The Atoms of Language
by Mark Baker
Linguists, I imagine, used to sit by the hearth at their club tossing roots back and forth and criticizing the younger generation for a drift toward taxonomy. It must have been quite cozy.
After World War II, in my fanciful picture, things got less settled. Noam Chomsky set the cat among the pigeons with conjectures about deep structure and universal grammar. Computational linguists began doing algorithms. Developmental neurophysiologists appeared on the fringes where the fire couldn't keep them warm, but they kept having to bring in extra chairs. People took sides and collegial silence no longer held.
Maybe there was a degree of calm in more recent years. It seemed Chomsky was onto something. Old alignments decayed and, while new ones came about, they now proved to be factions debating which is the true heir of the master. Steven Pinker, among many others, took Chomsky's principles as axioms rather than subjects for wrangling. Qualifying terms like "computational" became less important.
Mark Baker now brings to the world of pop linguistics a topic that's clearly attained ripeness in the profession, the theory of parameters in language. In a very inadequate summary: Children arrive ready to start learning languages. Which ones they learn will depend on which ones they hear. Their learning facility drops off over time, so that a 14-year-old (just beginning to be exposed to language courses in many American schools, alas) has a many-times harder task in picking up a second language than she had in acquiring her first between the ages of 18 months and 8 years.
More: Translation is possible. A Navajo/English bilingual has no great difficulty taking Navajo in (say over a battlefield radio link on a Pacific island) and creating English sentences that convey the same message. The languages have no similarity and the peoples who invented them are cousins only in the most extended meaning of the word, but everybody recognizes that the Code Talker is using language on both sides. This trivial-seeming bit of knowledge, that translation is possible, is crucial.
If the language-working part of your brain were a machine, how many knobs would it have on the front? The old-time linguists might have looked at Ancient Egyptian, Afrikaans and Chukchi and answered, "Lots." One bunch for nouns, another for word order, maybe one for selecting whether the object of a sentence is animate or inanimate. (They might even have said the machine that spoke Hungarian did not even have the same knobs as the one that spoke English.) Baker suggests that the answer is instead eight (or some other rather small number) and that they aren't knobs but plain up/down switches.
Well, there is a startling idea. Look: If the little tyke comes primed to learn either Mohawk or Spanish, maybe there is just one kind of machine. If that's so, then the machine must be capable of configuring itself on the fly. People who learn two languages in childhood must have machines that can flip from one state to another. What if you could find out how the machine sets itself up and how it handles the change from language to language? Could you count the knobs and put labels on them? This is exactly what Baker sets out to do.
You build up a phrase—say, the knobs on the machine—by stringing together words according to how they are related to each other. On governs machine (that is, forms a phrase that conveys a certain relationship) and knobs governs on machine. The governing word at each step in such a construction is called the "head" of its little phrase. Any reader of English will see knobs on machine as the natural way to make such a phrase: the head comes at the beginning every time. Not so in Japanese, which consistently puts the head at the end, machine-on and then machine-on knobs. A very large class of languages is split in two on this criterion, and there isn't a third choice. Write this head directionality down as the name of a switch with two positions, and you have made a start toward labeling the controls.
In an attempt to create a grand system of language, parallel with the chemist's periodic table of the elements, Baker and other parameter theorists have identified one or two dozen of these two-way switches. In The Atoms of Language Baker not only lists the names and describes the consequences of the settings, but also places the parameters in a hierarchy: If the "polysynthesis" parameter is set to No, then "head directionality" has to be set to First or Last. First leads to more choices, English coming about if "optional polysynthesis" is next set to No, "subject side" to Beginning, "verb attraction" to No, and finally "serial verbs" to No. Six up/down choices lead you from the undifferentiated machine to a real language. Flip a couple of the switches the other way and you get Navajo instead. The child learning two languages operates one machine but changes a few settings when changing over from one to the other.
Now you can build a kind of inverted tree diagram, each parameter generating two branches and each branch leading to either another choice point (parameter) or some languages reached by tracing down the tree. An important but puzzling point is that the tree has nothing to do with relationships between languages. The same parameter values that lead to English lead to Indonesian; Japanese and Choctaw fall together at the end of another branch. How do languages with a common ancestor, like French and Spanish, end up at different places in the diagram? As a start toward explaining (the point is really not the major concern of the book), Baker cites both a "just-so story" about linguistic drift and some anecdotal material about influences between geographical neighbors. The problem remains for someone else to deal with at length.
It's a nice theory, but does it have anything to do with the real world? We learn a lot about language by watching children learn it, and there is at least a suggestion (again, this is not Baker's focus) that children not only begin with no values assigned to their parameters, picking the right settings out of the air, but that they even learn the choices in a definite order: tracing from the top of the tree down the branches. If that's so, it lends strong support to the diagram and the theory as a whole.
The Atoms of Language is well worth the reading. It wants pretty close attention but repays the effort in some new understanding.
| Approved Ben Teague web site |
 |
The Atoms of Language