Sexual conflict can arise when males evolve traits that improve their mating success but in doing so harm females; by reducing female fitness, male harm can diminish offspring production in a population and even drive extinction

Male harm offsets the demographic benefits of good genes. Ewan O. Flintham, Vincent Savolainen, and Charles Mullon. Proceedings of the National Academy of Sciences, March 2, 2023, 120 (10) e2211668120. https://doi.org/10.1073/pnas.2211668120

Significance: Organisms vary in their biological condition. Those individuals in better condition show improved survival or fecundity, and so populations composed of such individuals should be more viable. However, high-condition individuals also express larger sexually selected phenotypes, some of which, in males, can harm females. The good genes hypothesis posits that sexual selection on condition-dependent traits indirectly increases mean condition and therefore population health. Here, using mathematical models, we show that this effect should rarely be expected when sexual traits cause harm: Instead, good genes selection leads to larger harming traits, reduced female fecundity, and potential population crashes.

Abstract: Sexual conflict can arise when males evolve traits that improve their mating success but in doing so harm females. By reducing female fitness, male harm can diminish offspring production in a population and even drive extinction. Current theory on harm is based on the assumption that an individual’s phenotype is solely determined by its genotype. But the expression of most sexually selected traits is also influenced by variation in biological condition (condition-dependent expression), such that individuals in better condition can express more extreme phenotypes. Here, we developed demographically explicit models of sexual conflict evolution where individuals vary in their condition. Because condition-dependent expression readily evolves for traits underlying sexual conflict, we show that conflict is more intense in populations where individuals are in better condition. Such intensified conflict reduces mean fitness and can thus generate a negative association between condition and population size. The impact of condition on demography is especially likely to be detrimental when the genetic basis of condition coevolves with sexual conflict. This occurs because sexual selection favors alleles that improve condition (the so-called good genes effect), producing feedback between condition and sexual conflict that drives the evolution of intense male harm. Our results indicate that in presence of male harm, the good genes effect in fact easily becomes detrimental to populations.

Discussion

Here, we have integrated key aspects from two groups of models: i) sexual conflict (e.g., refs. 18, 22–24, 27, 29) and ii) condition dependence (e.g., handicap models, 40–42, 48). In doing so, we have uncovered some surprising insights into how sexual selection shapes trait evolution and population demography.

Our analyses indicate that, like other sexually selected traits (30, 41, 42), male harm and female tolerance readily evolve condition dependence such that male and female investment into sexual conflict increases with condition. Populations in better condition thus experience more intense conflict, which impairs offspring recruitment and can jeopardize population persistence. In particular, if the severity of male harm increases more strongly with condition than does baseline female fecundity, we observe a counter-intuitive pattern whereby high mean condition is associated with low mean fitness. Such a negative association between condition and fitness is especially likely to emerge where condition has a genetic basis in males and females, as selection favors “good genes” that improve individual condition but also increase the intensity of male harm.

Our results contrast with the common view that sexual selection on good genes also improves mean fitness (43, 55, 65, 66) and mitigates the costs of sexual conflict (42). This is because for the good genes effect to work, that is, for sexual selection to increase mean condition, sexual traits must show appreciable condition-sensitivity (e.g. large κz in our model). In other words, sexual selection acts strongly on condition when good genes confer large increases in the size of sexual traits that males can afford to express. Therefore, the good genes effect is strong where a reduction in mutation load is associated with a significant increase in male sexual trait expression (e.g. in Figure 3A when κz = 1, an increase in condition, green points, is associated with a much larger increase in male trait size, blue points, see also figure 1a in (42)). Importantly, the repercussions of the good genes effect for male trait size are typically larger than for baseline female fecundity. This is because if condition genes greatly improve absolute baseline survival or fecundity, then natural selection should be intense enough to maintain them at relatively high frequency irrespective of the action of sexual selection (Figure 3C see difference between orange curves and black curve decreases with κα). The variation in genetic condition that is available for sexual selection to act upon is therefore limited, constraining the influence of the good genes effect on female fecundity relative to male trait expression. Altogether this means that, when male traits beget harm, the good genes effect has much greater scope to influence demography through sexual conflict than through baseline female fitness. Population benefits of good genes are thus easily reversed by their side-effect in exacerbating antagonistic male-female interactions.

In highlighting the relationship between condition and male harm, our results have implications for empirical work, in particular experimental evolution approaches to unpicking the consequences of sexual selection. Some studies have identified population-level benefits of sexual selection by tracking variation in condition (inferred through condition-correlated traits, e.g., body size, male mating success, offspring viability, or immune function) as a proxy of population viability (e.g., refs. 67–70 reviewed in ref. 71). Our results indicate such traits are poor indicators of population viability in species exhibiting male harm. Without measuring mean fitness, these studies therefore provide ambiguous evidence for an advantage to sexual selection. Indeed, when experimental studies do track more direct measures of mean fitness (such as female fecundity), they typically detect a weaker increase than for indirect measures such as body condition, mutation frequencies, or male mating success (71, 72). For example, a study in fruit flies found that sexual selection was associated with lower mutation load but also diminished offspring recruitment, which was attributed to the effects sexual conflict (73). Furthermore, a number of experimental studies have found evidence that sexual selection improves persistence in populations experiencing environmental stress (e.g., temperature variation in refs. 74–76, reviewed in refs. 77 and 43). This has been attributed to the idea that sexual and natural selection should be more closely aligned in such cases, i.e., that condition genes are more likely to benefit both sexual and nonsexual fitness when both sexes are poorly adapted (78–80). However, our analyses provide an additional explanation here: Environmental stress, by lowering male condition, may reduce the expression of male harm and so increase mean female fitness. More generally, we suggest that condition dependence provides a mechanism for the strength of sexual conflict to diverge in different environments a pattern, for example, observed in drosophila (81), and water striders (82).

To produce tractable results, our baseline model makes a number of simplifying assumptions. We assumed a well-mixed population (i.e., no effects of spatial subdivision); the absence of genetic constraints that affect condition- or sex-specific phenotypic expression; that males direct their mating attention indiscriminately toward high- and low-condition females; and that females express a tolerance trait that mitigates harm but does not impact male mating success. Relaxing these assumptions may alter the ecoevolutionary dynamics of sexual conflict through kin selection (27, 29), the strength of natural selection on females (70, 83), and the presence of coevolutionary intersexual arms races that escalate sexual conflict (24, 28, 84). Moreover, we considered two forms of condition variation, with condition either 1) a purely environmental trait or 2) purely genetic, while in nature, condition may frequently be an intermediate of these two scenarios (46). To relax some of these assumptions, in SI Appendix, Appendices C and D we also analyzed extended versions of our baseline model that allowed us to consider alternative bases of condition and multiple ecological settings for sexual conflict. In certain cases, such as where condition-dependent plasticity is prevented by genetic constraints, or where genotype × environment interactions depress male condition, we find that male harm can be diminished, so aiding population persistence (SI Appendix, Appendix D). However, our main results remain robust, that is, sexual conflict generates a negative relationship between average condition (e.g., good genes) and population size in the vast majority of the cases considered. In fact, we find that the costs of condition-dependent harm for population persistence may be strongly exacerbated in many common scenarios of sexual conflict, such as where female resistance drives evolutionary arms races, or where male and female conditions are encoded by different genes. We therefore suggest that the effects of condition variation found in our model will be general features found across most forms of sexual conflict (SI Appendix, Appendix D for a deeper discussion of extensions to our baseline model and of modeling choice and implications).

Finally, while our study was presented in the context of sexual conflict, our results are also more broadly applicable to the evolution of competitive traits with demographic effects. For example, antagonistic social interactions commonly drive the evolution of weapon phenotypes in both sexes (85, 86). These traits, which are often condition dependent, can influence the survival of interacting conspecifics (12, 13, 20, 33, 87, 88), and so their expression may also diminish mean fitness in high-condition populations.

The acute effects of alcohol on social cognition: High alcohol doses tend to have negative effects, low doses may have some positive effects, e.g., on emotion recognition

The acute effects of alcohol on social cognition: A systematic review of experimental studies. Isabelle Cristina Baltariu et al. Drug and Alcohol Dependence, March 2 2023, 109830. https://doi.org/10.1016/j.drugalcdep.2023.109830

Highlights

• This systematic review focused on social cognition changes after drinking alcohol.

• High alcohol doses (>0.8 g alcohol/kg body weight) tend to have negative effects.

• Low doses (<0.4 g/kg) may have some positive effects, e.g., on emotion recognition.

• Interpersonal characteristics may mediate the effects but remain understudied.

Abstract

Rationale: Alcohol effects on social cognition have been studied by measuring facial emotion recognition, empathy, Theory of Mind (ToM) and other forms of information processing.

Objectives: Using the PRISMA guidelines, we reviewed experimental studies that examined effects of alcohol on social cognition.

Methods: Scopus, PsycInfo, PubMed, and Embase were searched between July 2020 - January 2023. The PICO strategy was used for identifying participants, interventions, comparators, and outcomes. Participants (N=2330) were adult social alcohol users. Interventions consisted of acute alcohol administration. Comparators included placebo or the lowest alcohol dose. Outcome variables were grouped into three themes: facial processing, empathy and ToM, and perceptions of inappropriate sexual behavior.

Results: A total of 32 studies were reviewed. Studies measuring facial processing (67%) often found no effects of alcohol on the recognition of specific emotions, effects facilitated emotion recognition at lower doses but worsened emotion recognition at higher doses. In studies measuring empathy or ToM (24%), lower doses were more likely to lead to improvements while higher doses were generally impairing. Within the third group of studies (9%), moderate to high alcohol doses made it more difficult to perceive sexual aggression accurately.

Conclusions: Lower alcohol doses might sometimes help facilitate social cognition, but most data were in line with the idea that alcohol tends to worsen social cognition, particularly at higher doses. Future studies might focus on examining other moderators of the effects of alcohol on social cognition, particularly interpersonal characteristics such as trait emotional empathy and participant and target gender.

Keywords: alcoholsocial cognitionemotion recognitionempathytheory of mindaggression

4. Discussion

This systematic review summarized experimental studies on alcohol’s acute effects on social cognition, with the aim to gain a better understanding of alcohol-induced changes in social cognition as a mediator in the alcohol - aggression relationship. While people often drink with others to obtain positive interpersonal outcomes (Peele & Brodsky, 2000), changes in the processing of social stimuli may also have inadvertent negative consequences, at least in some people. To examine social cognition changes after drinking, we inspected past alcohol administration studies that assessed alcohol’s effects on facial processing (theme 1), empathy and ToM (theme 2), and perceptions of inappropriate sexual behavior (theme 3). We reviewed 32 studies involving mostly healthy adult volunteers. Methodologically, studies used a wide variety of alcohol doses (range 0.14-2.5 g/kg) and outcome measures.

With respect to theme 1, one key finding of our systematic review was that alcohol can increase the accurate recognition of happiness in emotional faces at both low and high doses (Dolder et al., 2017, Kano et al., 2003, Nagar et al., 2021, Sripada et al., 2011) while at moderate and high doses it can decrease the recognition of fear, sadness, and anger (Attwood et al., 2009a, Attwood et al., 2009b, Borrill et al., 1987, Craig et al., 2009, Dolder et al., 2017, Eastwood et al., 2020, Honan et al., 2018, Kano et al., 2003, Khouja et al., 2019, Nagar et al., 2021, Sripada et al., 2011, Stevens et al., 2008, Stevens et al., 2009). Different emotions are thought to communicate different intentions, with anger inviting confrontation and fear inviting protection, for example (Parkinson, 1996). On the one hand, this has been interpreted as alcohol inducing positive cognitive biases. On the other hand, it is possible that alcohol consumption impairs one’s sensitivity to signals of submission and distress from others, thus increasing the probability of responding aggressively (Attwood & Munafò, 2014).

The results for disgust recognition were less consistent, with both decreases and increases at moderate to high doses (Attwood et al., 2009a, Borrill et al., 1987, Eastwood et al., 2020, Felisberti and Terry, 2015). While alcohol does not always have effects on facial emotion recognition, this does not appear directly related to the administered alcohol dose, which implies that other factors may moderate the effects. For instance, gender (Craig et al., 2008), social anxiety (Stevens et al., 2009), and trait emotional empathy (Dolder et al., 2017) may have a moderating influence. The observed effects of alcohol on facial emotion recognition suggest that, in most individuals, alcohol increases social information processing in a way that may improve their mood (Dolder et al., 2017, Kano et al., 2003, Sripada et al., 2011) while particularly at higher doses the effects on social cognition may be more likely to induce a negative emotional state (Craig et al., 2008; Eastwood et al., 2020; Honan et al., 2018) thereby explaining a higher risk for aggression (Heinz et al., 2011). Besides impairing detection of negative expressions, drinking alcohol can increase the misclassification of negative expressions as neutral (Atwood et al., 2009a; Kamboj et al., 2013). While this may serve to reduce social anxiety (Stevens et al., 2008), thereby making alcohol a social lubricant, at the same time alcohol might reduce controlled, effortful cognitive processing, thereby creating a narrowing effect on attention. In this process, called “alcohol myopia”, the range of social cues that can be adequately processed becomes restricted (Steele & Josephs, 1990). Even though alcohol can have an anxiety- reducing effect at lower doses, at higher doses can negatively influence social judgment (Francis et al., 2019, Johnson et al., 2018, Herzog, 1999).

A second key finding for theme 1 was that while at moderate doses alcohol can increase facial scanning and improve facial memory (Colloff & Flowe; 2016) at high and moderate doses these cognitive processes are more likely to become disrupted. In general, studies were more likely to find positive effects on social information processing at lower doses and to find impairing effects at higher alcohol doses (Borrill et al., 1987, Harvey, 2014, Sripada et al., 2011). These findings are in line with the idea that dose moderates the effects of alcohol on social cognition. For example the findings in the study by Penton-Voak et al. (2012) suggest that after a moderate alcohol dose women perceive a threat in men but they can have increased sex-related cognitions; while men get worse at perceiving social signals this leading to escalating male-male aggression.

With respect to theme 2, while studies found that moderate doses had variable effects, high doses generally decreased empathy and ToM (Battista et. al., 2014; Francis et al., 2019; Hu et al., 2018; Karlén et al., 2019). The decrease in empathy may be closely linked to impaired recognition of negative emotions (particularly fear and sadness) as mentioned above. Moreover, decreased empathic ability has been linked to aggression before (Dethier and Blairy, 2012, Philippot et al., 2005).

Finally, three studies that focused on perceptions of inappropriate sexual behavior (theme 3) all found that, at moderate to high doses, alcohol negatively influences the accurate processing of cues of sexual aggression (Gross et al., 2001, Loiselle and Fuqua, 2007). These results could be taken to mean that alcohol increases the risk of sexual aggression. This idea is consistent with multiple previous studies that found increased hostility and aggressive behavior after alcohol, even though these studies generally did not focus on sexual aggression (Bushman and Cooper, 1990, Ogle and Miller, 2004). Indeed, while all reviewed studies focused on aggression between men and women, experimental studies on the effect of alcohol on aggression have generally not examined both participant gender and target gender.

The results for alcohol’s effects on the perception of inappropriate sexual behavior can be connected to the aforementioned results on facial emotion processing (theme 1): the reduced anger recognition accuracy (Khouja et al., 2019) and increased tendency to see happy expressions (Dolder et al., 2017, Khouja et al., 2019) could help explain why, after moderate doses of alcohol, women had worse risk detection while listening to a date rape (Loiselle & Fuqua, 2007). Similarly, men’s tendency to judge women’s friendliness and arousal as higher after a moderate or high alcohol doses may be due to impaired processing of sad, disgusted, and fearful (Felisberti and Terry, 2015, Honan et al., 2018, Kamboj et al., 2013) expressions, in combination with improved processing of happy expressions (Dolder et al., 2017, Kano et al., 2003, Sripada et al., 2011). Moreover, men’s decreased perception of antisocial behavior at high alcohol doses (Karlén et al., 2019) fits with the results about decreased anger recognition (Borrill et al., 1987, Khouja et al., 2019). Importantly, women but not men became better at perceiving antisocial behavior, both with the moderate alcohol dose and with the high dose (Karlén et al., 2019). These findings are in line with previous studies that confirmed the alcohol-aggression link: experimental studies showed that acute alcohol consumption can lead to aggression (Chermack and Taylor, 1995, Dougherty et al., 1999, Giancola, 2002, Giancola, 2004, Zeichner et al., 1994), while correlational studies found that even in the context of chronic alcohol use aggression is most often reported during episodes of acute intoxication (Chermack & Giancola, 1997). In line with our results, clinical evidence suggests that emotion recognition and ToM ability is indeed impaired in AUD individuals (Kumar et al., 2022, Le Berre, 2019), and this is thought to be due to the large amount of drinking.

Overall, our review suggests that although alcohol in lower doses may improve social functioning, at higher doses alcohol is particularly likely to impair social cognition, thereby it may increase the risk of aggression in some people. Specifically, while results were not consistent across studies, lower doses were able to increase facial scanning (Colloff and Flowe, 2016, Penton-Voak et al., 2012), happiness recognition (Dolder et al., 2017, Kano et al., 2003), empathy and ToM (Dolder et al., 2017, Johnson et al., 2018), and decrease the recognition of negative emotions (Atwood et al., 2009b; Craig et al., 2008; Eastwood et al., 2020; Honan et al., 2018; Kamboj et al., 2013; Nagar et al., 2021). Higher alcohol doses were able to decrease not only the recognition of negative emotions (Sripada et al., 2011) but also facial scanning (Colloff and Flowe, 2016, Honan et al., 2018), and happiness recognition (Kano et al., 2003), empathy and ToM (Francis et al., 2019, Herzog, 1999, Thiel et al., 2018). The fact that results were not consistent across studies might partially be due to moderating factors other than alcohol dose. Individual characteristics such as participant gender (Battista et al., 2014; Craig et al., 2008; Karlén et al., 2019, Penton-Voak et al., 2012) and social anxiety (Stevens et al., 2008, Stevens et al., 2009) may play a role, but so may situational variables such as target gender (Penton-Voak et. al., 2012) and social context (e.g., Dolder et al., 2017; Thiel et al., 2018). For one, we propose that more careful studies of (participant and target) gender as a potential moderator of the effects of alcohol on social cognition are needed. Importantly, researchers conducting these studies should decide on whether to vary doses by participant gender or not. Additionally, we propose to study relevant interpersonal characteristics as potential moderators, such as trait emotional empathy (Dolder et al., 2017, Giancola, 2003).

4.1. Limitations of the review

We previously noted that the studies reviewed in this manuscript used a wide variety of alcohol doses (range 0.14-2.5 g/kg). While categorizing alcohol doses into low, moderate, and high facilitates their interpretation, this categorization could lead to missing subtle differences between doses, as BAC would more accurately represent alcohol levels in the blood (Martin and Sayette, 1993). However, only two of the reviewed studies (Kamboj et al., 2013, Nagar et al., 2021) compared alcohol’s effects at lower and higher BAC.

Related, we also noted that the experimental procedures for alcohol administration differed substantially between studies. In this context, it is of relevance that alcohol has biphasic effects: stimulation is observed on the ascending limb of the absorption curve and sedation is observed on the descending limb (Pihl et al., 2003). To get to the peak of the ascending limb it takes 36 minutes and to get to the peak of the descending limb 109 minutes (Marczinski & Fillmore, 2005). In some studies, participants were asked to start completing performance measures of social cognition immediately after alcohol consumption (Francis et al., 2019, Stevens et al., 2009, Thiel et al., 2018), while in another study, participants were instructed to wait 40 minutes while completing baseline questionnaires (Kamboj et al., 2013) or to complete the task only when the desired level of intoxication was achieved (assessed using breathalyzer tests) (Colloff & Flowe, 2016). Participants in the former studies were thus more likely to complete tasks while on the ascending limb than participants in the latter studies. This between-study variation in the time between alcohol administration and social cognition assessment might help explain between-study variation in the effects of alcohol on facial emotion processing (theme 1), for example. In addition, not all facial emotion recognition studies report an overall alcohol effect on general emotion recognition, however in which studies this information was available, this effect also varied from small (Khojua et al., 2019; Padula et al., 2011), moderate (Borrill et al., 1987;Craig et al., 2008; Eastwood et al., 2020; Felisberti & Terry, 2015; Tucker & Vuchinich, 1983) and high (Honan et al., 2017; Nagar et al., 2021; Sripada et al., 2011).

Another limitation of our review was that we included blinded studies (involving double or single blinding, see Table 1) (Atwood et al., 2009a; Atwood et al., 2009b; Craig et al., 2008; Dolder et al., 2017; Felisberti and Terry, 2015; Kano et al., 2003), unblinded studies (Battista et al., 2014, Farris et al., 2010, Herzog, 1999), and studies for which blinding was not reported (Stevens et al., 2008, Stevens et al., 2009). Most of these studies included participants who at least expected to drink alcohol during the study. People with alcohol expectations tend to adapt their behavior to match expected alcohol effects (Martin and Sayette, 1993). Therefore, the use of an effective placebo manipulation is generally recommended (Babor, 2000) but our review included studies that differed in the way the blinding was organized and if it was successful. Also, studies did not report information on the blinding of researchers, and if this was checked.

A final limitation of our review is that the measures used to assess aspects of social cognition variables varied widely across studies. Some of the reviewed tasks offer more ecological validity in the assessment of social cognition thanks to the stimuli having similarities with real life situations during which a person might consider others’ emotions, thoughts, intentions, or behaviors (Hogenelst et al., 2015). Examples include the empathic accuracy task, or the video tape of interpersonal violence situation, and even the audiotape of the date rape. Conversely, results from basic tasks involving simple stimuli, e.g. black and white still images of facial expressions should be interpreted in light of their relative lack of ecological validity.

4.2. Implications of this review for future research

The present review shows that drinking alcohol can negatively affect social cognition, particularly at higher doses. This supports the idea that changes in social cognition can mediate alcohol’s effect on aggression. However, given substantial variability in results between studies, even at comparable (low-moderate-high) doses, we recommend that future studies focus on elucidating which person-level and situation-level variables might act as moderators. For example, at the person level, factors like alcohol tolerance, metacognitive processes, personality factors (Kalant, 1996, Kuntsche et al., 2008, Spada and Wells, 2006) might be interesting to study because the alcohol aggression relationship may not be present all the time and in all individuals. Also, situational aspects of a social interaction involving alcohol (or an interaction-like situation created in a performance-based measure) might be considered. For example, the next step would be studying the behavior in a social interaction situation (Zawacki, 2011) or employing several potential relevant social cognition measures after drinking (Schreiter et al., 2013).

Finally, the finding that alcohol can acutely impair social cognition, particularly at higher alcohol doses (this review) does not exclude the possibility of reverse causality. That is, by improving social cognition, social drinking may be reduced. In other words, social cognition training may diminish alcohol use in social settings, which may have clinical implications. Several such training options are available: micro-expressions training (Russell et al., 2006), example emotion and ToM imitation training (Mazza et al., 2010), and a dynamic interactive social cognition training in virtual reality (Nijman et al., 2020). While the effects of social cognition training on drinking in a social context have not been studied, to learn more about the role of social cognition in the alcohol-aggression link, this may be an important next step.