We agree that Sijilmassi et al. have identified an intriguing puzzle and developed a promising solution, though one with an important gap. We are persuaded that motivating narratives of common origin and common fate are sufficiently similar, widespread, and culturally salient as to call for explanation. True, there are no formal phylogenetic controls on the case studies. But the geographic spread and internal variation makes it unlikely that this pattern is an accident of history, widespread now as a result of inheritance from one or few ancestral cultures.

The proposed explanation – that these narratives support large-scale collective action by supporting belief in mutual mass interdependence – has considerable plausibility. As the authors point out, individuals have limited resources to invest in coalitional choice and the stakes can be high. Joining the right coalition can bring great rewards, but the wrong choice can be catastrophic. They further note that cooperative choices are selectively favoured when the cooperating individuals' fitnesses are aligned or interdependent. Indeed, as a culturally complex species for whom differential expertise and divisions of labour make environmental adaptation very much a collective enterprise, fitness interdependences are undeniably part of behaviourally modern human lifeways. Information about interdependence should therefore be vitally motivating, and a shared history of successful mutual interaction is good prima facie evidence of interdependence.

However, we should resist generalising this reasoning to the macro-scale of tribes, nation states and ethnolinguistic groups. First, group membership at this scale becomes more exclusive, as these are not the porous, fission–fusion groups of forager lifeways. So while belief in positively entangled interests across such groups may increase the investments that members are willing to make, it can rarely motivate joining. Second, as the authors are aware, there is a crucial distinction between actual interdependence, where success for Oscar really does depend on success for Max (and others), and the mere belief in interdependence. This is important because interdependence does not scale up well. Genuinely interdependent collective action becomes fragile at scale, because if any essential participants fail in their role then the efforts of all are undermined. Collective action at scale therefore demands structural redundancies, and the fungibility of individuals. But once Max's specific participation in the collective action project becomes fungible, Oscar's success does not depend on Max and their fitness interests (in this respect) are no longer interdependent. All else equal, collective action becomes more effective as one adds both headcount and role redundancy (hence God is on the side of the big battalions), but interdependence declines as the individual contribution of each agent becomes less critical to the overall outcome. At the macro-scale at which most of Sijilmassi et al.'s discussion is focussed, genuine interdependence has largely disappeared.

To some degree the authors fudge this, claiming that at the tribal and nation state level agents have a genuine stake in the general welfare of the collective of which they are a part (target article, sect. 3.2). In some cases (such as existential struggles) this might be true. But unless their hypothesis appeals to high-level selection (and the authors explicitly deny this), “general welfare” is an uncashed metaphor. For the most part, the fitnesses of any two individuals in large collectives are not that closely aligned, both for the reasons stated above, and because each should prefer that the other be in the front lines of any battle (real or metaphorical) instead of themselves. As Sijilmassi et al. also accept, interdependence almost always fails in one direction: even if Oscar depends on the general welfare, the general welfare is independent of Oscar and his deeds – and at large scales an individual's contribution to collective outcomes is vanishingly small (a familiar conundrum in collective action topics from climate action to the rationality of voting). Similarly, the second adaptive demand on coalition-joining agents – recruiting new parties to their coalitions – also evaporates, as almost none of us are positioned to increase the patriotic fervour of our fellow citizens in ways that would be significant with respect to (fitness-aligned) outcomes. Figure 2 is therefore misleading in large-scale contexts.

So this proposal faces a crucial problem. What makes narratively transmitted illusions of interdependence so seductive? If coalitional participation is a high stakes decision, we would expect well-honed mechanisms of epistemic vigilance to detect fraud and manipulation. The discussion in section 3.1 (target article) shows that in decisions about cooperation and alliance formation, agents are typically nuanced and canny. They are not easily imposed upon. Moreover, agents should be especially wary about committing to coalitions, because false positives (join when you should not) are more costly than false negatives (failing to sign up to a good deal). It is not always prudent to shirk or be a neutral, but it is nearly always better than throwing in with a lost cause (unless your joining would be the difference maker).

Sijilmassi et al.'s main explanation for the compelling character of myths of common origin and fate appeals to intuitive psychology; more particularly, to a commitment to the reality of social collectives as integrated wholes with deep shared histories. But in the absence of genuine fitness interdependence this is on similar ground to the account that they rightly reject in discussing rival views: elite manipulation and origin myths generating cooperative commitments through illusory shared kinship. Why would fictive kinship, generated by a terminological sleight of hand be credible? Indeed! But equally we need an explanation of why epistemic vigilance does not dissolve the illusion of common fate through shared history on the large scale. Because fate is not common, and history is not shared.

We suggest that Sijilmassi et al.'s model, elite manipulation, and fictive kinship are all implicit adaptive lag hypotheses. Agents are vulnerable to illusions of aligned fitness interests (albeit different illusions), because they continue to rely on cues of fitness alignment that were once reliable, because they were so in small social worlds, but are reliable no longer. If we are right, this locates the problem – an updating failure – but does not solve it.