This is definitely a fascinating topic.
I remember reading David Brin's work (the Uplift series) where he considers this subject in passing.
His take on it was quite interesting, if I remember correctly.
It goes something like this:
If you have an Oral tradition, language has to be highly structured, as it needs to provide in-built robustness against corruption.
Conversely, if you maintain written records, then the language can be looser because your recording media provides protection against corruption.
Now, what I find interesting about this is, suppose you look at Sanskrit. Sanskrit is highly structured with a very rigid grammatical structure. Its also a very old language, and was originally used by speakers with a strong Oral tradition.
Compare this to modern English, which is extremely loose in terms of structure (and arguably becoming more so, e.g. with new usages adopted for the internet and text messaging).
Now, where this ties in to the original qluestion: take this back to first principles. Oral traditions predate written ones, so earlier languages could be expected to be very formal. Rate of change of languages should also speed up over time, and be particularly slow in ancient (pre-literate) times.
From a common sense viewpoint, and bearing in mind that even apes can be taught the rudiments of (non-vocalised) language, homo sapiens' capacity for language surely predates the actual emergence of instances of vocal language.
Now, if you believe that all homo sapiens were concentrated at one place at a given time (and I've seen suggestions that there was a time that the global population of homo sapiens was less than 50,000), it suddenly becomes possible that there was one, common, proto-language.
So, how do we work out what this was? Well, I think we have to go back way before the last ice age, to whenever the population bottleneck was. If this occurred after the capacity for language was developed (highly likely imo, as language processing does not appear to be restricted to homo sapiens), then its likely that at a certain time, most/all homo sapiens were using a common language. If they had the capacity for vocalisation (and I'm guessing that all Homo Sapiens can vocalise), then this language would have been spoken. It would have been highly structured, and fairly immutable.
Over time, as populations diverged, the language would have changed, but at a much slower rate than we'd expect today. The upshot is that its quite possible there were far less languages in use in ancient (pre-Ice Age) times than we think. The downside is that its unlikely we'll ever know what they were, as they will have been based on Oral tradition.
Perhaps in this context, the tower of Babel is a pointer to a very distant past.