Paradigm Lost

Barcelona's Museum of Science hosted the Evolang 2008 conference that ran this past week from March 11-15.

The Barcelona conference on the evolution of language has ended and I want to digest what I have learned about the way forward. However, it is clear that the old paradigm of speech is the fruit of a generative grammar can no longer stand. (Generative grammar is the theory launched by Noam Chomsky that mathematical rules and algorithmic procedures can be written precisely enough to “generate” all the possible sentences of a natural, human language and only the possible sentences. In other words, it can generate a sentence like The boy took a swing at the ball, but not The ball took a swing at the boy.)

Regulars on this blog will know that I have been a steady skeptic of the generative paradigm, but now it is apparent that I am far from alone. Generative grammar will continue to be a source for practical work in making machines that can use human language, but it is no longer a source of biological or psychological ideas about the nature of language users or the role of speech in human affairs.

The first and primary reason for the old paradigm’s failure is that brains are not digital machines. (See reports on the presentations of Friedmann Pulvermuller, Brain Circuitry Challenges Linguistic Models, Gary Marcus, The Practicality of Studying Language Origins, and Chris Knight, Some Say versus Others Say) The brain cannot follow all the rules and procedures written for digital, generative machines. Beyond that, however, is the problem that generative grammar offers no insights into the history or function of language. The generative machine works or it does not. It is not designed to tell us why, say, a sentence like The king is a traitor was impossible in 1549 while in 1649 it was not only possible, but the king was executed for being a traitor.

Chomsky probably recognized this determination from the beginning and for decades he opposed any biological inquiry into how it happened that all humans and only humans have an innate mechanism for learning and generating particular languages. Eventually, however, the refusal to consider evolution’s role in producing a verbally competent species could not be denied. The fact that the Evolang conference just ended was the 7th such gathering in a dozen years is proof of that interest.

The generative solution to this growing scholarship was to make use of a series of more powerful generative machines, known as the Chomsky hierarchy. This term refers to a list of machines with ever more powerful rules and procedures for generating sentence strings. There may have been some evolutionary ladder moving species up the hierarchy from one machine to the next. Attention to this process up the ladder has focused on the leap to the last level. This step has its defenders (see: The Eternal Duality) but the last day at the conference saw two presentations on this very point about movement up the hierarchy (see reports about presentations by Derek Bickerton, Words Are More Human than Syntax, and Joris Bleys, Recursion Can Be a ‘Side Effect’).

Bickerton’s attack was the more radical because it denied recursion’s existence altogether and moved the focus away from syntax (what the rules and procedures of generative grammar are all about) to the semantic/symbolic/attentive demands of individual words. Most striking was the reception of Bickerton’s presentation received. The applause was the loudest and longest observed by this blogger after any of the presentations. Then none of the questioners defended either Chomsky or recursive syntax.

The absence of a defense does not mean generative grammar has lost its adherents. It is only the odd ones among them who are interested enough in the evolution of language to even think of attending such a conference, but the absence of all the leaders of the effort to understand the evolution of generative grammar—Steven Pinker, Marc Hauser, Tecumseh Fitch, Ray Jackendoff, and Paul Bloom — cannot be entirely a coincidence. It is somewhat like the day the rioters of St. Petersburg awoke to find all sign of the tsar’s retinue was gone. Something decisive had happened.

But any reader of Thomas Kuhn knows that failed paradigms do not go away. They persist until they can be replaced by something else. On the other hand, readers in the history of psychology know that Kuhn was an optimist. There have been many paradigms in the history of psychology, but they have all run into rocks. Instead of being replaced by a new, better paradigm, the earlier, chaotic condition returns. Thus, despite the fact that psychology is about the same age as evolutionary biology, it has made nowhere near the same progress.

The philosopher John Searle argues that psychology keeps running into the same rock: the mind/body puzzle. In every case a new paradigm attempts to explain some set of psychological phenomena in non-mentalistic terms, but is eventually wrecked when it becomes clear that seemingly mentalistic phenomena can no longer be shoved aside.

This story appears to have been repeated in the wreck observed in Barcelona. The generative-grammar paradigm attempted to explain all language in terms of syntax, a set of rules and procedures that could sidestep the mentalistic problem of meaning. But meaning has ultimately forced itself into the field. Bickerton raised the white flag, transferring control to the semantic demands of individual words. (It was, by the way, remarkably courageous of Bickerton to change his mind so publicly. As someone who has written many stories about the history of science, I know how rare it is to change one’s mind and prepare to move on.)

The question the conference has left unanswered is whether a new paradigm will appear or whether the study of language evolution will simply fall back into the chaos and confusion that once made taboo all inquiry into speech origins. I will discuss that in future posts (as I now return to my once-weekly blogging), but perhaps I should leave this one by acknowledging that the final mood of the conference seemed quite optimistic. There were about 170 registered participants, perhaps half of them students. The presence of so many students gave people hope that the field will continue to grow. But, says I, those students are going to have to cut themselves a new path.

Comments

[Neil]

First, I'd like to say that it has been a pleasure to find and read this blog, especially with the liveblogging of the conference. Thanks for doing it!

I just wanted to make the note, as you I'm sure know, that those adherents of generative grammar that you list mostly disagree with Chomsky's theories. Jackendoff, for example, has proposed a very different grammatical paradigm with a balance between syntax and semantics which has allowed him to address evolutionary perspectives. He and Chomksy have fairly publicly debated on the topic.

So, "generative grammar" is not necessarily the same as "Chomsky's generative grammar" (i.e. Minimalism). The latter is to be rejected, while the former may still have potential in it yet.

(Full disclosure: I am Jackendoff's grad student)
------------
BLOGGER: It is true that there are different generative grammers, Somewhere on my blog is a discussion of the Jackendoff system, but in this case those differences may not matter as the more general rejection of the generative paradigm:

[brian bayly]

That is a great report and analysis of Evolang! --- and provocative. Trying to be brief makes the following seem harsh. Please mentally flesh out a longer, more tentative version.

In “Some Say versus Others Say” EBB writes:
“Humans are unique, not just in having language, but in living in a world of digital symbols (i.e. something is A or not A) and the real question is how we came to be analog creatures who live in an artificial world of digital symbols . . .
. . . Knight and Powers are insisting that digital things are human artifacts and cannot have a place in the …world of brute facts.”

When a person sees a tree some neurons in her brain go through a dance. When she sees a second tree, again some neurons dance, including many of the same neurons and pulses as before (So far we remain close to the world of brute facts).
But the person’s brain may extract the common features of the two dances; some neurons may go through a third dance that represents the common features and lets go of or loses the accidental differences. By such a process the person forms an abstract concept (though not yet a symbol). The abstract concept is a human artifice. The person has created a category, and a third plant she sees either adds to the category or belongs outside it (“A or not A”).

How did we come to live in an artificial world … ? By categorizing.

Why are human systems of categories so elaborate compared with an ape’s? Because we get pleasure from categorizing. For pleasure we invent categories, many of which later turn out to have utility.

Now look for syntax: suppose a person observes someone walking, someone running, someone eating nuts and someone picking fruit. A person who gets pleasure from categorizing may put the first two together in one group and the latter two together in a second group --- thus laying groundwork for distinguishing transitive from intransitive verbs. Words arise through categorizing items and relations, and syntax arises through categorizing events. Apes do enough categorizing to get by. It is the uniquely human feature pleasure-from-categorizing that makes human culture more complicated, including human language.

To recognize the extensive consequences of pleasure from categorizing is not quite a new paradigm. But it is a step toward a fresh picture of where language came from, why we have it and apes don’t. It would be a step forward if pleasure from categorizing were given more attention. Perhaps readers of this blog will lead the way.

“… chimps mostly don’t WANT to name things just for the sake of it” – Jean Aitchison

“… (other animals) never extract characters for the mere fun of the thing, as men do.”
- William James.