Many years ago Bill Labov wrote an article called How I got into linguistics, and what I got out of it. It describes his path from college, through failed writing jobs and a stint as an inkmaker, up to his eventual discovery of the field of linguistics, Bill writes: "From what I learned about the small, new field of linguistics, it seemed to be an exciting one, consisting mostly of young people with strong opinions who spent most of their time arguing with each other."

Hey, sounds pretty fun!

He goes on to describe his surprise that most of the data linguists of the time seemed to be drawing from was more or less made up out of their heads, with linguists relying on their own intuitions about the grammaticality of the constructions they wanted to study. He explains that he thought he could do better, and that he wanted to study language as it is actually used by real people, in order to get not only at accurate description but also to use that description as the basis for explanatory theories about why Language works the way it does. The article is pretty modest in light of the fact that anyone who's ever taken a course or two in linguistics knows that Bill went on to do this and then some; he really created modern sociolinguistics and arguably empirical linguistics even more broadly construed.

Well, I wasn't the one who found Bill's article. My mother was. It was included in a recruiting booklet that Penn sent around to prospective undergrads. At the time I was planning on something like a double major in math and English or French, not wanting to choose between math and language. My mom read his article and thought, hey, this sounds like something Kieran might like. She passed it along to me, and I found it hugely inspirational. I spent the rest of the my senior year of high school reading everything about linguistics that I was able to get my hands on, with the result that I not only went to Penn and majored in linguistics, but ten years later I found myself graduating from the same institution with a PhD.

Even now I feel very grateful for my background in linguistics, and most especially for the kind of background in linguistics that the Penn department seems pretty uniquely able to provide. Equal measures formal/theoretical and empirical, Penn has world-class programs in both sociolinguistics and computational linguistics, with the result that even students focusing in other areas can't help but graduate with a pretty solid grounding in these approaches to the field. After college I seriously considered changing disciplines and pursuing the philosophy of science. But it seemed to me that one couldn't really do meaningful work in the philosophy of science without having first done meaningful work in some science, and linguistics felt like a perfect candidate: I was interested in it, for one thing, and as a science it seemed quite immature, still working out its methodologies. Sure, there was Penn, but there was also MIT, where most scholars relied primarily on the intuitions of themselves or their colleagues to gather the set of data on which theories would be based. It is a common joke among empirical linguists that you can look at the grammaticality marking of sentences in certain authors' papers over many years; invariably sentences that start out ungrammatical lose their asterisk over decades of consideration.

As history now shows, I ended up disappointed. Maybe it's because I had some background in math.

I found myself increasingly frustrated with work that conflated models of linguistic behavior with explanations of linguistic behavior. I found that most papers by very smart people in theoretical syntax explained word order by stipulating invisible features that words needed to satisfy by moving around. As a model, this is not crazy. It allows for clear taxonification of the set of possible factors that drive word order and other typological facts across languages. But most of the papers I heard or read included -- consciously or otherwise -- some statement of why, and why always boiled down to explanations than were really more like models. It further tickled my internal snark to note that the more Greek letters that a model included, the more likely it was that the author of the model would insist on its explanatory nature. So much for the philosophy of science.

So okay, halfway through graduate school I decided that pure theoretical syntax was not for me, although I still had enough common sense that I wasn't prepared to ditch it altogether. After all, theoretical syntacticians get jobs. I started focusing more intently on applying formal work to areas of discourse or pragmatics, working closely with Ellen Prince, who had made a very successful career in the same vein. I look to work like Ellen's or Bill's best, or some of the stuff Mark Liberman has done, and it's really amazing. It looks at real language used by real people, it stays away from hopelessly abstract formulations, and it starts off with problems that are scoped narrowly enough to be studyable but broadly enough to be interesting. But the more I looked into the literature, the more I talked to people at conferences, the more I discovered that work of that quality was really in the minority, even looking just at the set of work that purports to be empirical (which is itself the minority within the field of linguistics).

And so I finished my dissertation and I left. One reason I left, and one reason I ended up at Microsoft, is that I wanted to work in a context where I could not only look at real problems -- because let's face it, I could have done that in academia to the extent that I could have selected my areas of focus -- but where it would be expected that I do so. A really great dissertation is read by maybe 100, 200 if we're being generous, people. Maybe it is of use to 10 or 15% of them. And that's a really, really great one. So I wanted a context where my background in linguistics would be relevant, where I could still think about problems in that space, but where Greek letters wouldn't count as evidence of scientific theory. It turns out that in order to achieve all those things I had to come to a place that isn't really about science at all, but is instead about technology.

Which brings me up to the minute. I have been spending the last few days attending the annnual Linguistic Society of America meeting. It is my first fully academic conference since I started at Microsoft, and I expected all kinds of ambivalent and confusing feelings. What I have ended up with isn't confusing at all. I've gone to a bunch of talks and walked away with the same frustration with which I left theoretical linguistics two and a half years ago. Many people here are very, very smart. But the papers on the whole are unfalsifiable (back to the ol' model masquerading as explanation) or designed without reasonable experimental controls (in the case of work that is ostensibly empirical). There are exceptions. But there aren't many. It is not different in spirit than what I was reacting to a few years ago, but my recent stint as an engineer has me feeling it even more acutely. I met someone here from Google whose background is very similar to mine, down to academic subfield and current role at Google. He seems to share my bewilderment. But beyond that, it really is just me (or us?). The people with whom I previously identified professionally, people whose intelligence and diligence I really respect, do not share my reactions in the least. Rather, they react as a block, with the good judgment expected of high-calibre linguists. It is I who do not fit in.

The other strange thing about being here is that this is the main arena for first-round job interviews for students looking for assistant professor positions. I interviewed here a few years ago. I have spent a lot of time here talking to people about Microsoft, especially to students who seem to share some of the same perspective that I felt when I decided to go into industry rather than academia. Some of my friends from school have recently landed in tenure-track roles at good institutions, but many more of them are here interviewing. It is odd being on the other side. I don't envy them the anxiety, the self-esteem roller coaster, that saturates the entire process.

It has been very nice to see and catch up with people. I've made some good contacts here who may eventually become good hires for Microsoft. But the best part of this experience is the reaffirmation of the choice I made a few years ago to get out of academic linguistics -- and the clarity with which I can see the ways in which my approaches to the problem space have changed.