What I did in grad school

Last week I picked up Innumeracy by John Allen Paulos. It’s an old one, but a good one. In it, he talks a little about how mathematicians are partially to blame for innumeracy.

“It is almost always possible to present an intellectually honest and engaging account of any field, using a minimal of technical apparatus. This is seldom done, however, since most priesthoods (mathematicians included) are inclined to hide behind a wall of mystery and to commune only with their fellow priests.”

I don’t think jargon is necessarily a way to hide away from the outside world. It’s got a lot of uses, but not recognizing that it exists is a barrier to communication with non-initiates. Personally, though, I just find it very hard to explain things I know very well to people and part of the problem is thinking you need to start at the very beginning.

Now that I’m out of academics, I spend a great deal of time with normal people. When someone asks me what I wrote my dissertation on, I usually just mumble something about long consonants in the world’s languages and look embarrassed. I feel like without explaining the general assumptions of generative grammar and then how phonology fits in there and then how Optimality Theory works not to mention the representation of segmental length (moraic theory), my dissertation makes no sense. A lot of the advice I’ve seen on research say that the core idea of your research should be explainable to non-technical audience. I’ve never been able to get a hang of this skill.

Maybe I’m just slow, because lately I think I know what to say. Here’s my attempt.

There is good evidence that the brain uses very economical representations to store information about sounds in language. For example, in languages that have them, long consonants are basically twice as long as short ones—when someone says a word with a long consonant they hold the consonant twice as long as they would a short one. So, there are two possible ways the brain could represent them. They could be two short consonants put together or they one consonant that is marked “long.”

Suppose you use the symbol t to represent a short t sound. Then a long t sound could be made by two t’s in a row (tt) or one t that is somehow marked as being “long” (like t:).

All of the data from languages point to t: being the right representation. That is, human brains say, “don’t use two symbols to represent a sound when you can use one.”

How do we state this generalization in our model of how linguistics knowledge is stored in our brains? Some models just flat-out say, “don’t use two symbols when you can use one.” My dissertation showed that every component of the model has to be constructed so that none of them prefer two symbols to one. Which ends up being quite tricky. And we have to wonder if this is just a property of consonants.

I’d appreciate it if you could let me know if this is understandable or not. And if you want to read the gory details you can go here.