If there is one thing Chomsky has taught us about sentences, it is that they are unbounded. A clever person can always write a longer sentence. So I was interested when I read British anthropologist Robin Dunbar's latest paper on language origins (here) and found this sentence, "it may be no coincidence that the levels of intentionality that adults can cope with is the same as the level of embedding that we can cope with in sentences." [p. 56] With this concept of coping, Dunbar seems to be hinting that he believes sentences are bounded after all—maybe not by syntactic rules, but by psychological ones.
Chewing over Dunbar's remark has brought together many themes of this blog. In particular, I've been thinking about Dunbar in the context of a presentation James Hurford gave 4 ½ years ago in Poland (see: The Word-Sentence Continuum). The presentation was on the evolution of predicates. In it he argued that it is pragmatics— the study of language in its social context—that makes syntax interesting and not the other way around. I was in complete agreement with Hurford and have since adopted much of the terminology he used in that talk. However, syntax can explain things in terms of rules, while pragmatics has seemed ad hoc.
A second critical theme on this blog is the speech triangle: the speaker and listener pay joint attention to a topic. I've gone on endlessly about these interactions. Mostly I have focused on the evolution of speakers and listeners, but implicit in all this has been the importance of topics.
Another theme is the continuing quarrels I've had with Chomskyans who always are very bright (more learned than me in many cases) and yet they seem trapped in a world where rules trump experience. A sentence like
(1) The boy the man the woman loved saw ran.
follows a syntactic rule, but is so hard to follow (cope with) that any claims for it as acceptable strike me as absurd. Talk about coping offers me a way to understand the dispute without concluding that my opponent is crazy. I've decided that there is a fourth component of language, one that takes into account the need for speaker and listener to cope with the utterances.
Classically, there are only three components of language: vocabulary, syntax, phonetics, and they are used to analyze language according to rules. When we consider how language evolved, we think of it in terms of these components. Having read Dunbar, however, I now think there is a fourth component and that a sentence cannot be fully analyzed without examining it. I am tempted to use Dunbar's term and call the component copability, but dictionaries do not recognize the word and I do not like inventing jargon. So let's call it the social component, or sociality.
Sociality refers to the rules followed in order for the producer and the audience to cope with the task of visualizing or understanding what is said. Sociality differs from the performance/competence distinction long advocated by Chomsky. Performance deviations from syntax are assumed to be mere accidental frictions that get in the way of producing what our competence would have us say. Furthermore, performance refers only to the producer side of the utterance. Sociality, however, is not accidental, and its rules bind both a sentence's producer and audience.
To see what I mean, look at sentence (1). It is not some accident that gets in the way of understanding, but some limitation on our ability to cope with it. What is more, reading Dunbar persuades me that this limitation can be expressed as a rule. How many rules are there? That's to be determined, but I have already noticed one and Dunbar suggests a second. In this post I will stick with just one and save Dunbar's business about intentionality for later. The point of this post is simply to argue that sociality is a legitimate component of language, as worthy of study as syntax, the lexicon, or phonetics.
To make my case I will simply assert a rule of sociality: speakers and listeners can cope with a topic and a subtopic in a sentence, but not a second subtopic as well.
A topic combines a subject with news about the subject. In English, at any rate, news consists of a least a verb with other details being optional. Some may be surprised that I am analyzing topics, a feature of sociality, by referring to subjects and verbs, features of syntax; however, the components of language are hierarchical and use what lies below them. Sociality uses syntactic features, just as syntax uses vocabulary, and words use phonetics. The addition of a fourth layer to the language system suggests that a top-down analysis of how language works begins with sociality rather than syntax, and that observation justifies Hurford's preference for pragmatics over syntax. Sociality gets closer to explaining what we observe in nature.
English topics combine a subject with a verb and can also add one or more details.
Topic = Subject + verb [+ detail(s)].
(2) The boy ran.
Sentence (2) is a simple topic, consisting only of a subject and a verb. The sentence is fully social, meaning any normal speaker can cope with the task of understanding it.
Subtopics are topics that depend on other topics for their meaning. For example,
(3) The boy the man saw ran.
Investigation may determine that children under a certain age can cope with sentence (2) but not (3), but any normal, English-speaking adult can cope with both. In (3) the topic is the boy … ran and the subtopic is the man saw. The subtopic only makes sense by remembering the topic: the man saw [the boy]. The topic is remembered rather than repeated aloud because the listener can cope with the omission. Syntacticians have a number of alternate solutions to why phrases like the second the boy is omitted, but sociality rules are higher level explanations and do away with the need for syntactical explanations.
Now we can see what is wrong with sentence (1). It violates a sociality rule by adding a second subtopic, the woman loved. Understanding the second subtopic requires remembering that it refers to the first subtopic and understanding the first subtopic requires remembering the topic referred to. It is too much to cope with.
The rule can also explain the source of some ambiguities. Consider
(4a) The boy the man saw hit the ball.
(4b) The boy the man saw hit the ball ran away.
In (4a) we have got a topic and subtopic, but how can we be sure whether the ball is a detail associated with the boy or the man? Did the boy hit it, or did the man see it being hit? The answer is provided by the punctuation mark. If the topic is the boy and the subtopic is the man saw [the boy] hit the ball, then the topic is incomplete, telling us nothing about the boy. If it was the man who saw the ball being hit, a sentence like (4b) would be needed. (4b) has additional news about the boy: he ran away. (I suspect that some readers are unhappy with (4b) and had to read it twice to understand it. If the sentence were spoken aloud by a person bearing the news, the speaker's brief pauses would make it immediately clear. By losing language's primal component, written language sacrifices much that spoken language can provide.)
This analysis works fine at the sociality level, which treats of topics and news, but syntax has nothing to offer here. The meaning of (4a) is settled by the period and punctuation marks are not syntactic structures. Without the sociality rules, the puzzle of (4a) just seems like some sort of linguistic glitch. Of course, syntax can often help assign news to the proper subject:
(5a) The boy the man saw chased the ball.
(5b) The boy the man saw chase the ball…
Here the verb conjugation shows us that in (5a) chased the ball refers to the subject boy while in (5b) the man saw [the boy] chase and more news about the boy will have to follow.
So here we see a rule-based way of explaining meaning that, if left to the rules of syntax or lexicon must go unsolved. This success suggests that sociality really is a fourth component of language, and language cannot be fully analyzed without it. Come to think of it, Chomsky has identified a number of ambiguities that seem like mere glitches in the system. Might many of them be explained by discovering and applying rules of sociality? And if we had such rules, might we be able to teach better writing by being able to explain just what readers can and cannot cope with?
I have considered myself what makes sentences such as (1) so hard to understand. From the viewpoint of a computer scientist, the most straightforward explanation lies in computational complexity: Most language utterances can be constructed and understood by humans in real time. This very much limits how complex its structure can be. In practice, you will be hard-pressed to find recursive clauses with more than one embedding, maybe two in written language. Restricting the depth of such constructs, they can be dealt with by a low level parallelism and limited working memory. Indeed, I'd argue natural language can be described with regular-level grammars, at the cost of generality and elegance.
Shieber85 (http://www.eecs.harvard.edu/~shieber/Biblio/Papers/shieber85.pdf) may argue that "down this path lies tyranny", but I'm more inclined to see it as a path of liberation from over-generalized rule systems and toward computational tractability.
Posted by: Justin Kaeser | May 11, 2014 at 05:26 PM