People keep sending me links to Nicholas Wade’s latest ill-informed NY Times blort about a linguistic topic, in this case based on Quentin Atkinson’s Science paper “Phonemic Diversity Supports a Serial Founder Effect Model of Language Expansion from Africa,” whose abstract says:
Human genetic and phenotypic diversity declines with distance from Africa, as predicted by a serial founder effect in which successive population bottlenecks during range expansion progressively reduce diversity, underpinning support for an African origin of modern humans. Recent work suggests that a similar founder effect may operate on human culture and language. Here I show that the number of phonemes used in a global sample of 504 languages is also clinal and fits a serial founder–effect model of expansion from an inferred origin in Africa. This result, which is not explained by more recent demographic history, local language diversity, or statistical non-independence within language families, points to parallel mechanisms shaping genetic and linguistic diversity and supports an African origin of modern human languages.
But I get tired of wading through, and then whaling on, the ever-out-of-his-depth Wade (see, e.g., here, here, and here), so I decided to wait until I could link to a decent analysis, and thanks to marie-lucie, I hereby present Richard Sproat‘s “Science Does It Again,” a thoughtful discussion that pokes at some important holes in the theory. I will quote his summary and let you read his (quite brief) review for the details: “Atkinson’s thesis is striking, but as I said above such striking conclusions require striking support, and I believe that the paper in its current form does not provide enough support.”
Quoth Sproat:
“The Plymouth Bay Colony did not all of a sudden speak a variety of English with fewer phonemes merely because they migrated to the New World. Had they remained an isolated and small colony for many generations, perhaps they might have ended up that way.”
In fact, though, North American English actually does have fewer phonemes than its relatives elsewhere after a mere four centuries. The LOT phoneme has been lost essentially everywhere, merging with THOUGHT or PALM or both. What is more, the LOT = THOUGHT = PALM varieties are in the West, which was settled last.
I don’t believe for a minute that that actually counts as a data point for Atkinson’s thesis; for one thing, we know that these are not founder effects. But the coincidence is odd.
Sproat has a brief comment to the NYT article (#20) giving the link to his review, and since then George Starostin has also provided a shorter but very pointed rebuttal (#61).
“Blort”?
It’s as solid as the DNA model it looks to — i.e., a completely unverifiable house of cards. A hypothesis that’s not falsifiable may be fun for drunken conversation on a slow Saturday afternoon, but it’s not science.
I didn’t send the link because I don’t hate you that much. I just said to myself “Wait till Hat sees this!”
“Blort”?
An expressive bit of phonesthetic lexification. Surely you can tell the intended attitude toward the linked item from the sound of it.
I like the NYT map showing isopleths of declining phonemic diversity. Some highlights: the Caucasus apparently has fewer phonemes than Europe, which in turn is uniform, unlike Somalia, which, though monolingual, has three bands of phonemic diversity in the northeast alone. Good crackpottedry.
Ben: Before rejecting (or accepting for that matter) out of hand, I would suggest reading the article itself and Sproat’s initial reaction. I am not qualified to comment on the methodology, but it is an interesting proposition and should get lots of attention.
I’m in no position to comment on the substance of Atkinson’s research, so let me offer some general thoughts. First, Nick Wade’s article is by no means a ringing endorsement of the work. He quotes linguists who are evidently skeptical of what Atkinson done, and in several places makes clear that the conclusions are suggestive but far from certain. My take, after reading Wade’s piece, is that this sounds like an interesting idea but more research is needed, as they say, to decide whether it really holds up or not.
Second, journals like Science and Nature are quite overtly aiming to publish research that’s in some measure provocative. If you wait until everything is cut and dried, and all the reviewers are happy, you will publish mostly dull, mediocre papers, and you will publish them late. It’s a mistake (a common one, unfortunately, even among people who ought to know better) to think that scientific publication is some kind of Good Housekeeping Seal of Approval. On the contrary, the reason to publish papers is put ideas into circulation and generate further analysis and discussion. By which measure, publishing Atkinson’s paper has been a success, at least in this community.
@Dale Favier: Is the methodology flawed? If not, what would account for the geographic correlations?
Maybe I just don’t know enough about the subject, but it seems like the whole analogy with genes is based on a pun between two different senses of the word “diversity”. To say that a language with more phonemes has greater “diversity of sounds” is a clever way of putting it, but the genetic analog would be, say, more distinct mtDNA nucleotides. (Note that a language with more phonemes is a language each of whose speakers has more phonemes.) Conversely, the linguistic analog to genetic diversity is linguistic diversity: more languages and language varieties within a population. No?
When I posted a link to Atkinson’s supplemental data in the Romeyka thread, I wondered if the paper will start a cycle of mass media simplification and a wholesale rejection by the linguists. Sure enough it did, in a nick of time.
I just want to point out that the Science paper is not as sweeping in its analysis as one might conclude from the mass media accounts … and that Sproat’s critique may be unduly sweeping. Of course I may have the usual biases of a molecular anthropolgist, doh.
The genetic diversity is different in that it is lost on an individual level, when none of the living members of a group succeeds in passing a particular variant to their surviving offspring. The loss may happen equally well when a very small group outmigrates, or when a far larger population stays isolated and stagnant (not growing) over a number of generations. Flows of genetic material between groups prevent a loss, as does fast population growth.
Phonemes would be much harder to loose in a relatively large but stagnant population IMHO. The outmigrating groups would need to be really small to result in such loss, but this might well have been the case in the loss enroute from SE Asia to Hawaii? So Sprout’s supposed counterexample may not be disproving anything.
Interflows between populations are well known to enrich phoneme sets (the best known example is probably Xhosa picking Khoisan clicks), so here again, Sprout’s counterargument (“they should be photemically languages in Africa if the loss over time is inevitable”) is invalid.
The glacial age migrations out of Africa to New Guinea and Australia must have followed the now-submerged coastlines of South Asia in a very short timeframe, and the speed of migration effectively shrunk the distances between these ancient human diversity havens. Hard to build any good objections because of this quirk of the map, either.
The distance effect on phoneme diversity is significant but far less pronounced than its effect on genetic diversity, clearly conforming to the linguists’ belief that the matters of culture are far more complex than the mixing of the gene pools; nothing to object here.
Most intruguingly perhaps, the relative richness of mass languages may be not so much due to their sheer size today, but due to their variety of local dialects in the past. The interacting dialects might have contributed to preservation of phonetic diversity with much greater ease than any interacting but mutually unintellegible languages (like Khoi-San and Bantu). And conversely, if an outmigration involved a dialect group shifting from interaction into isolation, then this might have resulted in reduction of diversity as well.
I think the latter hypothesis is where the linguistic data may be most useful for disproving or supporting the Atkinson hypothesis. Phonetic difference between extant dialects are studied in exhaustive details, and the effects of dialect interactions, isolation, and migration on phonetic diversity may actually be evaluated with the help of exosting data. ‘Cuz of course we can’t do anything about the phonemes of the prehistoric times no matter how much excited are the jornalists about those 🙂 !
I suppose that Mr Edison’s phonograph was the earliest recording device for spoken languages; before that reliance is presumably on inferences from writing. Consequently we know very little, and that uncertain.
– David L: journals like Science and Nature are quite overtly aiming to publish research that’s in some measure provocative
Sure, but they are science journals (in the everyday definition of science: physics, biology, etc), not linguistic journals, and they are not trying to get linguists as reviewers along with scientists. Phonemes are not comparable to genes. No linguist has proposed using the number of phonemes in a language as a true measure of overall language complexity: two otherwise very similar languages (or even dialects) can have markedly different phonemic inventories (eg Spanish and Portuguese, or even Northern and Southern French).
In more general terms, the evolution of a language is subject to very different factors from the evolution of a species: for instance, biological species can adapt to their physical environment, but they cannot interact and adopt features from each other, while this happens all the time under conditions of language contact (ie a speech community’s learning and adoption of some features of another language, loss of some features of its own language, and partial or total language replacement together with retention of a few irreducible features of the original language, etc).
In addition to Sproat, again I recommend George Starostin’s comment on the NYT article (see above).
– Courtesy of MOCKBA on the ROMEYKA thread, this is the link to the language data which are appended to the otherwise inaccessible Science article:
http://www.sciencemag.org/content/suppl/2011/04/13/332.6027.346.DC1/1199295-Atkinson-SOM.pdf
(MOCKBA, I did not know you were posting while I was writing my comment).
Interflows between populations are well known to enrich phoneme sets
Actually, the well-known cases are not typical of what most often happens in situations of language contact, which is replacement of the foreign phonemes by the perceived closest equivalents in the recipient language (as in “speaking with an accent”). Adoption of foreign phonemes as additions to the inventory of one’s own language is most likely when there is widespread bilingualism in a population over a period of time and the use of foreign pronunciations has some positive social significance.
The interacting dialects might have contributed to preservation of phonetic diversity
That depends on the relative prestige of the dialects, whether speakers emphasize their differences (in small groups of relatively equal power) or want to downplay them (eg when rural individuals from various regions move to a big city). In any case, phonetic diversity is not the same as phonemic diversity (eg the varied pronunciations of the single phoneme /r/ in European languages).
Marie-Lucie, bothy points are very convincing, and probably the Xhosa example is so well known exactly because it’s so contrary to expectations. Of course smaller children exposed to nonoverlapping phonemic sets may be able to adopt them all?
In any case, if the data analysis is really statistically sound, I’d be led to believe that the mechanisms of loss and maintenance of phonemic diversity must exist for real, and to wonder how to detect them using present-day data.
But I don’t know if the dataset doesn’t have any serious flaws, such as omission of less diverse African languages or more diverse SE Asian or Amerindian languages. 504 languages is a small subsets, and fluctuations (as well as conscious dataset biases) are possible. Atkinson’s hypothesis held if Oceania and America were excluded, but he didn’t check if exclusion of South or Central Africa will be OK.
Also, as it often happens in genomic studies, the p-values may lack correction for multitesting. So I think the real scrutiny should come not from the indignation, but from the analysis of the language set (how unbiased and representative it is?) and statistics.
GeorgeW: Fair point. I should have made it clear that by “crackpottedry” I was referring to the map itself, not the paper, since the map looks extremely scientistic and yet is clearly hokum (for the reasons I point out, among others).
I would be curious to see how any of these results change if you drop the khoisan languages out of these calculations. This is one cluster of languages with an extremely high number of phonemes, which also happen to be in the far end of Africa. If they are merely a fortuitous outlier, they could skew the correlation quite a bit.
For example, in figure S5 of the supplement, without the leftmost 3 data points I really don’t see any significant correlation.
@YM: However, if the outlier cluster on the other end of the scale were also dropped, the effect might be largely offset.
JC: North American English actually does have fewer phonemes than its relatives elsewhere after a mere four centuries.
North American English speakers are not all descended from a population explosion among the founding groups, they have absorbed a high percentage of people from a wide variety of linguistic backgrounds, most of which did not have as many vowel contrasts as English (which still has more vowels than most languages).
Contrast this situation with the one in francophone Canada, where until recently population growth was within the group (high birth rates) and there was very little immigration from other francophone countries, let alone switch from English to French: Canadian French varieties are very conservative, it is European French (especially in cities) which is rapidly losing phonemic diversity because of both in-migration and immigration.
I have to agree with David L’s point: theories do not have to be completely and rigorously established to be published. In one sense that’s obviously true, since by their nature, theories can be overturned by new data.
That said, there surely should be a bare minimum of support for a theory, before one can publish them, especially in self-styled “prestigious” publications like Science.
My, and others’, critique of Atkinson, at its core, comes down to this: that the population/phoneme-set-size correlation discovered by Hay and Bauer has no known explanation; that it has not been demonstrated to hold for population ranges plausible for the Paleolithic (and indeed Atkinson’s own data suggest there would be no correlation); that there is no directionality to the correlation; and that therefore it is mere speculation to propose that serial founder effects involving this mechanism are at work in explaining distributions of phoneme systems today.
The size-complexity relationship could be simply due to the way that when expanding languages are learned by non-native speakers you usually have creolization (simplification), whereas when a language is only spoken by native speakers and as a mother tongue, those simplifications don’t happen and complexification can take place.
To the extent that this happens, it could wipe out the whole argument. Creolization happens very quickly, whereas complexification is presumably slower. But if complexification actually happens, and I’ve been assured that it does, then the contrast between two languages, one simpler and one more complex, could simply be the result of two historical events (one expansion and one isolation) rather than evidence that one language is older than the other.
I invented a toy model: a Japanese trade empire establishes a coastal enclave in Borneo. A Japanese based creole comes into being. The Japanese leave permanently, and the creole develops on its own with no further contact with Japanese. Then the coastal creoles are driven into a valley in the interior by a new empire, where they live in substantial isolation for centuries and complexify. After a certain point, it might have no creole qualities but seem ancient, and it would probably have only tenuous similarities to Japanese.
We mostly think on the historical scale of 1-3 thousand years, occasionally maybe 5,000. But when you look at prehistory, which this theory does, the scales are at least 10 times longer (history is 5-10,000 years, human prehistory is 50-100,000 years.) So our models of language change may be too cramped for the time scale of this theory.
Richard, it is unfortunate that the reviewers at Science didn’t point their attention to the obvious lack of correlation between the current population sizes (and the current scale of their differences) and the prehistorical population size distributions which would need to postulated for the Atkinsonian founder-effect hypothesis. The author might have addressed the issue if it has been raised in time…
I get your point, that the outsize differences between populations are so recent that they ought to be immaterial for any truly ancient language-evolution theory. I begin to suspect that the two empirical correlations (with the “distance covered” vs. the number of speakers) may be completely unrelated to one another. If, of course, the correlations are really as good as they appear to be in this (likely biased) subset of 504 languages.
@ John, serial creolization may be an alternative mechanism of loss of phonemic diversity upon outmigration. If the non-core groups of speakers of creolized languages are the ones moving out, then they will carry simplified phonemic systems with them. There are a number of historic examples of migrations disproportionately carried out by the members of the non-core subethnic groups. Basques, Moriscos and Crypto-Jews out of Spain, Cossacks into Siberia…
Another interesting difference between the language dynamics of present times vs. Paleolitic population spread is that the latter involved mostly nomadic or semi-nomadic groups. There are examples of recent semi-nomadic sweeps into new territories with mixed populations, resulting in fluid linguistic patterns. (Cossacks and Teptyari in the Great Steppe, Komi and Nenets sweeps in formerly Sami European subarctics, Hazara tribal consolidation in Afghanistan…) Do we know how these migrations affect phonemic diversity?
On a longer time scale, but not a lot, are the Turkish, Bantu, Arabic, and Malayo-Polynesian expansions. Even Persian and Aramaic were expansion languages, as I understand. I have no knowledge of the creolization or otherwise of these languages.
A propos MOCKBA’s point. So just to be clear, I *was* one of the reviewers for Science, and I made exactly this point in my review. Apparently that wasn’t convincing enough.
As to whether Atkinson addressed this in the final published version: not having seen the final published version, I can’t say. But I don’t see, frankly, how he could have addressed it.
Wow Richard, that’s a twist of fate! I wish I could double check the final version, but I am heading for a week-long break and won’t have the benefit of institutional subscription until I’m back.
Are there any ways to resolve the “size range issue”? If you notice I played with explaining it away, using a [fairly unconvincing] hypothesis about dialect richness and interdialectic exchanges. Better than nothing LOL.
Anyway, a little anecdote, apropos. Once I got a very paradigm-smashing publication into the very same Science. Also in a field which made most colleagues cringe, like, no more use than counting devils on a needlepoint. Well, years later, the paradigm is alive and well, my paper is pretty much forgotten, but … the field is blooming.
Liberman’s analysis on the Language Log
http://languagelog.ldc.upenn.edu/nll/?p=3090
has further very useful analysis of the phoneme-diversity issue.
Another interesting difference between the language dynamics of present times vs. Paleolitic population spread is that the latter involved mostly nomadic or semi-nomadic groups. There are examples of recent semi-nomadic sweeps into new territories with mixed populations, resulting in fluid linguistic patterns.
Movement into uninhabited areas can be very fast, too. People talk about the vast distances of Eurasia making diffusion slow, but Paris and Beijing are 5000 miles apart. An individual could easily walk that distance in a few years (3 years at 5 miles per day, about 14 years at one mile a day).
Slow diffusion is the result either of inhospitable terrain (mountains, deserts, swamps, and wide rivers) or else of resistance by people already there. These reinforce, since the most hospitable environments will also be the most settled. But absolute distance really isn’t even a factor. (The discomforts of travelling across Siberia make people overestimate the difficulty).
I don’t have an accurate number, but during the Mongol Empires during times of peace, the Mongol pony express could almost certainly go from the Crimea to Beijing in 2 or 3 months.
This may seem like a simple point, but I’ve found that it’s not intuitive to everyone.
Sorry for jumping in late, but this is just too interesting…
Marie-Lucie: the phonological history of French over the past few centuries on both sides of the Atlantic is somewhat more complex than you make it seem. Canadian French, by the eighteenth century, was phonologically neither more nor less complex than the upper-class Parisian standard.
Outside of Paris, however, many varieties of French born through dialect contact (especially in urban settings) were more complex than either. The recent (twentieth century) reduction in vowel phonemes in Europe is an innovation which I am not convinced is directly due to language shift to French on the part of many new speakers. My main problem is the timing: such shift had taken place from the late eighteenth to the early twentieth century without standard (Parisian) French phonology being affected.
John Emerson: creolization does indeed entail radical simplification of morphology, but phonology is by comparison relatively unscathed: thus most French Creoles have preserved the French consonant phoneme inventory intact, and indeed many creoles have borrowed new phonemes from their substrate languages (thus, many West Indian Creoles have such West African phonemes as phonemic tone, pre-nasalized and coarticulated stops).
Contact thus seems to be a two-edged sword: while language shift *can* indeed lead to the loss of (non-central) phonemic distinctions, dialect/language contact can also lead to the borrowing of new phonemes and the birth of contact varieties with more phonemes than the contributing dialects/languages. In this light Mockba’s comment, to the effect that language spreads are often initiated by “non-core subethnic groups”, is quite relevant.
Unfortunately we typically know all too little about the linguistic features of the various languages spoken by members of these nomadic/multilingual groups.
Etienne: I was talking about recent developments.
Canadian French, by the eighteenth century, was phonologically neither more nor less complex than the upper-class Parisian standard.
But three centuries later it is definitely more conservative.
Outside of Paris, however, many varieties of French born through dialect contact (especially in urban settings) were more complex than either.
If you look at rural dialects of Northern France, they are usually more complex than urban varieties. See for instance the survey work of Henriette Walter.
The recent (twentieth century) reduction in vowel phonemes in Europe is an innovation which I am not convinced is directly due to language shift to French on the part of many new speakers.
In-migration from rural areas to big cities (especially Paris) causes a switch to a more homogenized (and phonologically poorer) version of French on the part of French speakers from different regions, and especially their children.
My main problem is the timing: such shift had taken place from the late eighteenth to the early twentieth century without standard (Parisian) French phonology being affected.
“Standard” was defined by pre-WWI phoneticians as le parler de la haute bourgeoisie parisienne, but language evolution is rarely driven by the upper class. To a French person, l’accent parisien does not refer to the pronunciation of the upper class, but to that of uneducated Parisians, especially in the NE quadrant of Paris and the adjacent suburbs (a rough equivalent of “cockney” in England). But present-day educated Parisians (which include many people whose parents and grandparents were from the provinces, including those with non-French substrates – Breton, Occitan, etc – as well as those from other countries) are no longer speaking the way older phoneticians described the “standard”.
In any case, I was referring to recent phonetic and phonological evolution, which I have observed in my own lifetime: since I have not lived in France for many years, I have not gradually gotten used to changes in speech, so I find them quite striking when I encounter them even among my own relatives. I am not a Parisian, but I used to recognize most features of my own pronunciation (at least when deliberate) in older descriptions of “Standard” French phonology, and current descriptions do not fit my speech, which is now quite old-fashioned.
I too havo noticed a phonetic change in my own lifetime: the Northern Cities Vowel shift, which I can’t describe linguistically, but pronounces “bag” to rhyme with “vague”. I don’t remember it from growing up in Minnesota, but after leaving and returning 40 years later, I hear it now, and it sounds odd to me.
Well, a little digging, and problems in Atkinson surface. In the online materials, there is the sentence “However, recent work has shown that elements of language, such as commonly used words (S19) and ome phonemes (S20) can be highly stable, some persisting for tens of millennia (S19).” S19 is “M. Pagel, Q. D. Atkinson, A. Meade, Nature 449, 717”, and on fetching this article I find that it is about the rates of lexical evolution (replacement of words in Swadesh-200) in IE.
What this previous article shows is a set of projected half-lives of lexical items before they are lexically replaced; these range from 750 years for words replaced most often to a whopping 76,530 years for words never replaced at all in the 87-language sample. Of course this is a preposterous artifact of the mathematics, since the work assumes a conventional age for IE of 8700 years. By the same token, the isotope bismuth-209 has a half-life of about twenty billion billion years; that does not mean that atoms of it predate the observable universe. So the slippage from mathematical half-lives to actual lives happens between the Nature paper and the Science one.
Now of course this is not a direct traverse of Atkinson’s current article, but “it is like the thirteenth stroke of a crazy clock, which not only is itself discredited but casts a shade of doubt over all previous assertions.” (A.P. Herbert)
“What is more, the LOT = THOUGHT = PALM varieties are in the West, which was settled last.”
How far west do you have to go to find this? LOT != THOUGHT/PALM in my speech (Bay Area, ancestors settled during the Gold rush form upstate NY).
Do you mean LA (“the west coast of Iowa”, actually the west coast of Germany, see M-L’s comment above. Or do you mean Hawai’i, the east coast of Japan?)
Complexification – Salishan and Athapaskan languages all have a poop-load more phonemic complexity that Korean or Japanese or Mandarin or Mongolian, on and on, for no particular reason that I have ever heard of, including distance from Africa.
Diversity – Peak diversity as an indicator of point of origin is a reasonable general principle, but there are lots of perfectly good counter-examples. Sinitic is more diverse in Southern China, and for pretty obvious reasons. I-E is more diverse in Europe. No Siouan language is spoken in that group’s proto-homeland, wherever it may have been.
Jim: I’m very surprised to hear that. LOT=PALM (the father-bother merger) is pretty pervasive in North America, except in Eastern New England and Canada, and they both have LOT=THOUGHT (one variety of the cot-caught merger). How would you characterize the phonetic shapes of LOT and PALM=THOUGHT in your speech?
Jim: “Salishan and Athapaskan languages all have a poop-load more phonemic complexity that Korean or Japanese or Mandarin or Mongolian, on and on, for no particular reason that I have ever heard of, including distance from Africa.”
Deutscher (“The Unfolding of Language”) claims that complexity is a function of relative isolation of the speech community and simplification results from contact and interaction with other communities.
What Deutscher says seems to be pretty close to my surmise — that simplification and complexification are independent alternative processes, either of which can take place when the circumstances are right, regardless of time period. In which case the conclusion of the paper we’re discussing is invalid.
It reminds me of the linguist who applied glottochronology to Hawaiian Creole English, formed mostly in the 19th century, and concluded that it split from standard English about 1300 AD.
The assumption of regularity and unidirectionality of historical processes was a frequent blunder in 19th c. science.
John Emerson: It makes sense that when different languages interact that simplification would occur. Pidgins, as an example, are highly reduced forms of communication. Creoles arise out of contact situations and they typically have simplified structures.
John Emerson: It makes sense that when different languages interact that simplification would occur. Pidgins, as an example, are highly reduced forms of communication. Creoles arise out of contact situations and they typically have simplified structures.
JC: “However, recent work has shown that elements of language, such as commonly used words (S19) and ome phonemes (S20) can be highly stable, some persisting for tens of millennia …”
The perceived “stability” of some elements is a problem, because there is a difference between word stability and phonological stability. For instance, one of the examples of stability given by Merrit Ruhlen is the word for “nephew”: Latin nepos/nepotis, Italian nipote. Ruhlen appears to think that there is something particularly stable overall about this word, but why ‘nephew’ should be more stable than other words is not explained. In fact the uncanny resemblance of the Latin and Italian forms is not a function of the meaning of the word but of the phonological stability of the phonemes /n, p, o, t/ between Latin and Italian during a mere two and a half millennia. As a counter-example, the Spanish equivalent is nieto which means both ‘nephew’ and ‘grandson’ (both meanings of Italian nipote as well). On the other hand, a study of Latin to French lexical replacement which ignored known phonological evolution would consider French eau ‘water’ (pronounced /o/) an instance of replacement of Latin aqua rather than a continuation.
JE: It reminds me of the linguist who applied glottochronology to Hawaiian Creole English, formed mostly in the 19th century, and concluded that it split from standard English about 1300 AD.
This linguist did not accept the conclusion at face value! Glottochronology sounded like a good idea when it was first introduced, but it has been largely discredited because of the patent absurdities it led to when applied to many languages with a known history, as in this case. Some linguists are still working on a more refined version of the method, but I am quite skeptical of quantitative methods in historical linguistics: as an adjunct, perhaps, but not as basic methods.
GW, JE on Deutscher: “Complexification” is not the same as complexity. Some complexities in isolated populations could be remnants of earlier structures which have been simplified through language contact in less-isolated varieties of the language or family: but “contact” covers a multiplicity of social situations, including dialect mixture and language shift (to a different language). At any rate, it is well established from sociolinguistic studies that language evolution is much more rapid in large cities that attract diverse populations than in isolated rural areas whose populations are either stable or diminishing (through migration to cities). To a city dweller, the speech of those isolated areas may appear as bastardized varieties of the standard speech, while it is city speech which has been simplified through the interactions of speakers of various origins. Complexification does occur in pidgin/creole situations when speakers need to invent ways to express themselves in situations where the linguistic material they have at their disposal (the pidgin) does not provide them with models for more elaborate utterances.
Back to the Atkinson theory, it relies on simplified assumptions which are at odds with what is known about the way language change proceeds and how it is crucially affected by the non-linguistic history of speech communities, including situations where these communities are compelled to shift to another language and do not pass on their own language to their descendants.
Quentin Atkinson’s latest.
JC: thank you for the link, but it comes from Academia which is not immediately accessible. I sometimes get emails from Academia informing me of people wanting to read something I have written (but is not with Academia), but I have never tried to access their site myself until now (two references). They want my email and a password – I suppose it is a special password just for their site.
I don’t have anything published to my name, and I managed to register an account at Academia just fine. Just ignore or click away the exhortations to upload your research, and there’s a simple email/password registration underneath.
Me too.
I’m now an independent scholar.
Latest? That’s the (in)famous paper from 2012.
That’s true if you want to download the paper. If you merely want to read it, just scroll down. I’m shown the full paper, and I’ve never tried to log in.
Eh, so many infamous papers, so little time…. I think of the phonemic drift paper (dribbling phonemes as we go!) and the three-choices-makes-three-characters papers as the truly infamous ones.
Marie-Lucie, whenever a website asks you for your email address and a password, they always mean a password invented by you on the spot and specific to that site. No legitimate site will ever ask for your email password. (My wife was stymied by not knowing this convention just the other day, when she wanted to buy a pair of slippers and didn’t see why she should need to provide her email password for that.) Truth be told, I have a special not-very-secret password that I use for all such sites, since I don’t particularly care if someone masquerades as me on academia.edu. I save the reasonably good passwords for work and email, and the really strong password for my bank.
JC, thanks for clarifying the password business.
I am also registered at Academia.edu as independent scholar.
Sumerologist, at that.
And now every week Academia.edu emails me all the latest research on Sumer…