Introduction

Alcohol is one of the most widely used psychoactive substances in the world1,2, and it is most often consumed in social settings3,4,5,6. Acute alcohol intoxication has both positive and negative effects on social behavior. Often, alcohol intoxication induces strong social facilitative effects, including increased talkativeness, agreeableness, and social bonding7,8,9,10,11,12. These desirable social outcomes have been implicated in the development of alcohol-related problems, as individuals may escalate their drinking due to these potent social rewards3,13. Conversely, alcohol intoxication has also been associated with adverse social consequences, such as heightened aggression and interpersonal conflicts, highlighting the significant social costs that alcohol consumption can have11,14. Given robust evidence indicating the existence of both positive and negative social consequences of alcohol intoxication, there has been much interest in investigating the underlying mechanisms that may produce such varied social effects (e.g., 3,15,16).

A recent and growing body of research indicates that alcohol-induced changes in social cognition (e.g.,17,18,19) may contribute to these desirable and undesirable social consequences of alcohol use (e.g., 17,20,21). Social cognition is conceptualized as a set of mental abilities that permit successful social interactions, including perceiving, processing, interpreting, and responding to social stimuli22,23,24. Here, we focus on alcohol’s effects on two specific social cognition factors commonly examined in the alcohol administration literature (i.e., emotion recognition and empathy).

Emotion recognition

Emotion recognition, the ability to accurately identify emotions in others16,25, plays an important role in successful social communication26,27. Two meta-analyses demonstrated that individuals with alcohol use disorder (AUD) showed impairments in emotion recognition conveyed by facial expressions compared to healthy controls25,28, suggesting that long-term heavy alcohol use may impair the ability to recognize emotions in others. In addition, many controlled laboratory experiments have investigated the acute effects of alcohol intoxication on emotion recognition of facial expressions in social drinkers with mixed findings. A recent systematic review of this literature16, however, found no reliable evidence of alcohol-induced changes in the recognition of positive (i.e., happiness) or negative (i.e., sadness, anger, fear, disgust, contempt) emotions across 17 studies, regardless of the dose of alcohol administered or the type of emotion recognition task used (cf.29). This calls into question whether alcohol intoxication reliably induces changes in the ability to recognize emotions in others.

Notably, though, there are important methodological limitations to prior alcohol administration studies that have examined alcohol’s acute effects on emotion recognition (see16 for a detailed discussion of these limitations). First, many studies did not report conducting power analyses to determine adequate sample sizes, and it is likely that some of these prior studies were underpowered to find effects of alcohol on emotion recognition. Indeed, several alcohol administration studies included in the systematic review reported using fewer than 20 participants in their studies (e.g.,30,31,32). Second, several studies administered doses of alcohol (0.2 g/kg-0.4 g/kg) that may not have been sufficiently large to induce substantial intoxication, especially in the often heavy social drinkers recruited for such studies (e.g.,17,33,34). This is important because researchers theorize that alcohol’s impact on social cognition, such as emotion recognition and empathy, is mainly due to its interference with the cognitive appraisal and processing of social information35. Models like the appraisal-disruption model35 and the alcohol myopia model36 propose that this interference occurs only when intoxication is sufficient to significantly impair various cognitive domains, including perception, reasoning, and appraisal processes35. Third, all prior studies compared an alcoholic beverage to a placebo beverage, but many did not report conducting a placebo manipulation check (e.g.,17,33), and of the ones that did, only a few reported successful placebo deception37,38. Any meaningful conclusions drawn about the pharmacological vs. expectancy effects of alcohol on emotion recognition require successful placebo deception (i.e., if participants in the placebo condition do not believe they are consuming alcohol, it becomes difficult to disentangle expectancy effects from the pharmacological effects of alcohol on emotion recognition). Finally, many studies used static images of facial expressions in their emotion recognition assessments (e.g.,38,39), which fail to capture the dynamic nature of facial expressions conveying emotions and have been criticized for lacking ecological validity (e.g.,40,41).

This study was designed to address the methodological limitations of prior studies investigating the acute effects of alcohol on emotion recognition. Given previous mixed findings in the literature, we refrained from formulating specific directional hypotheses about alcohol’s effects on emotion recognition. Instead, our goal was to examine these effects while addressing key limitations identified in previous studies. Here, in a much larger sample of young adults (N = 156) than most prior studies (e.g.,31,32), we used an alcohol-administration method developed to simulate an episode of binge drinking (as defined by the National Institute on Alcohol Abuse and Alcoholism42) in a laboratory session, involving the administration of a dose of alcohol that was two to four times larger than the doses administered by most prior studies (e.g.,17,33,34) and sufficient to cause intoxication10,43,44. Further, we achieved near-perfect placebo deception, enabling us to disentangle alcohol expectancy effects from the pharmacological effects of alcohol on emotion recognition. Finally, we used a well-validated behavioral measure of emotion recognition based on dynamic, multimodal emotion expressions to capture emotion recognition abilities in a more ecologically valid manner (that is, the Geneva Emotion Recognition Test [GERT]45).

Empathy

In addition to alcohol’s effects on emotion recognition, we also examined alcohol-induced changes in empathy. Empathy lacks a universally accepted definition (see18,40,46), but is typically construed as having a cognitive component (i.e., the capacity to understand another’s perspective or emotional state) and an affective component (i.e., the capacity to respond emotionally to or share another’s emotional state47,48). Empathy plays a crucial role in how individuals relate to and connect with others, influencing the quality of their relationships and interactions49,50, and thus alcohol’s impact on empathy could have downstream implications on social behaviors.

Systematic reviews and meta-analyses have demonstrated that individuals with AUD show impairments in empathy18,51 and a related social cognition factor (i.e., theory of mind: the capacity to understand other people by ascribing mental states [e.g., thoughts, intentions, beliefs] to them28,52), suggesting that chronic heavy alcohol use is also associated with impairments in empathy and theory of mind. Few controlled laboratory experiments, however, have investigated the acute effects of alcohol intoxication on empathy or theory of mind. In a sample of 60 healthy adult drinkers (ages 18–50), one prior study tested the effects of consuming a low dose of alcohol (i.e., 0.24 g/kg in females, 0.29 g/kg in males) vs. a placebo beverage on empathy using the Multifaceted Empathy Test (MET53,54), which is a widely used computer task that purports to assess cognitive and affective empathy53,54,55,56 [Note: Despite the labels, the cognitive empathy component of the MET mirrors the methodology used in emotion recognition tasks (see16), and the affective empathy component of the MET simply asked participants how much they “empathized” with the person in each photograph. We agree with recent calls to use more specific descriptors for such task components57,58,59. As such, we refer to the cognitive empathy component of the MET as assessing emotion recognition and the affective empathy component of the MET as assessing perceived empathy]. Results demonstrated that alcohol enhanced perceived empathy for positive stimuli on the MET, particularly in participants with lower trait empathy17, as assessed by the Interpersonal Reactivity Index (IRI, empathic concern scale60). Two other studies using higher doses of alcohol (i.e., 0.64–0.82 g/kg) similarly demonstrated an enhancement effect of acute alcohol intoxication on social cognitive factors related to empathy (e.g., theory of mind61, common ground62).

Evidence is mixed overall, though, on the acute effects of alcohol on empathy and related constructs (i.e., theory of mind), as five studies reported null or opposite effects to those described above63,64,65,66,67, perhaps because many of the same methodological limitations evident in emotion recognition studies are also concerns here. That is, prior studies testing the acute effects of alcohol on empathy and related constructs have tended to use small sample sizes (Ns < 25; e.g.,63,65) and low doses of alcohol (≤ 0.56 g/kg; e.g.,64,67) [Note: One study64 included a higher dose of alcohol (0.8 g/kg), but only 16 participants consumed this dose]. Notably, prior studies have used empathy measures that have been criticized for lacking ecological validity (e.g.,41,68,69,70), as nearly all of them required social drinkers to consume alcohol while alone in a room and then complete third-person empathy tasks on a computer (also while alone), having participants merely observe others through pictures, videos, or written descriptions when making empathy inferences (see40). Alcohol administration studies that utilize more ecologically valid group drinking protocols and assess empathy using second-person paradigms (i.e., having participants engage/interact with another person when making empathy inferences), will likely enhance our understanding of alcohol’s effects on empathy18,40.

In the current study, we tested the impact of an intoxicating dose of alcohol (0.74 g/kg in females and 0.82 g/kg in males) vs. placebo, delivered in a multi-participant alcohol administration setting, on cognitive and affective empathy. Due to limited and mixed findings in this area, no directional hypotheses were proposed. We rather aimed to clarify alcohol’s effects on empathy using a more robust methodology than prior studies. To determine whether alcohol affects empathy differently based on the assessment method, we measured empathy using both third-person (i.e., the MET53,54) and second-person (empathy prompts regarding interaction partners in the lab) methods. Finally, we tested whether alcohol’s effects on empathy varied across participants with differing levels of trait empathy, as those with lower trait empathy might be more susceptible to alcohol-induced changes in empathy according to both theoretical models40 and empirical findings17.

Methods

This study was approved by the Carnegie Mellon University Institutional Review Board, and all methods were performed in accordance with the relevant guidelines and regulations for human participation alcohol administration studies. Informed consent was obtained from all participants prior to their inclusion in the study. This study includes a subset of young adult heavy drinkers participating in a longitudinal study aiming to predict the development of alcohol problems from alcohol’s acute effects in a laboratory social setting (see71 for a detailed description of the methods). The empathy and emotion recognition measures used here were integrated into the parent study while it was in progress, resulting in some participants not receiving the full set of measures.

Participants

Participant inclusion criteria included owning a smartphone (for ambulatory assessment methods included in the parent study), drinking at least one day per week, and binge drinking at least four times in the past month, defined as consuming 5 + /4 + standard alcoholic drinks (for male and female individuals, respectively) during one occasion. Participants also had to affirm that they could comfortably drink at least three drinks in 30 min. Participant exclusion criteria included current or past AUD or substance use disorder treatment, diagnosis of moderate or severe AUD, diagnosis of a serious psychiatric disorder (e.g., schizophrenia, bipolar disorder), substance use dependence (other than nicotine or caffeine), non-English speaking, a gender identity that did not align with sex assigned at birth (due to sex-based differences in alcohol dosages and power requirements related to sex-based analyses), and contraindications to consuming alcohol in the lab, based on the following criteria: a prior bad/allergic reaction to the type or amount of beverage used in the study; nursing, pregnancy, or currently trying to conceive; a medical condition or medication use that contraindicated alcohol use; and BMI < 18.5 or > 29.9 to control the alcohol dosage. The final sample included 156 participants (see below for participant descriptives).

Procedures

Briefly, participants were randomly assigned to groups of 3 unacquainted persons; each group was randomly assigned to drink an alcoholic beverage or a placebo beverage. To satisfy power requirements of the parent project, alcohol participants were oversampled, resulting in 117 alcohol participants and 39 placebo participants. After completing baseline assessments, groups were seated around a circular table and administered their study beverages. Groups were free to interact during the beverage consumption period (36 min), but were asked not to discuss their level of alcohol intoxication (see10). Alcoholic beverages were administered as a cranberry-vodka cocktail in three equal parts over the 36-min group-drink period. The dose of alcohol administered was adjusted for participants’ weight, sex, and height, intended to achieve a peak blood alcohol concentration (BAC) of approximately 0.08% (0.82 g/kg males; 0.74 g/kg females). For participants drinking the placebo beverage, the vodka bottle contained flattened tonic water and false BAC recordings were given (see71). Placebo deception was 99.35% effective; all but one placebo participant estimated that they drank at least 1 oz of vodka at the end of the study (see below). The non-deceived placebo participant and their group were omitted from analyses.

Immediately following the beverage administration, participants were brought into separate rooms where they reported their subjective intoxication (0–100 scale8) and BACs were recorded with a breathalyzer. These measures were repeated at 30- and 60-min post-beverage consumption. The empathy and emotion recognition measures were administered on the ascending limb of the BAC curve. As described in more detail below, the State Empathy Scale (SES72) began approximately six minutes post-alcohol consumption; the GERT45 was completed approximately 11 min post-alcohol consumption, and the MET53,54 began approximately 34 min post-alcohol consumption. After completing additional questionnaires/tasks unrelated to the current study (e.g.,73), participants ate a meal and completed a post-experimental questionnaire that asked them to estimate their total vodka consumption during the study. Placebo participants were then debriefed. Alcohol participants were debriefed when their BACs dropped below 0.04%. Following debriefing, participants were paid $100 and permitted to leave (they were not permitted to drive).

Measures

Demographics

Participants reported at baseline their sex, age, race, and highest education level attained (see73).

Alcohol use

Alcohol use was assessed at baseline using the 30-day Timeline Follow-Back Schedule (TLFB74), which asks participants to report the number of drinks they consumed each day in the past 30 days. TLFB data was used to create an index of past 30-day alcohol use by dividing the total number of drinks consumed in the past 30 days by the number of drinking days. This variable was square-root transformed for analysis to correct non-normality.

Trait empathy

Trait empathy was assessed at baseline using two measures; the Interpersonal Reactivity Index (IRI60) and the Toronto Empathy Questionnaire (TEQ75). The IRI is a 28-item self-report questionnaire that measures empathy using four subscales: Perspective-Taking; Empathic Concern; Personal Distress; Fantasy. A total score was used in analyses, with higher scores reflecting higher empathy. Cronbach’s alpha was good (\(\alpha\)=0.82). The TEQ is a 16-item measure that captures the frequency at which a wide range of attributes associated with theoretical facets of empathy (e.g., emotional contagion, emotion comprehension) occur. Items are rated on a 5-point Likert scale (0 = Never to 4 = Always), with higher scores indicating higher overall empathy. Cronbach’s alpha was good (\(\alpha\)=0.84).

SES

Participants completed the adapted SES72, a 12-item measure designed to evaluate participants’ subjective impressions of recognizing, experiencing, and relating to other group members’ emotions. Participants were asked to consider one individual they were just drinking with in the lab and respond to cognitive empathy items (e.g., “I could understand what this person was going through when interacting with them”), and items assessing two components of affective empathy (e.g., “ I experienced the same emotions as this person when interacting with them,” labeled affective empathy on the SES, and “I could relate to what this person was going through when interacting with them,” labeled associative empathy on the SES) regarding that person on a 5-point Likert scale (0 = Not at all to 4 = Completely; see Supplemental Tables 1 and 2 for the original and adapted SES). Participants did not choose who to rate; this was assigned by the research team. All participants were asked to rate the participant seated to the left of them during the group-drink period; therefore, all participants were rated by one other participant, and all participants rated one participant. In addition to individual scores for cognitive, affective, and associative empathy, a total score was calculated by averaging the three subscales. Cronbach’s alpha for individual scales was good: cognitive empathy (\(\alpha\)=0.80), affective empathy (\(\alpha\)=0.80), and associative empathy (\(\alpha\)=0.85). Cronbach’s alpha for the total scale was excellent (\(\alpha\)=0.93).

Table 1 Participant characteristics and alcohol use.
Full size table
Table 2 Correlations among study variables.
Full size table

MET

Participants completed the English-validated MET53,54, a computer-assisted test consisting of 40 photographs showing people in positively- or negatively-emotionally charged situations (e.g., young woman flirting in a café, man in a suit with his head in his hands). Emotion recognition (labeled in the test as cognitive empathy) was assessed by asking participants to infer the emotional state of each individual by selecting the correct option from a list of four responses (e.g., excited, desperate, proud, bored). An emotion recognition score was calculated by computing the percentage of correct responses relative to total responses. Perceived empathy (labeled affective empathy in the test) was assessed by asking participants to indicate on a 9-point Likert scale (1 = Not at all to 9 = Very strongly) “how much do you empathize with this person.” An empathy score was calculated by computing the mean of the items. Cronbach’s alpha was very low for the emotion recognition component (\(\alpha\)=0.35) and excellent for the perceived empathy component (\(\alpha\)=0.95).

GERT

The short-form GERT45,76 was also used to measure emotion recognition abilities. The GERT measures individual differences in people’s ability to recognize others’ emotions in the face, voice, and body. The GERT is based exclusively on dynamic, multimodal emotion expressions (i.e., short video clips) to measure emotion recognition ability in a more ecologically valid fashion than still photographs. It consists of 42 short video clips with audio in which ten actors express 14 different emotions. After each clip, participants were asked to choose which of the 14 emotions was expressed. Based on study hypotheses, we only analyzed a subset of 18 video clips depicting the 6 emotion outcomes commonly examined in the alcohol administration literature (i.e., joy, sadness, anger, fear, surprise, and disgust). The total emotion recognition score was derived from these six emotions. Cronbach’s alpha was very low (\(\alpha\)=0.25) [Note: Internal consistency remained low when calculating Cronbach’s alpha for all 14 emotions; \(\alpha\)=0.54]. Given the moderate correlation between the GERT and MET cognitive empathy (r = 0.34), and more importantly, the theoretical rationale suggesting that the MET cognitive empathy component more accurately functions as measure of emotion recognition rather than cognitive empathy (see16), the MET cognitive empathy scores were reported alongside the GERT results and interpreted as a component of emotion recognition rather than cognitive empathy.

Analyses

Data was analyzed using separate 2-level mixed models, featuring random intercepts accounting for the clustering of individuals within 3-person groups, using SAS analytical software77, to assess the effect of beverage condition (alcohol vs. placebo) on empathy scores (i.e., MET perceived empathy, SES cognitive empathy, SES affective empathy, and SES associative empathy) and emotion recognition scores (i.e., MET emotion recognition, total emotion recognition on the GERT, as well as the six individual emotions: joy, sadness, anger, fear, surprise, and disgust). When significant effects occurred, sensitivity models were then run with covariates included, as well as models testing for interaction effects between beverage condition (alcohol vs. placebo) and trait empathy (i.e., IRI and TEQ scores) on empathy and emotion recognition outcomes. Covariates included sex (0 = males, 1 = females), age (in years), group sex composition (0, 1, 2, or 3 females), race (0 = Caucasians, 1 = non-Caucasian), highest education level (in years), and alcohol use from the TLFB. Results are reported for uncorrected statistical tests. The current study provided 80% power to detect a small effect size (Cohen’s f = 0.096) of alcohol’s effect on these social cognition factors, assuming α = 0.05.

Results

As noted above, one participant did not believe the placebo manipulation, resulting in their and their group’s exclusion from analyses. The final sample included 156 participants (mean age = 22.51 ± 2.16 years; 31.41% female; 72.44% White; 16.67% Asian; 3.21% African American; 7.05% more than one race; see Table 1 for participant characteristics for the whole sample, as well as for the alcohol and placebo groups separately). The alcohol and placebo groups did not significantly differ on age, education, race, or alcohol use. However, the alcohol group had a higher percentage of females and different group sex compositions, with more groups containing a higher proportion of females, compared to the placebo group. As noted above, the SES, GERT, and MET were administered ~ 6, ~ 11, and ~ 34 min after the completion of drinking, respectively. The average BAC increased from 0.051% immediately after drinking to 0.058% at 30 min and 0.060% at 60 min post-drinking. There was no significant difference between the immediate and 30-min BACs (t = -1.83, p = 0.07) or between the 30-min and 60-min BACs (t = -0.57, p = 0.57). These results suggest that all alcohol participants were on the ascending limb of alcohol absorption for all study tasks.

Table 2 presents bivariate correlations among study variables. As can be seen, female sex had significant negative associations with age and past 30-day alcohol drinks per drinking day, and significant positive associations with the GERT, and both IRI and TEQ trait empathy. Education had a significant positive association with age and a significant negative association with past 30-day alcohol drinks per drinking day. The GERT had a significant positive association with MET emotion recognition. IRI trait empathy had significant positive associations with TEQ trait empathy, MET perceived empathy, and SES total empathy. Finally, TEQ trait empathy had significant positive associations with MET perceived empathy and SES total empathy.

Emotion recognition

Tables 3 and 4 display the results of the main mixed effects models examining effects of alcohol (vs. placebo) on GERT scores and MET emotion recognition scores respectively. Table 5 shows the results of sensitivity models with covariates included for significant main effects models for GERT scores. As can be seen, there was no significant effect of alcohol (vs. placebo) on total emotion recognition on the GERT (see Table 3) or on MET emotion recognition (see Table 4). Sensitivity models were thus not conducted. For specific emotions, there was a significant effect of alcohol (vs. placebo) on GERT anger recognition. Individuals in the alcohol group made more errors on anger recognition compared to individuals in the placebo group (B = -0.18, p < 0.01). When participants incorrectly identified anger, which accounted for 35% of total responses in the alcohol group and 21% in the placebo group, they most commonly selected irritation (72.58% and 80% of cases, respectively) followed by disgust (11.29% and 8% of cases, respectively). This effect of alcohol (vs. placebo) on anger recognition held when controlling for covariates (B = -0.23, p < 0.001). In these models, education was significantly associated with anger recognition, that is participants with higher education demonstrated better anger recognition (B = 0.05, p < 0.05). There were no other significant effects of alcohol (vs. placebo) on any other specific emotion (i.e., joy, sadness, fear, surprise, and disgust).

Table 3 Effects of alcohol (vs. placebo) on emotion recognition: Main model (n = 156).
Full size table
Table 4 Effect of alcohol (vs. placebo) on the Multifaceted Empathy Test (MET).
Full size table
Table 5 Effects of alcohol (vs. placebo) on emotion recognition: Sensitivity model (n = 156).
Full size table

Empathy

Table 4 displays results for the models investigating the impact of alcohol (vs. placebo) on MET perceived empathy. As can be seen, alcohol did not have significant effects on MET perceived empathy. Sensitivity models were thus not conducted. Table 6 presents the results for models investigating the impact of alcohol (vs. placebo) on SES empathy scores. As shown, there were no significant effects of alcohol on SES cognitive or affective empathy, but there were significant effects of alcohol (vs. placebo) on SES associative and total empathy. Participants who consumed alcohol reported higher SES associative (B = 0.31, p < 0.05) and total (B = 0.25, p < 0.05) empathy compared to placebo participants. Sensitivity analyses indicated that the significant effect of alcohol (vs. placebo) remained only for associative empathy after controlling for covariates (B = 0.27, p < 0.05). Group sex composition had a significant positive association with SES associative empathy (B = 0.67, p < 0.05). Groups that were all female reported higher SES associative empathy.

Table 6 Effect of alcohol (vs. placebo) on the State Empathy Scale (SES).
Full size table

In terms of moderation, trait empathy measured by the IRI moderated the association between alcohol (vs. placebo) and SES cognitive empathy (B = 0.02, p < 0.05). Those with higher trait IRI empathy scores at baseline reported higher SES cognitive empathy under the influence of alcohol (see Table 8 and Fig 1). There were no additional moderating effects of trait empathy as assessed by the IRI or the TEQ on alcohol’s effects on any other outcome variable (see Tables 7 and 8).

Fig. 1
figure 1

Interaction Between Trait Empathy and Alcohol Condition on SES Cognitive Empathy. IRI trait empathy was positively associated with SES cognitive empathy among participants in the alcohol condition.

Full size image
Table 7 Effect of alcohol (vs. placebo) on the Multifaceted Empathy Test (MET): Interaction effects.
Full size table
Table 8 Effect of alcohol (vs. placebo) on the State Empathy Scale (SES): Interaction effects.
Full size table

Discussion

Alcohol intoxication can lead to both positive and negative social outcomes, and there has been much interest in investigating the underlying mechanisms that may produce such varied social effects (e.g.,3,15,16). Recent research suggests that alcohol-induced changes in social cognition, namely emotion recognition and empathy, could play a role in these desirable and undesirable social consequences (e.g.,17,20,21). However, notable methodological limitations of prior studies make drawing firm conclusions difficult. This study aimed to investigate the acute effects of alcohol on emotion recognition and empathy, addressing some of the methodological limitations of prior research. By recruiting a large sample, administering an intoxicating dose of alcohol, achieving near-perfect placebo deception, and employing more ecologically valid emotion recognition and empathy tasks, the current study provides a robust assessment of how alcohol may impact these social cognition factors.

Emotion recognition

We found no significant differences between participants in the alcohol and placebo conditions on overall GERT emotion recognition or on the recognition of five specific emotions on the GERT (i.e., joy, sadness, fear, surprise, and disgust). We also did not find significant differences between participants in the alcohol and placebo conditions on MET emotion recognition. However, participants in the alcohol condition demonstrated a diminished ability to recognize anger on the GERT compared to those in the placebo condition, and this effect persisted even after controlling for covariates. Specifically, while the most frequent incorrect attribution for anger in both the conditions was irritation followed by disgust, participants in the alcohol condition were significantly more likely to make these errors than placebo participants. This suggests that alcohol participants may have recognized the overall negative affect of these expressions but were less able to pinpoint the specific emotion of anger. This error could lead intoxicated individuals to underestimate or misinterpret others’ anger, potentially resulting in inappropriate social responses during interactions. On the GERT, accuracy is measured as the proportion of correct responses, making it difficult to determine whether reduced anger recognition in the alcohol participants resulted from decreased sensitivity to anger expressions or increased misattributions to other emotions. Future studies are needed to determine this, as well as to clarify the potential impact of these errors in anger recognition on actual social behaviors21,31. It might also be useful to assess reaction times to correct or incorrect emotion recognition to further understand how these errors might influence decision-making and social interactions in real-time situations.

Given the very low reliability of both emotion recognition measures, though, these results should be interpreted with much caution. Our finding is consistent with two prior studies that showed that a lower (0.04 g/kg78) and higher (1.975 g/kg39) dose of alcohol (vs. placebo) worsened anger recognition on the Six Alternate Forced Choice and a task using 44 photographs constructed by Ekman & Friesen79, respectively. Neither of these studies provided internal reliability estimates for their emotion recognition measures, though. Indeed, it is important to note that no prior study investigating the effects for alcohol on emotion recognition used the measure we used here (i.e., the GERT), and only one prior study (out of 17) reported internal reliability estimates for their emotion recognition measure. Specifically, Kamboj and colleagues37 reported excellent reliability (\(\alpha\)=0.97–0.98) for response times to identify emotions on the Dynamic Emotion Expression Recognition Task (DEER-T80), but they did not provide reliability for correct responses. Future studies should routinely report reliability estimates for emotion recognition measures to aid in the interpretation of results, and efforts should be focused on developing emotion recognition measures with better reliability.

Empathy

We found no significant difference between participants in the alcohol and placebo conditions on perceived empathy on the MET, which is a traditional third-person empathy task used in the alcohol literature. Our null findings are in contrast to one prior much smaller (N = 60) study that found that a lower dose (0.24 g/kg for females/0.29 g/kg for males) of alcohol (vs. placebo) consumption increased perceived empathy on the MET17. Participants in the alcohol condition did, however, significantly differ from participants in the placebo condition on associative empathy (a component of affective empathy assessing how well participants related to another participant’s emotions) and total empathy, as measured by the SES—a novel second-person empathy task introduced in this study. Specifically, participants who consumed alcohol reported higher SES associative and total empathy in response to another participant they interacted with within the lab compared to placebo participants. However, this significant difference remained only for associative empathy, and not total empathy, after controlling for covariates. These results suggest that individuals under the influence of alcohol perceived themselves to be more able to relate to the emotions of interacting social partners, without necessarily experiencing the same feelings as them. Additional research is needed to explore this possibility.

Several factors could explain the discrepant findings between the alcohol and placebo conditions on perceived empathy on the MET vs. associative empathy on the SES, including differences in what the SES and MET were measuring, as well as distinctions between second- and third-person assessments. The SES, a second-person task, assessed participants’ retrospective subjective impression of having recognized, experienced, and/or related with another person’s emotions during direct interactions. In contrast, the MET, a third-person task, involved self-report ratings of perceived empathy in response to static images of emotions. These differences in scope and context might underly the observed effects: the SES associative empathy component captures a broader retrospective assessment of relating to another’s emotions, while the MET provides a more immediate, task-specific evaluation of empathy. Indeed, alcohol may bias recall of performance in certain contexts (e.g.,81), which could contribute to the observed effects on the SES but not the MET. More research is needed to explain these discrepant results across empathy assessments.

In terms of moderation, we found that individuals with higher baseline trait empathy scores on the IRI reported higher SES cognitive empathy under the influence of alcohol. This result contrasts with Dolder and colleagues17, who found that individuals with lower baseline trait empathy on the IRI showed larger increases in perceived empathy on the MET under the influence of alcohol. This discrepancy may be attributable to several factors: differences in the empathy tasks used, variations in the alcohol doses administered (0.74/0.82 g/kg for females/males in our study versus 0.24/0.29 g/kg in theirs), and the greater statistical power of our study, which included over two and a half times as many participants as the study by Dolder and colleagues. Further research is needed to explore whether lower or higher trait empathy moderates any effects of acute alcohol intoxication on empathy.

Limitations

Although we addressed many methodological weaknesses of prior alcohol administration studies examining social cognition, there are still limitations to consider. Most importantly, participants interacted in groups of strangers prior to completing the social cognition assessments used here. Differences in social experiences during these interactions across the alcohol and placebo conditions (see71) could have influenced participants’ levels of engagement, attention to interaction partners, and subsequent self-reported empathy. Given that alcohol alters interactive social behavior, it is possible that the changes we observed in social cognitive measures were secondary to these altered social interactions rather than a direct pharmacological effect of alcohol. Additionally, because alcohol was administered to all members of each group, the observed effects on social cognition could also reflect the influence of alcohol on interaction partners rather than, or in addition to, the direct effects of alcohol on each individual. This study is not able to disentangle whether alcohol affected social cognition directly or whether alcohol-induced shifts in one’s own behavior or in the behavior of interaction partners influenced social cognition. Future studies should investigate these mechanisms separately to clarify alcohol’s direct versus indirect effects on social cognition.

In addition, there is considerable conceptual overlap between emotion recognition and cognitive empathy, which both involve perceiving and understanding another person’s emotional state. While some researchers consider these to be independent constructs (e.g.,82), others view emotion recognition as a lower-level process that supports empathy (e.g.,41,69). The MET effectively measures emotion recognition and labels it as cognitive empathy, further blurring the distinction between these constructs83,84. Future research should aim to develop and use distinct measures that separately assess emotion recognition and cognitive empathy. This approach could elucidate whether alcohol intoxication primarily affects the foundational perceptual processes and/or the more complex empathic responses involved in social interactions. Multi-participant alcohol administration studies that permit real-time social interactions may be particularly helpful to disentangle emotion recognition from cognitive empathy.

Further, the GERT and the MET both relied on images/videos of other people (that is, were third-person tasks), which surely do not reflect the complexity of real-world social interactions, where contextual cues such as body language, tone of voice, and social feedback influence the recognition, interpretation, and response to others’ emotions85,86,87,88. More broadly, a significant limitation in the social cognition literature is the lack of emotion recognition and empathy tasks that require people to actually interact with others and assess their social responses in real-time. This is particularly relevant given our significant findings using a task that evaluated empathic responses to another individual during an interaction (the SES) and not for the third-person task (the MET). Future studies should focus on developing emotion recognition and empathy tasks in the context of real-time interactions, with live social feedback, to capture the dynamic and reciprocal nature of social exchanges in naturalistic settings. These types of tasks might also demonstrate better reliability, which is currently a concern with some empathy (see68,70) and emotion recognition (see16) tasks. With more ecologically valid and reliable tasks, future research can better clarify alcohol’s effects on social cognition.

Finally, like most studies exploring the link between alcohol intoxication and social cognition (see18,25,28,89), our sample was predominantly White and male. This may limit the generalizability of findings to people of color and females. Research suggests that males and females differ in empathy, with females typically reporting higher empathy than males in adolescence (e.g.90,91) and adulthood (e.g.92,93), and in their alcohol consumption, with males typically consuming more alcohol on average than females94. These differences warrant further examination into whether alcohol impacts social cognition differently across sex, especially given the limited research in this area.

Conclusions

In summary, we found that participants in the alcohol condition showed impaired recognition of anger, but no other specific emotions. Additionally, those in the alcohol condition reported higher associative empathy in response to direct interactions with others, but similar empathy responses to placebo participants on a third-person task using static images. Finally, individuals with higher trait empathy who drank alcohol reported higher cognitive empathy in response to another individual. Reliability for the emotion recognition tasks were very low, and nearly all prior studies failed to report reliability estimates for their emotion recognition measures. No prior study used real-time social interactions to investigate the effects of alcohol on emotion recognition or empathy. Findings highlight the need to focus efforts on developing more reliable and ecologically valid social cognition tasks and paradigms in order to clarify the effects of alcohol on emotion recognition and empathy.