January 09, 2005

Homo Hemingwayensis

Over the past couple of years, there's been renewed controversy about the role of recursion in human language. Steven Pinker and Ray Jackendoff, in an article entitled "The Faculty of Language: What’s Special about it?", put it this way:

In a recent article in Science, Marc Hauser, Noam Chomsky, and Tecumseh Fitch (2002) offer a hypothesis about what is special about language, with reflections on its evolutionary genesis. ... HCF differentiate (as we do) between aspects of language that are special to language (the “Narrow Language Faculty” or FLN) and the faculty of language in its entirety, including parts that are shared with other psychological abilities (the “Broad Language Faculty” or FLB). The abstract of HCF makes the extraordinary proposal that the narrow language faculty “only includes recursion and is the only uniquely human component of the faculty of language.” (Recursion refers to a procedure that calls itself, or to a constituent that contains a constituent of the same kind.) ... The authors suggest that “ most, if not all, of FLB is based on mechanisms shared with nonhuman animals…In contrast, we suggest that FLN – the computational mechanism of recursion – is recently evolved and unique to our species” (p. 1573). Similarly ( p. 1573), “We propose in this hypothesis that FLN comprises only the core computational mechanisms of recursion as they appear in narrow syntax and the mappings to the interfaces”.

There are lots of interesting issues here. To what extent does human language rely on mind or brain systems that are unique to language as opposed to shared with other human activities? Are language-specific mechanisms qualitatively different (like the difference between auditory and visual spatial localization), or just slightly specialized instances of more general systems (like the difference between touch perception on the arm and on the leg)? Are human linguistic systems qualitatively different from analogues in other animals (like elephants' trunks) or just quantitatively different (say, just the consequence of bigger brains)? Or have qualitative differences emerged from quantitative ones? Pinker and Jackendoff offer a range of arguments in opposition to Hauser, Chomsky and Fitch, and you can go read their paper and see what you think.

Now, though, I'd like to come at this question from a different and less serious direction.

A piece by James Thurber entitled 'A Visit from Saint Nicholas (In the Ernest Hemingway Manner)' appeared in the 12/24/1927 edition of the New Yorker. It starts like this:

It was the night before Christmas. The house was very quiet. No creatures were stirring in the house. There weren’t even any mice stirring. The stockings had been hung carefully by the chimney. The children hoped that Saint Nicholas would come and fill them.

The children were in their beds. Their beds were in the room next to ours. Mamma and I were in our beds. Mamma wore a kerchief. I had my cap on. I could hear the children moving. We didn’t move. We wanted the children to think we were asleep.

One of the defining characteristics of this style is that the writer doesn't tell you much about how the pieces of the story go together, at least not explicitly. The usual sorts of discourse relationships exist among the phrases, but very little of this structure is encoded by phrasal embedding within sentences. Thus

We didn’t move. We wanted the children to think we were asleep.


We didn’t move, because we wanted the children to think we were asleep.


The stockings had been hung carefully by the chimney. The children hoped that Saint Nicholas would come and fill them.


The stockings, which the children hoped that Saint Nicholas would come and fill, had been hung carefully by the chimney.

As a result, there's very little syntactic recursion. In the cited passage, the only examples are a couple of nouns post-modified by prepositional phrases ("the night before Christmas", "the room next to ours"), which you could analyze as involving noun phrases inside of noun phrases, and a couple of sentential complements of propositional attitude verbs ("the children hoped that Saint Nicholas would come", "we wanted the children to think we were asleep"). If we take such structures to be limited in principle to one level -- as they generally are in practice in such writing -- then no recursion is really involved. And these simple structures can easily be replaced by even simpler ones, even less suggestive of recursion. Thus the possible NP → NP PP patterns like "the night before Christmas" and "the room next to ours" can be replaced by [N N] patterns like "Christmas eve" and "the next room". The sentential complements could be replaced by pronouns: "Saint Nicholas was going to come and fill them. So the children believed, anyhow", and "We pretended to be asleep. We wanted the children to believe that."

Now suppose that next week, explorers in the Himalayan rhododendron forest find a new hominid species, H. hemingwayensis. These secretive creatures have been able to hide away from humans for millennia, partly because they are so well adapted to their dense, mountainous habitat, and partly because they are quite smart. Their cleverness includes the ability to amuse themselves while hiding by engaging in vocal displays, known anthropomorphically as "discourses", which they use to form and maintain social bonds and to compete for social prestige. These displays have a complex structure, in most respects just like human language. There is a small inventory of meaningless phoneme-like elements -- different for each troupe -- out of which a large vocabulary of more than 10,000 well-individuated "words" is formed. These "words" are combined into "clauses", just as in human languages, and are "inflected" according to their sentential function. Most surprisingly, the "words" seem to have "meanings" just like human words do, and the "sentences" and "discourses" are put together in familiar ways that we find it easy to believe we understand. In fact, when we record and translate some of these "discourses", we find patterns that remind us of many familiar human stories, suggesting that over the millennia, there has been a steady trickle of cultural contact between these creatures and humans.

The only thing is, the communication system of H. hemingwayensis appears to have absolutely no recursion whatsoever. "Nouns" can be modified, but only in the parallel "big bad wolf" sort of way. There is no (syntactic) clausal subordination, and no relative clauses. There are no sentential complements, though pronouns and some noun phrases ("that idea", "his argument") can be used to refer to explicit or evoked propositions. When you translate one of their stories, it comes out just like Thurber's imitation of Hemingway:

It was Christmas eve. The grove was very quiet. No creatures were stirring in the grove. There weren’t even any rodents stirring. The baskets had been hung carefully from the upper branch. Saint Nicholas would come fill them. The children had that hope, anyhow.

And so on.

How would we react to this discovery? Would we say that H. hemingwayensis has language without recursion, or no language at all? I think the answer is obvious. In fact, we can ask exactly this question about the historical reaction to the particular modernist style that Hemingway exemplified. In effect, the reaction was just "gee, these people are writing in a curious and interesting way." As far as I know, no one ever said anything like "these people are not using the English language", much less "these people are not human".

Let me say that I don't think my H. hemingwayensis scenario is very plausible. It's hard for me to believe that any creature could develop anything much like human language without at least some limited form of recursive compositionality. By my hypothesis, H. hemingwayensis does have compositional syntactic and semantic structures up to the clausal level, and analogous sorts of structures at the discourse level. Given that much, I'd be surprised not to see some use of syntactic embedding that goes at least a step or two in the direction of recursion (though whether this is implemented by fully recursive mechanisms is another matter).

But it seems preposterous to claim that syntactic recursion "is the only uniquely human component of the faculty of language". As we've observed in other contexts, sometimes it takes a really smart person to have a idea like that.

[Some other Language Log posts that are relevant to this topic:

Hi Lo Hi Lo, it's off to formal language theory we go
Cotton-top tamarins: on the road to phonology as well as syntax?
Humans context-free, monkeys finite-state? Apparently not.
Parataxis in Pirahã


Posted by Mark Liberman at January 9, 2005 09:58 AM