Continuations. It was actually completely wrong. I "proved" that you didn't want continuations in a programming language, because there were only a few possible uses of them, and they could all be more cheaply implemented using special-purpose operators. The good bits got re-used in On Lisp. The rest was crap.

What I should have written about was macro hacking and embedded languages and so on. That was what I really studied in grad school. But it didn't seem theoretical enough to be a thesis topic.