Blurb:

Since early in the twentieth century there has been a theory, put forward mainly by French-speaking linguists including André Martinet, which asserts that the sound changes that occur from time to time in any language are governed by the quantity of work done by particular phonetic contrasts in keeping words apart. This sounds a priori very plausible, but English-speaking linguists have queried whether it is in fact true. To my mind it is decisively refuted by the history of Chinese phonology, which has been a millennia-long history of repeated mergers of significant contrasts. Yet to my surprise, in the present decade a group of linguists around Abby Kaplan have offered strong new evidence in favour of Martinet’s functional-yield theory, deploying statistical reasoning in a more sophisticated manner than is common in linguistics. However, the Chinese counterevidence remains unrebutted. Thus we are left with an apparently insoluble paradox, which neither I, nor a number of other experts I led to consider the problem, have succeeded in resolving.