I wonder if there exists any summary or paper analysing the time it takes for the creation of a new language (with all reservations concerning definitions of languages et dialects etc.)?
Take, for example, the Romance languages, which are derived from Vulgar Latin. It would be (I suppose) around 400 years (between 4th and 8th century), maybe a little less, and for Yiddish some 500 years (between 10th and 15th century).
The question you're asking is probably not answerable, but if we just focus on one area where a lot of work has been done, maybe we can get closer to answering it.
Glottochronology as originally conceived by Morris Swadesh is one approach that tries to answer this question in a specific area. It's very controversial, but a lot of research in this area is built on variants of Swadesh's idea, sometimes in an evolutionary biological or computational context.
The underlying assumption in glottochronology is that there is a core vocabulary for any given language that is relatively stable and changes at a constant rate analogous to radioactive decay or gene replacement. Numbers are obtained by considering data from languages whose chronological dates and degrees of separation are fairly well-known (e.g. Indo-European). By mapping the rate of change or retention, language divergence can be measured. The accuracy and usefulness of the results, on the other hand, is much debated.
Some of these earlier papers might be of interest:
Bergland, K., & Vogt, H. (1962). On the validity of glottochronology
Chrétien, C. D. (1962). The mathematical models of glottochronology
Dyen, I., James, A. T., & Cole, J. W. L. (1967). Language divergence and estimated word retention rate
More recent works:
Gray, R. D., & Atkinson, Q. D. (2003). Language-tree divergence times support the Anatolian theory of Indo-European origin
Heggarty, P., Maguire, W., & McMahon, A. (2010). Splits or waves? Trees or webs? How divergence measures and network analysis can unravel language histories
Renfrew, C., McMahon, A., & Trask, L. (Eds.) (2000). Time depth in historical linguistics
My sentiments are in line with some of the commenters who point out that on its face the question is asking for too much. On the other hand, there is some recent work that might give you an idea of how similar questions are being approached. There is a good summary article (Croft 2006) outlining Croft's "evolutionary" approach to language change. This paper introduces some useful concepts in thinking about the issue of language "speciation." For a preview of some issues in applying quantitative models to language evoluation, have a look atthe article in press by Levinson & Gray. Those two authors have received mixed feedback from linguists on some of their recent work in historical linguistics: they introduce interesting new mathematical techniques to language change, but their underlying assumptions are often thought to be more than is warranted.
One of the fastest documented language forks is the case of Afrikaans: Systematic Dutch colonisation of South Africa started around 1679, and already 100 years later Afrikaans has developed enough to be considered a distinct language from standard Dutch. Standard Dutch remained the medium of writing for another 100 years, and it took until 1925 for official recognition as a national language of South Africa (replacing Dutch).
I'll give your question a go. First, in some cases, dialects and languages are for all intents and purposes the same. That does not mean in all cases, but in rare cases, they can be as different as either one is from another dialect or language. What distinguishes a language from a dialect in such cases is the cradle that embraces that language or dialect, made up of or lacking a military, a country, a poplulation, a history, a heritage; official recognition.
Sometimes, it's even more complicated depending on the theories of linguistics embraced or appealed to if to argue either or. For example, for some people, and indeed many, Chinese is a language. Point blank: it is its own language. It's not a dialect, it's not a family of languages, it is simply a langauge. Period. And in Chinese, there are several dialects, two of which are of course very well-known. Other people will argue, however, that no, this is wrong: that Cantonese is its own language, just as Mandarin is. And, from there, things start to get messy. I personally am of the view that Chinese is a language, and the latter, varieties or dialects therein. At any rate, for them to become a language, or for some language like Portuguese or Spanish or so forth to become languages, the factors I think are more relative to the period in which the languages came to be and what happened and in some cases has happened within and outside those languages (sciences, philosophies, etc) along the way.
In short, you might compare it to countries. When does a country become a country? Well, if you ask someone from Tawain if they have their own country, you can imagine that they will tell you that they do. They'll even give you all kinds of justifications: their own passport, their own government, military, etc. Embassies even. But if you look up the UN's list of countries, or that of the Department of Defense, you'll not see it listed as a country. So is it a country? Well, what makes a country a country is such that Tawain is about as close as you can get to a country without being officially recognized as such for political reasons. Languages are very much like this. There are all kinds of external factors that come into play when thinking of such in terms of language versus mere variety or dialect. It is not enough that internally the population should vie for such a status. For example, The Ivory Coast recently declared that it wishes to be called Côte D'ivoire, retaining its French identity. Or, for example, many people outside of Iran erroneously call Persian Farsi without knowing that Farsi is actually a transliteration of Arabic because Arabic doesn't have the P sound in its language, meaning that Farsi actually comes from Parsi, both of which are endonyms (or languag/place names) within the Persian language itself, the root of which is now used within the language to distinguish between the two dialects of Persian between Iran and Afghanistan but that nonetheless gave us the word in English that we already have to begin with: Persian. All too often people hear a country or place or language as pronounced or said in the original language and think that they have to say it in such a way if to say it correctly, but this is wrong. Farsi and Côte d'Ivoire are no exception. At any rate, as you can see, internal and external influences play a critical role in the determining of the status of a language. The rest of the world has to agree and for a long enough time, even if ultimately the status is merely a putative one.
So in short, even ignoring all the unstated factors for what makes a language an actual language versus something else, outside factors would seem to complicate the definition to so great a degree that trying to model language on a time line in the way that you ask could only best be done hypothetically such that through estimates from a couple hundred years, or in the longest cases, thousands, where given factors typically present in official languages could be observed and potentially later fade away or die out immediately, one could argue through induction that any given language could occur. In the minimal amount of time possible, like empires or rapid migrations or expansions etc, one might conclude that although considered a language locally or in surrounding regions, or even more broadly yet still hypothetically, only temporarily so, prior to dying or being overtaken, etc. In short, I don't think that 200 years is impossible depending on the initial sociolinguistic conditions, but I think that would likely be the absolute best-case scenario, where the only hurdle the protolanguage or dialect might be thought to have to overcome be something like state formation.
