Bart de Boer on The Origins of Vowel Systems
Bart de Boer explores why thousands of unrelated languages are so similar in certain respects
There are about 6000 languages spoken in the world today, and even though they can appear bewilderingly different at first, they do not consist of random sounds and grammatical rules. There are many ways in which unrelated languages are similar. An example is the limited repertoire of sounds that all human languages use for making words. The question then arises: why are all these different, unrelated languages so similar in certain respects?
Often such similarities are explained as the result of innate and genetically specified properties of the brain. However, this requires the evolution of complex and very specialized neural circuitry. Such evolution, although not impossible, would be slow and involve many complex steps. So it is desirable to explain regularities of human language as the result of mechanisms other than biological evolution. My book investigates such a mechanism: self-organisation in a population of language users. This mechanism is applied to the vowel systems human languages use.
Languages show remarkable regularity in the vowels they use. Although humans can produce and distinguish at least fifty different vowels, human languages do not use random repertoires of vowels. In fact, almost all languages, whether they have only three or up to fifteen different vowels, use /i/, /a/ and /u/. Furthermore, by far the most frequently occurring repertoire of vowels contains only five: /i/, /e/, /a/, /o/ and /u/. Languages such as English, French or German that have more than ten vowels are in fact exceptions. It turns out that the vowels and combinations of vowels that are most often used are the ones that are easiest to produce and easiest to distinguish from each other. Such systems are called optimal. But how have they become optimised? And who is doing the optimisation?
Individual language users do not optimise. When one looks at how children learn vowel systems, one finds that they try to emulate, as closely as possible, the vowel systems that their parents and peers use. This is illustrated by the way subtle differences in pronunciation can persist in closely related dialects or by the fact that some languages have vowel systems that are far from optimal. The question remains: how have vowel systems become optimised?
The hypothesis investigated in my book is that the dynamics of a population of language users automatically pushes the vowel systems in a direction that makes them optimal for the communication task at hand. This would result in systems of vowels that are easy to produce and easy to distinguish.
This hypothesis is implemented and tested as a computer model for two reasons. First, the research was conducted as part of an artificial intelligence project whose purpose was to investigate the origins of intelligence and language. The methodology of artificial intelligence requires the use of demonstrable and testable computer models. Second, the dynamics of self-organisation are so complex that they are hard, if not impossible, to understand without actually making a computer simula-tion of the phenomenon. Without such a simulation, the term self-organisation can only be used in a vague and metaphorical way. The computer model was not designed to be an accurate model of how language changes over time, but rather to represent production, perception and population dynamics as realistically as possible.
A large part of the book is devoted to a detailed description of these computer models. The results are very promising: vowel systems emerge that closely follow the universal patterns of human vowel systems. Perhaps even more interesting is the fact that the frequency with which different types of vowel systems emerge corresponds well to the frequency with which they occur in human languages.
Although the role of self-organisation is only investigated in detail for vowel systems, I propose that it can also play a role in shaping other parts of language. I give some simple examples of possible ways which this could occur.
I have tried, through the research described in this book, to enhance our understanding of vowel systems by proposing a mechanism through which they have become optimised: self-organisation in a population. The close correspondence between modelled vowel systems and those observed in human languages shows that computer models of self-organisation can be used successfully to investigate linguistic and cognitive questions.
Bart de Boer
| 
VIEW BASKET
