Depression and suicidality as evolved credible signals of need in social conflicts

Depression and suicidality as evolved credible signals of need in social conflicts. Michael R. Gaffney et al. Evolution and Human Behavior, February 25 2022. https://doi.org/10.1016/j.evolhumbehav.2022.02.004

Abstract: Mental health professionals generally view major depression and suicidality as pathological responses to stress that elicit aversive responses from others. An alternative hypothesis grounded in evolutionary theory contends that depression and suicidality are honest signals of need in response to adversity that can increase support from reluctant others when there are conflicts of interest. To test this hypothesis, we examined responses to emotional signals in a preregistered experimental vignette study involving claims of substantial need in the presence of conflicts of interest and private information about the signaler's true level of need. In a sample of 1240 participants recruited from Amazon Mechanical Turk, costlier signals like depression and suicidality increased perceptions of need, reduced perceptions of manipulativeness, and increased likelihood of support compared to simple verbal requests and crying without further symptoms. The effect of signaling on likelihood of support was largely mediated by the effect of signaling on participants' belief that the signaler was genuinely in need. Our results support the hypothesis that depression and suicidality, apparent human universals, are credible signals of need that elicit more support than verbal requests, sad expressions, and crying when there are conflicts of interest.

Keywords: DepressionSuicideEvolutionary medicineCostly signalingMental health

4. Discussion

As predicted, in vignettes involving conflicts of interest and private information about the need for help, costly signals of need increased participants' belief in the victim's claims and their likelihood of helping her, with the increase in belief and the likelihood of helping increasing monotonically with signal cost. As predicted, the increase in likelihood of helping was largely mediated by the increase in belief in the victim's claims. In an exploratory analysis, costlier signals also decreased perceptions that the victim was manipulative. These results provide evidence that, contrary to the influential “interpersonal” view that depressive behaviors are socially dysfunctional (reviewed in Hames et al., 2013), they in fact outperform verbal requests, sad expressions, and crying in providing benefits to victims when there are conflicts of interest.

Signal effects were largest in the “brother-in-law” and “romantic partner” vignettes, both of which involved claims of assault against participants' imagined daughters, and smaller in the “basketball coach” and the “thwarted marriage” vignettes. The smaller effect in the “basketball coach” vignette might have been because in the role of athletic director, participants did not value their relationship with the star player as much as we anticipated (e.g., due to lack of relatedness), or how participants weighted the costs of suspending the coach vs. punishing a potentially innocent person (for discussion of suicidal signaling to kin vs. nonkin, see Syme & Hagen, 2018). The US vignettes also had different degrees of evidence against the victim beyond just denial by the accused, ranging from strong evidence in the basketball coach vignette (a negative police report) to moderate evidence in the romantic partner vignette (no physical injuries) to weak evidence in the brother-in-law vignette (nothing beyond denial by the brother-in-law), raising the possibility that credible signals are more effective when negative evidence is lacking (Dylan Tweed, personal communication).

The small signal effect in the “thwarted marriage” vignette, which involved the Indian sample, could indicate that our results do not generalize across cultures, undermining our adaptationist hypothesis. It could also reflect our poor understanding of contemporary Indian culture regarding dowry (the effect was larger in older participants). Baseline belief in the older daughter, and likelihood of helping her, was relatively high at baseline (58%) compared to victims in the other vignettes. Private information and conflict therefore probably played a smaller role and thus costly signals were less necessary. We observed a similar pattern in our pilot study, in which baseline belief in the victim's need was high, and costly signals had smaller effects than they did in the current study. Additionally, supporting the older daughter came at the cost of one's younger daughter, which may also help explain the relatively small signal effects. A final consideration is that data from the Indian sample appeared to be of lower quality, limiting our confidence in any of these interpretations (see the Limitations section for more information).

In the “brother-in-law” and “romantic partner” scenarios, Crying had little effect on the magnitude of pro-victim responses relative to Verbal request, suggesting it was not costly enough to serve as a reliable signal in times of substantial conflicts of interests. In contrast, both Depression conditions increased support, albeit to similar degrees. One potential reason for the similar effects of the Depression conditions is the increase in costs from Mild depression to Depression was small (e.g., grades dropping from As to Bs in Mild depression vs. Cs in Depression). Such small changes may be less impactful in vignettes than in real-life, where the effects of signaling may increase in severity as they persist over time.

The effect of Suicide attempt on T2 Belief was similar to the Depression conditions across vignettes, but it resulted in greater T2 Action. One interpretation is that although some participants did not believe the victim's story, her signal nevertheless convinced them that she needed help. For example, maybe the brother-in-law did not assault her, but the presence of his family in her home was causing genuine distress. Support for this interpretation comes from our mediation analyses, which showed that the likelihood of help was largely, but not entirely, mediated by signal's effect on belief in need.

There were minor associations of age and sex with T2 Belief and T2 Action in the US participants, with both being higher among females and younger individuals. The US vignettes all involved assaults against young women, which might have been more salient to female and younger participants. In the Indian sample, T2 Belief and T2 Action were somewhat lower among those with more education and among females, respectively. Costlier signals, suicidality in particular, had a larger effect among older individuals, perhaps because older individuals were more likely to have children of marriageable age, like the victim in the vignette.

Contrary to our adaptationist hypothesis, and supporting the mainstream view that depression is a psychopathology, participants' perceptions that the victim was mentally ill increased with signal cost. However, there have been extensive media campaigns to convince the public that depression is a mental illness with the laudable goal of reducing stigma (Corrigan, 2012; Rüsch, Angermeyer, & Corrigan, 2005). Even so, in the Depression conditions across vignettes, no more than 25% of participants thought the victim was mentally ill, and in the Suicide attempt condition the proportion of participants perceiving mental illness exceeded 50% only in the basketball coach vignette. Although perceived mental illness was associated with somewhat lower T2 Belief and T2 Action, this effect was mainly evident in the Verbal request and Crying conditions.

Finally, after the T3 evidence that the victim was telling the truth, likelihood of helping by the US participants increased to near ceiling, an effect that helped validate our vignettes. Among Indian participants, in contrast, participants only slightly increased their likelihood of helping from their T2 level. One interpretation of the latter is that Indian participants tended to believe the older daughter anyway, so their decision to help was not changed by additional information.

4.1. Limitations

This study has less ecological validity than real-world observations of depressed individuals interacting with their social partners, which might have biased results in a pro-signaler direction if the lack of real costs of helping made support feel less costly or if there was a social desirability bias toward helping (Grimm, 2010). It may have also biased participants against helping if they could not fully imagine the characters in the story as kin or interdependent partners, and the survey's short duration may have weakened the strength of the costlier signals as bargaining tools.

Our design did not include vignettes with male signalers. For this reason, we have no data on the possibility of sex differences in the effectiveness of the signaling strategies examined. Although not predicted theoretically, such differences are possible if the costs of signaling vary between the sexes due to differential access to alternative bargaining strategies (Hagen & Rosenström, 2016) or if one sex tends to suffer greater negative reputational effects when displaying the emotions and behaviors in the vignettes. It is also possible the costliness of the situations presented in the vignettes differ by sex. This study therefore most clearly demonstrated the effectiveness of costly signals of need by females, leaving open the question of the effectiveness of costly signals of need by males.

Compared to the US sample, far more Indian participants failed our attention checks, which is consistent with botting, unfamiliarity with English, or low-effort responses (Kennedy et al., 2020). If this high failure rate indicates lower-quality responses among those who passed the attention checks, the weak signal effect in the thwarted marriage vignette may simply be due to greater noise rather than differences in the scenario or the effectiveness of the signals compared to those in the US. Another concern relevant to all vignettes is that our decision to anchor the T1 sliders at 0 may have resulted in participants being more likely to report extreme values.

Finally, we adopted game theory models of bargaining with incomplete information as our theoretical framework, but there are many other models of credible signaling (e.g., Számadó, 2011), including for need (Számadó, Czégel, & Zachar, 2019) and suicidality (Rosenthal, 1993). If depression and suicidality involve signaling, they might be better explained by a different model.