Alan Grafen spoke to the Marschak Colloquium about his project of showing rigorously that adaptive optimization is a necessary consequence of natural selection. For intellectual-history buffs, the talk featured a fascinating account of how Fisher’s Fundmental Theorem has been misunderstood by population geneticists.
But a puzzle arose about a side-issue, and I’m soliciting proposed solutions.
Grafen referred to what is apparently a well-known case of an evolutionary signaling equilibrium in a predator-prey situation.
There is a species of starling that is preyed on mostly by merlins. When a merlin chases a healthy, fast starling, the starling sings as he flies for his life.
The behavior seems nonsensical, but makes sense in signaling terms. Only a healthy starling has the extra lungpower to sing while flying fast. Thus merlins learn that chasing singing starlings is mostly a waste of time, and they will stop chasing when the target sings.
Apparently the singing is hard-wired in the starlings, but the merlins have to learn not to chase singers; young merlins give chase, but older merlins back off.
The result is a Nash equilibrium: the behavior of the merlins makes sense, as long as singing is a reliable signal of speed and endurance, and the behavior of the starlings makes sense given the behavior of the merlins. (Being chased uses up energy, so there’s a benefit to the starling in having even a chase sure to eventuate in escape end sooner rather than later; thus the starling wants to emit the signal, given that he can, and the merlin wants to receive it.)
So it’s easy to see how starlings would be selected for the impulse to sing while flying. And it’s not too hard to understand why the system hasn’t been destroyed by some starlings’ evolving the capacity to sing even if they’re not in fact capable of escaping the merlin; to sing while flying fast requires excess energy capacity, and may be impossible to fake.
Why merlins have to learn, rather than having evolved a reflex to back off from singing targets, is something of a puzzle, but not a really daunting one; maybe in this case the learned behavior is harder to fool than an evolved response would be.
But there’s a much harder puzzle: How could the signaling adaptation, on the starlings’ side, have arisen in the first place?
In a world where enough fast starlings sing so that the merlins learn not to chase singers, the impulse to sing while dodging a merlin is advantageous. So the behavior is likely to be selected for, once it’s somewhat prevalent.
But how could the game possibly have gotten started? The first starling to emit that signal could have derived no benefit from it, since the merlin chasing him wouldn’t have learned not to bother with singers. If the chase is going to happen anyway, singing is a potentially fatal waste of breath. So if singing while escaping arose as a mutlation, it should have been selected against, not for.
As I say, apparently this is a well-known phenomenon, so the question must have been asked and answered before. However, there were no ornithologists in the audience, so no one knew the answer. Neither the lecturer nor anyone else had any ideas. How about you?
Update: Lots of responses, for which my thanks. Two ideas seemed like good candidates:
1. Perhaps singing while being chased is sexual display, or singing while flying fast is sexual display and it gets carried over.
2. Perhaps the singing is a warning to others, selected for as part of inclusive fitness if the warning differentially favors the singer’s relatives.