowledge, to examine the interactive effects of social and physical context on subjective responses to alcohol. Results suggest that both positive and negative high arousal effects of alcohol may be particularly prevalent when drinking in groups in party or bar settings. On the other hand, a group setting in a less stimulating environment such as a living room may be the most appropriate context to detect low arousal positive effects of alcohol. Finally, it appears that low arousal negative effects of alcohol are relatively stable across social and physical contexts. These results not only lead to interesting and novel future research directions, but also have important public health implications. Gaining a greater understanding of contextual influences on alcohol response may serve to inform prevention and intervention efforts and ultimately lead to reductions in risky drinking patterns and alcohol-related consequences.

Public Significance Statement.

The current study shows that drinking context affects subjective experiences of alcohol effects. These results have important implications for scientists studying alcohol effects and may lead to a better understanding of the role of context in transitions from social to problem drinking.