The impact of alcohol on brain response in social context: A hyperscanning alcohol-administration trial

Abstract

Evidence for the integration of alcohol into social life dates to the beginning of recorded history. Humans’ tendency to combine social interaction with alcohol has been attributed to alcohol’s ability to shift social perception, with behavioral research suggesting alcohol fosters social connection and diminishes perceived social threat. Yet the acute effects of alcohol on brain responses in social context are as yet unexplored. Combining experimental alcohol-administration with an EEG hyperscanning paradigm, the current study examines the effect of alcohol on evaluation of self- and other-linked performance. Social drinkers (N = 128) were administered either an alcoholic (target BAC 0.08 %) or control beverage in pairs. Dyads engaged in a gambling task while event-related potential Feedback Effects (FEs) to wins and losses were assessed simultaneously in both participants. Findings indicated a significant correlation in FEs among players and observers. Results further revealed alcohol effects that emerged specifically in the social domain, with alcohol intoxication significantly reducing the magnitude of FEs among observers paired with a stranger. In contrast, alcohol’s impact on FEs was non-significant when participants observed a familiar partner, as well as when participants were actively engaged in playing. Taken together, findings provide evidence for core social (e.g., observational) dimensions of human cognition and further offer clues surrounding neural pathways supporting the widespread integration of alcohol into social life.

1. Introduction

“The culture of drink endures because it offers so many rewards: confidence to the shy, clarity for the uncertain, solace to the wounded and lonely, and above all, the elusive promises of friendship and love”

—Pete Hamil, A Drinking Life: A Memoir (Hamill, 1994)

“An intelligent man is sometimes forced to be drunk to spend his time with fools.”

—Ernest Hemingway, For Whom the Bell Tolls (Hemingway, 1987)

The human capacity for navigating expansive social networks is arguably among the species’ greatest strengths (Aronson, 2012; Dawkins, 1982). Human social structures have facilitated the pooling of labor and exchange of ideas on an unprecedented scale, permitting us to conquer environmental challenges, enlarge the species, and bring into being marvels of technology (Gulati, 1995; Social Structures: A Network Approach, 1988; Lin, 2007). Yet the same social world that permits international space stations simultaneously presents formidable challenges (Aronson, 2012). Whereas humans’ close primate relatives coexist in small societies (Dunbar, 1998; Greene, 2014; Morrison, 2017), humans are confronted daily with a comparatively expansive social group, exposed to the unpredictable actions of a network of unknown others (Beck, 1992; Taleb, 2007). Some researchers have gone so far as to posit that the human brain has evolved as a uniquely social organ (Dunbar, 1998; LeDoux et al., 2020; Ramachandran, 2011). In light of this, core to the understanding of human strength is the understanding of factors that promote humans’ ability to cope in an environment comprising unfamiliar people with whom they are nonetheless enmeshed and interdependent (Baumeister, 2005; Goleman, 2007; Wasserman and Faust, 1994).

Among the most widely employed tools for coping in these heterogeneous social contexts can be found in the drug alcohol (Alcohol and Humans: A Long and Social Affair, 2020; Steele and Southwick, 1985). Evidence for the integration of alcohol into social life dates to the beginning of written history (McGovern, 2013; Standage, 2006; Slingerland, 2021), with current studies indicating that between 13 % and 26 % of social interactions conducted during leisure time feature alcohol consumption (Rot et al., 2008). Reward from alcohol is intensified when consumed in company versus alone (Doty and de Wit, 1995; Kirkpatrick and de Wit, 2013; Pliner and Cappell, 1974), social enhancement is the most widely endorsed reason for consuming alcohol (Cooper, 1994), and contexts featuring unfamiliar individuals (e.g., bars and large parties) have long been associated with particularly heightened—and sometimes hazardous—patterns of consumption (Brown, 1985; Casswell and Zhang, 1997; Fairbairn et al., 2018; Senchak et al., 1998; Shih et al., 2015). Indeed, alcohol’s longstanding and pervasive role in human socialization is intertwined with the fact that it has also been identified as the drug causing the most aggregate harm across both individual users and society (Nutt et al., 2010). Thus, identifying brain mechanisms promoting alcohol’s as yet mysterious socially cohesive properties has emerged as a research priority.

Yet a range of challenges has emerged in the study of acute drug effects on the human brain, particularly for research seeking to model these as they might manifest in social context. In particular, direct vasoactive effects of alcohol confound signals associated with benchmark imaging modalities such as functional magnetic resonance imaging (fMRI), which examine blood flow as a proxy for brain activity (Bjork and Gilman, 2014; Rickenbacher et al., 2011; Strang et al., 2015; Marxen et al., 2014). Further, alcohol has psychoactive properties that exaggerate reactivity to immediate environmental cues (Steele and Josephs, 1990; Fairbairn and Sayette, 2014; Curtin and Fairchild, 2003; Sayette, 2017; Fairbairn and Kang, 2025), thus presenting challenges for modalities such as fMRI and positron emission tomography (PET) that feature atypical recording environments (e.g., noisy, supine, enclosed). Thus, no research to date has explored brain effects of alcohol in the context of in-vivo social contact, limiting understanding of addictions’ underpinnings as well as our broader knowledge of how humans navigate the social world surrounding them.

Over the past two decades, neuroscientific methods featuring multi-participant recording arrays have emerged, permitting the scanning of two brains concurrently—a technique referred to as hyperscanning—in social contexts (Dikker et al., 2021; Hari et al., 2013; Hasson et al., 2004; Montague et al., 2002; van Schie et al., 2004; Zhang et al., 2019; Jin et al., 2020; Yu and Zhou, 2006). This advancement offers critical insights into the neural mechanisms of social perception and coordination that single-subject recordings cannot provide. In fact, evidence for the utility of this socially informed neuroscience has accumulated, with studies suggesting that the “mere-presence” of another individual significantly alters both resting state and task-evoked brain activity (Fairbairn and Kang, 2025; Hari et al., 2013; van Schie et al., 2004; Zhang et al., 2019; Jin et al., 2020; Yu and Zhou, 2006; Barker et al., 2015). These studies have demonstrated that interpersonal contexts, such as cooperative gambling tasks and competitive prisoner’s dilemma tasks, reliably evoke shared neural dynamics linked to social coordination (Barker et al., 2015; Verbeke et al., 2014; Holroyd, 2022; Balconi and Vanutelli, 2017; Tugin et al., 2015). Among the most widely employed recording modalities within this literature is electroencephalography (EEG), which has the advantage of offering a relatively unconstrained recording environment, permitting in-vivo social contact between participants, while further circumventing alcohol-specific recording artifacts by offering a direct measure of brain activity (versus an indirect measure based on blood flow) (Dikker et al., 2021; Hari et al., 2013; van Schie et al., 2004; Zhang et al., 2019; Dikker et al., 2017; Djalovski et al., 2021; Hasson et al., 2012). Segments of the continuous EEG signal can further be yoked to elements of the task environment, yielding components known as event-related potentials (ERPs), which provide temporally precise and functionally meaningful measures of brain activity (Fabiani et al., 2007; Münte et al., 2000). Hyperscanning approaches that specifically measure ERPs permit in-vivo social contact while also holding constant key elements of the stimulus space, so providing interpretable individual brain responses in the context of social activities that mirror real-world drinking settings (e.g., television watching; gaming). Thus, while early hyperscanning studies historically focused on inter-brain coupling metrics derived from free-form EEG signals (Kinreich et al., 2017; Czeszumski et al., 2020; Kayhan et al., 2022; Nam et al., 2020), hyperscanning research has since expanded to include research examining ERP metrics derived from two brains in tandem, enhancing the interpretability by holding constant task environment and thus context for assessment. As described next, hyperscanning ERP research has begun to examine how feedback-related processing and performance monitoring unfold in cooperative and competitive tasks, demonstrating shared neural responses between actors and observers. Yet, how psychoactive substances such as alcohol may modulate these neural dynamics in social contexts remains unexplored.

Of particular utility for studying the neural underpinnings of how we evaluate both ourselves and others in social context are ERP responses associated with performance monitoring. These include inherent responses to our own behavior (e.g., making a motor error; the so-called Error-Related Negativity (ERN)) as well as responses to feedback about performance or task outcomes, such as losses and wins in a gambling task (Feedback-Related Negativity or FRN; Reward Positivity or RewP) (Falkenstein et al., 1991; Gehring et al., 1993; Holroyd and Coles, 2002; Hajcak et al., 2005; Holroyd and Krigolson, 2007; Proudfit, 2015; Krigolson, 2018). Performance monitoring ERPs have been linked to phasic changes in mesencephalic dopamine signals affecting activity in a brain network, including the anterior cingulate cortex, medial frontal cortex, and striatum, which is responsible for updating behavior in response to feedback that deviates from expectations, whether in a positive or negative direction (Holroyd and Coles, 2002; Gehring and Willoughby, 2002; Holroyd et al., 2004). The difference between the response to losses and negative feedback, which elicits more negativity (FRN), and the response to gains and positive feedback (RewP) is maximal over medial frontal scalp locations around 250 ms after the presentation of the feedback/outcome. Strikingly, multiple ERP hyperscanning studies have reported that this same type of feedback effect (FE) (Dikker et al., 2017) can be elicited not only in those actively involved in tasks but also among observers with no direct role in task performance, showing that, in social contexts, people engage their own reinforcement-learning systems in response to others’ performance (Tugin et al., 2015; Proudfit, 2015; Marco-Pallarés et al., 2010; Balconi and Vanutelli, 2017).

The degree to which people engage in performance monitoring, as either an actor or an observer, is affected by the task context and the nature of the interpersonal relationships represented therein (Fiske, 2019). Important, the behavior of unfamiliar interaction partners (i.e., strangers) tends to evoke a heightened level of scrutiny (Heuristics and Biases, 2002), particularly in situations involving mutually-dependent losses and gains. For example, when an unfamiliar colleague makes a visible mistake during a shared task, we may react with internal scrutiny or discomfort, whereas a similar mistake from a close friend might be met with amusement or empathy. Reflecting the context-dependent nature of performance monitoring processes, FEs are larger for outcomes with more motivational significance (Gehring and Willoughby, 2002), and FE size is positively correlated with dispositional anxiety, consistent with claims that these responses arise from the engagement of adaptive control processes in the face of uncertainty about actions and their outcomes (Cavanagh and Shackman, 2015). Correspondingly, a number of ERP hyperscanning studies have found that the presence of an observer increases FEs in actors (Tian et al., 2015), especially under competitive conditions (Czeszumski et al., 2019). FEs are on average smaller for observers than for actors (Ma et al., 2011; Koban et al., 2012), at times emerging as pronounced (van Schie et al., 2004; Yu and Zhou, 2006; Koban et al., 2010) and at times minimal (Ma et al., 2011; Koban et al., 2012), and these observer FEs (oFEs) have been shown to reflect the individual’s evaluation of the outcome from their own perspective (van Schie et al., 2004; Itagaki and Katayama, 2008; Koban and Pourtois, 2014). Carp et al. (Carp et al., 2009) further found that lower perceived similarity between observer and actor is correlated with larger oFEs, consistent with the idea that we have evolved to judge the actions of strangers through a framework of vigilance (Heuristics and Biases, 2002), aiding us in the compilation of evidence surrounding the unknown other’s characteristics to help us inform decisions about whether to cement new social ties or instead invest resources elsewhere (Tversky and Kahneman, 1974).

These dynamics align with tenets of the social-attributional model (Fairbairn and Sayette, 2014), which proposes that individuals engage heightened social evaluative vigilance when interacting with unfamiliar others as part of an adaptive strategy to manage the uncertainties and potential social threats posed by strangers. This vigilance involves increased monitoring of others’ behaviors and heightened sensitivity to social cues, which can contribute to social anxiety or evaluative concerns in novel settings. Building on this, a later expansion and elaboration of this framework (the “social-cognitive model” (Fairbairn and Kang, 2025)) posits that alcohol reduces perceived social threat and attenuates evaluative monitoring, particularly in contexts of social uncertainty, such as interactions with strangers. Together, these models suggest that alcohol may play a socially buffering role by dampening the cognitive and affective processes underlying social vigilance, thereby facilitating smoother social interactions and reducing evaluative stress in unfamiliar social environments. The widespread integration of alcohol into social life, particularly spaces featuring stranger interaction, may therefore be attributable to alcohol’s ability to silence the internal judge, reducing the cognitive demands associated with performance evaluation in social context and thus leaving open the potential for a more cohesive social sphere (Aronson, 2012; Axelrod, 2006; Bradford et al., 2013; Fiske, 2004; Hefner and Curtin, 2012). This theoretical perspective informs our hypothesis that alcohol’s effects on performance monitoring will be most pronounced in stranger interactions, where evaluative concerns are elevated, and will be diminished in interactions between friends, where trust and familiarity often preclue the need for such vigilance.

Although alcohol’s general dampening effect on performance monitoring ERPs (e.g., ERN) has been established (see Fairbairn et al., 2021, for a meta-analysis (Fairbairn et al., 2021)), all of the previous studies used a single-person paradigm, and whether these effects vary across social contexts remains unknown. Given the impact of social factors on attention and other aspects of cognitive functioning, effects of alcohol on processes like performance monitoring may well be different in social compared to non-social contexts, with important implications for understanding patterns of alcohol use and corresponding risks. Thus, ERP hyperscanning offers a unique advantage to the study of alcohol’s effects by (a) incorporating in-situ social contexts, in which participants can play multiple roles, and (b) allowing the simultaneous measurement of functionally well-characterized, task-evoked responses. In this regard, the present study is the first to combine drug-administration methods with an ERP hyperscanning paradigm to concurrently examine FEs in actors and observers, exploring alcohol’s influence on evaluative processes in social contexts. Uniquely, while prior research has investigated the effects of cannabis in naturalistic social settings (Fachner, 2006), to our knowledge, no studies have applied EEG hyperscanning to examine alcohol’s impact on brain processes during live social interactions.

The primary objective of the current study was twofold: (1) to examine whether acute alcohol intoxication selectively attenuates FEs in social contexts characterized by heightened evaluative demands—specifically, during observation of a stranger’s performance; and (2) to assess whether this modulation differs across social roles, by evaluating alcohol’s effects on neural responses in both active task performers (players) and passive observers wihin dyads composed of either friends or strangers. We hypothesized that alcohol’s impact on FEs would be non-significant among friends, where broader social evaluative effects are subdued. Conversely, we anticipated that alcohol would substantially reduce the magnitude of the FE among strangers, influencing both task performers and observers.

2. Materials and methods

Ethics Statement.

Written informed consent was obtained from all participants prior to their inclusion in the study. The privacy rights of human subjects were strictly observed throughout the research process. The study protocol was reviewed and approved by the Institutional Review Board (IRB) of the University of Illinois at Urbana-Champaign, under the approval number 16,263.

Participants consisted of 128 healthy drinkers ages 21–30 recruited from the local community. A sensitivity power analysis using the simr (Green and MacLeod, 2016) package in R revealed that this sample size was large enough to detect a small-medium interaction effect assuming an alpha level of 0.05 and 80 % power. These participants constitute a subset of those enrolled in a large project focused on alcohol reinforcement in unfamiliar social context who also underwent hyperscanning EEG procedures during the time period from 10/2021–2/2023 (ClinicalTrials NCT03449095). To meet eligibility, participants had to identify at least one same-gender friend—someone with whom they considered themselves to have been friends of at least 6 months—who also met inclusion criteria for the study. For the present study we selected dyads wherein both members yielded at least 30 usable trials for all critical conditions (player win and loss and observer win and loss). Dyads were fully counterbalanced for alcohol and relationship condition. Exclusion criteria included medical conditions for which alcohol consumption is contraindicated; recent use of psychoactive substances (e.g., use of hallucinogens more than five times or any use of opioids, benzodiazepines, or other depressants/sedatives in the past 30 days); current use of medications—excluding prescription eye drops or ADHD medications—that could not be withheld for 24 h; pregnancy in females; a history of major psychiatric diagnosis or severe alcohol use disorder; and a history of skull fracture. Please see Table 1 for full descriptive statistics of the participant sample, including sociodemographic characteristics and alcohol and other substance use history.

Table 1.

Descriptive statistics of the participants.

Variable	Descriptive Statistics
	M (SD)
Age	23.02 (2.11)
Gender	n	Percentage
Female	64	50 %
Male	64	50 %
Race/Ethnicity	n	Percentage
White	69	53.91 %
Black	4	3.13 %
Asian	52	40.63 %
Native Hawaiian/Other Pacidic Islander	2	1.56 %
More than one race	2	1.56 %
Hispanic/Latino	34	26.56 %
Alcohol use (past 30 days)	M (SD)
Drinking Days	7.48 (4.71)
Number of Drinks per Occasion	3.64 (1.75)
Other substance use (past 30 days)	n	Percentage
Hallucinogen (e.g., psilocybin, LSD)	2	0.02 %
Heroin	0	0.01 %
Opioids	0	0.00 %
Benzodiazepines	0	0.00 %
Other Depressants/Sedatives	0	0.00 %
Stimulants	1	0.01 %

Following successful completion of screening, two friend dyads were invited to the laboratory for simultaneous beverage-administration sessions. On the day of the experimental session, half of participants were randomized to complete experimental procedures in the company of their own friend whereas the other half were assigned to complete procedures in the company of the friend of the other participant. All participants were casually and informally introduced at study outset to ensure no prior acquaintance in the stranger condition, with procedures carefully designed to maintain unfamiliarity prior to beverage administration. Among participants assigned to the friend condition, friends reported knowing each other for an average of 2.53 years (SD = 2.54), spending time together 2.61 h per week (SD = 1.54) throughout their acquaintance. In response to the question, “How would you describe your current level of friendship with this individual (1–9),” the mean rating was 7.27 (SD = 1.55). All friendship questions were completed independently and in private to reduce potential bias.

Upon entering the lab, participants’ height and weight were taken (required for precise alcohol dosing). All participants were required to provide 0.00 % BAC (Intoximeters Alco-Sensor IV) reading and female participants were required to take a pregnancy test (Human chorionic gonadotropin [hCG] urine test strip). Following the completion of baseline questionnaires—including measures of demographics, alcohol use history, and other assessments unrelated to the current study (for a full list, see ClinicalTrials.gov entry NCT03449095)—participants consumed their assigned study beverages (Alcohol or Control) in pairs over the course of 36 min. The dose of alcohol was adjusted based on each participant’s estimated body water content, calculated using an updated version of the Widmark formula (Watson et al., 1981; Widmark, 1932) that accounts for weight, height, sex, and age, and calibrated to achieve a peak BAC of approximately 0.08 %. Participants in the alcohol condition received the study beverage in the form of a vodka-soda cocktail, whereas participants in the control condition received the volumetric equivalent of a soda beverage. Due to unanticipated compensatory reactions from placebo manipulations in prior laboratory alcohol research (Fairbairn et al., 2015; Testa et al., 2006), and to replicate use conditions in the real-world contexts we were seeking to understand, no deception was employed regarding the content of participants’ beverages in the control condition (i.e., control participants were aware they were not consuming alcohol).

Following beverage-administration, participants were brought into separate rooms to provide BAC readings and complete self-report questionnaires, including an Eight-Item Mood Measure (8 MM) (Fairbairn et al., 2015; Fairbairn and Sayette, 2013), the Inclusion of Other in the Self (IOS) Scale (Aron et al., 1992), and the Perceived Group Reinforcement Scale (PGRS) (Creswell et al., 2014) (see Table 2). Upon completion of the self-report questionnaires, each dyad member was fitted with a stretchable electrode cap (Easycap, BrainVision LLC), to which 13 silver/silver-chloride electrodes were attached; these encompassed left and right prefrontal, medial frontal, medio-lateral frontal, medio-lateral central, and medio-lateral parietal locations, as well as midline central (equivalent to Cz), posterior, and occipital sites. To record eye movements and blinks, electrodes were placed on the outer canthus of each eye and the left orbital ridge. All scalp electrodes were referenced online to the left mastoid and re-referenced offline to the average of the right and the left mastoids. The continuous EEG was amplified using Sensorium amplifiers through a bandpass filter of 0.02–100 Hz and recorded at a sampling rate of 250 Hz.

Table 2.

Full correlation matrix on the associations between fes and self-report measures.

	PGRS	8MM—Positive	8MM—Negative	IOS	Player FE	Observer FE
Social Bonding (PGRS (Creswell et al., 2014))	1	.401**	−0.227**	.529**	−0.091	−0.045
Positive Mood (8 MM◦)	.401**	1	−0.264**	.258**	−0.149	0.066
Negative Mood (8 MM◦)	−0.227**	−0.264**	1	−0.204*	−0.012	−0.192*
Self-Other Overlap (IOS (Aron et al., 1992))	.529**	.258**	−0.204*	1	−0.012	.192*
Player FE	−0.091	−0.149	−0.012	−0.137	1	0.122
Observer FE	−0.045	0.066	−0.192*	.192*	0.122◦◦	1

^**p < .001.

^*p < .05.

^◦The 8 MM assesses four negative mood states (annoyed, sad, irritated, bored) and four positive mood states (cheerful, upbeat, happy, content), encompassing all quadrants of Russell’s affective circumplex. (Russell, 2003).

^◦◦Significant associations were found between a player’s FE and the FE of an observing individual during simultaneous play, but the correlation became non-significant when linking a player’s FE with their own FE as an observer, emphasizing the role of shared settings over individual factors in shared FE variation (see Social Effects: Player-Observer Correlations).

Dyads next completed an ERP hyperscanning task while seated side-by-side (30 cm separation) in the EEG recording chamber (see Fig. 1a). A gambling task was chosen as one that elicits a strong and reliable FE (Holroyd and Coles, 2002; Gehring and Willoughby, 2002; Miltner et al., 1997; Yeung and Sanfey, 2004) while also demonstrating sensitivity to social observation (Koban et al., 2010). In the course of this game, participants are presented with two values (combinations of $10, $20, $30, $40, and $50) on a computer screen, each of which represents a possible win or loss amount (e.g., “$20 or $50″). The player selected one of these values by pressing a button on a handheld response device, and, following this selection, they were informed whether they won or lost the amount of money they selected (see Fig. 1b). After a practice block, dyads engaged in 4 rounds (i.e., blocks) of the game, alternating roles of player and observer (order randomly assigned), with each round consisting of 80 consecutive trials. In each block, participants were shown 32 stimuli with a value difference of 10, 24 stimuli with a difference of 20, 16 with a difference of 30, and 8 with a difference of 40. Specific pairings of values were presented 8 times in each block, with position of the numbers on the screen counterbalanced. Order of presentation of the value pairings was randomized within each block. The feedback on each trial (i.e., whether the choice was a win or a loss) was generated randomly with the constraint that, across the experiment, half the trials were losses and half wins. Feedback consisted of the chosen value presented in the center of the screen, with losses presented in red with a minus sign in front of the value and gains in green with a plus sign.

Fig. 1.

Experimental set-up and ERP task design of the hyperscanning procedures.

Fig. 1a (left). Dyads completed the ERP task while seated side-by-side in the EEG recording chamber. The player made monetary selections using a button press on a handheld response device. Fig. 1b (right). Experimental stimuli and task design. In the course of this game, participants are presented with two values (combinations of $10, $20, $30, $40, and $50) on a computer screen, each of which represents a possible win or loss amount. Participants are required to select one of these values, and, following this selection, they are informed whether they won or lost the amount of money they selected.

Participants were informed that the goal of the game was to maximize the amount of “money” earned as a team, with the potential to trade money earned in the game for prizes. A cooperative prize structure was chosen for this task in light of our focus on potential social-cohesive effects of alcohol. Dyads were asked to refrain from talking or moving during the course of the game, and EEG output was continuously monitored to ensure compliance. To further maximize observer engagement, participants not actively playing were instructed to keep mental track of wins and losses. Each trial began with an array of fixation crosses for 500 ms, followed by 300 ms of blank screen. Value choices were presented for 1200 ms, followed, 300 ms later, by the feedback screen for 400 ms. The next trial began 1100 ms later.

EEG processing followed established guidelines (Picton et al., 2000). Data were digitally filtered from 0.2 to 30 Hz. To measure FEs, which are observed 200–350 ms after stimulus onset, 1000 ms epochs were extracted, from 100 ms prior to stimulus onset to 900 ms post-stimulus onset, with the 100 ms pre-stimulus period used as a baseline. Thresholds for exclusion of trials contaminated by blocking, drift, large eye movements, and excessive muscle activity were established for each participant using condition-blind visual screening. Eye blink activity was corrected (Dale, 1994). Average trial loss to artifacts was 11 % (range 0–58 %), and all available, artifact-free trials were included in the computation of the mean amplitudes. Based on previous research, the mean amplitudes of evoked potentials to negative (loss) versus positive (gain) feedback were extracted separately for each trial and role (player and observer), averaged across four medial frontal sites (see top of Fig. 2) (Holroyd and Coles, 2002; Gehring and Willoughby, 2002; Holroyd et al., 2004).

Fig. 2.

Player-observer feedback effect (FE) correlations.

Top: Red dots indicate the scalp positions of the four electrodes used in the current FE analyses for both players and observers Bottom: Scatter plot showing significant associations between the magnitude of a player’s FE and the corresponding FE magnitude elicited in the observer during the same phase of the gambling task.

Multi-level modeling was used to assess the impact of alcohol on FEs, accounting for the nested structure of the data, with individual gain and loss trials (Level 1) nested within participants (Level 2), who were nested within dyads (Level 3). For within amplitude level analyses, fixed effects in the models included Feedback (gain vs. loss), Beverage Condition (alcohol vs. control), and Social Context (friend vs. stranger), as well as their interactions. Random intercepts were specified at both the participant and dyad levels to account for between-subject and between-dyad variability. Importantly, ERP responses to gain and loss feedback were modeled separately within this analytic framework. FE scores—created at the level of the subject by subtracting evoked potentials for loss from gain trials—were used only in models assessing the correlation between player and observer ERP responses during simultaneous gameplay.

Data and code availability statement.

All data and code required for replicating analyses can be accessed at https://osf.io/rzexg/. The analysis code developed for this study, including the multilevel models employed to assess the impact of alcohol on FEs, is also available via the OSF repository. The code is written in SAS (SAS Institute Inc., Cary, NC, USA), and detailed instructions for replicating the analyses are provided within the repository.

3. Results

Beverage Manipulation Check:

Participants assigned to receive alcohol achieved an average BAC of 0.068 % (SD=0.019) following the group drink period, ultimately rising to a peak BAC of 0.076 % (SD=0.012) measured immediately following the hyperscanning task.

Self-Report Associations:

Analyses revealed associations between player FE values and self-reports of social bonding and mood immediately prior to the gambling task. Player FEs were significantly associated with both self-other overlap (Aron et al., 1992) with the individual observing them (their co-player), b=0.261, t = 2.57, p=.0124, as well as negative mood measured after interacting with this individual, b=−.149, t=−2.27, p=.0264. Higher player FE values were linked with greater self-other overlap and lower negative mood. Whereas associations between oFEs and self-reports did not reach significance, effects tended in the opposite direction, with higher levels of social-emotional reward being associated with lower oFE values (see Table 2 for a full correlation matrix).

3.1. Social effects

Player-Observer Correlations:

Analyses indicated significant associations between the magnitude of a player’s FE and the magnitude of the FE elicited by the individual observing that player simultaneously during the same portion of the gambling task (see Fig. 2). Specifically, as the size of FEs among players increased, the size of oFEs also significantly increased, b=0.194, t = 2.20, d = 0.255, p=.0312, an effect that remained significant after controlling for trial loss, b=0.191, t = 2.21, d = 0.251, p=.0306, as well as for the total amount won/lost in usable trials, b=0.197, t = 2.30, d = 0.256, p=.0249. In contrast, when players’ and observers’ FE values were linked at the level of the participant vs. simultaneous play (i.e., when a player’s FE was linked with their own FE as an observer measured later/earlier in the same session) the player-observer FE correlation emerged as non-significant, b=0.132, t = 1.54, d = 0.027, p=.1289, as it did when observer values were shuffled across dyads, b=0.050, t=0.67, d= −0.076, p=.5027. Thus, in this case, (shared) setting appeared to trump person-level factors as a driver of shared variation in FEs. Finally, the player-observer FE association did not appear to be driven by similarity/selection effects within friend dyads, as the association between player and observer FEs emerged as significant selectively among strangers, b=0.290, t = 2.87, d = 0.386, p=.0073, and did not reach significance among friends, b=0.124, t=0.97, d = 0.162, p=.3394.

3.2. Alcohol effects

Player FEs:

In line with pre-registered planned comparisons (NCT03449095), primary models explore feedback by alcohol interactions separately within unfamiliar (stranger) dyads, with friend dyads offering a secondary point of comparison for understanding generalizability of effects across variable social contexts. For player FE values, a robust (expected) effect of feedback emerged across experimental conditions. For those assigned to complete study tasks in the company of a stranger, there was no significant interaction between alcohol condition and feedback, b=0.316, t=0.66, d = 1.359, p=.5171, with the effect of feedback emerging as significant and strong across both alcohol, b = 1.659, t = 5.16, d = 1.481, p<.0001, and control conditions, b = 1.975, t = 5.52, d = 1.607, p<.0001. A similar pattern of findings was observed among participants assigned to complete the task with a friend, with no significant Alcohol by Feedback interaction, b=−.187, t=−.39, d = 1.129, p=.698, and a significant effect of feedback across both alcohol, b = 1.801, t = 6.24, d = 1.481, p<.0001, and control, b = 1.614, t = 4.24, d = 1.607, p=.0002, conditions. See Fig. 3 and Table 3.

Fig. 3.

Alcohol’s impact on player and observer FEs.

a. (top): Grand-averaged EEG responses recorded across medial and lateral frontal electode sites (F3, F4, F7, F8; see Fig. 2), separately plotted for players and observers across all experimental conditions. a. (bottom): Bar plots visualizing the sizes of FEs for each condition. Error bars represent the standard error (SE) of the mean.

Table 3.

Descriptive statistics of FEs (mV) across conditions.

Conditions			Mean	sd
Strangers	Control	Observer FE	0.79	1.62
		Player FE	1.97	1.74
	Alcohol	Observer FE	−0.14	1.87
		Player FE	1.66	1.58
Friends	Control	Observer FE	0.01	2.41
		Player FE	1.61	2.04
	Alcohol	Observer FE	0.23	1.92
		Player FE	1.80	2.01

Observer FEs:

For FEs focused specifically in the social (i.e., observational) domain, effects emerged selectively according to participants’ intoxication level and social surroundings. Specifically, for participants assigned to complete tasks with a stranger, there was a significant interaction between alcohol condition and feedback, b=0.935, t = 2.54, d = 0.642, p=.0165 (see Fig. 4). A significant oFE emerged selectively among unfamiliar dyads assigned to consume no alcohol, b=0.793, t = 4.11, d = 0.693, p=.0003, while this effect was eliminated with alcohol intoxication, b=−.142, t=−0.45, d=−0.107, p=.6549. In contrast, irrespective of alcohol group assignment, oFEs among participants who were already friends more closely resembled effects captured among intoxicated strangers. Within friend dyads, no significant interaction between alcohol condition and feedback emerged, b=0.216, t=0.39, d = 0.007, p=.6991—oFEs were non-significant within both control, b=0.011, t=0.03, d = 0.006, p=.9801, and alcohol, b=0.227, t=0.71, d = 0.167, p=.485, conditions. See Figs. 3 and 4, and Table 3.

Fig. 4.

Interaction between Beverage Condition and Feedback on ERP Amplitudes among Observers.

Left: Among participants assigned to complete tasks with a stranger, there was a significant interaction between alcohol condition and feedback, b=0.935, t = 2.54, d = 0.642, p=.0165. A significant oFE emerged selectively among unfamiliar dyads assigned to consume no alcohol, b=0.793, t = 4.11, d = 0.693, p=.0003, while this effect was eliminated with alcohol intoxication, b=−.142, t=−0.45, d=−0.107, p=.6549. Right: Among friend dyads, no significant interaction between alcohol condition and feedback emerged, b=0.216, t=0.39, d = 0.007, p=.6991—oFEs were non-significant within both control, b=0.011, t=0.03, d = 0.006, p=.9801, and alcohol, b=0.227, t=0.71, d = 0.167, p=.485, conditions.

3.3. Exploratory analyses

Although our primary analyses examined gain and loss trials together to assess FE responses, analyzing them separately allows detection of asymmetrical effects—such as selective attenuation of reward responses (e.g., RewP (Proudfit, 2015)) or amplification of loss signals (e.g., FRN (Hauser et al., 2014))—that may be masked in a difference score. Therefore, we conducted exploratory analyses examining the main effect of alcohol separately on gain and loss trials. Results indicated that alcohol did not produce statistically significant main effects on either gain (b = 1.28, t = 1.83, d = 0.162, p=.072) or loss (b = 1.07, t = 1.64, d = 0.145, p=.107) ERP amplitudes. Similarly, social context showed no significant effects on gain (b=−0.32, t=−0.44, d = 0.039, p=.660) or loss (b=−0.48, t=−0.71, d = 0.063, p=.478) trials when modeled independently. These findings suggest that the observed alcohol- and context-related effects are most evident in the contrast between gain and loss responses, i.e., the FE, rather than in absolute amplitude changes within either condition.

4. Discussion

By employing experimental alcohol-administration methods and an EEG hyperscanning paradigm, the present study sheds light on the interconnectedness of brains in social context. The study strategically focused on a class of event-related brain potentials—performance monitoring ERPs, or feedback effects (FEs)—which have been linked to reward processing and reinforcement learning. These responses have been linked to phasic shifts in mesencephalic dopamine signals in a brain network involved in reinforcement learning, observed when outcomes deviate from expectations, whether favorably or unfavorably. (Holroyd and Coles, 2002) We explored the impact of alcohol on performance monitoring processes, both among those actively engaging in tasks and among those observing them. Replicating prior work (Jin et al., 2020; Jin et al., 2020; Koban et al., 2010; Koban and Pourtois, 2014), we observed that FEs are shaped by social and emotional processes, with positive correlations between the size of player FEs, self-other overlap, and reduced negative mood. Novel to this work, we discovered a significant in-the-moment correlation in the magnitude of FEs among players and observers, offering support for core social dimensions of human cognition. Critically, findings further revealed alcohol effects that emerged specifically in the social domain, with alcohol intoxication significantly reducing the magnitude of FEs among observers paired with a stranger. In contrast, alcohol’s impact on FEs was non-significant when participants observed an individual who was already a friend, as well as when participants were actively engaged in playing, pointing to unique capabilities for alcohol in blunting social-evaluative processes in unfamiliar social settings. Overall, results of this study not only shed light on how alcohol might interface with neural processes linked with performance evaluation in social context but appear to further bolster the view of the brain as an organ steeped in not only the intra- but also the inter-personal domain.

One potential mechanism underlying this “social buffering” effect is alcohol’s modulation of the neural circuits that support social evaluation. Specifically, FEs are thought to originate primarily from the medial prefrontal cortex (mPFC), with strong contributions from the anterior cingulate cortex (ACC) (Holroyd and Coles, 2002; Gehring and Willoughby, 2002). These regions form part of a broader performance monitoring network that tracks action outcomes, signaling when adjustments are needed following errors or unexpected feedback. Importantly, in social contexts, this same network plays a central role in social evaluative processing, with the ACC particularly implicated in detecting, monitoring, and responding to social conflict and feedback (Somerville et al., 2006; Lavin et al., 2013). Increased ACC activation has been associated with heightened sensitivity to social rejection, negative evaluation, and the perceived need to regulate one’s behavior under social scrutiny (Somerville et al., 2006; Eisenberger et al., 2003). In interactions with strangers—where social uncertainty is high—this evaluative network is likely to become more active, increasing vigilance to and monitoring of not only one’s own behavior but the actions of others. Alcohol may buffer this response by dampening activity within the ACC and related prefrontal regions, reducing the salience of negative cues in social situations and decreasing the cognitive demands associated with social vigilance. This dampening could, in turn, lead to observable reductions in ERP amplitudes, including both FEs and oFEs, as observed in our study, particularly during performance monitoring in socially evaluative situations involving strangers.

The differential effects of alcohol in stranger versus friend interactions observed in our study can be understood within the ramework of social-cognitive theories (Fairbairn and Sayette, 2014; Fairbairn and Kang, 2025) governing interpersonal vigilance. Interactions with strangers are typically marked by heightened social uncertainty and an increased need to monitor social cues, as individuals work to assess the intentions, reliability, and trustworthiness of unfamiliar others (Chang et al., 2010; Fareri et al., 2012). This increased vigilance is adaptive, serving to protect against potential social threats and to guide decisions about whether to invest in new social relationships. In contrast, interactions with friends involve established trust and familiarity (Baumeister and Leary, 1995; Clark and Lemay 2010), reducing the cognitive demands of social monitoring. In these familiar contexts, individuals can relax evaluative processes, confident that their behaviors are less likely to be negatively judged or socially penalized. Within this framework, alcohol’s social buffering effects may be most pronounced when evaluative demands are highest—such as in interactions with strangers—by dampening the neural systems that support social vigilance (e.g., the ACC and mPFC). When interacting with friends, however, the need for such vigilance is already low, leaving less opportunity for alcohol to further reduce evaluative processing.

Several prior studies have also found that measures of empathy or self-other overlap can influence oFEs, with larger effects in more empathetic individuals and in dyadic contexts with higher levels of perceived closeness (Jin et al., 2020; Ma et al., 2011; Thoma and Bellebaum, 2012). However, it is important to note that these studies employed task situations in which observer and player outcomes were independent. Given that oFEs have been shown to reflect an egocentric perspective on outcomes (van Schie et al., 2004; Itagaki and Katayama, 2008; Koban and Pourtois, 2014), our results suggest that, in the absence of task demands to do so, people’s willingness to voluntarily “take on” others’ actions may be enhanced by empathy or self-other overlap. Indeed, Ma et al. (Ma et al., 2011) found that the increased oFEs for friends compared to strangers only held if the observer was never actively involved in the game. Results in the present study then highlight important variability in social appraisals across relationship types and contexts, as well as a potential role for alcohol in moderating these appraisals. Notably, in a cooperative context, we found a different pattern, in which oFEs were present in stranger dyads but diminished (statistically not attested) in friend dyads. Other studies have also found oFEs in dyads made up of strangers when, as here, players’ performance matters for the observer (Yu and Zhou, 2006; Marco-Pallarés et al., 2010; Koban et al., 2010). Moreover, Yu et al. (2019) manipulated social closeness among strangers by establishing an “in-group” (a partner in a previous task) and an “out-group” (a previous competitor) and found larger oFEs for out-group than in-group members. Thus, oFE magnitudes can also signal the observer’s level of vigilance toward players’ behavior, elicited by cooperative task demands and/or social appraisal. In a cooperative task, friends seem more likely to reduce vigilance – to “trust” their partners to perform the task in a manner that will benefit the dyad and therefore to divert resources away from performance monitoring (Carp et al., 2009). In contrast, tasks that create interdependencies among strangers may give rise to a state of heightened vigilance for observers. This state of alertness assists in navigating situations featuring uncertainty surrounding outcomes with potential implications for the self, increasing attentiveness and so potentially impacting learning (Marco-Pallarés et al., 2010).

We did not find an effect of alcohol on player FEs, which aligns with some prior research (Euser et al., 2011), although not all, as suggested by our meta-analysis (Fairbairn et al., 2021). Of note, previous studies have focused on alcohol’s effects on ERPs within solitary settings. Alcohol’s impact on attentional and performance monitoring processes has been shown to vary across task conditions, with alcohol often failing to yield effects in situations featuring highly salient stressful stimuli and/or singular task demands (Steele and Josephs, 1990). One possibility is that the social context and interdependent nature of outcomes in our task served to “up the ante,” inducing a myopic focus on the task at hand—an attentional state over which alcohol typically exerts muted effects (Steele and Josephs, 1990; Steele et al., 1986; Steele and Josephs, 1988). This underscores the necessity for future research to integrate the social dimension when examining alcohol’s neurocognitive effects, as social factors may modulate the extent to which alcohol affects performance monitoring processes.

Although previous studies have independently captured FEs separately among both players and observers (Yu et al., 2019; Yang et al., 2022; Peterburs et al., 2017; Huberth et al., 2019; Kang et al., 2010), our current study, which also leverages a notably large sample, is (to the best of our knowledge) the first to show that there is synchrony in FEs between players and observers within the same task context. This finding has methodological as well as theoretical implications, as it presents a promising avenue for exploring the interconnectedness of brains within social contexts, through the use of hyperscanning methods in tandem with well-established EEG paradigms. Typically, hyperscanning paradigms have measured correlations (synchrony) between individuals in overall patterns of brain activity across time periods of minutes or longer, revealing the impact of factors like joint attention for learning and interpersonal evaluation (Dikker et al., 2021; Hasson et al., 2004; Kawasaki et al., 2013; Redcay and Schilbach, 2019). Nevertheless, as with any methodology, concerns have been voiced about hyperscanning paradigms, particularly pertaining to the complexities of interpreting these large-scale inter-brain metrics while also considering environmental and behavioral factors (Holroyd, 2022; Redcay and Schilbach, 2019; Novembre and Iannetti, 2021). In light of these considerations, our study instead combined hyperscanning with a well-studied ERP paradigm, facilitating in-vivo social interaction while simultaneously controlling for crucial aspects of the stimulus environment. Extracting ERPs while hyperscanning provides discernible individual brain responses linked to specific cognitive and neural functions, while still affording opportunities to capture synchronized brain activity among multiple individuals, thereby deepening our insight into the interplay between social contexts and the impact of alcohol on modulating social brain functioning. In this case, supplemental analyses highlighted the strength of situational (vs. individual) factors for understanding player-observer FE correlations: Although the FE of an individual engaged in playing the gambling task was significantly linked to the oFE of the individual simultaneously observing them, the magnitude of a given player’s FE was un-linked with the magnitude of their own oFE as captured later in the experiment. Moreover, these situational effects persisted even with control for task outcomes (i.e., amount won/lost), pointing to an important social dimension underlying the effect pattern.

Indeed, we observed that player FE magnitudes are positively related to self-other overlap. Although the literature contains a limited number of studies with adequate power for between-subject analyses, a relatively consistent pattern across studies is the finding that performance monitoring effects can be enhanced by social motivations. For example, players’ responses to errors (Error Related Negativities) have been found to be larger for people who score higher on measures of empathy (Larson et al., 2010; Santesso and Segalowitz, 2009). In our cooperative gambling task, where players’ choices mattered for their co-present observers, we found that player FEs were larger when they felt more connected to their partner and thus presumably were more motivated to make choices that benefitted the dyad. The further correlation between player FEs and reduced negative mood resonates congruently with previous literature highlighting the link between diminished FE values and negative affective states, as observed within contexts including depression (Keren et al., 2018), although when modeled together it appeared that social connection explains more variance in FE size than does emotional state. Future studies could consider examining other factors such as trait empathy, mood, and anxiety in adequately powered samples to further explore these preliminary results.

Several limitations of the present study warrant consideration. First, although our sample included substantial racial and ethnic diversity—addressing a significant gap in alcohol research—it included a relatively high proportion of demographic groups among whom subpopulations have historically demonstrated low levels of drinking (e. g., Asian participants). This diversity represents a strength, as it extends the generalizability of alcohol research beyond the predominantly White samples that have historically characterized the field (Chartier and Caetano, 2010). However, cultural factors are known to influence drinking patterns, social behaviors, and sensitivity to alcohol’s effects (Heath, 2000; Chrzan, 2013), and these differences may also extend to performance-monitoring processes examined in the study. For instance, certain Asian cultural contexts (e.g., those influenced by Confusian heritage) tend to emphasize interpersonal harmony and concern for others’ evaluations and, as a result, individuals from these backgrounds may experience heightened self-monitoring in unfamiliar social situations, where maintaining face and avoiding social errors are prioritized (Huang and Chang, 2017; Nichols, 2015). Such effects may differ from those observed in Western cultural contexts, where direct communication and individual assertiveness may be more normative (Guo et al., 2025; Ma and Jaeger, 2010). However, to our knowledge, whether cultural norms interact with alcohol and social context to influence neural processing has not yet been directly tested. Future research should explore whether alcohol’s social and neural effects vary across racial/cultural groups with differing drinking norms and practices, ideally in even larger and more demographically diverse samples. Second, one design consideration that may have influenced the findings was the instruction for observers to mentally track wins and losses during gameplay. This instruction was intended to ensure that observers remained attentive to task outcomes. However, this requirement may have introduced a mild cognitive load that could have detracted from participants’ full emotional engagement with the task, particularly across a high number of trials. While we did not observe any clear attenuation of feedback-related ERP components indicative of disengagement, future studies might consider alternative strategies to sustain observer attention without imposing additional cognitive demands. Third, while the decision not to use deception in the control beverage condition was intentional and aligned with real-world drinking contexts, it is important to acknowledge that this approach may have introduced expectancy effects, potentially influencing participants’ social evaluations and neural responses. Future studies might consider incorporating a placebo condition to assess whether the current findings replicate under blinded conditions.

Taken together, the results highlight the critical interplay of alcohol and social settings on performance monitoring processes. Social cohesion tends to increase performance monitoring for players, and we showed here for the first time that, at least in cooperative tasks, FEs between players and observers become synchronized. However, alcohol dampened the oFEs that were otherwise elicited in stranger dyads, pointing to reduced vigilance and possibly increased cohesion and trust. This interaction of alcohol with social dynamics suggests implications for societal functions and tendencies toward addiction. (Nutt et al., 2010) Considered together with research pointing to a role for hazardous consumption in contexts featuring social novelty, our research indicates that alcohol may release us from a state of hypervigilant social evaluation in the early stages of relationship formation, so potentially releasing resources for shared pleasures.

Data and code availability statement

All data and code required for replicating analyses can be accessed at https://osf.io/rzexg/. The analysis code developed for this study, including the multilevel models employed to assess the impact of alcohol on FEs, is also available via the OSF repository. The code is written in SAS (SAS Institute Inc., Cary, NC, USA), and detailed instructions for replicating the analyses are provided within the repository.

Data availability

Data will be made available on request.