Would-Be Revolutionaries Strike Again

Fify years ago there was a scientific revolution that overthrew the tectonic ideas of the time and made sense of many old mysteries and new ones too. Is it time for a similar revolution in linguistics?

Steve Levinson and Nicolas Evans have become the bomb-throwers of linguistics. Last year they published a paper titled "The Myth of Language Universals" that caused an uproar merely by citing the facts of linguistic diversity. (See: Is Anything Universal in Language?) It was a direct challenge to the idea of a Universal Grammar (UG). Now the two authors have published a further paper, this one in Lingua, titled, "Time for a Sea-Change in Linguistics," that hurls another bomb into the UG den (abstract here). They propose a different focus from the one that generativists have pursued lo these past fifty years, and they do not couch their argument in polite language.

Levinson-Evans try energetically to kill the old king, which they call "Chomskyan-inspired" C-linguistics. They offer a D-linguistics (D for diversity) as a replacement. The table below shows some of the differences between the two styles:

	C-linguistics	D-linguistics
Says language comes from	Internal processes in individual brains.	Interaction of culture and biology.
Investigates language as	Form	Function
Linguistic facts framed in terms of	Universal grammar	Capacity for diversity
Theoretical results	List of universals	Statistical generalizations

I happen to be deeply sympathetic with all the D-linguistics points and I appreciated the many digs at "C-linguistics" that are salted throughout the paper. Yet, much as I love revolution, and much as I believe that language studies desperately needs to escape the box canyon it has entered, I do not see the latest Levinson-Evans paper as providing the map out.

They argue like this:

The issues matter enormously to a discipline which, if it makes the right moves, can regain its centrality as the study of what makes us human, thus linking the humanities, the social sciences and biology in a way that will guarantee its long term disciplinary success. … Or it can remain an armchair discipline, where all the interesting new developments will be made by invaders from other disciplines working on our turf. Consider the fate cultural anthropology: faced with many of the same opportunities, it has turned inward into an exclusively humanities subject without a future; cross-cultural empirical reference is now done largely by biologists, psychologists, behavioural ecologists and other social scientists. [pp 2-3]

That's a political candidate's argument, appealing to hope for success and fear of being powerless and ignored. During my days as a middle-school science teacher I told my classes that in science we did not determine answers by voting; I hope that boast still holds.

By an odd coincidence I know something of the way sciences go through sea changes. I have written a couple of books telling the stories of upheavals in physics and geology, and when I assembled pieces for my anthology of science writing I read both material that led to great changes and material that hoped to be followed by revolution but was not.

Revolutions in science do three things simultaneously: they kill off an old theory; they replace it with a new understanding; and they preserve the achievements of the old theory by putting its facts into a new context.

A good example of such a change is in the revolution wrought by plate tectonics at exactly the time Chomsky was hoping to revolutionize linguistics. Geologists had a tectonic theory of how the earth's crust became deformed by mountains, volcanoes and earthquakes. That theory was discarded in broad outlines, but the mass of data and detailed account of the changes has survived and been given a new context, plate motions and their enormous pressures.

Revolutions like that are rare, and I'm not sure that one has ever been managed in the human sciences. More often a project just runs out of steam and investigators go on to something else.

A good example of that failure is the history of Chomskyan linguistics itself. Generativists moved from looking for an empirical UG that would stand up to linguistic diversity and explain how children learn language so rapidly, to looking for a set of parameters and principles that could account for learning any language, to defining a process of "merge" that allows the generation of sentences. In no case was the switch the result of a great new understanding. It just became apparent that the old system was not going to work and people moved on to something more abstract and general. Personally, I'm not in the mood for any more half revolutions.

A full revolution demands discovering an account of language that absorbs the achievements of C-linguistics while explaining them in terms of a new approach. If this replacement theory turns out to be some type of D-linguistics, the change will mean explaining the formal discoveries of C-linguistics in functional terms. (Note: Terrence Deacon has done a decent job of explaining formal syntax in symbolic terms; see, Protolanguage Was Symbolic) C-linguists have long denied that there are such functions, an attitude that explains their anti-adaptationist, anti-gradualist approach to evolution. It also suggests that, if one is to happen, any D-linguistics revolution will be a Darwinian one.

Since this blog often grumbles about the Chomskyan approach, some readers may be surprised that I say there have been any achievements to absorb, but I hope I have mentioned several times that in their long search for a UG the C-linguists have managed to identify a series of constraints that limit language. Basically these limits have to do with how we can move words and phrases around in sentences. The limits vary more than proponents of a UG wish, but every language has its rules. Some languages (e.g., Latin) permit extensive movement, some are more constrained (English), and some are still more rigid (Swahili), but none are so free that any combination of words is possible or so fixed that no movement is ever allowed. Why is that?

The Levinson-Evans paper gives no hint of an approach to this question. They do say that they don't want to throw out any C-linguistic baby with its ample bathwater, but they focus entirely in their terms of diversity. They are ready to retain the descriptions of various languages developed over the decades by C-linguists, but they are not interested in the constraints the C-linguists talk about.

The authors say "the distribution of attested structural types across the design space reflects the likelihoods of the evolutionary pathways that would engender them, rather than specific constraints on the structures themselves." [2] They could be right, of course, but they are explaining away an observation rather than explaining it. A full revolution doesn't just dismiss observations as illusions, it predicts the illusion.

Comments

[gameswithwords]

Thanks for the synopsis of this article. I read the BBS article some time back but haven't gotten through all the commentaries. Did it really cause an uproar? My lab read it with some interest, but I don't recall it causing an uproar ... or really convincing anyone, either.

I recall one of the issues in that paper being a focus on rare, poorly-described languages. Language description is hard. Despite thousands of studies and thousands of native-speaking researchers that have worked on English, there is a lot about English that is still not fully understood and a lot of mischaracterizations have been made along the way.

Therefore evidence from a language that has been studied by at best one or two non-native speakers is highly suspect. It would not be the first time the characterization of such a language was deeply wrong (I have some nice examples around somewhere, but not at the tips of my fingers).

Where I'm going is that not all evidence is created equal. Some is more believable than others. So when you have a solid theory supported by a lot of solid data but happens to clash with some fairly weak data, it's not clear that you should keep the weak data and toss out the theory.

I went through some of these issues here as well.

[Jerry]

Don't you think, that many researchers are trying helplessly to derive Language Universals from Speech practices by simply miss-presenting paradigm of Language as Speech?
Language is 1 or 1.5 million years older then any its Speech manifestation. Speech practice is only 50 - 45,000 old. Those entities are not the same as they viewed by C-and D- ppl.
Why do not consider third position:
http://linguistlist.org/pubs/papers/browse-papers-action.cfm?PaperID=31640