Those of you who have read this weblog for a while know that ASPM is one of the genes that was once a major topic of interest. But the 2000s turned into the 2010s, and I kind of lost interest. There was some really strange result though that ASPM and tonal languages had some association. But there are all sorts of weird correlations. Nevertheless, in 2012 a study of Europeans found those with a particular ASPM allele were better at recognizing particular tones.
Curiouser and Curioser. Now in Science, a new study of Cantonese speakers, ASPM-lexical tone association in speakers of a tone language: Direct evidence for the genetic-biasing hypothesis of language evolution:
How language has evolved into more than 7000 varieties today remains a question that puzzles linguists, anthropologists, and evolutionary scientists. The genetic-biasing hypothesis of language evolution postulates that genes and language features coevolve, such that a population that is genetically predisposed to perceiving a particular linguistic feature would tend to adopt that feature in their language. Statistical studies that correlated a large number of genetic variants and linguistic features not only generated this hypothesis but also specifically pinpointed a linkage between ASPM and lexical tone. However, there is currently no direct evidence for this association and, therefore, the hypothesis. In an experimental study, we provide evidence to link ASPM with lexical tone perception in a sample of over 400 speakers of a tone language. In addition to providing the first direct evidence for the genetic-biasing hypothesis, our results have implications for further studies of linguistic anthropology and language disorders.
One thing that needs to be made clear, and is obvious in the above figure: the SNP within ASPM is statistically associated with a better perception of tones but has a smaller impact than IQ, and in particular, musicality. In other words, ASPM is not the “tone gene.” It’s just a gene that has a nontrivial (detectable) effect on one’s ability to recognize tones.
On the first reading of this paper I thought back to Terrence Deacon’s The Symbolic Species: The Co-evolution of Language and the Brain. Deacon takes a different tack that Steven Pinker in The Blank Slate, emphasizing that language emerges out of a gradual co-evolutionary dynamic, rather than being a distinct cognitive module that is “purpose-built” for the task. Deacon (and, to be fair Pinker and the Chomskyite tradition in general) emphasizes that the nature of linguistic phenomena is constrained and shaped by our neuro-cognitive architecture.
The authors seem to think their result has evolutionary implications:
In conclusion, we found direct evidence of gene-tone association, providing the critical direct evidence for the genetic-biasing hypothesis of language evolution. We hypothesize that ASPM is expressed in the pitch center of the auditory cortex (20), which would, in turn, enhance the lexical tone perception of carriers of the favored allele. Populations with a higher frequency of the favored allele would be more likely to have lexical tone in their language.
Al well and good. But we need to take a step back, as I did after reading the paper. First, the tone-enhancing T allele is the major allele in humans, and seems to be the ancestral one. In fact, the T allele is at a much higher frequency in Africa than it does in East Asia. You can confirm this in the 1000 Genomes Browser, but I also generated some frequencies from the HGDP:
| CLST | A1 | A2 | MAF | MAC | NCHROBS |
| Kalash | C | T | 0.64 | 28 | 44 |
| PapuanSepik | C | T | 0.63 | 10 | 16 |
| Palestinian | C | T | 0.53 | 48 | 90 |
| Druze | C | T | 0.50 | 41 | 82 |
| PapuanHighlands | C | T | 0.50 | 9 | 18 |
| Sardinian | C | T | 0.48 | 27 | 56 |
| French | C | T | 0.48 | 26 | 54 |
| BergamoItalian | C | T | 0.42 | 10 | 24 |
| Sindhi | C | T | 0.42 | 20 | 48 |
| Adygei | C | T | 0.41 | 13 | 32 |
| Orcadian | C | T | 0.40 | 12 | 30 |
| Russian | C | T | 0.40 | 20 | 50 |
| Pathan | C | T | 0.40 | 19 | 48 |
| Bedouin | C | T | 0.39 | 36 | 92 |
| Tuscan | C | T | 0.38 | 6 | 16 |
| Brahui | C | T | 0.36 | 18 | 50 |
| Makrani | C | T | 0.32 | 16 | 50 |
| Basque | C | T | 0.32 | 14 | 44 |
| Mozabite | C | T | 0.31 | 17 | 54 |
| Uygur | C | T | 0.30 | 6 | 20 |
| Balochi | C | T | 0.29 | 14 | 48 |
| Burusho | C | T | 0.27 | 13 | 48 |
| She | C | T | 0.25 | 5 | 20 |
| Yi | C | T | 0.25 | 5 | 20 |
| Han | C | T | 0.24 | 16 | 66 |
| Dai | C | T | 0.22 | 4 | 18 |
| Tu | C | T | 0.22 | 4 | 18 |
| Hazara | C | T | 0.18 | 7 | 38 |
| Maya | C | T | 0.17 | 7 | 42 |
| NorthernHan | C | T | 0.15 | 3 | 20 |
| Bougainville | C | T | 0.14 | 3 | 22 |
| Japanese | C | T | 0.13 | 7 | 54 |
| Naxi | C | T | 0.13 | 2 | 16 |
| Yakut | C | T | 0.12 | 6 | 50 |
| Tujia | C | T | 0.11 | 2 | 18 |
| Miao | C | T | 0.10 | 2 | 20 |
| Biaka | C | T | 0.09 | 4 | 44 |
| Mbuti | C | T | 0.08 | 2 | 26 |
| Lahu | C | T | 0.06 | 1 | 16 |
| Daur | C | T | 0.06 | 1 | 18 |
| Hezhen | C | T | 0.06 | 1 | 18 |
| Mongolian | C | T | 0.06 | 1 | 18 |
| Oroqen | C | T | 0.06 | 1 | 18 |
| BantuKenya | C | T | 0.05 | 1 | 22 |
| Pima | C | T | 0.04 | 1 | 26 |
| Mandenka | C | T | 0.02 | 1 | 44 |
| Yoruba | C | T | 0.02 | 1 | 44 |
| BantuSouthAfrica | C | T | 0.00 | 0 | 16 |
| Cambodian | C | T | 0.00 | 0 | 18 |
| Colombian | C | T | 0.00 | 0 | 14 |
| Karitiana | C | T | 0.00 | 0 | 24 |
| San | C | T | 0.00 | 0 | 12 |
| Surui | C | T | 0.00 | 0 | 16 |
| Xibo | C | T | 0.00 | 0 | 18 |
Basically, the C allele is a West/South Eurasian allele. But, it’s also found at low frequencies within Africa. It could be that the C and T allele have been “balanced” in some way in human populations for al ong time. Who knows, perhaps the C allele is benefit, or hitchhiked with something else? It’s interesting that the Karitiana and Surui don’t even carry the C allele.
This is a functionally interesting gene, but I’m not sure what the SNP tells us about tonal vs. non-tonal in a deep way.
I doubt it since tones, like morphology, can be lost or gained in a language family.
The case for deep ancestral human languages being tonal is strong, but the analysis of the paper which focuses primarily on linguistic adaptation, as opposed to co-evolution in general of hearing to the climate and terrain of a local environment, is weak.
Another theory of tone languages focuses on the environment.
Tone languages (and especially those with complex tone systems as opposed to simple ones) tend to be more vastly common in places that with tropical climates and very rare elsewhere. The locations have temperatures and humidities that influence sound transmission through the air, and have terrain influences (e.g. tree density) that impact how far away you would need words you speak to carry best. See https://wals.info/feature/13A#2/19.3/152.8 So, one theory is that tone languages arise in places where the sound transmission qualities of the air and terrain favor them.
Tonality appears to be a part of a total phoneme set for a language and should be looked at as part of an evaluation of points like a languages consonant and vowel and click inventory. Tone appears to be more common in languages with a smaller amount of other phonemes, and basically substitutes for additional vowels. There have been suggestions in the literature that the local climate and ecology can make certain phoneme sets better in some places than in others, that the nature of one part of a phoneme set influences the nature of other parts of the phoneme set, and that there are specific non-random factors that favor particular subtypes of phonemes in particular conditions.
An environmental explanation is supported by the observation that tonality in language seems to be more of an areal effect than one that tracks language families. For example, within the Afroasiatic language family, tonal languages appear in the Omotic, Chadic, and Cushitic branches of Afroasiatic (the Southern tier of Afroasiatic languages, mostly in Ethiopia and the African Sahel), according to Ehret (1996), but the Semitic, Berber, and Egyptian branches do not use tones phonemically. There is a fair amount of circumstantial evidence, when you look at patterns of semantic tone use globally in all sorts of languages, to suggest that tonality is more of an areal feature than it is an indicator of the ancestral source of a language. Neighboring languages that come from different families often share the feature of semantic tonality, while languages within the same language family often differ in their use of semantic tonality.
Incidentally, the geographic distribution of languages with tone systems is similar, although not identical, to the geographic distribution of languages with glottal consonants. Both are most common in sub-Saharan Africa, Southeast Asia, and the subtropical and tropical regions of the Americas (although the Americas are far from uniform despite all except the Na-Dene and Inuit language families probably having a common ancestor ca. 14kya). But, the Chinese dialect family uses tone, while it does not utilize glottal consonants. As used in this sense, tone is the term used to describe the use of pitch patterns to distinguish individual words or the grammatical forms of words, such as the singular and plural forms of nouns or different tenses of verbs.
It could be that the ancestral type ASPM gene “tunes” ones hearing system to better distinguish sounds in a certain pitch range, in general, in places that that the temperatures, humidities and terrain most conducive to tonal languages, while the derived type ASPM gene loosens to focus of the hearing system so that it isn’t so primed to maximizing hearing of sounds in particular set of conditions, which would be adaptive elsewhere.
The case that your ASPM variant enhances fitness primarily by making your hearing system better adapted to your primary environment makes more sense to me in an evolutionary selective fitness sense. If people with the region appropriate variant hear subtle slight sound differences better than people who lack it, that could increase the ability of a hunter-gatherer to locate prey, to detect predators, to locate lost children who have wandered far away, to hear your enemies coming to get you, to detect a fire that has gotten out of control or something that you are standing on that is about to break, and cumulatively, that could produce a gradual, put persistent selective fitness advantage in the evolutionary sense.
I find it harder to believe that the tone language specific application of this trait would have much of a selective fitness effect. An inability to distinguish by sound alone two words that would both make contextual sense in a tonal language that you and the speaker share might tweak one’s social status in the community a little, but it seems less likely to have a big impact on mortality or lifetime reproductive success. It’s not impossible, but it would seem like a weaker explanation.
Against this backdrop, the natural question to ask is one that wouldn’t otherwise be obvious, which is “why aren’t there more tonal language in South Asia?” which has substantial linguistic diversity and climate features in part of the region that are very similar to places in Africa, Southeast and East Asia, and the Americas where tonal languages are predominant.
One partial answer to this is that the Indo-Aryan languages developed in places that did not have this climate and didn’t spontaneously pick up this feature upon arriving in the subcontinent. But, this doesn’t explain why we don’t see tonal Munda and Dravidian languages in South Asia. The urheimat of the Austroasiatic languages of which the Munda languages are a family member, is Southeast Asia, which the vast majority of languages are tonal. And, the Dravidian languages, as far as anyone knowns are autochthonous in South Asia.
In both cases, I think that the likely explanation is a language learner effect.
The Munda languages, at least initially, seem to have had a fairly northerly distribution within South Asia where hearing well suited to tonality wouldn’t have been advantageous to the local people who probably accounted for all or most of the women in the community at the time of first contact when the Munda languages would have been adopted by people integrated into the early Munda communities. If half the people had trouble hearing the tones, that feature which was almost surely present in an ancestral pre-Munda language probably didn’t survive.
In the case of the Dravidian languages, which also probably have an ancestral version that was tonal under the environmental hypothesis, the pertinent fact is that there are no meaningful communities in Dravidian India that do not have substantial ANI admixture dating to the last 2000-3500 years. The language learner affect at the time of ANI-ASI admixture could have stripped the Dravidian languages in existence at the time of their tonal features for the same reasons. The ubiquity of the Hindu religion in Dravidian India which has clear Indo-Aryan and Harappan synthesis origins, likewise suggests that the language learners were not just anybody, they were culturally influential elites whose language choices tend to influence whole communities by cultural imitation.
How would pitch-accent languages fit into any of these scenarios?
See e.g., https://en.wikipedia.org/wiki/Pitch-accent_language
“How would pitch-accent languages fit into any of these scenarios?”
WALS Online appears to classify pitch-accent languages as “simple tone systems.” Or, at least, pitch-accent systems have a very similar geographic distribution to pitch-accent systems and the distinction between the two may be largely a subtle and subjective one.