Language Log

Unchanging Pronouns?

A while back John McWhorter was discussing here an article on distant relationships that emphasized the popular claim that personal pronouns are super-stable -- that they can confidently be expected to persist in languages over very long periods and are thus reliable indicators of genetic relationships among languages (that's `genetic' in the historical linguist's metaphorical usage). One reason historical linguists are skeptical of this claim is that it's so easy to find languages in which personal pronouns have undergone a lot of change. One source of instability in pronouns is borrowing: in an earlier post Bill Poser pointed out that in some parts of the world, for instance Southeast Asia, pronouns are borrowed freely, so that pronoun borrowing is by no means out of the ordinary (that is, it's not the marked case). Another source of change in pronoun systems is analogy of various kinds. Below are two fairly typical examples of pronoun systems that have changed enough to make detection of historical links difficult or impossible through simple inspection.

First, consider the Indo-European language family, almost everyone's favorite source of illustrations of how languages are related. Here are three sets of Indo-European pronouns, from three different branches of the family:

In the Latin set, other choices could have been made for the third-person forms; these comprise one of several sets of masculine forms. None of the other possible choices resembles any of the Russian or English forms either, however. In the Russian transliteration, the vowel y is a high back unrounded vowel (not found in English); it's not like English y.

The point about these three sets of forms -- which are the forms you'll find if you look up the English meanings in a bilingual English/Latin or English/Russian dictionary (namely, the method used by Greenberg and others who search wordlists for similarities) -- is that not one pronoun obviously matches across all three languages. The three words for `I' are in fact etymologically connected, but a casual inspection certainly wouldn't pick them out as similar enough to count as potential matches. Latin and Russian share a second-person singular pronoun, but English doesn't, because thou is now obsolete, replaced by you by analogy to the second-person plural. None of the words for `he' or `they' match, and it'd be a big stretch to claim a match (on casual inspection) for `we'; one might or might not hypothesize a match for Latin vos and Russian vy, but English you wouldn't fit, if one is using Greenberg's method. So if you were looking for a relationship among just these three languages, the pronouns would give you little or no basis for a `yes' answer.

Now consider the three sets of forms below, from three of the 26 or so Salishan languages (spoken by rapidly dwindling numbers of elders in Washington, British Columbia, Idaho, and Montana):

Salishan languages are more closely related than Indo-European languages are. The standard estimate for the time depth of the Indo-European parent language, Proto-Indo-European, is perhaps 6,000 years BP. (This estimate is unshaken by recent dramatic but not very convincing claims of different time depths.) Latin, 2,000 years older than Modern English and Modern Russian, is of course closer in time to Proto-Indo-European than the modern languages are. Proto-Salishan has been estimated at about 4,000 years; Montana Salish and Coeur d'Alene are considerably more closely related than either is to Squamish, because these two belong to the same branch of the family (Southern Interior Salish). But even with the shallower time depth for Salishan, no single pronoun obviously matches across all three languages, and three or four of the pronouns (1sg., 2sg., 1pl., and probably 2pl. ) clearly don't match in Montana Salish and Coeur d'Alene.

So what does this mean for the super-stable-pronouns hypothesis? What it means is that pronouns, like all other words, are changeable, via regular sound changes, analogic restructuring, and borrowing. Sound changes can and often do make the connections between historically related forms unrecognizable, as with the Indo-European words for `I' in Latin, Russian, and English. Analogic change introduces (often restructured) old pronouns into new places in the system, and borrowing brings entirely new pronouns into the system; in both cases the original pronoun is replaced by a different one. This is no surprise to historical linguists. It may be a surprise to those who still believe that it's possible to investigate language history, including language relationships, by taking shortcuts, without doing any actual historical linguistic analyses.

One more question: do pronominal systems change less than other parts of the vocabulary? I don't know the answer. But I don't think anyone else does either.

Posted by Sally Thomason at July 4, 2004 12:31 AM