Why these three explanations are not really rivals
Altruism is puzzling because helping another at your own cost seems disadvantageous. The classic answer is kin selection: if the beneficiary shares your genes, helping them can still spread those genes. Multilevel selection keeps that logic but widens the unit of selection beyond genes and individuals to include groups, so that group-level cohesion and competition between groups can also favor altruistic behavior [2]. In that framing, kin selection is not replaced; it becomes one ingredient inside a broader multilevel account.
Costly signaling answers a different question. Instead of asking why someone helps a relative, it asks why someone would pay a visible cost to help or share when the benefit is not obviously returned. A costly display can be credible precisely because it is expensive, so observers can read it as a signal of quality or status [4]. That is a separate explanatory job from kinship, which is why the two ideas can sit side by side rather than fight over the same ground.
When group-level selection does the explanatory work
The clearest case for extending kin selection comes from large-scale human groups. Nations are often treated as artificial constructs, but a sociobiological view treats them as extended kin groups whose cohesion arises from genetic relatedness, culturally evolved institutions, and intergroup competition [2][5]. Here multilevel selection is doing real work: it explains cooperation among people who are not close relatives by pointing to shared ancestry at a broader scale plus institutions and competition between groups.
The scope matters. This argument is about nations and other large cultural groups, not about a single alarm call between siblings. Within that scope, the claim is that kin selection alone is too narrow, and multilevel selection refines rather than overturns it [2][5]. The two papers make the same move, so the complementarity is consistent across them rather than a one-off proposal.
Where costly signaling adds something kinship cannot
Costly signaling enters when altruism is public and status-relevant. In a spatially explicit model of food sharing in fragmented environments, the assumptions include that others notice or attribute status to generous donors, which is the costly-signaling channel [4]. That model also invokes kin selection and reciprocal altruism, so signaling is one strand in a mixed explanation rather than a standalone theory of all sharing.
Human studies point the same way. A program testing predictions from several evolutionary theories treats costly signaling in a sexual-selection context as a distinct hypothesis alongside kin selection, reciprocity, and multilevel selection, and notes that few works directly compare them [3]. The useful takeaway is that these theories generate different predictions about who helps and when, so they can be tested against each other even while operating together in real populations.
What kin selection still explains on its own
Some altruism needs no group-level or signaling story. Alarm signaling is the classic case: a caller pays a direct cost by drawing predator attention while nearby conspecifics benefit, and kin selection plus reciprocal altruism are the mechanisms used to explain how such signaling persists [1]. That account also describes how groups protect signaling integrity through receiver discrimination, physical policing, and game-theoretic constraints, which is about keeping the system honest rather than about group competition or status display [1].
So the division of labor is fairly clean. Kin selection covers altruism among relatives and the stability of risky signals within social groups [1]. Multilevel selection covers cooperation at the scale of larger groups, especially where institutions and intergroup competition matter [2][5]. Costly signaling covers public generosity that advertises quality or status [3][4]. They compete only when they make conflicting predictions about the same behavior, and the human work suggests the more productive approach is to test those predictions side by side [3].
Sources used in this answer
The Evolutionary Architecture of Social Alarm Signaling: Altruism, Kin Selection, and the Systematic Mitigation of Deception
Alarm signaling persists through kin selection and reciprocal altruism, with receiver discrimination, policing, and game-theoretic constraints maintaining signaling integrity.
From Kin Selection to Multilevel Selection
Multilevel selection refines kin selection by treating nations as extended kin groups whose cohesion comes from relatedness, culturally evolved institutions, and intergroup competition.
Predictors and correlates of altruism in humans from the evolutionary viewpoint
Kin selection, reciprocity, sexual selection, and multilevel selection make different predictions about altruism, and few studies have directly compared them.
Multilevel selection and the evolution of food sharing in fragmented environments: a spatially explicit model and its implications for Early Stone Age behavior
A spatially explicit model of food sharing assumes that generous donors may be noticed or granted status, a costly-signaling channel alongside kin selection and reciprocal altruism.
Nations as Natural Families: From Kin Selection to Multilevel Selection
Nations can be viewed as extended kin groups, with multilevel selection extending kin-selection explanations of altruism to include institutions and intergroup competition.