Ecology and phonology
« previous post | next post »
Ian Maddieson and Christophe Coupé, "Human spoken language diversity and the acoustic adaptation hyothesis", ASA 2015
Bioacousticians have argued that ecological feedback mechanisms contribute to shaping the acoustic signals of a variety of species and anthropogenic changes in soundscapes have been shown to generate modifications to the spectral envelope of bird songs. Several studies posit that part of the variation in sound structure across spoken human languages could likewise reflect adaptation to the local ecological conditions of their use. Specifically, environments in which higher frequencies are less faithfully transmitted (such as denser vegetation or higher ambient temperatures) may favor greater use of sounds characterized by lower frequencies. Such languages are viewed as “more sonorous”. This paper presents a variety of tests of this hypothesis.
Data on segment inventories and syllable structure is taken from LAPSyD, a database on phonological patterns of a large worldwide sample of languages. Correlations are examined with measures of temperature, precipitation, vegetation, and geomorphology reflecting the mean values for the area in which each language is traditionally spoken. Major world languages, typically spoken across a range of environments, are excluded. Several comparisons show a correlation between ecological factors and the ratio of sonorant to obstruent segments in the languages examined offering support for the idea that acoustic adaptation applies to human languages.
No doubt there'll be a more compete publication soon, but so far, all we have is that abstract for a presentation at the recent meeting of the Acoustical Society of American in Jacksonville FL.
The linguistic data is from the Lyon-Albuquerque Phonological Systems Database. I'm not sure where the "measures of temperature, precipitation, vegetation, and geomorphology" come from, but presumably in the full publication, the authors will provide that dataset as an appendix, or point to its online location.
There's been a quite a bit of mass-media pick-up, e.g.
Kiona Smith-Strickland, "Languages Are Products of Their Environments", Discover 11/4/2015
Zoë Schlanger, "Differences of Human Languages Driven by Climate and Environment", Newsweek 11/4/2015
Emily Underwood, "Human language may be shaped by climate and terrain", Science 11/4/2015
"Environment and climate helped shape varied evolution of human languages", phys.org 11/4/2015
"Climate, environment influenced language evolution", Business Standard 11/5/2015
"The environmental link to language evolution", Language)( magazine 11/5/2015
Angus Chen, "Did The Language You Speak Evolve Because Of The Heat?", NPR 11/6/2015
There is clearly serious work behind that abstract, several qualitative steps up from the recent claim that the evolution of Australian English was shaped by drunkenness. (Though I can foresee some downward steps as the tabloids get hold of it…)
But in the case of correlational studies like this, I worry about the fact that geographical diffusion means that the true number of degrees of freedom in such datasets is much smaller than it might seem to be. For some concrete expressions of this concern, see e.g.
"Cultural diffusion and the Whorfian hypothesis", 2/12/2012
"High-altitude ejectives", 6/14/2013
"Linguistic diversity and traffic accidents", 8/15/2013
Maddieson and Coupé may have allowed for this problem, and their results might hold up even if they didn't. But we'll have to wait for a fuller report to find out for sure.
Update — See also Sean Roberts, "The effect of climate on individual speech: Does Larry King use more vowels when it's warmer?", a replicated typo 11/7/2015.
Daniel E Johnson said,
November 7, 2015 @ 7:07 pm
I'm surprised this post doesn't make any reference to the climate/tone controversy of early 2015.
[(myl) Are you referring to Caleb Everett, Damián Blasi, and Seán Roberts, "Climate, vocal folds, and tonal languages: Connecting the physiological and geographic dots", PNAS 10/24/2014? (See also this FAQ by Roberts.) I'm not aware of any serious controversy about that case.]
Y said,
November 7, 2015 @ 11:06 pm
myl: Do you mean Everett et al. is generally accepted, or generally not accepted?
[(myl) I mean that I haven't seen any significant argument against accepting the results of Everett et al. 2014. And in the other direction, it's meaningful that one of the co-authors, Sean Roberts, was previously the source of serious and well-founded skepticism about studies of this kind, e.g. Sean Roberts & James Winters, "Linguistic Diversity and Traffic Accidents: Lessons from Statistical Studies of Cultural Traits", PLOS ONE 2013, and "Uncovering spurious correlations between language and culture", a replicated typo 8/15/2013.]
Daniel E Johnson said,
November 9, 2015 @ 9:46 am
http://arstechnica.com/science/2015/06/louder-vowels-wont-get-you-laid-and-other-tales-of-spurious-correlation-2/4
http://dejonedge.blogspot.com/2015/01/my-first-post-publication-review.html
I assume that Bob Ladd is adjudged "serious" or "significant", even if I'm not myself.
[(myl) I don't understand. The Ars Technica post doesn't mention Everett, Blasi & Roberts (2014), though it does mention Everett (2013) — which Roberts criticized in the work I cited. And Bob Ladd is mentioned there as a proponent of nomothetic research, not a critic of it. Your post is mildly (and I think appropriately) critical of Everett et al. (2014):
In summary, I found the argument for the geographic correlation itself to be fairly strong, although I did not really look into the details here. The link between the proposed phonetic effect and language change was plausible, but needed more grounding in research on language change in general and the loss of tone in particular. But I was less convinced that the physiological (or phonetic) effects of dry and cold air are really an obstacle to producing phonological tone. Like Everett et al., I too hope "that experimental phoneticians and others examine the effects of ambient air conditions on the production of tones and other sound patterns, so that we can better understand this pivotal way in which human sound systems appear to be ecologically adaptive."
But I wouldn't describe this as "the climate/tone controversy of early 2015" — you seem to be largely in agreement with the paper, in fact maybe more convinced than I am.]
Y said,
November 11, 2015 @ 1:35 am
Everett's works on tones and humidity and on ejectives and elevation, and perhaps also Maddieson and Coupé's recent work, suffer from several common issues. First, control for independence. In Maddieson's map as shown on the NPR site, there are four areal clusters of high-vowel languages: equatorial Africa, PNG, Australia and Amazonia. For Africa, most of the languages are Bantu or Niger-Congo, an areal if not genetic cluster. Australia, PNG and Amazonia are uncontroversially areal clusters, though each contains many language families not known to be related; these clusters are clearly evident through even by non-phonological characteristics alone. I don't know how M&C account for genetic and areal factors. Everett only accounts for genetic factors.
Second, the mechanisms aren't convincing. Maddieson and Coupé appeal to the effects of communication through thick vegetation; but habitation in all these areas are in clearings, and communication is carried on at short distances, as elsewhere. Everett et al. appeal to an elevation effect to explain ejectives, though the actual spread of the languages they study is mostly up to modest elevations (<1000 m), where any such effect is minimal. The synchronic effect of dry air on tone production could be easily confirmed in consultation with speech therapists, say of Chinese in Tibet, but they have not done so. As in the other two papers, I think they appeal to extreme and atypical situations.
Lastly, all these studies take no account of the time dimension. If the environment exerts a pressure toward loss/gain of tone or consonant clusters, the effect should be noticeable within language families which have spread over a large range of climates in the not-so recent past. But the phonologies of Rajasthani, Assamese and Kashmiri are not so different, nor are those of Northern Paiute and Pipil, etc.
Andreea Calude said,
November 11, 2015 @ 3:47 pm
Sorry I missed the link I was trying to give in my previous post:
http://twentytwowords.com/funny-graphs-show-correlation-between-completely-unrelated-stats-9-pictures/
Sean Roberts said,
November 13, 2015 @ 8:24 am
I'm very interested to see how Maddieson & Coupe control for relatedness – I only had the slides to go on, and I couldn't see an obvious control, though Coupe is adept at ecological analysis, so I would be surprised if there is NO control.
@Y: Everett et al. (2015) do account for areal relationships as well as genetic relationships (and we also cite a clinical paper which reports speech problems in tone language speakers with laryngeal problems). But your point about lack of evidence of diachronic change is quite right. We're run some preliminary phylogenetic analyses, but it's not straightforward (climate change, branch length of the trees etc.). Work in progress, I'm afraid.
There'll be a special issue in the new Journal of Language Evolution on climate and language, so perhaps the tone/climate controversy of 2015 is yet to come!
Y said,
November 13, 2015 @ 3:40 pm
Sean, I don't see how you control for areal relationships in Everett et al. (2015). Most of the old-world complex-tone languages in the map (fig. S1) come out of only four clusters: 1. (±humid) Niger-Congo languages and neighbors throughout Africa, with simple or no tone pepered throughout. 2. (arid) 'Khoisan' languages 3. (±humid) Sino-Tibetan and the SE Asian language area 4. Some languages of the (humid) PNG language area. So we have just three unquestionably independent humid complex-tone data points, balanced by three non-humid ones (Khoisan, Ket, Shakavian). In addition there are thousands of Austronesian, Australian and Papuan languages lacking complex tone. In the New World there are ('humid') Amazonian languages from 5 families showing CT, one (mostly humid) family of Mexico (Oto-Manguean), and 10 N. Am. languages belonging to 8 families, 4 from humid areas (Sm'algyax, Creek, Caddo, Tuscarora), and 4 arid ones (Kaska, Achomawi, Kiowa-Tanoan, Keres). In sum, when areal and genetic relations are taken into account, the number of truly independent samples is much smaller, and the correlation is not apparent (to me at least).
The physiological mechanism suggested calls for a fairly extreme condition, not a typical one. The study quoted (Hemler et al.) shows an effect on phonation of inhaling very dry air, but does not find a difference between phonation at normal room humidity and in humidified air. For the purposes of this study, this mechanism does not explain why tonal languages are more abundant in SE Asia than in Europe.