Everett, Blasi & Roberts (2015) review literature on how inhaling dry air affects phonation, suggesting that lexical tone is harder to produce and perceive in dry environments. This leads to a prediction that languages should adapt to this pressure, so that lexical tone should not be found in dry climates, and the paper presents statistical evidence in favour of this prediction.
Below are some frequently asked questions about the study (see also the previous blog post explaining the statistics).
- What about Swedish / Tibeto-Burman languages / Language X?
- Why did you do the fancy statistics?
- What's the prediction about historical change? Do humid languages gain tone or dry languages lose tone?
- Why did you break languages down into two tonal categories, complex and non-complex categories?
- All languages use pitch, is there evidence that lexical tone especially affected?
- How did you assign a humidity value to each language?
- What about languages spoken over large areas?
- Are tone languages harder to learn in dry countries?
- What are the implications for linguistics?
- Does this theory apply only to tone, or to other aspects of language, too?
- Does the climate affect other aspects of culture, like singing?
There are certainly exceptions to the prediction. But we're not expecting a hard and fast rule, just a statistical tendency. We were aware of such exceptions before even embarking on the analysis of the database. Our account is not simply deterministic, but suggestive of gradual pressures operating at the same time as other pressures known to impact the evolution of sound systems. Occasionally such influences may even be at odds. We note in the SI, however, that even in a language family like Tibeto-Burman, in which there are exceptions, the overall pattern holds in the predicted direction. Tibeto-Burman languages spoken in more desiccated regions are less likely to employ complex tone.
In the paper, we used Monte Carlo tests (explained in this post), rather than more traditional methods such as regression. We did this because we predicted that languages in dry areas would be less likely to have lexical tone, but this says nothing about the distribution of languages in humid places. That is, we're only expecting the effect in dry areas, not across the board. In this case, regression is not suitable (see Blasi & Roberts, submitted). Another aspect that was addressed by the Monte Carlo tests is the relatedness of languages.
Note that we also did straightforward tests where possible. Our analysis of isolates is one such example, as are the intracontinental and intrafamilial analyses in the supporting information. These results are consistent with those of the "fancy" MC tests that more carefully control for potential confounds.
Languages lose and gain tone over time. We don't yet know whether this is supported by historical change. Our expectation is that languages moving into dry areas will be less likely to gain tone contrasts in the first place, rather than dry air leading to loss of tone or humid air leading to the adoption of tone.
This facilitated our analysis, particularly regarding the Monte Carlo tests. But note that we also tested the association without this binning strategy, by examining languages according to number of tonal contrasts (see the supplementary materials) within families and within continents. Again, the results of these tests were consistent with the hypothesis.
We recognize that all languages use pitch contrasts for various purposes, often pragmatic. We note this in the paper. Furthermore, we recognize that tonal contrasts (in languages that have them) are in many cases not simply pitch-based but rely on other factors such as laryngealization. This point is also noted in the paper. We also recognized that F0 modulation can be as extreme in non-tonal languages as in tonal ones. None of these factors contravene our guiding assumption--one well buttressed by the linguistic literature: tonal languages, particularly those with complex tone, require that a generally higher burden be placed on the maintenance of precise pitch patterns in order to contrast meaning. Nevertheless, we do not rule out the possibility that the effects of desiccation may impact other kinds of pitch patterns observed in many languages. We do not claim in the paper that desiccation only impacts tone, though it may well be given the special semantic load of pitch in tonal languages. However any other predictions associated with such pitch patterns, were they fully developed, would be impossible to test with extant databases. We stress that our prediction is simply and directly motivated by the laryngology data we survey in the paper.
The humidity of each language is the mean humidity over the year for the language's geographic location.
We used monthly diagnostic above ground specific humidity data from NCEP/NCAR 40-Year Reanalysis Project. This lists the mean monthly specific humidity for points around the earth from about 1960 to 2014. For each geographic point, we calculated the mean over all years for each month, then took the mean of these 12 values as the overall mean specific humidity. In the linguistic database, each language is assigned a geographic location based on the historical or socio-political center. For example, English is located in London and Mandarin in Beijing. We linked each language location to the closest mean specific humidity value.
We assume the climate of each language can be approximated by a single point. Some global languages (English, Mandarin, Spanish etc.), are currently spoken over a wide range of climates.
However, while there are a few languages spoken by many people, the vast majority of languages have fewer than 10,000 speakers. This makes it more reasonable to assume a single point.
Our paper does not address this question. This is an interesting question, but may not be a straightforward prediction of the theory. There are two possible questions here - learning a native language (e.g. from birth, "L1 acquisition") and learning a language later in life ("L2 acquisition"). First, we'll address L1 acquisition.
There are two main mechanisms by which language can change. The first is through differences between the language that adults speak and the language that children learn. Languages must adapt to be learnable by children (Christiansen & Chater, XXXX). In this case, if dry air affects speaking in a way that makes it more difficult for children to learn a particular sound contrast, then it might affect how the language changes over time.
However, we think this is not very likely (and very difficult to test). We were thinking more in terms of change on an utterance-by-utterance level. Contrasts that are hard to produce or perceive would be less likely to be reproduced from conversation to conversation (rather than generation to generation). It's therefore possible that the selective pressure of dry environments could apply and lead to the patters we see, without there being effects on language learning. In other words, language learning by children could be completely unaffected by climate, but we might still expect the effects we see.
The second type of learning is adult language learning. This case might be different, because adults find learning distinctions in sounds harder overall. L2 learning is also sensitive to psychological aspects such as confidence and motivation. So if sounds are harder to produce or perceive due to dry air, adult learners may find them harder to learn. In theory, this is testable by looking at learning performance over a range of climates. However, with a difference in climate comes a difference in culture, socioeconomic status, motivation and so on, which would complicate the answer.
When we look at the world's languages, we see a lot of variation. Some aspects, like lexical tone, can seem completely alien to speakers of many European languages. Similarly, the variable stress patterns of languages like English can seem strange to speakers of other languages. However, rather than seeing these differences between languages as odd or due to chance, our research suggests that languages might actually be well adapted to their environments. That is, each language is apparently well designed for the communicative needs of its speakers.
The idea that climate can affect culture is not particularly new in the field of anthropology. However, linguistics has a tradition of researching effects on language that humans share universally, and assuming that every language has essentially the same selective pressures acting on it (see however, the literature on linguistic relativity). We suggest that it's possible to research the differences in language based on climatic differences (what one might call 'geophonetics').
We concentrated on lexical tone, because it relies on control of speech sounds that are particularly affected by humidity. However, in principle there are many other aspects of the sounds of a language that could be affected. Also, the sounds of a language can, in principle, have a knock-on effect on other parts of language like morphology or syntax. We hope to investigate these possibilities in the future.
The general theory might predict differences in music or singing syles. However, there are differences in the function of singing to language. Singing is often performative, while language is communicative. In this case, there may be less pressure on singing to adapt to the environment. In fact, performative aspects of culture may deliberately go against selective pressures on language. For example, poetical styles are often have complex rules, going against a general pressure for simplicity and efficiency in communication.