Most contemporary evolutionary explanations of sex differences in aggression involve inter-male competition, a component of sexual selection (Darwin, Reference Darwin1871/1901; Trivers, Reference Trivers and Campbell1972). Campbell (Reference Campbell1999) departed from this by examining selection pressures on females, proposing that sex differences in aggression and associated attributes arose from selection pressures on females to facilitate staying alive, and consequently being more likely to rear their offspring to maturity. The target article widened Campbell's “staying alive theory” (SAT) to include a range of self-protective reactions to various kinds of threat. Extended SAT provides a new way of considering attributes that are at present viewed individually (often as pathologies), enabling them to be viewed as a set of female adaptations.

The extended SAT raises the question of whether the specific attributes covered, such as immune responses, pain thresholds, emotions, and personality, form a coherent adaptive complex, and if so whether they are linked to one another at an individual level, so that someone with a strong immune response also shows a low pain threshold and more sleep disturbances. Alternatively, the various attributes covered by SAT may represent alternative ways of achieving the adaptive end-result of facilitating female longevity.

In psychology, discussions and syntheses of psychological sex differences have typically focused on individual attributes rather than coherent groups of attributes. This follows earlier syntheses rooted in individual differences research (e.g., Garai & Scheinfeld, Reference Garai and Scheinfeld1968; Maccoby & Jacklin, Reference Maccoby and Jacklin1974), culminating in more recent times with meta-analyses of specific attributes. These meta-analyses were synthesized by Hyde (Reference Hyde2005) and Zell, Krizan, and Teeter (Reference Zell, Krizan and Teeter2015), both presenting sex differences as a list and calculating the percentage that were of different magnitudes (large, medium, small, or null). Conclusions were drawn on the basis of how many of these attributes fell into the different categories. Links between the individual attributes were not part of this approach, which is essentially atheoretical.

In contrast, broad theoretical explanations, whether they be in terms of social roles (e.g., Eagly, Reference Eagly1987; Wood & Eagly, Reference Wood and Eagly2012) or evolutionarily based adaptive complexes, such as sexual selection (Archer, Reference Archer2019), or the extended SAT, raise the issue of how the various attributes might be related to one another. The statistical underpinning of sex differences viewed as sets of attributes, rather than individual items, has only recently begun, following the work of Del Giudice (Reference Del Giudice2009). For example, from an evolutionary perspective, sex differences in aggression can be viewed as being linked to other attributes that are the evolutionary consequence of male competition, such as bodily strength (Sell, Eisner, & Ribeaud, Reference Sell, Eisner and Ribeaud2016; Sell et al., Reference Sell, Cosmides, Tooby, Sznycer, von Rueden and Gurven2009). Likewise, the attributes considered in the revised SAT could be regarded as an adaptive complex or as a series of alternative adaptations whose common function is to minimize mortality in the sex that is the primary caretaker of the offspring.

The meta-analyses and large-sample studies that form the bases of existing syntheses of studies of sex differences are usually concerned with specific attributes, such as aggression, or depression, and typically the summary statistic used is Cohen's d, the standardized difference between the mean values for males and for females. This is appropriate for summaries where attributes are viewed as independent items (e.g., Hyde, Reference Hyde2005; Zell et al., Reference Zell, Krizan and Teeter2015). However, where groups of coherent attributes are being considered, it may be appropriate to consider such groupings together and to apply the multivariate D statistic (Mahalanobis distance), which represents the distance between the two sexes in multivariate space. Despite some persuasive arguments that this type of analysis is the appropriate way forward (e.g., Del Giudice, Reference Del Giudice2009; Del Giudice, Booth, & Irwing, Reference Del Giudice, Booth and Irwing2012; Eagly & Revelle, Reference Eagly and Revellein press), it has seldom been carried out in practice (for exceptions, see Lippa [Reference Lippa2001], for masculinity–femininity; Conroy-Beam, Buss, Pham, & Shackelford [Reference Conroy-Beam, Buss, Pham and Shackelford2015], for mate-choice criteria).

Both the modified SAT theory, and my own evolutionarily based review of sex differences (Archer, Reference Archer2019), relied on the more conservative approach of using d vales as summaries of sex differences in individual attributes. This underestimates the size of differences, so that when these are considered as groups of attributes using the multivariate Mahalanobis D (Eagly & Revelle, Reference Eagly and Revellein press), much larger values are typically found than when individual d values are averaged: One important caveat is that larger Mahalanobis D values are obtained where the individual attributes show lower intercorrelations. These observations lead back to the empirical question of the extent to which individual attributes covered by SAT form a coherent grouping, or represent alternative responses that all have the consequence of reducing mortality.

A final point is the extent to which the extended SAT complements or replaces sexual selection as an evolutionary explanation of human sex differences. In view of coherent sets of male attributes (musculature, aggression, impulsiveness), it is likely to be complementary, although this could be subject of future study. To begin with, it would be a simple matter to compare the scope of the present article to the attributes associated with sexual selection in previous syntheses (e.g., Archer, Reference Archer2019; Daly & Wilson, Reference Daly and Wilson1988), and the extent to which these two groupings were related or independent. The extended SAT could also be applied to relevant within-sex variations: We would predict that men showing greater degrees of paternal care would score higher on attributes covered by the modified SAT.