Tuesday, May 27, 2008

Chaos in language, a review

Mitchener WG, Nowak MA. Chaos and language. Proc Biol Sci. 2004;271(1540):701-704.

Language is a stream of sound—phonemes, the individual units of speech like the /s/ sound in stream and the /t/ sound that follows it—words, phrases, sentences, and blogs (or, if you prefer, books). Language users manipulate the stream by means of an internal computational system—grammar. Research shows that languages and their grammars, acquired from generation to generation, are stable for about a century. Languages change, of course, but they are marked by relative equilibrium over the course of about 100 years.

Personalize that: One hundred years ago, my grandfather would have been two years old. There’s no question that the English language spoken in his house at that time generally equates with the English language spoken in my house today. Such is English expressing its equilibrium. But it would be wrong, I think, to describe those two instances of English as equable. Even incremental change in language adds up over time. At the end of 100 years, the difference can make quite a difference.

Grimm’s and Verner’s Law detail changes to certain sounds from Indo-European to Germanic languages, changes that are especially relevant for English. Sound shifts can have dramatic effects on a language and can be seen, for example, in the shift from the voiceless stop of the /p/ in Greek pyr (translated = fire) to the voiceless fricative of the /f/ in English fire. The initial consonant changed because, for some reason, the way people pronounced the consonant shifted. Thus, despite a very different look, pyr/fire, this is the same word except for the change from p to f. The similarity is striking if we go back 1300 years or so to Old English, which spelled our modern fire as fyr.

Systematic changes in language, like those described by Grimm’s and Verner’s Law, are unpredictable and unavoidable. Language is, after all, necessarily dynamic, reflecting the nature of its users. Where different languages come into contact with one another, big changes in one or both languages can be expected. Over time, Old English lost case endings on nouns, for example, and this is likely due to contact with Old Norse. Changes may also oscillate: Mitchener and Nowak note, for example, that changes in language morphology follow a pattern:

isolating < agglutinating < inflecting < isolating…

English is coming full circle, changing from inflecting to isolating.

Changes can also enter a language through acquisition errors, or learning errors, in which a child’s acquired grammar does not match the parents’ grammar. Perfectly natural variation in speech pattern or the use of multiple languages in the home environment might trigger learning errors.

In short, Mitchener and Nowak use mathematical models to study language change. What they find is that some language change—such as lention, vowel shifts, and morphology type—arises from reanalysis and variation among speakers and follows regular patterns. But language change is also unpredictable and highly sensitive to perturbations of learning error, borrowed vocabulary, language contact, and so on. Simple errors in learning can lead to complex changes over time. Language, in the Mitchener and Nowak study, is sensitive and stochastic, showing key characteristics of chaotic dynamical systems.

No comments: