Cladistic Methodology and West Germanic

For about a century and a half now one of the usual tasks of historical linguistics has been reconstructing the prehistory of linguistic communities. (That is, of course, the side of reconstruction that non-linguists are interested in.) A perennial question is what sort of model of linguistic diversification best fits the data. This paper will explore that question in the light of data supporting a West Germanic subfamily of languages. The earliest model of language diversification, with obvious points of contact in evolutionary biology and manuscript stemmatics, was the Stammbaum or " family tree " model, and it is virtually the only model used by non-linguists (such as computer scientists or population geneticists) who try to make use of linguistic data. Linguists, by contrast , spent much of the 20th century downplaying its usefulness in favor of other models (such as the " wave " model) suggested by dialect geography or sociolinguistics. However , it's noticeable that the most rigorous theoretician of historical linguistics, Henry Hoenigswald, continued to work with the tree model throughout his career. That alone would be reason enough to keep it in our toolkit. Other reasons will presently become clear. There seems to be a general belief that a model of linguistic diversification should try to capture what actually happens as speech communities expand, diversify, disintegrate , and split up. It's certainly true that one can't ignore that without violating the Uni-formitarian Principle (A.1 on the handout). In a paper published about a decade ago Malcolm Ross worked out a model of diversification based on the tree model but incorporating additional features to handle network-like diversification of dialects and even reintegration into a single speech community of historically related communities that have not diverged too much. This is necessary because (as he points out) the tree, wave, and convergence models are not really alternatives: " one cannot take the same set of language data and, by the application of different models, reconstruct alternative sets of linguistic events " (p. 211). Rather, different models are appropriate for different datasets generated by different historical processes. Ross works on Austronesian languages in western Oceania, where a very large number of languages have diversified at various