Raising the bar

Courtesy of Arts & Letters Daily, an article in the L.A. Times about how good a year Albert Einstein had in 1905. The article caught SC's attention for two reasons:

1) There's a bit of terribly fallacious statistical reasoning in there:

It also helped that he was struggling with these problems at a very propitious time. Stephen Jay Gould has pointed out that the reason it's so hard to hit .400 in major league baseball today is that the whole level of play has been raised. In the 1920s and 1930s there were many weak teams, against which it was easy for top hitters to pump up their averages. Today though, there are fewer consistently weak teams. Batters have a higher standard against which to try to stand out.

This isn't entirely wrong, but it is misleading. It's true that the level of play has improved considerably from the '20s to now. But a statement like this suggests that ERAs and batting averages have just been trending straight down, and that pitching and defense have unilaterally gotten better while hitters have not. This is just flat-out wrong. In 1876, the National League batting average for all players was .265, and the league ERA was 2.31; in 2003, those same figures were .262 and 4.28. Of course, it's not correct to compare those numbers directly -- in between, there have been changes to the construction of the ball, considerable expansion in the number of teams, and plenty of other factors that are even harder to measure the impact of directly, like the rise of short-relief specialists. But it isn't really accurate to say that nobody's hitting .400 simply because the level of play has gotten better.

2) Reading statements like this inevitably brings up comparisons to fields that readers know more about. Since you're reading about this at a linguistics blog, the obvious comparison is to linguistics. Could someone in the field have a year like Einstein's 1905? Has anyone had a year like that? And has the field progressed to the point where no enormous breakthroughs could be had in a single year anymore?

Instinctively, SC's answers to these questions are: no, no, and no. In the realm of past accomplishments, Chomsky's publication of Syntactic Structures stands out as an earthshattering moment for the field, in that it riveted many people's attention on the idea of transformational grammar. In both cases, some of the ideas involved had already been floating around: the Lorentz transformation for physicists, and Zellig Harris' dependency grammars for linguists. But Einstein's achievements spanned multiple branches of physics, where Chomsky only addressed syntax. Perhaps combining Chomsky's 1957 with the later publication of The Sound Pattern of English, we get a similar result, but there is still an enormous difference in the remaining impact of the work. Einstein's papers remain largely true for everyone in physics; the esoterics of string theory and dark matter are needed to explain deviations at levels of detail that linguistics might be centuries from addressing. Chomsky's late '50s/early '60s output has met with a wide variety of responses, some of which merely revise the same basic architecture, and some of which refute it altogether. A more recent example might be the string of successes enjoyed by John McCarthy and Alan Prince in the latter half of the '90s in phonology, but even Optimality Theory doesn't strike SC as really having conquered the field the way that relativity or Einstein's accounts of the photoelectric effect or Brownian motion did. So nobody's done it yet.

For a variety of reasons, SC doubts that anyone will have an "Einstein 1905" in linguistics, either. Under the best of circumstances, gathering adequate data and proposing analyses can be done in a few months, especially if you don't have to go do fieldwork to get said data. But the work doesn't lend itself to having leaps of intuition while standing at a blackboard. Even if it did, the publication process -- at least through journals -- is tedious. Perhaps the Internet provides adequate relief for that, but we're still left with the last problem: are there even enough questions for someone to have a huge year like this?

Fortunately, yes, so your host could happily be proved wrong on whether or not anyone could yet have a great year like this. One of the downsides of modern syntactic theory is that in all frameworks, it's too easy to keep proposing new features to handle odd cases. SC is fairly sure he read this lament on another linguistics blog recently, but can't recall who to credit for it. Nevertheless, a massive overhaul of any of the existing theories -- or a new one that does away with the lot of them -- is very much still in order. Optimality Theory has similarly embarrassing gaps in dealing with "opaque" derivations, and an account that eliminates the need for Sympathy would probably be as close to a True Theory of Phonology as we could ever hope for. Despite the similar nature of the problems, it's not clear to your host that the solutions would actually have all that much in common, and so while there are all sorts of major breakthroughs to be had, the odds of anyone getting them all at once strike SC as pretty low.

Comments

[Daniel]

In fairness to Mr. Gould, it sounds like the article is misconstruing what he actually wrote. The article is included in his compilation - Triumph and Tragedy in Mudville - and what he more points out is that in the past you had more outliers (a wider bell curve) but there is probably some human limit on how well a ballplayer can perform and since pitching has also improved and the rules have changed (raising and lowering the mound, changing strike zones, etc.) in ways that keep the average batting average around the same that today there are less outliers in hitters (a smaller bell curve) and thus, less likely a chance for a player to hit .400