Over the years this blog has reported a fair amount about the collaboration of Morten H. Christiansen and Nick Chater, a pair of enterprising researchers determined to challenge the Cartesian theories of Chomsky et al. Let me summarize a few old posts:
- The Richness of the Stimulus: Discussed the two authors paper on how language is shaped by the brain. Echoing some ideas that I first encountered in Deacon’s work, they argued that instead of evolving brains that could use language, we evolved languages that were adapted to our brains.
- Language Adapted to Us: Continuing the previous discussion, the authors suggest that the world’s languages are easy to learn because over time languages are shorn of their hard-to-learn features. If a language is too hard for a child to learn, it disappears. This idea marks a break with Chomsky’s notion of a Universal Grammar, but did not mark a return to Skinner’s ready-for-any-lesson blackboard. Instead our brains make some writing on the blackboard easy, others more difficult. The basic unit of selection in this account is the word or a combination of words.
- Could Language Modules Have Evolved?: Adding to the previous work, the authors report that language changes too fast for its features to become genetically fixed in a specialized region of the brain. For example, nouns in Old English were declined, but today only a few pronouns are still declined. Noun case is expressed through word order. The change took place too quickly for genetic evolution to keep up. The point is that, according to the rules of natural selection or even genetic drift, there is no way for a Universal Grammar to evolve. Language structure must reflect something other than internal, linguistic rules.
- Natural vs Coordinated Challenges: The authors distinguish between the ability to manipulate and use the natural (N) world, and the ability to coordinate (C) one’s actions with others. An example of an N-based skill is throwing a ball. The would-be thrower has to practice the motions, release points, and power needed to make a ball travel to different spots. There is no way to avoid extensive practice and many people never become very good at it. A C-based skill can be something like nodding one’s head in a particular manner to indicate agreement. Greeks and Norwegians use contrary head signals, but that only confuses tourists. Locals learn and get the point quite easily, so easily that we never speak of a head-nodding skill. Language usage is a C-based skill, but many linguists assume that it is as hard to learn as an N-based skill.
- No Universals in this Theory: Building on their previous N/C distinction, the authors distinguish between arbitrary and functional features of language. Arbitrary features are those that, given the laws of nature, could be different. They change too fast for a population’s genes to catch up. So particular words and sentence structures cannot be fixed in a gene pool. They will change before evolution has time to act. Functional features are more general, like the ability to differentiate between speech sounds. Thus, languages cannot have begun with a genetic mutation that established an arbitrary feature of language. Since the whole of generative grammar is arbitrary, this argument says generative grammar is either impossible or the result of a miracle.
- Chunk and Pass: The authors now turn to another argument. According to generative grammarians, sentences can be infinitely long, and they are parsed as a whole. Christiansen and Chater say, no. Words flow out of a mouth like water going over a fall. Once gone they are not to be retrieved. This means the sentences must be actively parsed while they are still being generated. The method is one of chunking (combining words into larger units) and passing the chunk on to a holding pen where they can be combined into a complete idea.
The authors’ new book, Creating Language: Integrating evolution, acquisition and processing presents and enriches the whole of their work. It proposes that the three time scales commonly used to study language—the moment of generating an utterance, the lifetime of the individual language user, and the multi-generational changes to a particular language—are each subject to the same processes of comprehension and production. Those processes are the ones listed above: adapting language to the brains of the users, coordinating what speakers and listeners can agree on, and producing sentences that can be chunked and passed most readily.
Over the next several posts I will discuss this book in more detail. Should you read the book on your own? Yes, if you are okay with reading technical literature. It does a good job at subverting Chomsky’s approach, examining the assumptions that support generative grammar and showing that they are not Euclidean in either their profundity or their certainty.