I got an email asking why I decided to use Terry Crowley's Introduction to Historical Linguistics instead of Larry Trask's Historical Linguistics, so I thought I'd answer it here. When I had the choice, I went for Crowley's book. I have a strong interest in Austronesian languages, and at the time of first studying the subject I was very interested in New Guinea and Australia as well, so Crowley's was the natural choice. Trask, an expert on Basque, drew primarily on Indo-European and Basque examples, which are certainly interesting but not exactly what I was looking for. Having read Trask's as well - an excellent book, and actually somewhat more readable than Crowley's - I still think Crowley's is the better option for the beginner. There are a couple examples of editorial choices that put Crowley's above Trask's, if only slightly. For instance, if I remember correctly, Trask puts such things as final devoicing in his section on fortition, instead of in the section on assimilation/dissimilation. Final devoicing certainly is an example of fortition, but it's best thought of as an assimilatory change (assimilating the voicelessness of the silence following the word). At least, assimilation is the best explanation for why it occurs, rather than the more mysterious fortition.
So that's the answer. Crowley focuses on languages from the Pacific, which come from several language families and are (in many cases) morphologically simpler, making linguistic rules easier to understand. Trask, meanwhile, went a more classical route, which was less optimal for my needs. In addition, Crowley's lay-out, while less readable/humourous, is marginally easier to understand for the beginner. They're both excellent books, though, as is Lyle Campbell's, so take your pick based on your own criteria. If your interest is in Eneolithic Eurasia, or learning Grassmann's Law, Grimm's Law, etc, in quite a classical environment, then Trask's is the book for you.