We explored if the disposition to react with aggression while alcohol intoxicated was moderated by polymorphic variants of the oxytocin receptor gene (OXTR). Twelve OXTR polymorphisms were genotyped in 116 Finnish men [aged 18–30, M = 22.7, standard deviation (SD) = 2.4] who were randomly assigned to an alcohol condition in which they received an alcohol dose of 0.7 g pure ethanol/kg body weight or a placebo condition. Aggressive behavior was measured using a laboratory paradigm in which it was operationalized as the level of aversive noise administered to a fictive opponent. No main effects of the polymorphisms on aggressive behavior were found after controlling for multiple testing. The interactive effects between alcohol and two of the OXTR polymorphisms (rs4564970 and rs1488467) on aggressive behavior were nominally significant and remained significant for the rs4564970 when controlled for multiple tests. To the best of our knowledge, this is the first experimental study suggesting interactive effects of specific genetic variants and alcohol on aggressive behavior in humans.
To conclude, OXTR polymorphisms are candidates for having interactive effects together with alcohol on aggression as OXT is associated not only with directly aggressive traits, but also with other socially important skills and behaviors, such as generosity (Zak et al. 2007), trust (Kosfeld et al. 2005), envy and schadenfreude (Shamay-Tsoory et al. 2009), social recognition (Rimmele et al. 2009) and the ability to interpret the emotions of others from facial expressions (Domes et al. 2007b); many of them skills that according to Giancola (2000) are hypothesized to be impaired by the effects of alcohol on executive functioning, and thus increase the likelihood for aggression.
Besides replicating the effects of alcohol on aggression, we explored both main effects of OXTR polymorphisms, as well as interactive effects between the polymorphisms and alcohol on aggressive behavior.
Materials and method
The sample consisted of 116 Finnish- or Swedish-speaking male university students aged 18–30 [M = 22.7, standard deviation (SD) = 2.4]. All participants were Finnish citizens. The control and experimental groups were balanced considering language of the participants (χ2 = 2.62, P = 0.11). The participants were recruited through university mailing lists and posters or personal referral. Participation was compensated with two movie tickets. Interested participants were screened for contraindications prior to the experiment. Those with previous adverse reactions to alcohol, drug-related problems, poor general health, a body mass index (BMI) exceeding 30, medication contraindicating alcohol consumption, diabetes and/or tinnitus were excluded from the sample. Also, they were screened for excessive use of alcohol using the Alcohol Use Disorders Identification Test (AUDIT; Saunders et al. 1993). A cutoff was chosen in order to eliminate the 5% with the highest (extreme) levels in the population under study, on the basis of the analyses of a large population-based sample (age range 18–48, M = 26.2, SD = 4.8) (von der Pahlen et al. 2008 for a detailed description of the sample). All participants were asked to refrain from consuming alcohol 24 h prior to the test session and not to eat, drink or use nicotine 2 h before. Written informed consent was obtained from all participants stating the voluntary nature of the study. The research plan was approved by The Ethics Committee of the Abo Akademi University in accordance with the 1964 Declaration of Helsinki.
The design of the experiment was a mixed models design with aggressive behavior as the dependent variable and the between-subject variables genotype, alcohol manipulation and general level of alcohol consumption. Level of provocation was a within-subject variable. The participants were randomly assigned to either a group receiving alcohol or a group receiving placebo. The alcohol group consisted of 63 individuals, and the placebo group of 53 individuals.
Measurement of the general level of alcohol consumption
To measure the general level of alcohol consumption, the three first items of the Alcohol Use Disorders Identification Test (AUDIT; Saunders et al. 1993) were used. A composite score was calculated by summing the scores of the following items: (1) How often do you have a drink containing alcohol? (2) How many drinks containing alcohol do you have on a typical day when you are drinking? and (3) How often do you have six or more drinks on one occasion? Responses were given on a Likert-type scale from 0 to 4. Higher values on the composite score indicated higher alcohol consumption. According to a review by Fiellin et al. (2000), sensitivity of the items ranged between 54% and 98%, and specificity between 57% and 93%. Cronbach's α for the three first items was 0.55 in the present sample.
Measurement of aggressive behavior
A version of the Response Choice Aggression Paradigm (RCAP; Zeichner et al. 1999) was used to measure aggressive behavior. The RCAP is an adaptation of the Taylor Aggression Paradigm (TAP; Taylor 1967), which is together with its modified versions one of the most commonly used laboratory paradigms to measure aggression (Hoaken & Pihl 2000). The RCAP is a supposed reaction time test against a fictitious opponent in which punishments can be received from and administered to the opponent. As previously performed in some studies (Bond & Lader 1986, Krämer et al. 2008), aversive sound delivered through headphones was used as a substitute for electric shocks as punishments. After each reaction time trial, the participant was presented with the opportunity to punish his ‘opponent’ by executing an aversive sound on an ascending scale from 1 to 10, or to refrain from giving any punishment. Likewise, the participant also received punishments from the ‘opponent’ as provocation on predetermined trials. Aggressive behavior was measured as the level of aversive noise administered by the participant to the fictive opponent per trial. Provocation was defined as the level of aversive sound administered to the participant by the fictive opponent. The RCAP has shown acceptable internal consistency (Zeichner et al. 1999) as well as convergent validity (Zeichner et al. 2003).
The participants were informed that they would compete in two different computer games (a Go-NoGo paradigm and RCAP; Zeichner et al. 1999) of which the latter would be played against a male opponent, sitting in another room, undergoing the same procedure. Results concerning the Go-NoGo paradigm, an attention and response control paradigm, were not analyzed in this study and will thus not be discussed further. All participants were made to believe that they were going to receive alcoholic beverages leading to a moderate intoxication. In addition, it was emphasized that they could discontinue the experiment at any time. Informed consent was obtained from all participants. A cover story was used to conceal that the true purpose of the experiment was to measure aggression, stating that the aim of the study was to investigate the effects of alcohol on reaction time, interpersonal communication and temperament.
To ensure that the participants were sober at the beginning of the experiment, the first of five blood alcohol content (BAC) measures were taken using the Mark X Electrochemical Breath Alcohol Analyzer (Alcovisor®, C4 Development Ltd., Kownloon, Hong Kong) at arrival to the experiment. BAC levels will be reported in 1 cg alcohol /ml blood. The weight and height of the participant were measured in addition to a DNA saliva sample collected. Next, a dose equivalent to 0.7 g pure ethanol per kg body weight mixed with cranberry juice to an alcohol concentration of 10% by weight was given to the participants in the alcohol condition. Participants in the placebo condition were given a corresponding dose consisting of only cranberry juice. To enhance the perception of it being an alcoholic drink, the edge of the glass was wiped with vodka. Both groups consumed the drinks in the same manner during a period of 35 min, followed by a 15-min wait.
Before the RCAP, each participant chose their maximum level of aversive noise tolerated (on a scale from 1 to 10). The loudest possible sound level of choice (10) did however not exceed 80 dB, in accordance with the World Health Organization's safety recommendations for noise (WHO 1999). The instructions were presented both on the computer screen and verbally by the experimenter. All together 30 reaction time trials were administered of which 14 trials were ‘lost’ and 16 trials ‘won’. The order of the trials was first randomized and then presented in identical order to the participants. Upon each trial the participant had the opportunity to deliver an aversive sound to his ‘opponent’. The first sound-provocation by the ‘opponent’ to the participant occurred on the sixth trial and the participant was altogether provoked 15 times at 81.25–100% of their personal maximum sound tolerance. Each provocation lasted 200 milliseconds. After completion of the test, participants in the alcohol group had to wait for their BAC to descend to 0.0318% or below before leaving the experiment. At the end of the experiment, the true purpose of the study, as well as the fact that the opponent was fictive, was shown and all participants were thoroughly debriefed. The average time for the entire test session was 2 h, with an additional 1–2 h of waiting afterward for those who had been administered alcohol.
Oragene™ DNA self-collection kits (DNA Genotek, Inc., Kanata, Ontario, Canada) were used when collecting saliva samples from participants. Genotyping of the single nucleotide polymorphisms (SNPs) was made by the KBioscience laboratorium in the UK (http://www.kbioscience.co.uk) using the KASPar chemistry – a competitive allele specific polymerase chain reaction (PCR) SNP genotyping system performed with Fluorescent Resonance Energy Transfer (FRET) quencher cassette oligos (http://www.kbioscience.co.uk/genotyping/genotyping-chemistry.htm). The OXTR SNPs chosen for genotyping are relatively evenly distributed across the gene and have shown associations with behavioral traits in previous studies (Fig. 1).
The effects of the allelic variants and their interactive effects with alcohol consumption on aggressive behavior were calculated using the Generalized Estimating Equations method of SPSS 17.0 with the robust variance estimator. This procedure takes into account the dependent structure of the repeated measures. The dependent variable was the level of aversive noise administered by the participant to the fictive opponent. The effect of the polymorphisms (analyzed separately for each polymorphism), the effect of alcohol manipulation as well as their interaction were added as predictors, and the level of provocation and the general level of alcohol consumption as covariates. The experiment-wide significance threshold required to keep type 1 error rate at 5% was calculated using the effective number of independent variables (which in the current analysis was 10 according to the procedure suggested by equation 5 of Li & Ji 2005) in an approach utilizing a linkage disequilibrium (LD) correlation measure (Nyholt 2004) resulting in a corrected α-level of 0.0051. One-tailed significance tests were used to test the hypothesized aggression increasing effects of alcohol and provocation.
Descriptive statistics of genetic data
The distribution of genotypes in the sample of this study was not significantly different from what would be expected if the population was in Hardy–Weinberg equilibrium. For four SNPs (rs11720238, rs2254298, rs75775 and rs4686302) the frequency of the rare homozygotes was too low to be analyzed separately, and therefore, those homozygotes were grouped together with the heterozygotes for these SNPs for all subsequent analyses (Table 1). A schematic picture of the OXTR gene can be seen in Fig. 1 together with a LD plot.
Table 1. The 12 analyzed oxytocin receptor gene single nucleotide polymorphisms
SNP/ nucleotide composition
Minor allele frequency in sample (%)
Common homozygotes (n)
Rare homozygotes (n)
Common homozygotes (n)
Rare homozygotes (n)
SNP, single nucleotide polymorphism; G, guanine; T, thymine; C, cytosine; A, adenine.
*The rare homozygotes were grouped together with the heterozygotes for the remaining analyses. As the genotype at some SNPs could not be identified for all individuals, the N per row may in some cases not add up to N = 116 (alcohol n = 63, placebo n = 53).
Eleven participants (total N = 116) did not show any aggressive behavior during the test. Five of these participants belonged to the placebo and six to the alcohol group. For those who responded aggressively at least once, the mean percent of trials responded aggressively to was 55.26% (SD = 24.83, range 3.00–100.00%) with the mean level of noise administered to the fictive opponent being 3.17 (SD = 2.03, scale range per trial 0–10, range of individuals' means over all trials 0.03–8.30). Six participants showed aggressive behavior on each trial, with five of them having received alcohol and one having received placebo. There was no difference in general level of alcohol use between the alcohol and the placebo group t(114) = 1.75, P = 0.084.
In order to confirm that the task deception had been successful, participants were asked a few questions about their impression regarding the opponent and their interplay (e.g. ‘Did you perceive your opponent as fast?’). Finally, the participant was verbally asked whether he had believed that he was playing against another opponent or not. The deception manipulation appeared successful as all but five participants reported that they believed they were competing against another person. Also, all but three participants in the placebo group believed that they had been drinking alcohol. The eight participants who did not believe that they had been given alcohol and/or that they played against another opponent were not included in the final sample (N = 116). Participants in the alcohol group had a mean BAC of 0.062% (SD = 0.012) just before the task and 0.055% (SD = 0.009) after the task.
Effects of alcohol manipulation, level of provocation and general level of alcohol consumption on aggressive behavior
There was a trend for alcohol to increase aggression, Wald χ2 = 2.34, df = 1, one-tailed P = 0.063. Participants in the alcohol group showed higher levels of aggressive behavior [M = 3.15, standard error (SE) = 0.30] than participants in the placebo group (M = 2.51, SE = 0.28). A significant main effect of provocation on aggression was also found, Wald χ2 = 28.30, df = 1, one-tailed P < 0.001, B = 0.106, SE = .03 with higher levels of provocation being associated with higher levels of aggression. The interaction between provocation and alcohol manipulation was not significant, Wald χ2 = 0.72, df = 1, P = 0.396, neither the effect of the general level of alcohol consumption of the participants, Wald χ2 = 0.18, df = 1, P = 0.669.
Effects of oxytocin receptor gene polymorphisms on aggressive behavior
One significant main effect of the OXTR polymorphisms on aggressive behavior was found (when α = 0.05, i.e. when multiple testing was not controlled for) (Table 2). Participants who were carriers of the G:G genotype on the rs1042778 (M = 3.01, SE = 0.36) and the T:G genotype (M = 2.97, SE = 0.28), showed higher levels of aggression than individuals who were homozygous for the T allele (M = 2.02, SE = 0.28). When the α-level was corrected to account for multiple testing (α = 0.0051) the main effect of the rs1042778 did not remain statistically significant.
Table 2. Main effects of the tested oxytocin receptor gene single nucleotide polymorphisms as well as their interactions with alcohol
Main effect of SNP
Interaction between SNP and alcohol manipulation
The level of provocation and the general level of alcohol consumption were modeled as covariates.
SNP, single nucleotide polymorphism.
*The rare homozygotes were grouped together with the heterozygotes.
As can be seen in Table 2, the effect of alcohol intoxication on aggressive behavior was nominally moderated by two of the measured OXTR polymorphisms: rs1488467 (P = 0.006) and rs4564970 (P = 0.004). Although the interaction between rs1488467 and alcohol did not remain significant when using the corrected α-level of 0.0051, the interaction between rs4564970 and alcohol remained significant. The interaction resulted from the fact that alcohol increased aggressive behavior for those who were heterozygous on the two polymorphisms, whereas it did not have any effect on aggressive behavior for those who were homozygous for the G allele (Fig. 2). This was true for both the rs4564970 and the rs1488467 polymorphisms.
In this study, no participant was homozygous for the C allele for the rs1488467 or the rs4564970 polymorphisms. The frequencies of genotypes across the two alcohol manipulation groups were balanced for both SNPs (Table 1). There were no outliers in the measure of aggressive behavior. As seen in Fig. 1, the rs1488467 and the rs4564970 polymorphisms were in high LD.
The aim of this study was to experimentally investigate main effects of twelve OXTR polymorphisms, as well as their interactive effects with alcohol on aggressive behavior. To the best of our knowledge this is the first study to experimentally examine interactive effects between acute alcohol intoxication and measured genetic variants on aggressive behavior in humans.
Although the effect of alcohol on aggressive behavior did not reach significance, it showed a trend toward increasing aggressive behavior, in line with previous findings (Exum 2006). The results replicated the findings that provocation (Bettencourt & Miller 1996; Hoaken & Pihl 2000; Ito et al. 1996) significantly increased aggressive behavior. A main effect of one polymorphism (rs1042778) on aggressive behavior was found when α = 0.05, showing that carriers of the G allele had higher levels of aggressive behavior than those homozygous for the T allele. This effect did not, however, remain significant after correction for multiple tests. In a study by Israel et al. (2009), the G allele was associated with lower levels of prosocial behavior assessed with the dictator game. The rs1042778 polymorphism has also been associated with autism spectrum disorders (Campbell et al. 2011; Lerer et al. 2008).
Furthermore, nominally significant interactions between rs1488467 as well as rs4564970 and alcohol on aggressive behavior were found, of which the interaction between the latter and alcohol remained significant after taking into account multiple tests. Alcohol seemed to have a facilitating effect on aggressive behavior only for individuals heterozygous (C:G) on these polymorphisms. The two SNPs were in high LD with each other, suggesting that these SNPs would not be inherited independently of each other. None in the sample had the allelic constellation C:C on either of the SNPs; therefore, it remains unclear whether or not the interaction effect might be linear or result from a dominance effect of the C allele. The rs4564970 and the rs1488467 SNPs have not been studied as extensively as some other OXTR SNPs; however, in a study by Tansey et al. (2010) the rs4564970 SNP was nominally associated with risk of autism in one sample but not in a replication sample.
The 5’-non-coding region upstream in the OXTR, where both rs1488467 and rs4564970 are located close to each other, is likely to contain several transcription factor sites (Lerer et al. 2008), thus, it is possible that these polymorphisms are relevant for the expression of the gene and the amount of receptors in the brain. Further, the current associations may also be due to LD between the rs1488467/rs4564970 and other functional variants affecting OXTR function.
An interaction between alcohol and OXTR polymorphisms could be understood at different levels of action. It is hypothesized that alcohol affects aggressive behavior by disrupting executive functioning, which in turn affects key abilities that influence if a person will act aggressively or not, such as the ability to take the perspective of others, be vigilant to the facial expressions and body language of others, and the ability to correctly interpret these social cues (Giancola 2000; Heinz et al. 2011). As OXT has been associated with numerous similar socially important traits (Domes et al. 2007b; Rodrigues et al. 2009; Zak et al. 2007), one possibility would be that alcohol has a larger effect on aggressive behavior for those who, due to altered OXT signaling, already in a sober state have more difficulties with the above-mentioned social abilities.
Moreover, animal studies have shown that alcohol decrease levels of OXT in rat dams exposed simultaneously to ethanol and nicotine (McMurray et al. 2008), and suppress the release of OXT from the isolated hypothalamo-neurohypophysial system (Hashimoto et al. 1985; Knott et al. 2000). In humans, alcohol decreases the levels of OXT at least in nulliparous and/or lactating women (Menella et al. 2005; Menella & Pepino 2006) as well as during labor (Gibbens & Chard 1976). In addition, it is known that OXT can, through its receptor, stimulate its own release (Neumann et al. 1996). Thus, alcohol could affect the change in OXT levels differently depending on the genotype of the individual on OXTR polymorphisms. It was, however, not possible in this study design to identify the exact route through which an interaction would affect aggression. If replicated, further research would be needed to identify the routes behind the interaction effect.
Before concluding, some reservations should be noted. The participants were between 18 and 30 years old. In this age group, crime rates drop from having peaked in the late adolescence (Bloningen 2010). On the other hand, aggression seems to be a moderately stable behavior (Huesmann et al. 2009; Kokko & Pulkkinen 2005). The results can not be generalized to women without explicitly testing both genders, based on the previous studies indicating potential sex differences related to the effects of OXT (Domes et al. 2010; Kirsch et al. 2005). Although the validity of laboratory experiments measuring human aggression has been debated over time, the construct validity of laboratory measures has been supported (Giancola & Chermack 1998), and Anderson and Bushman (1997) have concluded that they have high external validity and can therefore be generalized to real-life aggression. Aversive noise was used instead of electric shocks in this study. It could be argued that giving electric shocks is a more explicit measure of aggressive behavior than administering aversive noise; however, the use of aversive noise is likely to capture more of the interindividual variation by a lower threshold to respond aggressively. The language of the participants was not measured unequivocally. It should also be pointed out that the negative results could be due to limited power of the study owing to the sample size.
In conclusion, it appears that alcohol could affect individuals in different ways, partly depending on the person's genotype. On the basis of the results from this study, we suggest that the rs1488467 and rs4564970 OXTR polymorphisms are of particular interest for interactive effects with alcohol on aggressive behavior. However, the results need to be replicated in independent samples.
We would like to express our gratitude toward Prof. Zeichner for kindly providing us with the details of the Response Choice Aggression Paradigm (Zeichner et al. 1999). This research was financed by the Finnish Foundation for Alcohol Studies, the National Graduate School of Psychology, a Center of Excellence Grant from the Stiftelsen för Åbo Akademi Foundation (Grant No. 21/22/05), Grant Nos. 136263 and 138291 from the Academy of Finland, the Swedish Medical Research Council, Märta Lundqvist Stiftelse, Åke Wibergs Stiftelse, Åhlén-stiftelsen, Jeanssons-stiftelsen, Söderström-Königska Stiftelsen and the Swedish Brain Foundation.