Erez Lieberman et al. at Harvard are looking at the rate of change in words to see if words evolve:
Lieberman was struck by this idea when he learned that the ten most common verbs in English (be, have, do, go, say, can, will, see, take, get) are all irregular. Instead of their past tenses ending in ‘-ed’, as do 97% of English verbs, they take the peculiar forms of was, had, did, went, said, could, would, saw, took and got.
Researchers suppose that this is because often-used irregulars are easy to remember and get right. Seldom-used irregulars, on the other hand, are more likely to be forgotten, so speakers often mistakenly apply the ‘-ed’ rule. The most commonly used word that they found this happened to was the verb ‘to help’ – the past tense used to be ‘holp’, but is now ‘helped’.
This could be seen as analogous to the way that crucially needed genes tend to stay the same throughout biological evolution, whereas those for less-often-used or specialist traits have more freedom to evolve.
Lieberman wondered whether he could quantify the effect. He and his team looked at 177 verbs with varying frequencies of use that were irregular in Old English, and examined how many had been regularized into the ‘-ed’ past tense by the eras of Middle and Modern English. They found that an irregular verb used 100 times less frequently is regularized 10 times as fast. For the more mathematically inclined, this can be expressed as: ‘The half-life of irregular verbs is proportional to the square root of their frequency.’
Read the whole thing.
The research is interesting, but I still think the analogy between cultural change — including linguisitic change — and natural evolution is facile. I have some remaining mechanistic concerns. What is the mechanism of natural selection in cultural evolution? Ideas may go out of use, but they are hardly ever completely eradicated. Can you demonstrate that increased usage constitutes some kind of reproductive fitness?
It may be an good parlor game to apply evolutionary principles to cultural trends, but until those principles can be rigorously and uniformly applied I remain unconvinced that is anything more.
Here is the actual paper.