One thing a lot of traditional historical linguists seem to get bent out of shape about is this: biologists using phylogenetic, cladistic and other models to look at genetic or family tree relationships among languages. The basic issue is that many linguists find that the quant folks are using models without fully understanding the language-particular complexities of the data they throw at the formulas. This is the old "gigo" or "garbage in, garbage out" problem. Some members of Team Verb have talked about that here in the past and will again in the future, no doubt.
Last week, political polling got the same treatment, from astrophysicists, with a paper to appear in Mathematical and Computer Modeling, as reported here. See here for a reaction from the very fine and Wisconsin-based pollster.com. Linguists have complained, among other things, that papers are getting published in Science or Nature without any review by people who understand linguistics at all. That is, the data are taken are obvious or trivial and the cool math warrants publication. One commenter at pollster.com makes the equivalent comment about this.
Part of the problem is perhaps that linguists and political scientists aren't yet comfortable with what this kind of modeling is really about, but that's a topic for another day.