Natasha Abner, Grégoire Clarté, Carlo Geraci, Robin J. Ryder, Justine Mertz, Anah Salgat, and Shi Yu have a paper in Science (1 Feb 2024, Vol. 383, Issue 6682, pp. 519-523; DOI: 10.1126/science.add7766) that studies family structure among sign languages; the abstract:
Sign languages are naturally occurring languages. As such, their emergence and spread reflect the histories of their communities. However, limitations in historical recordkeeping and linguistic documentation have hindered the diachronic analysis of sign languages. In this work, we used computational phylogenetic methods to study family structure among 19 sign languages from deaf communities worldwide. We used phonologically coded lexical data from contemporary languages to infer relatedness and suggest that these methods can help study regular form changes in sign languages. The inferred trees are consistent in key respects with known historical information but challenge certain assumed groupings and surpass analyses made available by traditional methods. Moreover, the phylogenetic inferences are not reducible to geographic distribution but do affirm the importance of geopolitical forces in the histories of human languages.
In their conclusion, they say “most notably, we found a closer relationship between the Western European sign languages and British and New Zealand SL than has been previously assumed and present a Western European sign language family tree that better reflects the broad scope of influence of French SL.” I don’t know enough to have any idea whether they’ve done a good job (although previous experiences with phylogenetic analysis, e.g., have jaundiced me); I hope better-informed Hatters will be able to say more. Thanks, Y!
I knew NZ SL is regarded as ‘close’ to British, and quite different to U.S. I’m surprised the survey didn’t include Australian, to validate their methodology: BANZSL is a sort of ‘dialect continuum’ — if I might abuse terminology.
Fig 1: In what way do European SLs fall closer to Brit/NZ than to U.S.? And why is Italian an outlier?
Conclusion
This seems weak. There’s plenty of spoken languages with no historical records. We can still detect phylogenetic relatedness.
The limitations of traditional historical linguistic analysis are well documented
“Well documented” indeed, and thus a feature, not a bug. With “traditional” historical linguistics you can see where the facts do not permit you to go any further while still having any claim to rigour.
With “phylogenetic methods”, this irritating “limitation” appears to have been removed. The limitations are now poorly documented (and buried deep in the opaque methodology.)
A competent and honest doctor will admit that they cannot always arrive at a firm diagnosis (no matter how important it might be to do so.)
A doctor who can always give a firm diagnosis is thus either incompetent or dishonest or both. (Also probably rich and on television a lot.)
Well said, and pretty much what I suspected.
To be fair, polytomies and low posterior probabilities are the method admitting the lack of a firm diagnosis on something, even if careless human writers chaperoning it might not admit this in the prose (conscientious ones even can and will run a few alternate analyses to also see and present what tree topology options such results are competing with).
Echoing what I noted at Twxttxr, I actually find here the tree dating more immediately suspicious than its topology: the French / Italian and French / American splits somehow end up at just about the absolute top that their calibration ranges allow, the British / NZ at the bottom. Seems that the actual differences are not in line with what the calibration might suggest; e.g. perhaps there’s been later convergence for either of the former two pairs, or perhaps NZ does not descend from standard British after all but from a longer-extant dialect variant. Could any of this be checked or caught with historical data? I have no idea how far back any sign languages are well-documented. [edit] Come to think of it, I also don’t know how well is the dialectology of sign languages known and I wouldn’t be surprized if something like “British SL” was actually somewhat heterogeneous, and just assumed to be homogeneous by some lingering influence of the “well it’s British English but signed” folk misconception. “These people are Brits / living in Britain, so the language they speak must be British SL”?
I suspect the systematic differences could have major impact on this: sign languages lack regular phonological change (as far as I’ve seen!) which makes detecting loans more difficult; they’re heavier on iconicity which makes accidental convergence a much bigger risk; lack of any long-running historical records at all means we have no starting hypotheses on what is core vocabulary or what are likely directions of change.
Oh, open access.
Fig. 1 is pretty close to “all language relationships are either immediately obvious or forever unknowable”. The paper is best regarded as a proof of concept – worthy of Science for that reason, not for its results.
So they’re taking a language as bag-of-words approach, and measuring relatedness of signing for a Swadesh list [Methodology, first para]. No comparison of syntax nor morphology.
Am I right in thinking that applying such an approach for English would find it’s closer related to French than Dutch?
@JP they’re heavier on iconicity which makes accidental convergence a much bigger risk;
They’ve excluded body parts from their Swadesh list, on grounds of likelihood of iconicity.
Not if you go by basic vocabulary. All the language-phylogeny-with-words-only papers have contained English and French in their datasets, and none found them together.
But see above on not actually knowing what is and is not basic vocabulary in sign languages.
Right, and it’s a good move, but in general this just means less data to worth with, not that there should be also other concepts for balance which end up more stable in sign languages than spoken languages. Seems to me that the bottom line is that sign languages simply evolve faster and routinely end up mutually unintelligible in like one century.
I wonder if is it primarily faster in phonological or faster in lexical development though? With handsign space being way bigger than phoneme space and sign languages’ segmental word length being way shorter, it might be hard to even tell the difference without detailed historical records.
Echoing and amplifying David Eddyshaw’s remark on the limitations of traditional historical linguistics being a feature, not a bug: one of its “limitations” is that, EVEN WHEN vocabulary alone is compared, the normal operating procedure is to remove from comparison borrowed words: thanks to the core discovery which made historical linguistics a science, namely the regularity of sound change, we can detect, even in the case of languages with no written tradition, or in the case of periods before the first written records were produced, borrowed elements. Thus we know that English “path” (and its cognates throughout Germanic) is a borrowed word and “mother” is a native English word (despite “path” having been borrowed before any Germanic language was written).
Because we have no way of telling whether any number of signs in a given sign language are indigenous or borrowed, any attempt to apply the comparative method to sign languages must be undertaken with extreme caution, and it should go without saying that none of the historical conclusions drawn from such a study should be taken too seriously. Especially since the “scholars” producing such “comparative” work can rest easy, knowing that, because sign languages have never been represented in writing, no archeological discovery is liable to demonstrate that their conclusions are nonsense (Again, this is unlike Classical historical linguistics: the discovery of Tocharian, for instance, showed that the scholars who had assumed the centum/satam split had originally divided Indo-European along a West/East axis were unambiguously wrong).
It’s pretty common for linguistic phylogeny papers to claim they can do something that traditional historical linguistics can’t. But I have never seen HL explicitly (and ignorantly) attacked as it is here:
AHEM! So the Comparative Method is “hindered” when working on marginalized and unwritten languages? Historical linguistics of such languages, has existed for, IDK, 150 years?
The reference is to Dunn, Terrill, et al. (2005), which use typological characters, not vocabulary, to classify Papuan languages. Aside from that their claims are unverifiable and unfalsifiable, their method has nothing to do with this paper’s.
That said, since I don’t know anything about historical linguistics of sign languages, I glanced at three references mentioned (Power; Reagan; and an article in the Routledge Handbook of Historical Linguistics). The main challenge I see for applying the comparative method to SLs is that no one has come up with an easy SL equivalent to segments or phonemes, and thus equivalents to sound correspondences and sound laws. I don’t see that anyone has tried really hard to find them, either. Most of the research on SL HL seems to be on language change from a sociolinguistic perspective, or on non-phonological phenomena: lexical borrowing, grammatical change, etc.
P.S. Ref. 4 is a detailed and enlightening discussion on deaf vs. Deaf, here.
That isn’t quite true, but close enough for this purpose.
which use typological characters, not vocabulary, to classify Papuan languages
I knew it! Kusaal is more closely related to Swedish than to Swahili! *
Welsh is more closely related to Arabic than to Latin!
Good grief. Where have these people been while scientific historical linguistics was being developed and practiced? This is basically astrology.
The actual paper is not open access, but the summary tells you all you need to know …
https://www.science.org/doi/10.1126/science.1114615
These are people who can’t take “We don’t actually know” for an answer. You know: non-scientists (at least when dabbling in this domain.)
I’m surprised at Angela Terrill, who wrote a very nice grammar of Lavukaleve. Non omnia possumus omnes, I suppose …
* Scandi-Congo Vindicated by Science! (Technically unnecessary, as it has of course been proven already by the finest Chomskyan reasoning known to humanity. And that’s the best reasoning. Oh yes.)
Like probably all commenters here I endorse the radical-to-some idea that deaf communities should foreground their SL as L1 and background the local oral language as L2. However, the most radical extension of that, refusing to use oral language at all, would seem to render a deaf community illiterate, at least until such time as technology allows auto translation from written~oral languages to CGI SL.
Yoruba is obviously a Sinitic language, and thus a close relative of Thai and Goemai.
I expect Zheng He had something to do with it all.
@mollymooly:
Lots of communities routinely use a different language for writing from the one they use for normal communication, without prejudice to which one they regard as their L1. (You could perhaps even make an argument along the lines that this has historically been the norm rather than the exception.)
DE: you can read Dunn et al. here for free, gratis, and at no charge.
Foley and Reesink are also experienced descriptive linguists working on Papua languages. I suppose they and Terrill were in charge of the data, Dunn and Levinson in charge of the sausage-making.
Ta, Y.
It doesn’t seem to be that Foley. This one seems to be a moonlighting geneticist, come to Solve Historical Linguistics by Science, the way they do.
The Real Linguists involved (who you might have hoped would know better) have probably been driven mad by their efforts to do Papuan comparative linguistics properly over the years, and are perhaps more to be pitied than censured. Any of us might begin to develop uncontrollable twitches after that kind of experience.
I don’t judge those who were tempted by This One Weird Trick Classifes Papuan Languages (Try It Tonight).
we have no way of telling whether any number of signs in a given sign language are indigenous or borrowed
[…]
…because sign languages have never been represented in writing
neither of these claims makes much sense to me.
on the first one: even without having done any systematic study, i know that there’s quite a bit of documented history of at least some sign languages, quite a lot of accumulated teaching materials for many of them, and the same kind of range of ages among their speakers as for any other language. which seems quite likely to make it possible to trace histories of borrowings over at least the past century. i’ve heard accounts of other kinds of language change in ASL, just from knowing Deaf folks; people who use a language (and even more so, native speakers who do language teaching) are generally good sources for changes in its vocabulary, as damn near every thread on here shows. but that depends on actually taking Deaf people seriously as sources of knowledge, rather than as mines of raw material to throw into an algorithmic GIGO box.
on the second: just plain not true. again, without any systematic study – just from knowing some Deaf folks and ASL interpreters and having learned a bit about the theatrical work of companies like the National Theater of the Deaf – i’ve seen teaching materials and transcriptions of ASL produced over at least a fifty-year period. sure, there isn’t a standard writing system for ASL, because as DE said, ASL signers generally read and write in english (the analogies to various similar uses of persian, classical arabic, mandarin, and hebrew are obvious). but there’s plenty of representation of ASL in writing, using a wide variety of approaches, some of which aspire to linguistic rigor. Y pointed out some of what’s been lacking in the scholarly research that contains some of that material; again, it seems to me that the basic problem remains disrespect for the people who use sign languages (just as it is, for example, with the various partial or just plain badly-made accounts of many indigenous oral languages of the americas that have been produced over the past few hundred years).
And there was desking of heads and palming of faces.
Oh, it’s much worse. These are people who are like “we want to do this research; it would have to build on other research that hasn’t been done yet because it’s too boring to get funded; so we’ll just come up with something that vaguely looks like a proxy, pretend it is the research our research has to build on, and then just do our research”.
That is widespread – and highly encouraged by that whole “publish or perish” thing: if you don’t publish something breathtakingly groundbreaking often enough and early enough, good luck finding a job ever.
I will say that for computational phylogenetics: their standards of graphics are much, much higher than those of historical linguistics. Those beautiful (if sometimes meaningless) graphs were unimaginable in linguistics pre-2000.
I need to put more effort into my pictures of trees and waves.
More foliage. More spume.
if you don’t publish something breathtakingly groundbreaking often enough and early enough, good luck finding a job ever.
Quite. Or in general, once the truth of some set of scientific results is well-established, scientists (or “scientists”) must, in order to survive, strive to establish falsehoods instead. From this follows all the usual methods, from baffle-’em-with-bullshit to domain imperialism.
What is domain imperialism?
I take JC to mean roughly ‘the attempt to impose the methods or models of a specific disciplinary domain as the interdisciplinary norm”.
What physicists, economists and psychologists have been trying to do for decades, essentially, having been joined in this game more recently by ‘data scientists’.
Near enough, yes.