Monday, October 7, 2013

The Progressive Loss of Syntactical Structure in Bird Song along an Island Colonization Chain

Posted by Machteld

A long, long, time ago, in a land not that far away from here, I worked on birds. Bird song, to be exact; learned bird song, to be even more exact. Why do songbirds sing the way they do? Why is it so consistent in some species, and why do other species have dialects?

It may be more easy to understand how geographical variation occurs in learned birdsong, than to understand the absence of it. Mistakes can occur in the learning process (cultural mutations), the transmission of the sound is different in some places putting natural selection on the song frequencies, and perhaps other species interfere with the vocal communication. These are all good reasons why the song of a species might differ between places.  However, most songbird species – all of which learn their song, show a high consistency of species specific song.  How do they do that?

Although the particular notes in songs differ between songs, one individual may even have several songs, and between males within a species there will be variation; there is a certain overall structure, which makes a great tit song recognizable from a blue tit’s song or a chaffinch’s song. This is called the ‘syntax’ of bird song, where each note is a ‘syllable’ in the analogy with human speech.
One explanation for how birds maintain their species’ typical song, even if the particular notes (syllables) in the song differ over time and distance is that birds have a filter, or bias, for what kind of song they learn. If you think about it, this makes some sense, since young birds learning their song will hear lots of sounds and bird song from other species. A pre-set bias will give them some guidelines of which songs to learn.  

But clearly, there is variation between species in syntax, so how did this evolve? One would think that, after so many years of biologists studying bird song (and they’ve been at it for decades – ever since the Bell Telephone Laboratories made it possible to make sounds visible on paper in the 40’s) there would have been some good handle on this issue. Not so, unfortunately. One particular technical problem that has bothered progress is that syntax is hard to quantify. There have been attempts, but this always ended in semantic debates about definitions, and well, very little progress is made when that happens.
A chaffinch male, picture taken from
Well, there is where this paper comes in. This is a study on the song of chaffinches. A bird that has been the focus of bird song studies ever since the beginning (1954).  This is a very common bird throughout Europe (and also in other parts of the world, having been deliberately exported on a few occasions), and lives on the mainland of Europe, but also on almost all of the islands in the periphery of the continent in the Atlantic, such as Britain, the Azores, the Canary islands). The nice thing is that the colonization route of the chaffinch to those islands had been figured out already quite a while ago, so that we had a repeated evolutionary experiment at our hands, not unlike the finches at the Galapagos. They first colonized the Azores, then went on to the Canary Islands, with Gran Canaria the last island to get colonized by the finches. 

Many, many, many, many, recordings of chaffinches later, on all those wonderful locations (which I was lucky to be part of on some of those locations, such as the little gem of an island El Hierro), we created a database of chaffinch song in Europe.

That was the practical part. Then the hard statistics came in. I am not going to try to explain this here in detail, but in essence, per population, the songs were analyzed to see which were the ‘atoms’, or the parts that always occurred as a unit, such as a syllable (song note) or group of syllables. Then, zoom out, and repeat this process: which units (classes of syllables) always occur together. You can see how you can start to quantify things this way. The more you can cluster units within units within units, the higher the redundancy in the syntax of a song, i.e. the more stereotypical a song is.
Mainland Europe chaffinch song is highly stereotypical. Songs sung by birds in Holland and Spain follow very much the same syntax. This starts to differ once you get on to the islands, first the Azores, then the Canary Islands, where this structure starts to fall apart, until at the last island, Gran Canaria, there is almost no syntactical structure to be found. Clearly, with every colonization event, syntactical structure in the song disappeared a little.

Why did this happen? There are a few possible explanations, and read the paper for those, but one thing that we argue is that populations that go through bottle necks, such as at each colonization event, there is strong selection to recognize anything that might possibly be a conspecific. The learning biases that I mentioned in the beginning were thus selected to become less restrictive, i.e. less biased. A wider range of songs passed for species’ specific song. If this happens a enough times, you end up with no structure in your song. Which is what you find in Gran Canaria. 

And now this work has resulted in a wonderful publication, in Current Biology, available online, but in press still: 

No comments:

Post a Comment