March 08, 2004

Suppose generative syntax was born in Nigeria?


French: JEAN EMBRASSE SOUVENT MARIE. ("John kisses often Mary.")

Generative syntacticians have told us that the reason for the difference in the order between verb and adverb here is that in languages with what the layman knows as conjugational suffixes, the verb moves to the left of the adverb in order to be marked by "inflection " writ large. Presumably, this is part of Universal Grammar, encoded by neurons in the brains of humans on Baffin Island, in car washes in Cleveland, hoeing gardens in Sumatra, getting wasabi-highs in Osaka, and everywhere in between. But does this really hold up?

It's an interesting idea -- but over the years contradictions have poured in. Scandinavian languages like Danish and Swedish are almost as poor in conjugational suffixes as English, and yet in some dialects the verb moves. Then little Faroese, close sister to Icelandic and bristling enough with inflection to give Latin a run for its money, often leaves the verb right where it is. Portuguese creoles like Cape Verdean and the one in Guinea-Bissau barely know a suffix from their elbow -- and yet in them, verb movement is not unknown.

Adherents of the verb movement analysis have come up with some intriguing "fixes" here, but one wonders how valid a theory can be based largely on a few languages spoken in Western Europe.

In this, a thought experiment has always haunted me. Imagine that modern theoretical syntax was founded by southern Nigerians speaking Edoid languages like Edo, Urhobo, and Degema, and were familiar only with these and Mande languages (like Mandinka, Mende, Susu) spoken further up the west African coast.

In Edoid languages, various tenses are encoded solely by a tone change on the verb rather than prefixes or suffixes. In Edo, when the A in IMA has a low tone, the word means I SHOW, but when the A has a high tone, it means I SHOWED. But Mande languages are tone-shy as subsaharan African languages go. They do not use tone to express tense, nor even do they use it much in the "Chinese" way, that is, distinguishing words from one another.

Now, as it happens, Edoid languages have subject-verb-object order, while Mande languages put their verbs at the end of a sentence.

Let's imagine that our Edoid-speaking linguists hypothesized that the difference in word order was because the verb in Edoid languages had to move to the left of the object to be marked by tone, as verbs in French are assumed to move leftward to be marked by a suffix. For them, Mande verbs stay where they are because there is no tone to be marked by.

For us, this looks kind of wacky, because marking tense is a side-dish function of tone in the typical subsaharan African language. In most of them other than the Bantu ones southside, differences in tone make different words from the same syllable: in Yoruba, FO on a high tone is "float", while on a low tone it is "fly". This seems the "main course" of tone in these languages; tone that marks tense seems to be "other," just as we assume that mammals have hair mainly to keep them warm rather than to keep them dry in the rain.

But Edoid languages are special. Tone does not distinguish verbs like in Yoruba. In Edoid, tense marking is a prime function of tone (which is why linguistics textbooks often use Edo to show how tone can mark tense). This is what our pioneering southern Nigerian linguists would be working from, and hence their hypothesis.

Nevertheless, for linguists in the real world, a theory that we are innately specified for verbs moving to the left to be marked by tone looks quaint. Theories of language change show that tones arise by accident, such as when consonants at the beginning or end of a word wear away and leave a difference in pitch as a remnant like the Cheshire Cat disappears and leaves his smile, to quote the masterful analogy of Jim Matisoff. We think of tone as merely one of many developments that a natural language may drift into over time, and its marginality in most European languages only reinforces that assumption.

Yet theories of language change also show that prefixes and suffixes, too, are almost always the result of whole words glomming onto other ones over time by accident. It is generally assumed, for example, that the -ED past ending in English (and its equivalents in other Germanic languages) arose as a form of DID -- "I WALK-DID" -- that got stuck onto the end of verbs and devolved into a homely suffix. And after all, legions of languages in the world have no conjugational endings or gender prefixes or suffixes at all. Mandarin is not strange -- it's just another way of speaking human.

As such, one imagines that the southern Nigerian Ur-linguist, confronted with Indo-European languages, would see prefixes and suffixes as beside-the-point accidents just as we see tone.

And I suspect that both they, over in their alternate universe, and we, in our real one, are correct. Just as only a narrow data set could lead us to suppose that tone drives verb movement, we must entertain that the idea that verbal conjugation (to the linguist, INFL) driving verb movement may well be an accidental notion that will not stand up to a broader examination.

Posted by John McWhorter at March 8, 2004 09:59 PM