Linguists recognize ecological, socio-economic, and historical factors apparently contributing to the answer. Language diversity—e.g., the number of native languages per 1,000 square miles of area—correlates with numerous potentially explanatory factors, but these factors are in turn correlated with each other. Hence one has to resort to statistical methods, such as multiple regression analysis, to tease out which factors have primary effects actually causing language diversity to be high or low, and which other factors have just apparent effects mediated by their correlations with those primary factors. For example, there is a positive correlation between Rolls-Royce car ownership and lifespan: Rolls-Royce owners live on the average longer than do people who don’t own Rolls-Royces. That’s not because Rolls-Royce ownership directly improves survival, but because Rolls-Royce owners tend to have lots of money, which enables them to pay for the best health care, which is the actual cause of their long lifespans. When it comes, though, to correlates of linguistic diversity, there isn’t yet a corresponding agreement about the actual underlying causes.
The four closest ecological correlations of language diversity are with latitude, climate variability, biological productivity, and local ecological diversity. First, language diversity decreases from the equator towards the poles: all other things being equal, tropical areas hold more languages than do equivalent areas at higher latitudes. Second, at a given latitude language diversity decreases with climate variability, whether the variability consists of regular within-year seasonal variation or of unpredictable between-year variation. For instance, language diversity is higher in tropical rainforests that are wet all year round than in adjacent more seasonal tropical savannahs. (This factor of seasonality could account at least in part, through the correlation between latitude and seasonality, for the higher language diversities in the less seasonal tropics than in strongly seasonal high latitudes.) Third, language diversity tends to be higher in more productive environments (e.g., higher in rainforests than in deserts), though again at least some of that effect could be because of a tendency for deserts and many other unproductive environments to be strongly seasonal. Finally, language diversity is high in ecologically diverse areas and tends especially to be higher in rugged mountainous areas than in flat areas.
These four ecological relationships are just correlations, not explanations. Suggested underlying explanations involve human population size, mobility, and economic strategies. First, a speech community’s viability increases with its number of people: a language spoken by only 50 people is more likely to disappear, due to its speakers all dying or abandoning their language, than is a language spoken by 5,000 people. Hence regions with a lower biological productivity (supporting fewer people) tend to support fewer languages, and to require more area for the speakers of each language. A viable population in Arctic or desert regions needs tens of thousands of square miles to support itself, while a few hundred square miles would be ample in productive landscapes. Second, the more constant is the environment between seasons and between years, the more self-sufficient and sedentary can a speech community be within a small area, without much need to move periodically or to trade for necessities with other peoples. Finally, an ecologically diverse area can support many different language communities, each with its own specific subsistence economy adapted to a different local ecology: for instance, a mountainous area can support mountain herders, hill farmers, lowland river fishermen, and lowland savannah pastoralists at different elevations and in different habitats.
Ecological factors thus already tell us several reasons why small New Guinea has 5–10 times more languages than does huge Russia, Canada, or China. New Guinea lies within a few degrees of the equator, so its people experience only slight variations in climate. The New Guinea landscape is wet, fertile, and productive. New Guineans don’t move much or at all with seasons or from year to year; they can meet all of their subsistence needs within a small area; and they don’t have to trade except for salt, stone for tools, and luxuries like shells and feathers. New Guinea is rugged and ecologically diverse, with mountains up to 16,500 feet, rivers, lakes, seacoasts, savannahs, and forests. One could object that China and Canada have even higher mountains and offer a larger range of elevations than does New Guinea. But New Guinea’s tropical location means that New Guineans can live year-round and farm at high population densities up to elevations of 8,000 feet, while high elevations in China and Canada are seasonally freezing and support only low human population densities (in Tibet) or no people at all.
In addition to those ecological factors, there are also socio-economic and historical factors contributing to differences in language diversity around the world. One such factor is that hunter-gatherer speech communities consist of fewer individuals but may cover larger areas than farmer speech communities. For instance, Aboriginal Australia was traditionally inhabited entirely by hunter-gatherers occupying an average of 12,000 square miles per language, while neighboring New Guinea supported mostly farmers occupying only 300 square miles per language. Within Indonesian New Guinea, I worked in areas supporting nearby both farmers (in the Central Highlands) and hunter-gatherers (in the Lakes Plains), with about two dozen languages for each lifestyle. The average hunter-gatherer language there had only 388 speakers, while the average farmer language had 18,241 speakers. The main reason for the small speech communities of hunter-gatherers is low food availability, hence low human population densities. Within the same environment, population densities of hunter-gatherers are 10 to 100 times lower than those of farmers, because much less food is available to hunter-gatherers, able to eat only that tiny fraction of wild plant species that is edible, than to farmers, who convert the landscape into gardens and orchards of edible plants.
A second socio-economic factor related to language diversity is political organization: language diversity decreases, and language communities increase in population and in area, with increasing political complexity from bands to states. For instance, the United States today, a large state with a single dominant coast-to-coast language, has a population about 30 times what was the population of the entire world at a time when the world still consisted entirely of hunter-gatherer bands and tribes with thousands of languages. The dominant U.S. language of English has largely replaced the hundreds of different local languages formerly spoken five centuries ago in what is now the national territory of the U.S. when it was divided among Native American bands, tribes, and chiefdoms. Underlying this trend is the fact, discussed in the Prologue, that increasing political complexity becomes necessary as a society increases in population—because a society of a few dozen people can make decisions in a group meeting without a leader, but a society of millions requires leaders and bureaucrats to operate. States expand their own languages at the expense of the languages of conquered and incorporated groups. That language expansion is partly a matter of state policy for the purposes of administration and national unity, and partly a spontaneous matter of individual citizens adopting the national language in order to obtain economic and social opportunities for themselves.