Are pigeons lazy bosses?

Revision as of 18:12, 22 June 2010

I was simultaneously excited and suspicious of this paper, which made the cover of Nature: http://www.nature.com/nature/journal/v464/n7290/abs/nature08891.html http://hal.elte.hu/pigeonflocks/index.html

Watching the videos I got the sense that the pigeons that were identified as "leaders" were also the most boring. They didn't seem to turn as much as the other pigeons. What if their "leadership" is just "laziness"? Take a spring model, in which every bird wants to be a little close and a little far from every other bird. Now get a lazy ("massive" in the analogy) bird, and the flock's behavior will be determined by his stubborness.

How to test this? Well if, from moment to moment, a given bird doesn't turn very much, I predict that he will appear as the leader of the group. To test this, I calculated the differences between the angles at each time step, and found the variance in these differences. straight flying bird will have low variance, wildly turning bird will have high variance. This variance should vary inversely with leadership position.

To avoid recalculating their hierarchies for each flight of the pigeons, I used the hierarchy the the authors published for the whole flock, over all recorded flights. It is on page 892 of the article. I did a paired test of ranked sums for each of the (in this case) 9 birds, comparing their laziness and their calculated position in the pecking order. My results were null: by this analysis they are unrelated.

But I like a few things about having done this:

• I got the data from the experimenters less than a day after emailing them for it.
• It took about three hours to do the analysis, and should take less and less as I become fluent in R.
• I was able to answer a question I had.
• My blog is a nice outlet for a quick report of the [null] result. Might be interesting to someone.

Null results ("ummm, nothing happened") have a history of being underreported in science. Naturally, people only want to share (and can usually only get published) when something Happened. There is actually a really neat history here, and there are some classic results that got published due to this bias.

As for lazy pigeon bosses, it still might be the case: I didn't compare one individual flights (because I'm avoiding the crazy calculations they did to infer hierarchies for each of the 15 flights constituting the data set). I cut some other corners too (ignored third dimension, velocity and other data). Also, I used a pretty low power statistical analysis, one that I was forced to use because of the coarse hierarchy data published in the paper. But I'm not going to keep trying, I'm pretty satisfied.

In summary, I was unable to support my pet hypothesis that leader pigeons are lazy pigeons. The authors propose a number of more exciting possibilities instead. For example: leadership role corresponded to accuracy in individual navigation back home, though they point out that the causation could go either way (leadership -> accuracy or accuracy -> leadership).