What would happen if the birds were, for example,
being bred by hand by somebody else, and it didn't
matter who they mated with? Well, in the first phase,
after Stage 2, the songs look a lot like those of the
feral cousin. What happens when we mask selection
is that everything starts to fall apart. Song linearity
decreases; the complexity of the song increases; the
number of transitions increases; the filter that the
bird needs to pick out somebody else degrades. This
is a filter degrading. It doesn't actually recognize anybody.
This is what you would predict is going on.
One of the surprising features is that the load on
the learning mechanism actually increases in this process,
even though there's not a cost, because there's so
much variability. What we see in these songs is something
very similar to what we see in this domestic
species. We see that the song transition probabilities
have become much more complex.