This month’s issue of PLoS Computational Biology contained an interesting article entitled “Signalling and the Evolution of Cooperative Foraging in Dynamic Environments“. Authored by Colin J. Torney, Andrew Berdahl, Iain D. Couzin (all of Princeton University), the article seeks to understand the ecological conditions under which costly signaling can evolve.
Many animals emit signals to conspecifics in response to finding food. At first pass, this behavior seems maladaptive: letting others know that there is food nearby might lead to faster depletion of the resource, or at least reduce the signaler’s competitive advantage, especially if some of the beneficiaries of the signal do not themselves signal when food is present. The only way that costly signaling could be an adaptive behavior is if others reciprocate when they find food, thus increasing the overall search area of all signalers.
The model used to investigate whether costly signaling can evolve relies on a spatial representation of both food and individual organisms. Food is assumed to be ephemeral, so that finding a food source now does not guarantee successful foraging in the near future. Both “full” and “reduced” forms of the model are used to determine which ecological conditions (if any) favor costly signaling. Perhaps the easiest way to understand the manner in which this model functions is to look at the video posted here. It shows a run of the reduced model, with green “signalers” and red “non-signalers” searching for an ephemeral spot of food.
This study finds that if resources are sufficiently ephemeral, costly signaling is advantageous. This emerges as a spatial effect: signalers are more likely to find themselves near other signalers, and are thus more likely to mutually benefit each other. Non-signaling freeloaders are not eliminated: an equilibrium ratio of altruists-to-cheaters emerges, with the exact value of that ratio depending on how large and how ephemeral the food patches are.
There are some interesting assumptions in this model that make me worry a bit about its applicability to real animal systems. Resource depletion is not explicitly modeled, so the number of other individuals attracted to a patch of food by a signal has no bearing on how costly that signal actually turns out to be: only the fixed cost specified in the model determines the cost, and rather than depletion the disappearance of the resource limits its exploitation. The only real system that I could imagine this model depicting well would be a fruit-bearing tree that provided massive amounts of fruit for a very short period of time. The movement of individuals in this model is also very odd: individuals are driven by the waves of the same function that disperses the food, which means that they cannot stop moving. This leads to the strange behavior of ‘blowing past the food source’, a maladaptive behavior that is hard to contextualize in a real system. The fluid motions associated with movement in this model seem best approximated by aquatic environments, so perhaps I need to replace my ‘ephemeral fruit supplies’ with ‘ephemeral phytoplankton blooms’.
It is also interesting how this article treats group selection. In the Introduction we learn of two hypotheses that might explain costly signalling in social groups, but both are shelved because they rely on “an implicit group selection argument”. But here is the major finding of this model:
“Effectively a signaller increases the local density of conspecifics in its immediate vicinity, meaning it is more likely to subsequently benefit from the behaviour of other signallers in the population.”
As is true in many other models demonstrating that cooperation can viably evolve, it is the spatial nature of this model that makes the evolution of cooperation possible. While clearly this is not classic group selection, it is also not classic kin selection (or simply selection on individuals) that allows the altruistic behavior to persist. It is the emergent spatial dynamics of the model that allow signalers to recoup costs associated with signaling: call those spatial aggregations “groups” or not, depending on your scientific and social politics.
The findings of this research are important, but unfortunately this is not an article that I would ever assign to my students: the explanation of the model is far too confusing and opaque. This is a problem that dogs numerical and individual-based modeling, and as a community we need to be more vigilant in demanding clear explanations of how our models work. I am a bit surprised that the review process at PLoS Computational Biology did not produce a clearer explanation of what is going on in this model. Nonetheless, it is great to see this journal publishing more articles related to ecology and evolution, as it is dominated by molecular biological work.