Manny Rayner's book reviews

In The Trouble With Physics, a book that really deserves to be better known, Lee Smolin draws our attention to a remarkable and worrying dog that didn't bark in the night. There was a revolution in our understanding of the physical world, every thirty years or so, from the time of Newton up until about 1970. We were due for our next one ten years ago; alas, string theory hasn't lived up to its early promise. It's a little early to start panicking, but something appears to have gone wrong. And Smolin thinks it's not just physics.

Well, I wonder if this book isn't another example. Semantics used to be a pretty exciting subject. The question of how to describe the concept of "meaning" was making steady progress. I'm not sure I could pick out milestones as confidently as Smolin does, but I think of Boole and Boolean algebra (mid 19th century), Frege and predicate calculus (late 19th century), de Saussure and the "sign" (early 20th century), Gödel and self-reference (1930s). Then Richard Montague, in the mid-60s, finally put together a coherent theory explaining in precise terms how sentences in English could have meaning. There was no hand-waving, just logic: a breakthough! For my money, Montague was as great a man as Chomsky. Unfortunately, he was gay and very promiscuous; he had a wild lifestyle that ended up with him being murdered in his own home, by person or persons unknown, when he was only 41. Please don't interpret this as criticism of the gay scene. I'm just sad about what it did to semantics.

Situations and Attitudes was supposed to be the next revolution in the subject. Barwise and Perry claimed they had fixed up all the important technical problems in Montague. It came out in 1983, just as I was myself getting into the field, and I read it with huge interest. At the time, I felt I only half understood it, but in retrospect "half" was an overestimate. Now, I would say it was more like a quarter: some of that was my fault, but quite a lot of it was the authors'.

I wasn't the only one getting confused. The Japanese government launched a major research program, the "Fifth Generation Computer Systems project", that was supposed to create a series of "planned breakthroughs" in various areas. One of them was understanding of natural language using Barwise and Perry's Situation Semantics. Many foreign researchers, including me, were invited to Japan on FGCS money - I spent a couple of months hanging out with them in 1989. But when I turned up, it was clear something was wrong. Situation Semantics hadn't delivered, and the revolution wasn't happening.

Though several other groups did much better than the Japanese, they also failed to create another decisive breakthough in semantics. We're still waiting for it. I remember visiting the home of one of my semantician friends in 2007. He had several shelves of books on the subject, and I remarked on them. He made a face, and expressed regret that they no longer seemed to have much relevance to his actual work - he does natural language processing for a major US software company, and he's pretty successful at it, but he was right. We've somehow lost the dream. What happened?