He states a thesis early on in the first chapter I find a bit strange:
[...] the search for the causes of language prevalence is not usually so easily resolved. In the historic record of contests between languages—when we have eyewitness testimony to keep us honest about what really went on—we often cannot point to cultural differences that were clearly curcial. Then we may have to look deeper: not just into the perceived associations of the different communities, how they looked to each other, the language communities' subjective reputations as well as their objective advantages, but even—and this is deeply unconventional, especially among linguists—to the properties of the langauges themselves.(p. 23)
Bizarrely, linguists almost universally assume that the basic properties of languages which they study—the kinds of sounds a language uses, its basic word order, whether it works by stringing together short and indpendent words or by coordinating systems of prefixes or suffixes—are irrelevant to languages' propspects of survival. After all, they reckon, every language is by definition learnable by children: that's what makes it a human language. If a community has problems propagating its language, there must be a social cause, not a linguistic one.
But for us, viewing the language as distinctive of the community that speaks it, we can only wonder what all that linguistic structure is there for. Perhaps a language's type even has survival value, determining whether a new population that has long spoken another language can readily take it up or not. This is one of the innovations of this book: to suggest ways in which it might actually matter what type of a language a community speaks. [...]
One thing that sounds strange to me is the claim of originality. Way back in the late 19th century it seemed indeed more fashionable to use unequivocally normative words about languages, to claim that Latin was intrinsically "advanced", that the isolating Chinese languages were "primitive", and more or less effective at expressing abstract concepts, which you'd easily grant might have something to do with the survival of the culture or language.
The other thing that I don't get is why it's so bizarre that linguists might believe language differences to be neutrally adaptive. I mean, it's my understanding that they claim empirical support for this, but even without that support in hand, I can imagine that the incredible diversity of languages isn't particularly for anything, that all languages are about as good as one another. (This proposition indeed being the currently fashionable one) For imagine rerendering the second paragraph above as:
Bizarrely, anthropologists almost universally assume that the basic features of people which they study—their skin color, the shape of their nose, their average height—are irrelevant to peoples' propspects of survival. After all, they reckon, every race is by definition capable of reproduction: that's what makes it a race. If a community has problems propagating itself, there must be a social cause, not a biological one.
I don't think anthropologists completely deny that having birth defects that render you sterile will put a damper on your dynastic ambitions, or that linguists think a language with a completely idiosyncratic 5,290,013-case noun declension system will survive long. However, I think they might well believe that the effects of the amount racial and linguistic variation as found in the world right now is mostly trumped by other factors, a signal lost in a lot of noise.