Ejectives, High Altitudes, and Grandiose Linguistic Hypotheses
[This post is collaboratively written by Martin W. Lewis and Asya Pereltsvaig]
As GeoCurrents has noted in several previous posts, leading scientific journals and influential media outlets often favor research in linguistics that makes strong claims that resonate with the general public. A number of these favored studies claim to find a correlation between a linguistic feature and a non-linguistic social or cultural trait. For example, a recent paper by Gay, Santacreu-Vasut and Shoham from the Berkeley economic history laboratory claims to have found a correlation between linguistic gender systems and female economic and political empowerment: women in countries with languages that make gender distinctions are supposedly less likely to participate in the labor market or politics and have a reduced ability to get credit or own land. An earlier paper by Boroditsky, Schmidt & Phillips (2003) reported a link between linguistic gender and attributes ascribed to various inanimate objects. Yet another paper, this one by behavioral economist Keith Chen from Yale University, highlights a correlation between tense marking and financial behavior: people whose native language makes fewer distinctions between the future and present purportedly think differently about the future and therefore make different financial decisions. These and similar studies go back to the strong version of the Whorf-Sapir hypothesis, all but discredited by professional linguists, which states that the language one speaks determines how one thinks.* Many such alleged correlations between the linguistic and social realms can be, and have been, explored; as Mark Liberman notes in his LanguageLog post,
“many relevant linguistic and non-linguistic datasets are now pre-compiled and available for easy download, and the software needed for fitting various sorts of statistical models can easily be run on your laptop. So if you have a bright idea — maybe alcohol consumption correlates with phonotactic complexity? really, it could — the chances are that you test a model within a few hours. If it doesn’t work out, there are plenty more to try — maybe coffee consumption helps to preserve morphological inflection?”
Another related strain of research seeks connections between linguistic attributes and physical geography, based on the idea that certain kinds of terrain or climate favor certain structural features in languages. A number of Asya Pereltsvaig’s earlier Languages of the World posts explored several such hypotheses. One hypothesis explored there is that rich systems of case marking expressing fine spatial distinctions are related to the complex topography of the mountainous landscape that the speakers of these languages inhabit; for example, many languages of the Caucasus have particularly complex case systems. Some other scholars have proposed that the presence of nasal vowels correlates with cold and damp climates, noting that Modern French, based on the dialect of allegedly cold and damp Paris, has nasal vowels, whereas Spanish and Italian, spoken in warmer and drier climes, do not. As Asya discussed in detail in those posts, such correlations, while seemingly plausible for a small set of languages, fail to apply if a global language sample is used. For example, as illustrated in the WALS-based map posted on the left, many languages with nasal vowels are found in area that as not cold and damp, including West Africa, Dagestan (Hunzib), northern Pakistan (Burushaski), and northern India (Hindi, Mundari).**
A recent paper by anthropologist Caleb Everett published in PLOS ONE, “Evidence for Direct Geographic Influences on Linguistic Sounds: The Case of Ejectives”, dodges the challenge of cross-linguistic typology by basing its claims “that the geographic context in which a language is spoken may directly impact its phonological form” on a large-scope typological study considering the 567 languages in the WALS sample pertaining to glottalized consonants. The specific correlation supposedly found by Everett involves a relatively rare type of sound called ejectives. Unlike “plain” stop sounds, such as [p], [t], or [k], pronounced with a closure in the mouth, ejective stops involve an additional closure of the glottis (the space between vocal folds), which creates the dramatic burst of air when the oral closure is released, giving ejective sounds a certain “spat out” quality:***
Ejective sounds are found in 92 of the 567 languages in the WALS sample (see map on the left). Perhaps the best-known examples of languages with these sounds come from the Caucasus region, between the Black and Caspian seas. Languages from all four families limited to the Caucasus feature ejectives, including Abkhaz (Northwest Caucasian family), Ingush (Nakh family), Dido (Northeast Caucasian family), and Georgian (South Caucasian, or Kartvelian, family). Ejectives are also found in languages of non-Caucasian families that are spoken in the region, most notably (some dialects of) Armenian and Ossetian. Outside the Caucasus region, ejective sounds are heard most often in Athabascan, Siouan, and Salishan languages of North America; in Aymara and southern varieties of Quechua, spoken in the Andes; in Amharic, one of the major languages of Ethiopia; in Hadza and Sandawe, two Khoisan languages spoken in Tanzania; in Khoisan**** languages of southern Africa; and in Itelmen, an endangered language spoken in Kamchatka. Ejectives were also used for the constructed language Na’vi, the language of the aliens in the film Avatar.
A number of these ejective-using languages, including those of the Andes, the Caucasus, and the Ethiopian Plateau, happen to be spoken in areas of relatively high elevation. According to Everett, this linkage is systematic:
“Languages with phonemic ejective consonants were found to occur closer to inhabitable regions of high elevation, when contrasted to languages without this class of sounds. In addition, the mean and median elevations of the locations of languages with ejectives were found to be comparatively high.”
Everett also argues that these patterns “surface on all major world landmasses”, and are not a result of the influence of particular language families. He concludes by specifying “a significant and positive worldwide correlation between elevation and the likelihood that a language employs ejective phonemes”. Everett proposes two “plausible motivations for the correlation”, thus suggesting that the implication works both ways: the presence of ejectives implies higher elevation AND higher elevation implies the presence of ejectives. One of Everett’s explanations is that:
“ejective sounds might be facilitated at higher elevations due to the associated decrease in ambient air pressure, which reduces the physiological effort required for the compression of air in the pharyngeal cavity–a unique articulatory component of ejective sounds.”
Everett second hypothesis is that “ejective sounds may help to mitigate rates of water vapor loss through exhaled air”, based on the fact that high elevation areas are often characterized by dry air. His first proposed explanation—that it is easier to pronoun ejective sounds under conditions of low air pressure—would lead to a prediction that (most) languages with ejectives would be found in areas of high altitude. His second theory—that ejective sounds are a biological adaptation to pervasive water stress in high altitudes—would lead to a prediction that many if not most languages spoken in lofty elevations would feature ejectives.
Everett claims that these predictions are generally borne out by his analysis. He does, however, hedge this claim to a considerable extent by noting that many languages with ejectives are spoken not in by rather near areas of high altitude, a feature particularly prevalent in North America’s Pacific Northwest. He argues that the correlation still obtains, however, because the lowland peoples living in these areas traditionally spent a considerable amount of time engaged in subsistence activities in adjacent high-elevation zones, and hence would have experienced conditions conducive to the development and use of ejective sounds.
In actuality, this particular thesis is spurious. According to Everett’s calculations, “the force required to produce [an idealized ejective] gesture at 2500 m would be roughly 26% […] less than the force required at sea level”. Yet in the Pacific Northwest of North America, areas at this elevation are ice-covered or support at best meager alpine tundra, and hence were seldom frequented by indigenous people. From southern Alaska to western Washington—one of the world’s major “ejective zones”—the connection between altitude and the presence of these distinctive sounds simply does obtain.
But is the correlation of ejectives with high-altitude languages found elsewhere in the world? Our analysis suggests otherwise. To illustrate the general lack of correspondence, we have taken the WALS data on the locations language with ejectives (given as “dots” placed on the spatial center of each given language) from four parts of the world (Ethiopia, southern and eastern Africa, South America, and the US) and overlain them on elevation maps.
As can be seen, most of the languages with ejectives (marked by white dots) are spoken in areas below the 2,500 meters mark. Even if we place the cut-off for “high altitude zones” at a much lower level, such as 1,500 meters (where the force differential is much less than 26%), many “ejective languages” still miss the mark.
Quite a few of these tongues are found in lowland areas, many of which are quite far removed from any lofty ranges. In southern Africa, languages with ejectives actually cluster in zones of moderate elevation. In Ethiopia and environs, half of the relevant languages are found off the high plateau. In South America, more languages with ejectives are found in the lowlands than in the highlands.
It could be argued that languages with ejective sounds developed in highland zones but subsequently relocated to areas of more modest elevation as their speakers migrated. But the same argument can also be applied to the reverse situation. Consider, for example, the Caucasus, the world’s most “ejective-rich” environment. Although the Caucasus, as a cultural-linguistic region, includes many highland areas, it also encompasses large expanses of low elevation, both to the south and the north of the Great Caucasus Range. Significantly, languages with ejectives are found in both the Caucasian highlands and lowlands. Several prominent linguists, most notably Joanna Nichols, have argued that the general historical tendency has been for languages of the north-Caucasus lowlands to move, under pressure from newcomers, into the higher-elevation zones of the south. As ejectives are found in many languages of the Caucasus regardless of the elevation at which they are or were previously found, it is difficult to view these sounds as an altitude-linked phenomenon. To the extent that ejectives correlate with altitude, this can also be a result of pre-modern population movements between areas of similar elevation; similar patterns of migration into areas of similar climate are well-attested in historical record.
Another objection to Everett’s proposal lies with the fact that most known and documented cases of a language acquiring ejectives occur via linguistic contact with an ejective-using language: like other types of sounds, ejectives can be “borrowed” through an intake of a sufficiently large number of vocabulary items that contain such sounds in the source language. Through this process Ossetian acquired ejective sounds from neighboring Caucasian languages, according to Ossetian linguist Vaso Abaev. Similarly, it has been argued that southern varieties of Quechua, particularly the Cuzco and Bolivian dialects, acquired ejectives via vocabulary borrowing from Aymara; northern varieties of Quechua, which have not been in contact with Aymara, do not have ejectives.
Other instances of a language developing distinctive (i.e. phonemic) glottalization, linguistically similar to ejectives, not only involve no change in altitude but occurred in areas of remarkably low altitude. One such case involves Danish, which has a phonemic feature called stød in traditional Danish grammar (literally, ‘push; thrust’). In phonetic terms, stød refers to laryngealization or glottalization. This phenomenon is found in most dialects of Danish (shown in pink in the Wikipedia map on the left) and in standard Danish—all of which are spoken in very low-lying areas. Our second example of a glottalic phenomenon found in a sea-level environment pertains to Estuary English, spoken in South East England, especially along the River Thames and its estuary. In this variety of English, the first element of a consonant cluster, such as [t] in bottle, is replaced by a glottal stop (similar phenomenon is found in Cockney as well). Though neither Danish nor Estuary English have ejectives in the strictest sense of the term, these phenomena are sufficiently similar to ejectives in their articulatory physiology that one would expect them to pattern with ejectives with respect to altitude, if Everett’s explanations are on the right track.
There is an additional catch: as Everett himself notes, his account is open to an objection concerning the acoustics of ejectives: while “the aforementioned lower pressure differential would in theory make ejectives easier to produce”, it also makes them “less perceptually salient at higher altitudes”. Everett’s counter-argument runs as follows:
“Given that one of the key acoustic characteristics of ejectives is their impact on the acoustic structure of adjacent vowels, it seems quite possible that they are preponderant at high altitudes due in part to articulatory ease, even though lower atmospheric pressure might reduce the salience of their associated burst of air.”
As Daniel Ezra Johnson in his comment on the LanguageLog post justly questions, “why don’t the people at sea level just make an ejective with 26% less compression, 26% less effort, 26% less air burst, and apparently still perfectly distinguishable effects on neighboring vowels?”
Let’s now turn to Everett’s second proposed explanation: that ejectives are a biological adaptation to high altitude in that they “may help to mitigate rates of water vapor loss through exhaled air”. As Everett notes, speech itself is costly in terms of water expenditure, and mountain climbers sometimes correspondingly minimize talking to conserve water. But alpinists engage in activities vastly more strenuous and demanding that those generally carried out by indigenous peoples of high-elevation zones. In actuality, water is readily available in most highland zones—much more so than in extreme deserts—and there is no evidence that most highlanders have ever been routinely subjected to water stress. As a result, evolutionary pressure to developed water-conserving ejective sounds would have generally been nil. As it turns out, few if any languages spoken in the world’s most extreme deserts, such as Tuareg, feature ejective sounds.
Perhaps the largest problem for the Everett hypothesis is the complete lack of ejective sounds in the WALS database in the world’s most extensive highland area, that of the Tibetan Plateau and its associated mountain ranges, including the Himalayas, the Pamir Range, the Hindu Kush, and the Tien Shan. This zone encompasses a vast range of distinctive highland environments, and it is populated by diverse groups of peoples speaking languages in a number of language families. If Everett’s theory accords with reality, one would then expect that ejective sounds would have emerged in a number of the languages spoken over this huge expanse of territory. Admittedly, Everett does mention the conspicuous lack of ejectives in the languages spoken on the Tibetan Plateau, but sidesteps the issue by noting that the Tibetans “have adapted to high altitude in distinct ways”, particularly by “breath[ing] at a faster rate than tested control populations”. Such an alternative adaptation to elevation, however, has not been documented among the large number of non-Tibetan peoples who live in high-elevation areas of central Asia—and whose languages likewise do not contain ejectives.
It appears that the correlation between absolute elevation and the presence of ejectives, needed to sustain Everett’s arguments, is weak at best. This supposed linkage is probably merely a spurious correlation of the sort discussed by Sean Roberts and James Winters in “Social Structure and Language Structure: The New Nomothetic Approach”. As Roberts notes in his ReplicatedTypo post, two other linguistic variables appear to be even more strongly correlated with altitude: Order of Object and Verb and the Relationship between the Order of Object and Verb and the Order of Adjective and Noun.
But maps of OV and VO languages, posted on the left, do not reveal any correlation between the order of object and verb and altitude: OV languages are spoken in the mountainous Andes of South America, but also in the lowland steppes of Central Asia; likewise, VO languages are spoken at both high and low altitudes. And if the correlations between these two word order patterns and altitude do actually exist, no physical or social explanation is available, suggesting that all three correlations are meaningless. As Mark Liberman points out,
“In every pair of datasets, for each variable in one of the datasets, we’ll see a distribution like those shown in the plots above, showing an especially strong statistical connection to a few of the variables in the other dataset. And sometimes these connections won’t make any sense — verb-object order and altitude, or velar nasals and savings rates, or lexical tone and acacia trees — while others will suggest a plausible causal story in one direction or the other”.
It remains to be seen whether the ejective-altitude connection belongs to the former or the latter category. For the time being, it is perhaps best to view this imaginative explanation as a mere just-so story, one that is superficially appealing, but which fails to withstand sustained scrutiny.
* A weaker version of the Whorf-Sapir hypothesis asserts that the language one speaks “gently nudges” one in the direction of thinking in certain habitual ways.
** The hypothesis that French has nasal vowels because of the cold and damp climate of Paris is discredited by other problems as well. One issue concerns the timing when French acquired nasal vowels, which is known from historical record to have happened by 1100, at the time of the Medieval Warm Period. Another issue goes to the core of the explanation for the alleged link: a cold and damp climate is said to cause habitual head colds which in turn are said to lead to nasality. Ironically, nasal congestion (whether caused by a cold, an allergy, or any other factor) leads to a lack of nasality, as the articulation of nasal sounds involving the air passing through the nasal cavity is made impossible by nasal congestion.
*** Ejective stops are by far the most common type of ejective sounds, but ejective affricates and fricatives, pronounced with a similar glottalic closure, are attested as well.
**** “Khoisan” is in all likelihood not a genuine language family, but rather a group of unrelated language families that share some prominent features (see Dimmendaal 2008).
Boroditsky, L., Schmidt, L., & Phillips, W. (2003) Sex, Syntax, and Semantics. In: Gentner & Goldin-Meadow (Eds.) Language in Mind: Advances in the study of Language and Cognition.
Chen, Keith (2013) The Effect of Language on Economic Behavior: Evidence from Savings Rates, Health Behaviors, and Retirement Assets. American Economic Review.
Dimmendaal, Gerrit J. (2008) Language Ecology and Linguistic Diversity on the African Continent. Language and Linguistics Compass 2(5): 840–858.
Gay, Victor; Estefania Santacreu-Vasut; and Amir Shoham (2013) The Grammatical Origins of Gender Roles. Berkeley Economic History Laboratory (BEHL) Working Papers. Online
Roberts, Sean and James Winters (2012) Social Structure and Language Structure: The New Nomothetic Approach. Psychology of Language and Communication 16(2): 89-112.
« The Growing Importance of the Arctic Council
Freedom of the Press Index for 2013—The... »