Language Evolution and Computation Bibliography

All human languages perform the same function, and the set of distinctions that they use to do so is probably highly constrained. The constraints come from the universal architecture of the human mind, which influences language form through the way it hears, articulates, remembers, and learns. However, within these constraints, there is latitude for variation from language to language. For example, the major categories of subject, verb, and object vary in their typical order, and some languages signal grammatical distinctions primary by syntax, or the combinatorics of words, whereas others achieve this mainly through morphology, or the internal mutation of words. What determines the historical evolution of any particular language across the possibility space formed by these different options? In this issue of PNAS, Dediu and Ladd (1) present evidence suggestive of an answer that has seldom been considered before, which is that interpopulation genetic differences may play a role.

Language is often thought of as the crowning human adaptation, the one that allowed Homo sapiens sapiens to conquer the globe. The assumption underlying such ideas is that verbal transmission of information provides unalloyed benefits, by reducing the costs of learning about the environment. However, this raises the question of why no other species has discovered such a good trick. I argue that verbal transmission is only likely to be adaptive in a restricted range of circumstances. Even then, it cannot be exclusively relied on, and it causes problems of deceit and instances of maladaptation. We should expect natural selection to have made us discriminating evaluators of verbal information who ultimately trust the evidence of our senses. Nonetheless, once language has become widespread, it can increase human adaptability, by increasing the efficiency of individual learning.

The computer simulation of language change in a finite, structured population which was presented in an earlier paper ('Using Social Impact Theory to simulate language change', Lingua 108, 95-117, 1999), is here extended to speech communities of different sizes. On the basis of the results it is proposed (a) that language change may be faster in small communities; (b) that linguistic borrowing is one sense more likely in small communities; and (c) that the evolution of linguistically marked structures is more likely in small communities. It is argued that these three generalisations could be used to make sense of the different patterns of linguistic diversity observed in the Old and New Worlds, and the distribution of marked word orders in the world's languages.

The Americas harbor a very great diversity of indigenous language stocks, many more than are found in any other continent. J. Nichols [(1990) Language 66, 475-521] has argued that this diversity indicates a great time depth of in situ evolution. She thus infers that the colonization of the Americas must have begun around 35,000 years ago. This estimate is much earlier than the date for which there is strong archaeological support, which does not much exceed 12,000 years. Nichols' assumption is that the diversity of linguistic stocks increases linearly with time. This paper compares the major continents of the world to show that this assumption is not correct. In fact, stock diversity is highest in the Americas, which are by consensus the youngest continents, intermediate in Australia and New Guinea, and lowest in Africa and Eurasia where the time depth is greatest. If anything, then, after an initial radiation, stock diversity decreases with time. A simple model is outlined that predicts these dynamics. It assumes that early in the peopling of continents, there are many unfilled niches for communities to live in, and so fissioning into new lineages is frequent. As the habitat is filled up, the rate of fissioning declines and lineage extinction becomes the dominant evolutionary force.

This paper presents a framework for simulating language change in social networks derived from Social Impact Theory. In this framework, the language learner samples the speech of individuals from right across his speech community, though he may weight their input differentially according to their social position. This conceptualisation is argued to be more realistic than that provided by other models. Computer simulations are used to investigate the effects on language change of different social structures and biases in language acquisition. From the results of these simulations, it is argued that the fundamental engine driving language change is the combination of inherent variation in language acquisition and differences between individuals in local social influence. Functional biases attaching to different linguistic variants influence the direction of language change.

Synergetic models of language structure predict that the length of a word will depend upon various parameters such as its frequency and the number of phonemes in the language. This prediction has been used to explain word length differences within languages, but less often to explain the differences between languages. Here I show that average word length across 12 West African languages is related to the size of the phonological inventory. This is an apparent example of the adaptation of language structure to the efficient communication of information. The hypothesised mechanism by which the relationship evolves are outlined.

The six and a half thousand languages spoken by humankind are very unevenly distributed across the globe. Language diversity generally increases as one moves from the poles toward the equator and is very low in arid environments. Two belts of extremely high language diversity can be identified. One runs through West and Central Africa, while the other covers South and South-East Asia and the Pacific. Most of the world's languages are found in these two areas. This paper attempts to explain aspects of the global distribution of language diversity. It is proposed that a key factor influencing it has been climatic variability. Where the climate allows continuous food production throughout the year, small groups of people can be reliably self-sufficient and so populations fragment into many small languages. Where the variability of the climate is greater, the size of social network necessary for reliable subsistence is larger, and so languages tend to be more widespread. A regression analysis relating the number of languages spoken in the major tropical countries to the variability of their climates is performed and the results support the hypothesis. The geographical patterning of languages has, however, begun to be destroyed by the spread of Eurasian diseases, Eurasian people, and the world economy.

Analysis of a linguistic atlas reveals an ecological gradient in the diversity of languages in West Africa. As one moves south from arid into lusher ecoclimatic zones, the average size of ethnolinguistic groups decreases. Various factors are considered which may have contributed to this distribution. I argue that the ethnolinguistic map is primarily a reflection of the systems of generalized exchange and mutual dependence into which people enter. It is hypothesized that such social networks function to reduce subsistence risk due to variations in the food supply. If this hypothesis is correct, the average size of ethnolinguistic groups should be inversely proportional to the degree of ecological variability they face. This prediction is tested and found to hold strongly for a large part of West Africa. There is also limited evidence of a correlation between linguistic diversity and topography. It is concluded that ecological risk has been a key historical force in West Africa and that the ethnolinguistic mosaic can be used as a valuable 'fossil record' of people's adaptive social and economic strategies.