Alcohol is favored for its acute ability to induce positive affect by activating neural reward systems (Fromme et al., 1999), as well as reduce stress and anxiety (Curtin and Lang, 2007). However, alcohol consumption, especially at high doses, is also associated with severely impaired social abilities, such as more aggressive behavior (Bushman and Cooper, 1990). Denson et al. (2008) has hypothesized that alcohol might increase aggression via a reduction in the ability to identify with and vicariously share the feeling, pain and thoughts of others, i.e., empathy (Denson et al., 2008), as the level of empathy is negatively correlated with the intensity and frequency of violence in sobriety (Miller and Eisenberg, 1988). Empathy is believed to be a key motivator and the proximate mechanism of altruistic and prosocial behavior (Preston and de Waal, 2002; Singer et al., 2006; Moriguchi et al., 2007). Thus, any change in empathy by alcohol consumption will have social implications.

In recent years considerable efforts have been made to investigate the neural correlates of human empathy. The majority of studies used experimental paradigms in which participants were exposed to stimuli depicting or indicating that other people were in pain. Recent imaging studies revealed that empathy for pain recruits a distributed network, including the anterior cingulate cortex (ACC) and bilateral anterior insula (AI), major components of self-pain related “affective pain matrix” (Jackson et al., 2005; Hein and Singer, 2008; Singer and Lamm, 2009). Alcohol acts mainly as a central nervous system depressant. Previous neuroimaging studies indicated that alcohol dampened activity in the dorsal anterior cingulate cortex (dACC) (Ridderinkhof et al., 2002; Marinkovic et al., 2012) and the bilateral anterior insula (AI) (Padula et al., 2011) in tasks other than empathy. As these three regions that were affected by alcohol consumption are considered the core affective pain matrix of empathy (Lamm et al., 2010a; Fan et al., 2011), we hypothesized that empathic responses for pain in these regions may be reduced, and the three regions that affected by alcohol intoxication (dACC, bilateral AIs) were taken as regions of interest (ROI).

The second question we want to address is whether the effects of trait empathy on the empathic neural responses are altered by alcohol consumption. Interestingly, the majority of empathy studies with sober participants, if not all, have not found any correlation between trait empathy and the activity of the affective pain matrix (Decety, 2011). However, Giancola (2003) demonstrated that trait empathy, the tendency for people to imagine and experience the feelings and experiences of others, has moderating effects on alcohol-related aggression in men and women but not in men and women in sobriety (Giancola, 2003). These results suggest that the effects of trait empathy on behavior was affected by alcohol intoxication, as well as underlying neural correlates.

More important, attention plays an important role in empathy. Previous studies revealed that activity in the affective pain matrix is reduced, even absent, if attention was diverted away (Gu et al., 2010; Lamm et al., 2010b). Thus, activity in the affective pain matrix may be more affected by trait empathy if attention is impaired. The capacity to divide and sustain attention is indeed impaired by alcohol, even at blood-alcohol concentration (BAC) levels of 0.02–0.03% (Koelega, 1995). Therefore, in accordance with Giancola's finding, we predicted empathic neural responses were more affected by trait empathy in the intoxication condition instead of the placebo condition. Our results confirmed this prediction and found that brain activity in the rAI was correlated with trait empathy only in the alcohol condition.

Another intriguing point is that as the affective pain matrix, but not a single brain region, is involved in empathic processing of pain, there are likely interactions between these different brain regions. Noted that functional connection between brain regions that are engaged simultaneously in a task (Rogers et al., 2008) can be affected by task-related parameters (Friston et al., 1997), i.e., psychophysiological interactions (PPI). For example, Friston KJ et al provided an example of physiological interactions in the visual pathway were modulated by attention (Friston et al., 1997). As attention was also impaired by alcohol consumption (Steele and Josephs, 1990; Ridderinkhof et al., 2002; Calhoun et al., 2004; Van Horn et al., 2006; Marinkovic et al., 2012) and that empathic responses for pain were modulated by attention (Gu and Han, 2007), the functional connection between brain regions could be affected by alcohol consumption. To test this hypothesis, we further conducted a psychophysiological interactions (PPI) analysis (Friston et al., 1997) with rAI as the seed in both alcohol and placebo conditions.

The present study was carried out to investigate the effects of alcohol consumption on empathy for pain. A significant interaction was found in the dACC. As the dACC is considered to be one of the core regions of the pain empathy-related network (Lamm et al., 2010a) to code the value of affective stimulus (Fan et al., 2011), our results suggest that the perceived affective value of empathic concern was impaired by alcohol intoxication (Steele and Josephs, 1990), which is in line with our hypothesis. Note that the dACC also plays a key and necessary role in voluntary and cognitive control (Shackman et al., 2011), such as resolving conflicts (Botvinick et al., 2001) or monitoring errors (Yeung et al., 2004; Marinkovic et al., 2012). Thus, reduced perceived affective value should result in an underestimation of others' pain as well as poor cognitive control, and then lead to poor control of aggressive behaviors (Steele and Josephs, 1990).

One of the most interesting findings was that the empathic responses in the rAI were positively correlated with trait empathy in the alcohol condition, but not in the placebo condition. This finding is not due to changes of affective states or behavioral responses after drinking, as these confounding factors are ruled out in the regression analysis. The insula has long been considered one of the key regions that related with traits. For example, activation in insula was also found closely related to various trait measures, such as anxiety (Simmons et al., 2006) or intolerance (Simmons et al., 2008). Thus, our results indicate insula is also related with trait empathy, and intoxication leads to trait empathy inflation. These findings significantly extends the self-inflating phenomenon (i.e., enhancing feelings of self-appraisal) (Steele and Josephs, 1990) to the social cognition domain of trait empathy in intoxication.

Moreover, we found that the functional connectivity between the rAI and attention network, which occurred during empathy for pain in the placebo condition, was missing in the alcohol condition. It is believed that attention plays an important role in empathy. For example, empathic responses in the AIs were significantly reduced when the participants were distracted (Gu and Han, 2007) or cognitively busy (Rameson et al., 2011). AIs activity during a pain perception task was also significantly reduced when higher activity was observed in the right dorsolateral prefrontal cortex (Lorenz et al., 2003). Considering that activities in the fronto-parietal attention/control network, including the SMA/ACC, bilateral dorsolateral frontal cortices (middle/inferior frontal gyri), and bilateral parietal areas, were significantly decreased by alcohol intoxication, we conclude that crosstalks between the rAI and attention network is impaired by alcohol consumption, resulting in impaired top-down modulation of the rAI.

Taken together, the most likely explanation is that empathy for pain is shaped by other factors (de Waal, 2007), such as social norm. Thus, people with both high and low trait empathy may display relatively normal empathy responses for pain in sobriety. However, as alcohol is a central nervous depressant, it seems that alcohol intoxication not only inhibits local neural activity, but also inhibits neural signal flow between brain regions. The communication between the attention network and rAI was downregulated, and influences of other factors on empathy were thus reduced, resulting in trait empathy inflation. These results provide a possible neural foundation for the finding that empathy plays a moderating role in alcohol-related aggression behaviors (Giancola, 2003).

Despite the fact that alcohol intoxication led to those neural changes, participants could still differentiate painful stimuli from non-painful stimuli correctly as well as their intensities after drinking. These findings not only undermine the possibility that participants were disengaged from our task after alcohol intoxication, they also suggest that there is disassociation between cognitive appraisal and neural empathic activation for pain after drinking. Preserved cognitive appraisal but impaired brain function in other domains such as affective information has been reported during alcohol intoxication (Padula et al., 2011). A recent lesion study revealed that only damage to the AIs rather than the dACC might be more necessary for affecting pain empathy and causing impaired behavioral empathic responses (Gu et al., 2012). Hence, the ability to correctly judge other's affective state may be preserved because of the relatively preserved empathic responses in the AIs. It is also possible that other factors (e.g., past knowledge) may be involved. These disassociations suggest that humans have an inability to acknowledge their social dysfunction after drinking (Fromme et al., 1999).

There are two limitations. The first one is that although the double-blindness procedure used in our manuscript is widely used in previous studies such as Ridderinkhof et al. (2002), it remains possible that participants became aware of the experimental condition given the strong behavioral effects of alcohol over water placebo and our results can be affected by participant awareness. However, we think that our results are robust as the observed significant correlation between empathic neural activity in the right anterior insula (rAI) and trait empathy cannot be interpreted by participant awareness. Second, our sample size is relatively small (data from 16 participants were valid). We also note that a significant stimuli type×beverage interaction was only found in the ROI analysis, but not in the voxel-wised whole brain analysis, and no interaction was found in the AIs, possibly due to the small sample size. These results are in line with a previous suggestion that an ROI approach yields higher sensitivity than whole brain analyses in pharmacological fMRI with prior hypotheses (Mitsis et al., 2008). Further studies using a larger sample size are desired.

In summary, we provide the first piece of evidence that the empathic neural response for pain in the dACC is impaired by acute alcohol consumption in a group of healthy social drinkers. Moreover, we also show that alcohol intoxication leads to inflated trait empathy in the rAI. The neural correlates underlying trait empathy inflation are likely due to the impaired crosstalk between the rAI and attention neural network, i.e., impaired top-down modulation. Whether similar mechanisms are involved in other alcohol-related self-inflation deserves further investigation.

YH, MF, ZW, and ZC: designed the study; YH, ZC, and MF: collected the data; YH: analyzed the data; YH, YP, and ZC: wrote the paper.