I've read this book a few times, and it's always a good read. It was fun to pick up the updated version and see what's changed (all carefully annotated in an appendix.) And I'm waiting on his new book about writing from my library queue, so I thought I'd revisit this one
In fact, now that I already mentioned the appendix, this book has a ton of appendixes, the actual text finishes around 75%, after that is non-academic sources, the academic sources, updates on changes to the text, and then two sources lists for the updates. And an index, and a glossary, and then a suggested reading list.
I love that kind of thing, pushes all my "oh really now" buttons when I read something I find difficult to believe, when I can just go look up the sources and make my own evaluation.
tl;dr: Chomsky is right, but he's not god. Sapir & Whorf maybe never even met, but they're right too, to a point. And Pinker is a really good read. And the rest of my ranting is below the cut.
Also despite really liking Stephen Pinker, and as I mentioned elsewhere, finding him really accessible for a linguist even when he starts diagramming sentences, I've always thought since my first reading of this book that he was something of a hardcore Chomskyian (i.e.: Language is innate, Sapir-Whorf are both going to hell.) Turns out, he's really not, and that is clarified somewhat in this updated version, not least by his own mock horror at being labelled a Chomskyian.
I, meanwhile, have modified my own position somewhat. I'm well aware that the original Sapir-Whorf hypothesis has it's issues. Like most all of the examples it's based on being utterly wrong. For those of you with eyes glazing over, that is, briefly, the idea that languages affect thinking, the old "Eskimoes have fifty words for snow, and this means they think about snow quite differently than we do". Obviously, this falls down in a billion ways: Eskimoes aren't a thing, anyone you might call eskimo doesn't have fifty words for snow, and English certainly doesn't get by with just one, we just do it by attaching adjectives. And as Pinker points out, you could as well say "Printers have fifty words for fonts and parts of fonts, that the rest of us don't" and the entire point becomes so mundane as to be irrelevant. Eskimoes (if there were such a thing) think about snow differently to most people because it's important to them, and the language reflects that. And because language provides the ability to distinguish more features of snow, this is something they learn early and young.
I've come to the conclusion they're both useful ways of looking at things. Language is likely quite innate, and Pinker's book is a really good, accessible, readable and often very witty entry point into how we can tell. The parts about pidgins and creoles, and about sign languages are utterly fascinating.
But from a cognitive psychology point of view, language is a sociocultural artefact, and it most certainly does affect how one thinks, just not really in the way the hardcore Whorfians say. You can, of course, think about a concept your language doesn't have words for. So we tend to react somewhat strongly to those who throw the baby out with the eskimo bathwater.
Anyone who's travelled or worked with people from other cultures can tell you "later" and "soon" and "immediately" have very different meanings, often culturally based. And anyone who's ever been labelled with a word like "retard" or "fag" or "nigger" can tell you that erasing words from our vocabulary is a long term positive - not because the word, itself, an inert conglomeration of letters, does the damage. It's because the thought attached to just that conglomeration of letters is invariably a specific and hurtful one, and raising that into public consciousness, putting it under the spotlight, helps everyone see it for what it is, and that is what we want to see erased.