Thursday, August 2, 2012

The selfish herd

So, one of the most interesting questions in evolutionary biology is the origin of collective behaviors. This can be the complex division of labor that we see in social insects and human societies, flocking behavior in migratory birds, or microbial formation of biofilms. It can be predators engaging in collective hunting, or prey engaging in collective being hunted. It's this last one that we're going to be talking about today.

As with many questions in evolutionary biology, there are a couple of dimensions that people are interested in untangling: proximal and ultimate causation. Proximal explanations focus on the "how" part of the solution, as in, "what are the molecular, genetic, etc. mechanisms and environmental cues that result in this behavior?" Ultimate explanations focus on "why," in the evolutionary sense of "what were the selective pressures that led to the evolution of this behavior?"

Herding or flocking behavior is a classic case. For example, why do sheep hang out in a big group, in contrast to say, leopards, which tend to be pretty solitary? There are a number of possible (and not mutually exclusive) ultimate explanations, but the most talked about one is probably defense against predators.

Best URL for sharing:
Permanent image URL for hotlinking or embedding:
Back in the mid-twentieth century, it was common for biologists to talk in fairly loose terms about collective behaviors having evolved as a result of their benefits to the group. Then, in 1966, G. C. Williams published Adaptation and Natural Selection, which dropped a lot of truth into the community. In particular, it emphasized the gene-centered view of natural selection that hit the public consciousness with Richard Dawkins's 1976 book, The Selfish Gene, and which has remained the dominant paradigm in evolutionary biology ever since.

Williams demonstrated that group selection, while possible, will generally be a much weaker force than selection acting on the individual. Therefore, it is good practice to look for evolutionary explanations at this lower level. Given plausible adaptive stories at the individual and group levels, one should favor the individual-level story. While the two stories might not be mutually exclusive, individual-level selective pressures are more likely to have played an important role in  the evolution of any particular trait than group-level selective pressures (all else being equal, of course).

In 1971, W. D. Hamilton published a theoretical analysis that brought this individual-level perspective to herding behavior. Hamilton argued that all you need is for animals to be trying to evade predators as individuals. If there are other individuals of their type around, they just need to try to position themselves between other individuals. Here's how Hamilton draws it:

This frog wants to position itself between the two frogs on the right. That way, when the sea snake comes up, it will eat one of the frogs at the edge, and the one in the middle will be safe.
All you need is for everyone to follow one simple rule: when a predator comes, position yourself between two other individuals. What you get then is a tight cluster of individuals.

You can actually try this at home. You probably need about eight or ten people. So, most of you might not be able to try this at home, but you could maybe try it at school or work. Have each person pick two other people in the group (but don't tell who your picks are). Then, everyone tries to get between the two people they picked. What you'll get is something a lot like a cluster of frogs climbing all over each other to get away from a sea snake.
Frogs maneuvering to get between other frogs results in the formation of a clusterf**k of frogs. I know, right? I was surprised, too, but my herpetologist friends assure me that "clusterf**k" is the official collective noun for a group of frogs. Don't even ask about sea snakes. You don't want to know.
Bonus activity: after you've disentangled yourselves from the frogpile, try this one. Each person picks two people again, labeling them "A" and "B" (in your head). Again, no one needs to say whom they picked. Now, each person should position themselves so that their "A" person is between them and their "B" person. If it helps, imagine that "A" is Mitt Romney, that "B" is the American People, and you are Mitt's tax returns. Your job is to position yourself so that Mitt keeps the American People from seeing you. I won't spoil how it comes out.

So, Hamilton's model provides a nice, simple model that can produce the observed behavior. The model is attractive because (1) it requires selection only at the level of the individual, and (2) it requires each individual only to follow a very simple behavioral rule. The collective behavior is an emergent property requiring no coordination at the group level.

Now, there's a new paper out that is attempting to look at this empirically, in sheep. The study involves strapping adorable GPS backpacks on a bunch of sheep (Figure 1c, below) and then letting a sheepdog chase them around.

You can look at the movies here. It's only a brief communication, and does not really nail anything down, but the authors interpret their results as broadly consistent with the selfish herd model. In particular, they are able to see that individual sheep seem to be trying to get to the center of the flock.

The cool thing is more the potential for this type of experiment. Yes, Hamilton's model is attractive and parsimonious, but if we want to understand the rules that actually govern the behavior of sheep when they are faced with a predator (or, in this case, an annoyator), we will need to get good quantitative data on individual behaviors in a variety of contexts.

Plus, look at that little GPS backpack!

I'll leave you with this.

Best URL for sharing:
Permanent image URL for hotlinking or embedding:

King AJ, Wilson AM, Wilshin SD, Lowe J, Haddadi H, Hailes S, & Morton AJ (2012). Selfish-herd behaviour of sheep under threat. Current biology : CB, 22 (14) PMID: 22835787

Hamilton, W. D. (1971). Geometry for the Selfish Herd Journal of Theoretical Biology, 31, 295-311 DOI: 10.1016/0022-5193(71)90189-5


  1. My recollection is that these experiments don't work so well in Karaoke halls in run-down resorts on the edge of Beijing....

  2. I have to say, that was in my mind when I went for the tax return metaphor.

  3. However, you have failed to explain why Australians are that horny.

  4. Ooh, I'll pass that question along to Guillaume!