Why can an arbitrary tag work as a kinship cue at all?
A tag is just a marker — a smell, a color, a call — that carries no inherent meaning about family. It becomes useful for kin selection only through coevolution: if individuals who share a tag tend to interact with relatives, then responding altruistically to tag-sharers can pay off even though the tag itself says nothing about genes [1]. Axelrod and colleagues frame this by extending Hamilton's rule, the classic condition under which altruism spreads when it benefits relatives, to cover kin recognition built on such arbitrary tags [1].
The key point is that the tag and the altruistic response are evolving together. The tag is not a reliable label that altruism simply reads off; the association between tag-sharing and relatedness is itself a product of the joint evolutionary process [1]. That is why the strategy can work despite the tag being arbitrary in origin.
What happens to altruism when the cue stops being trustworthy?
When relatedness cues are unreliable, the case for discriminating altruism weakens. Lehmann and Perrin modeled phenotype matching — recognizing kin by resemblance to oneself or to familiar others — in island-model populations, where groups are somewhat isolated and dispersal shapes who meets whom [2]. Their analysis indicates that as cues carry less reliable information about kinship, the benefit of discriminating shrinks, and with it the sociality that recognition can support [2].
Their model also points to a partial remedy: increasing cue heritability, meaning the degree to which the cue is passed down and tracks actual relatedness, improves recognition [2]. So unreliability is not simply fatal; it shifts the conditions under which altruism can be sustained. The same logic applies to arbitrary tags, where the tag's usefulness depends on how tightly it stays coupled to real kinship over evolutionary time [1].
Dispersal changes who counts as a reliable match
Recognition does not happen in a vacuum; it depends on population structure. In the island-model setting, dispersal patterns determine how often individuals encounter relatives versus unrelated look-alikes, and that in turn shapes whether phenotype matching is worth using [2]. Environmental factors in the model affect both discrimination and the emergence of sociality, so the same recognition rule can pay off in one dispersal regime and not another [2].
This matters for the unreliable-cue question because unreliability is partly a matter of ecology. A cue that is noisy in a well-mixed population may be informative in a structured one where relatives cluster. The models treat recognition and sociality as coevolving, so the answer to "does altruism survive unreliable cues?" depends on the surrounding dispersal and environmental context rather than on the cue alone [2].
Where the coevolutionary picture is still incomplete
The tag-based framework establishes conditions under which kin recognition can support discriminating altruism even as coevolution shifts cue reliability, but the analysis is framed around those conditions rather than around a full account of what happens when cues degrade badly [1]. The phenotype-matching model addresses unreliable cues directly but uses resemblance-based recognition rather than arbitrary tags, so its conclusions transfer to tag systems only by analogy [2].
Put together, the two lines of work agree on the direction — altruism tracks cue reliability — while leaving open how quickly tag-based systems collapse or re-stabilize when reliability falls. That is a question about the dynamics of coevolution, not just its endpoints, and it is where the evidence is thinnest.
Sources used in this answer
Altruism via kin‐selection strategies that rely on arbitrary tags with which they coevolve
Kin-selection strategies based on arbitrary tags can support discriminating altruism when tags and altruism coevolve, and Hamilton's rule can be extended to state the conditions for this even as coevolution changes cue reliability.
Altruism, dispersal, and phenotype-matching kin recognition
In island-model populations using phenotype matching, less reliable kinship cues reduce the benefit of discrimination and the sociality it supports, while higher cue heritability improves recognition.