Practice makes symbolic?
Well, not yet "symbolic" in the sense of symbolizing anything, at least "symbolic" in the sense of congealing into sequences of elements drawn from a discrete alphabet of distinct "symbols".
Here's an animation,supplied by Ofer Tchernichovski from CUNY, showing what happens as a zebra finch chick practices its song over a period of about 60 days:
You can see the initially-diffuse distribution of behavior gradually congealing into a half-a-dozen well-defined clusters, corresponding to what bird-song specialists call "syllables".
Here's what the end product sounds like:
You can see what it looks like here (in the acoustic sense of "look", anyhow). The different instances of a given syllable class are by no means identical, any more than all the pronunciations of a given human syllable type are identical. On the other hand, it's clear that it makes sense to think of these behavioral sequences as performances of strings of entities drawn from a finite (and small) set.
As far as we know, zebra finches don't mean anything in particular by choosing one sequence of syllables or another -- beyond the usual "Here I am, deal with it".
But Alex the parrot clearly made the step to associating properties of actions and objects with arbitrary choices of vocal gestures, just as our evolutionary ancestors presumably did. It's plausible that he associated emotional states with vocal gestures in the same way.
Whether "meaning" (at least at the word level) is something beyond this is a question for the philosophers and the semanticists and the cognitive psychologists. While they're considering the matter, it'd be a step forward for those of us who work on language sound systems to understand how gestural patterns come to be equivalent to symbol strings, and how symbol strings come to be associated with salient -- but arbitrary -- referential categories.
The way is open for someone to characterize in detail -- physiologically, behaviorally, developmentally -- the "neural automata" that generate and recognize birdsong. (As far as I can tell, there are no serious candidate theories out there at the moment.) And since the engineering problems and the physiological raw materials are similar, it's plausible that this might give us some hints about speech as well. Then again, maybe the crucial insights will come from the speech side first.
Now Dan Koditschek will explain how this looks from the perspective of a roboticist.