The Ramsey hypothesis

« previous post | next post »

Chris Button writes:

I’ve been working on adding Japanese readings to my dictionary*. I decided to add pitch accents on the kun readings, and started getting interested in the history there. I came across some amazing work by Bob Ramsey—notably this one**.

[*VHM:  Comparative historical dictionary of Sinitic and Indo-European.]
[**"The Old Kyoto Dialect and the Historical Development of Japanese Accent", Harvard Journal of Asiatic Studies, 39.1 (June, 1979), 157-175.]
Clearly, to my novice eyes, he is absolutely correct. I’m staggered no-one really accepted it! I suppose it’s that age-old issue with academia around it being very difficult to disrupt the old guard with their vested interests. In any case, it looks like this recent article adds some nice typological data to Bob’s brilliant proposal.
I wonder what Bob thinks of it nowadays?

I asked Bob what he thought of his 1979 hypothesis, and he said that he still believes that it is correct, though he has scarcely done much in the way of follow-up research on this subject.  Someone who has, though, is Elisabeth de Boer, the scholar whose paper is cited above and is the person who's done the most pertinent recent research on the Ramsey Hypothesis. She looks in depth at the many implications of the theory, and what this basic reassessment of Japanese linguistic history upends and changes.

Although I cannot be sure of all the details, what I know about the Ramsey Hypothesis is that Bob came up with his theory of Early Middle Japanese pitch accent and later sound changes around the time he was working on his dissertation (on Korean) for Sam Martin, which included a discussion of Middle Korean “tone” markings.  He was surprised by the hostility with which Japanese historical linguists reacted to a presentation he gave of this theory, which he formulated around that time.  The reason for the hostility was that, back in the 1940s, there had been a dispute between followers of Tōjō Misao and Yanagida Kunio about whether Japanese dialects were distributed in clumps all over the map or in concentric rings, with Kyōto at the center.  Japanese linguists at the time cobbled together a compromise, acknowledging that Yanagida’s concentric-ring theory (supported by lexical data) was correct but that documentary records of Kyoto Early Middle Japanese accent proved that it preserved old features that had been lost in peripheral dialects.  By pointing out a better theory of accent could be had by treating Late Middle Japanese and later Kyōto pitch accent as innovative (which makes sense from a cultural and historical standpoint), Ramsey was, in effect, upsetting a big apple-cart of the kokugogaku (Japanese linguistics) establishment.


Bibliographical notes by Jim Unger

DE BOER, ELISABETH. 2010. The historical development of Japanese tone: From proto-Japanese to the modern dialects. The introduction and adaptation of the Middle Chinese tones in Japan. Wiesbaden:  Harrassowitz.

RAMSEY, S. ROBERT. 1979a. Language change in Japan and the odyssey of a teisetsu. Journal of Japanese Studies 8.97–131.

RAMSEY, S. ROBERT. 1979b. The Old Kyōto dialect and the historical development of Japanese accent. Harvard Journal of Asiatic Studies 39.157–75.

Ramsey 1979b explains his theory of Middle Japanese pitch accent development.  Ramsey 1979a explains the politics of Japanese kokugogaku research responsible for the antipathy to his theory in Japan.  De Boer 2010 explains in great detail why Ramsey’s historical theory of Japanese accent is superior to others.  Additional pertinent articles by de Boer are available as PDFs here.

There is little for me to add except that, in

UNGER, J. MARSHALL.  2011.  Review of A History of the Japanese Language. By BJARKE FRELLESVIG. Cambridge: Cambridge University Press, 2010.  Language, vol. 87 no. 4 pp. 951–15.

I explained why Frellesvig’s dismissal of Ramsey’s theory was hasty and ill-advised.  I would only add that I expect future research on Korean-Japanese comparisons, using Ramsey-de Boer suprasegmental phonemics on the Japanese side, will prove fruitful.


Selected readings


  1. David Marjanović said,

    June 27, 2022 @ 3:07 pm

    Elisabeth de Boer has put her entire book on the historical development of Japanese tone here on her page. I highly recommend it even to people who, like me when I started reading it, barely know that Japanese "has pitch accent" of some sort.

    Comparative historical dictionary of Sinitic and Indo-European

    …that's… very ambitious for a single person, judging from the history of comparative historical dictionaries of IE alone.

  2. Antonio L. Banderas said,

    June 27, 2022 @ 3:10 pm

    Japanese readings to my dictionary

    Was it not a Chinese dict. though?
    Can you elaborate a bit on the purpose of adding Jap. readings with some examplification?

  3. Chris Button said,

    June 27, 2022 @ 4:35 pm

    @ Victor

    Thanks for posting this.

    @ David and Antonio

    Comparative historical dictionary of Sinitic and Indo-European

    The title given by Victor could easily be misinterpreted. I’m actually not quite sure what to call it yet.

    The purpose of the dictionary is to outline the etymological relationships between the “words” the Chinese characters represent while also incorporating graphic analyses of the base characters that go back as far back as we have evidence.

    To support my etymological speculations, I am drawing comparisons with semantic shifts elsewhere—usually Proto-Indo-European.

    So I am not compiling a dictionary designed to specifically show any formal connections between Chinese and Indo-European, but I do end up doing that for Indo-European and other language families as I come across words that do not have internal etymologies. So “cassia” in the recent LLog post is listed as a loanword, and not one from PIE.

    Regarding character readings, I’m including modern mandarin, Pulleyblank’s Late and Early Middle Chinese, and my reconstruction of Old Chinese based heavily on Pulleyblank’s work with some fairly significant changes. Other existing reconstructions of Old Chinese simply don’t seem to work very well in elucidating the etymological connections. Rather than providing some long treatise on Old Chinese reconstruction, the proof will be in the pudding here!

    I’m also including modern Japanese readings (with pitch accents for the kun readings) and am including reconstructed Sino-Japanese forms for the On-readings based on Marc Miyake’s work (with caveats on how reliably things can be reconstructed across go-on and kan-on time depths). Miyake makes good use of Pulleyblank’s work (while noting the circularity of using the Japanese and Chinese evidence for each other) since Pulleyblank’s work took into account the Japanese.

  4. David Marjanović said,

    June 28, 2022 @ 2:33 am

    That makes perfect sense, thank you! I'm looking forward to it. :-)

  5. ~flow said,

    June 28, 2022 @ 9:19 am

    In the second of the linked articles, the author writes "phonologically distinct H tones are present in all the modern dialects. [T]his means H tones must have been present […] in the proto-system".

    No, it does not. The dialects could have acquired this feature at a later time, similar to how both Romance and Germanic languages in Europe have articles when their common ancestor did not.

  6. Elisabeth de Boer said,

    June 28, 2022 @ 10:30 am

    The passage quoted in full is: "In several tone classes (of nouns, as well as inflected forms), such phonologically distinct H tones are present in all the modern dialects. While this means H tones must have been present in these tone classes in the proto-system, H tones are nonetheless lacking in the standard reconstruction of the MJ tone system."

    So I am not talking about some general feature ("H tones") that could have developed later in time in several dialects independently, but about concrete words that have /H/ tones in all the modern dialects in corresponding locations in the word: In Kyōto type tone systems like that of Kōchi one syllable earlier in the word than in Tōkyō type tone systems.

    This is the regular correspondence between these tone systems, and it is clear that one of the two systems derived from the other; either Tōkyō shifted the tones to the right or Kyōto shifted the tones to the left. This shift is what made these tone systems split. How could the regular correspondence be present in these words if there were no /H/ tones there to be shifted, at the time when this split occurred? Ergo, /H/ tones must have been present in these words in the proto-tone system at the time of the tone split.

    Examples are tone classes 2.3 and 3.4:
    Compare MJ (standard reconstruction), the Kyōto type tone system of Kōchi, and Tōkyō
    Tone class Gloss MJ (standard) Kōchi Tōkyō
    2.3 ike wa ‘pond’ LL-H /HØ-Ø/ /ØH-Ø/
    3.4 otoko wa ‘man’ LLL-H /ØHØ-Ø/ /ØØH-Ø/

    In Ramsey's MJ reconstruction tone classes 2.3 and 3.4 have HH-L and HHH-L tone, respectively. H followed by L in MJ is preserved as /H/ tone in the modern dialects, so the presence of /H/ tones in the modern dialects is explained. The location of the HL sequence in MJ agrees with Tōkyō, not Kyōto. This is one of the many arguments in favor of Ramsey's theory.

  7. David Marjanović said,

    June 28, 2022 @ 3:59 pm

    No, it does not.

    Indeed not; but it's the most parsimonious hypothesis if all else is equal.

    In your examples, not all else is equal. First, we know older stages where articles are decidedly rarer or even optional; second, the North Germanic suffixed article isn't a cognate of the West Germanic preposed article, and the Sardinian article isn't a cognate of the Continental Romance one.

  8. Chris Button said,

    June 28, 2022 @ 9:54 pm

    @ ~flow

    I think p.9 provides a nice explanation of what the author is getting at with that statement.

  9. Chris Button said,

    June 29, 2022 @ 8:44 am

    Or just read what Elisabeth de Boer has posted above :)

    For some reason, her post was not appearing in the thread when I posted my comment above.

  10. Jonathan Smith said,

    July 1, 2022 @ 12:24 pm

    Re ~flow's comment and implications, as far as tone languages proper are concerned (take say Sinitic), the problem is that AFAIK there are no proposed principles of diachronic change on which basis one could reconstruct proto-tone-values in a meaningful way; instead it seems all we can do is speculate on a "majority rules" basis in light of modern tone values (which is of course totally unprincipled.)

    but Japanese is totally different in employing (pitch-realized) accent, and I have no idea to what degree research in this area really leverages cross-linguistically derived reconstructive principles. I have my amateurish doubts… I think modern systems can be described as having (or lacking) a single "accented" mora, with a tightly restricted array of pitch patterns falling out on this basis. This means that, e.g., reconstructing Ramsey's *HHH based on Kōchi ØHØ : Tōkyō ØØH looks odd as there is no such thing as "HHH" on a modern system or any similarly behaving one (whereas "LLL" is indeed considered the realization of ØØØ).

    Further re: the relationship between tone languages proper and Japanese etc., I'm puzzled by statements in de Boer (2010) like "Middle Japanese, […] although containing a limited number of contour tones, was basically a register tone language," but maybe this isn't relevant to the principles underlying the proposed reconstructions.

  11. David Marjanović said,

    July 2, 2022 @ 3:30 am

    This means that, e.g., reconstructing Ramsey's *HHH based on Kōchi ØHØ : Tōkyō ØØH looks odd as there is no such thing as "HHH" on a modern system or any similarly behaving one (whereas "LLL" is indeed considered the realization of ØØØ).

    As de Boer's book makes clear, most modern varieties of Japanese allow only one H per word (including clitics), so yes, they can be described as having classic pitch accent (i.e. distinguishing tones only in the stressed syllable – though positing a stressed syllable in H-less words, the majority of all words AFAIK, is contrary to the phonetic facts on the ground).

    But, as the book also makes clear, MJ with written tone marks was different. It started out allowing almost any sequence of the two by far most common marks. (One of them is H, the other is L – the question of which is which is where Ramsay & de Boer differ from their predecessors.) Over time, more and more of these sequences disappeared by tone sandhi until only a maximum of one H per word happened to be left almost everywhere (and nearly so in the few surviving exceptions).

    In other words, historically attested MJ was a tone language. Just read the book. :-)

  12. Jonathan Smith said,

    July 2, 2022 @ 9:17 am

    Thanks, interesting indeed. I am doubtful the history can be as you describe (and the precise sense(s) of 'tone language' remains a problem if so), but you are right that I should finish the book (and probably a few others) before developing strong opinions…

RSS feed for comments on this post