 For today I thought let's look at kind of the pre-history of old Chinese. So what was old Chinese like, you know, from the perspective of comparative data? So that's what we'll be looking at and that's why I've called it before old Chinese, the lecture. So to start with there are ambiguities in old Chinese that kind of cannot be easily resolved because of subsequent sound change and this creates slightly a methodological question of like when do we draw a line between old Chinese and something that comes before old Chinese? And methodologically that you would say well if we use Chinese internal data then it's old Chinese and if we use comparative data then it's something before old Chinese. But it has to be recognized that that can be a little artificial, right? Like maybe there was some merger kind of that can only be recovered from comparative data but that for let's say structural reasons it would make more sense to imagine that that distinction that was later collapsed in the history of Chinese was there at the same synchronic stage as old Chinese. I hope it's clear what I'm saying but in any case this presentation is about the contribution of comparative data. So I include some sound changes here that back strength cigar treat as within the history of old Chinese. And then this I just want to make the point that it just so happens that all of the changes that I found from what I'm calling transient land same thing as sign or to them to old Chinese are mergers and they're mergers that don't interact with each other. Which is to say it's not easy to figure out their relative chronology because there's not a feeding bleeding relationship between them. Now I still have to organize them in some way so I sort of have my intuition about you know which ones seem like they happened earlier and which ones seem like they happened later. So which I mean I don't know that mostly has to do with how general the changes. So I sort of treat kind of finicky changes as later and more general changes as earlier but I just want to kind of point out that the relative chronology is hard to figure out and I haven't done it. So now into the details something we mentioned I think yesterday was the merger of velars and dentals after a very difficult to distinguish in in old Chinese. But with comparative data it's I don't know there's there's good reason to see that there's a merger going on here. So on this slide I have I have examples that end with like year and firewood and long and kindness. So just to look at one you know to kind of draw your attention to it let's look at wood. So middle Chinese is sin and actually Baxter and cigar reconstruct sin but I wonder whether it was sing because Tibetan has shing as the word for wood but there are also cases where the Tibetan cognates suggest a dental final in particular these two. So the comparison of love or pity in Chinese with kindness in Tibetan and then pungent or painful with liver that may seem like a like a stretch semantically but there's there it's very common for words for liver to mean like bitter things like that around the world. So in these cases the Tibetan points to a dental final. Okay the next one is a labial neutralization. Old Chinese it's very difficult to distinguish among up up and up on the one hand and um um and um on the other hand and that's because there was very early on a sound change where um where rounded vowels sort of de-rounded in um before labial finals. If we look at comparative evidence we can see some reason to tease this part. So noop in Chinese means enter and noop in Tibetan means sink or set of the of the sun. So that looks like it's one that can be reconstructed as up whereas uh to stand or something like to stamp in in in Tibetan we have a we have an ah in Tibetan and and Tibetan doesn't distinguish shua and ah it seems like they had a kind of across the board merger between the two. So we can propose that the the Chinese here had a shua. Yeah so there you go. Now uh same kind of contrast with u so we have um really clear evidence that three should have an u vowel from a comparative perspective. And then in this uh in this example hold in the mouth uh there's there's uh the Tibetan has an ah so probably all Chinese had a shua vowel. It's also hard to distinguish u and shua before a cutes but here we we're not so we forget about oh right this doesn't have to do with labials it has to do with the cutes uh so it kind of pair wise if you like uh un is hard to tell from un or from ur and a from ui. Okay so now I just want to point to evidence of the contrast from a comparative perspective. So we have move which back from cigar reconstruct with a shua vowel uh in all Chinese compared to uh this word gul which means to move in Chinese so that suggests an u and similarly in the word for eyebrow the the Burmese uh suggest an u and and you know I will get into this in the in the talk about Burmese but this way you see in written Burmese is quite late and in old Burmese uh all the all the ways are us but um I just happened not to have found the word body hair attested in old Burmese probably because of my failings as a old Burmese philologist um but in any case there you go which is to say the Burmese does support uh an u. Now on to our next case of ambiguity which is uh again because of mergers in the history of old Chinese it can be very hard to tell where there's a medial r so so this is just a notational thing but I'm going to use k for vealers uvealers labial vealers and labial uvealers p for labials and c for any consonant uh in type a syllables it's very hard to tell whether old Chinese had an r in in this circumstance which is basically before oval and in type b syllables there's a lot of circumstances in which it's hard to tell whether old Chinese had a medial r so now I'll look at comparative data and say like well it seems like maybe there was an r in this case and maybe there wasn't an r in this case okay so some comparisons that suggest no medial r in old Chinese uh there's this word geek which can geek which can reconstruct to guk or gruk which is the highest point on a roof compared to the Tibetan word for sky which has no r and similarly this this word who which means behind uh if we compared to Tibetan hok which means under and maybe it didn't have an r and uh might as well just talk with you the the last two the word for uncle ku in Tibetan doesn't have an r so probably maybe it was gu in old Chinese and not gru and b like a b the buzzes yeah is uh bungwa in uh Tibetan and maybe it was pong rather than prong in old Chinese comparisons but there are other comparisons that uh suggest that old Chinese does have the r so sticking with insects we've got gukmo in in Tibetan so probably this uh Chinese was gukro you know gukro sorry I have trouble pronouncing you dealers and in the case of feather we have gukro which probably comes from where we have sorry that was me pronouncing it in a kind of modern Tibetan way so sgro in old Tibetan uh would be from something like sgurah we'll get to that later which is the Tibetan uh changes uh auto all after levy dealers so maybe there was an r in in in the chinese word for feather as well you get the idea which is that we can uh distinguish maybe uh where to reconstruct an r partly based on comparative data there's a couple different ways I could sort of comment the question just to say that here the point is that we can't tell in old chinese whether or not the r was there that's what the brackets mean right so on this slide I'm saying we have no idea whether chinese had an r or not but if we look at comparative data we think probably it didn't and in this case we have no idea whether chinese had an r or not but when we look at comparative data we think it probably did but what I will say is that uh old chinese has reconstructed by baxter and cigar has a lot of medial r's and that's a consequence of the r hypothesis which we heard about in the discussion of the six-file theory which is that all uh division two syllables had some medial r so um maybe that should bother you maybe it shouldn't in terms of you know this is often the difficulty with kind of reconstructing a vocabulary is like how many you know it let's say in the 90s there was a feeling that uh chinese had too many j's in it so we or uh you know medial ya so the reconstructions were altered to get rid of all the medial yas but you know what point is it too many is a matter of taste basically right so um what you can say is that if uh the r hypothesis leads you to reconstruct a medial r in old chinese that can be confirmed in tibetan and vermice no problem right but if it's if you can stand up reconstructing based on the r hypothesis a medial r in old chinese that is not confirmed in the tibetan and vermice you need to somehow explain it and it might be an innovation in tibetan and vermice or uh which would say a loss of medial r in some circumstance or it means that an r has appeared in chinese now if the r's appeared in chinese then one reason it may have appeared is because of morphology and baxter and sagar for a long time have had the idea of an infix r which they use in those circumstances um i feel very skeptical about it um so i i think it would be better to to have another way to get rid of these r's old chinese okay so first we're going to look at this uh proposal of um and coblins which i have written as capital t lowercase r comes from lowercase r capital t which looks like some kind of metathesis but i don't think we have to understand it that way instead we can just understand it as that one of the sources of division two syllables and uh and um retroflex initials is not the medial r but is a pre-initial r so there are these things in chinese which we call demediated clusters i always have trouble saying that word but there's there's various kinds of um two character words that seem like they're somehow derived from a single morpheme and in many cases it's reduplication so there are you know you can say bang bang in chinese or something like that uh but we're looking at at that subset of these pairs where the second one starts with in middle chinese and l i think you can think of it this way that if you're going to reduplicate something that has a medial r in it one way to do that is to sort of break the cluster up so if you have a word like krang and you want to sort of expressively reduplicated you could say kangrang yeah whereas if you had a syllable like bang you would just say bang bang yeah and then we notice that uh we have pairs where you have an initial velar in the first syllable and then an l in the second one like this kang long and we also have ones uh where there's a labial in the first one uh and uh and an l in middle chinese in the second one like mung mung um but there there are no examples where uh one starts with a dental noticing this pattern uh zev handle says it looks like maybe uh this is evidence that there were no clusters like tra in old chinese so instead those cases which is actually kind of uh all of the retroflex initials uh where we have uh so far been reconstructing a t followed by an r maybe we should be reconstructing an r followed by a t and if we look at tibetan evidence we see some uh reason for accepting this conjecture in my old chinese reconstructions on this slide the middle reconstruction is the actual back through cigar reconstruction with the medial r and then the one on on the right so one step further back in time if you like is how the reconstruction would look if we accepted coven's suggestion so just to talk through a couple of them we have something like drang meaning strike so they would take that back to but if we uh believe that there were no trs in old chinese and instead they should be rt's then we would instead get rtong or maybe rtong i don't know but anyhow uh the tibetan rtong compares very nicely right yeah let's look at the the next one so dust they're in in middle chinese drun in old chinese but maybe not maybe it's rtun in in old chinese um and that compares quite nicely with tibetan the dual which means dust you see that this is our our kind of first pass at removing those old chinese medial rs where we don't want them um and if we left them we would probably have to explain them as some kind of uh morphological innovation but coven's conjecture doesn't get rid of everything we want to get rid of many examples of medial r in old chinese have no r in them in the corresponding tibetan or bermes word and let's look at these three examples specifically so laos in uh chinese is srik and it's shik in tibetan with no r yeah kill in old chinese is srat and the root is sat in tibetan and sat in bermes so no r and three is um srum in old chinese but it's gusum in um or gusum maybe in old tibetan and sum in bermes and it's particularly in these cases uh and in particular in kill and three that baxter and cigar propose this medial r in fix and they say like oh well three you know is like a it's like a it's like three things is a few things it's maybe some kind of repetition or uh something and then and then they make a similar argument or a kill it i i haven't been able to follow it it feels a little bit kind of deus ex machine to me but luckily we have a way to solve this problem as well which is a conjecture of uh pulley blanks and i want to be clear here that that um pulley blank is not proposing that it's a pre-initial r that caused this it's some kind of pre-initial so that's why i write it with a capital r whereas here in in in koblin's conjecture the proposal really is that you know uh that old chinese had no medial rs after dentals so those should be rs before the dental whereas in in pulley blanks uh conjecture we can instead say that some of the uh division two syllables come from some kind of pre-initial and we would have to think more about you know well which pre-initials do we need to reconstruct a different pre-initial than baxter and cigar reconstruct so i just sort of mechanically want to index it with this capital r which can mean you know a pre-initial source of division two whatever it may be and we see that that uh this attack is oftentimes let's say buttressed by evidence from tibetan uh where the tibetan does not necessarily have a pre-initial r but it does have some kind of pre-initial so looking at a few examples hollow uh is croc in old chinese so maybe you know we we can think the r actually comes from some kind of pre-initial and tibetan has a pre-initial s so skok yeah silver which frankly probably isn't cognate it's probably some kind of vandervoort you know is is i can't say in all chinese so maybe that's a pre-initial that r and in tibetan there is a pre-initial d so and then i'll just dump jump down to kill and three where you see that um tibetan in all tense forms except the imperative has a pre-initial so in the present it's a pre-initial g so maybe that's some kind of evidence of a cognate to whatever caused division two in all chinese and similarly for the word three tibetan has this g pre-initial so maybe it wasn't surum in all chinese but something sum you know maybe g sum maybe r sum can't be sure okay so that's polybank blanks conjecture and clearly you know more work needs to be done in this area but i think it's a promising conjecture in my book there were i think two counter examples but now one of them has been solved which i will get to in just a moment but there's only one real counter example then which is this word for sand compared to earth where tibetan you know we want we want to have some kind of pre-initial if if polyblank's proposal is correct but it it doesn't in this case so maybe the words just aren't cognate or maybe it's an innovation in tibetan who knows but i just want to draw your attention to that counter evidence to the proposal there is some you know breaking news since since my book came out that i think pushes very much in the direction of polyblank's conjecture which is comparisons between giao rong and olchonese in an article by giam zhuk lai yun fun i think and gong shun so let's look at the word for ice so the giao rong languages you know clearly have something like rpam whereas backstroke cigar would have reconstructed this prang in this paper the author has reconstructed rpam let me just reassure you that reconstructing a final m instead of a final nga here is okay even in backstroke cigar system now moving on to the next one um the giao rong languages i think jopuk is the most helpful here have a rm initial in their word for body hair compares very nicely to bermes muy where there's no medial r right so um um so backstroke cigar reconstruct for eyebrow but these authors suggest reconstructing it so basically it's very solid comparative evidence that polyblank's conjecture is on to something which is that uh pre-initial in this case pre-initial r which is sort of the easiest in a way in explaining division two is one of the origins of division two so some of these medial rs should instead be pre-initial rs um it's not a point that uh that the authors in this particular paper draw attention to so yeah so i can't i can't really say anything more about it i think the the way i feel most comfortable phrasing polyblank's conjecture is some pre-initial has caused division two retroflex initials and chongyu division three in some cases you know that we're saying that let's say i think it was yachun toff who came up with the r hypothesis and the r hypothesis very mechanically says in those three circumstances reconstruct a medial r so so polyblank saying that's that's treating those three circumstances as if they only have one origin whereas in fact we're trying to say they have uh at least two origins one medial r and one some kind of pre-initial i think that suggests that um we're only talking about medial r so we're we're replacing the analysis the proposal to reconstruct medial r in all chinese in some cases with an alternative proposal which is to to reconstruct a pre-initial which is to say you're right that uh we should really look at uh this word for ice more closely because uh baxter and cigar are not reconstructing a medial r they're reconstructing an initial r in the same paper they point out cases uh two of them where um the giao rong evidence sort of conflicts with polyblanks conjecture so in the first case uh to fry or roast there's no evidence for for an r of any kind in in in chinese but giao rong has a pre-initial r now i i just want to point out that in this case the giao rong evidence uh this is not mentioned in the paper uh is supported by tibetan cognate who from a sort of sino-tibetan perspective the edamon for roast did have a pre-initial r in in proto-sino-tibetan and then we need to figure out why evidence for it is not there in chinese maybe there's some very specific uh sound change in chinese having to do with i don't know type a vealer initial syllables or something like that you know it needs more thought and it is a chinese problem that's the point that i want to make whereas when we look at the word for name uh giao rong has a pre-initial r and there's no evidence for an r in chinese but if we look at tibetan there's there's also no r in tibetan so maybe this we don't need to worry about maybe this we don't we can just say is kind of well we'll worry about that let let the let the giao rong insists worry about that one uh it's not maybe relevant to uh the history of chinese whereas the first one because it's confirmed by tibetan clearly is a real sign of tibetan phenomena that then needs to be handled in terms of the history of chinese just to sum up in some phonotactic environments it's very unclear whether or not two reconstruct r in old chinese for that we can maybe supplement by a peak at uh comparative data in some cases chinese seems to have a medial r where we don't want one but we can get rid of those in two ways one is koblin's conjecture which is basically a phonotactic analysis of old chinese where we say that there were no medial rs after dentals they were only pre-initial r before dentals which is how it works in in tibetan instantly yeah so so maybe it makes a certain amount of sense and then polyblanks conjecture which is that some of these apparent medial rs in fact come from pre-initials not just pre-initial r before dentals but maybe pre-initial r in other circumstances and maybe pre-initials other than r you know you can look at a giao rong grammar to see kind of what they are but i would say um for our purposes the ka has something to do with the verbal system and the ta has something to do with the nominal system cautiously ever so cautiously ignore these giao rong prefixes for for a while but uh potentially they actually hold you know the keys to the kingdom the initial correspondences across signative and languages are a total mess so it's really like um you know generally we can work out the finals but the initials are a mess and um some people who are who aren't neo-grammarians say well you know this is just what happens languages change in crazy ways that that are not governed by rules whereas those of us who are neo-grammarians would say it's a sign that there used to be a lot of morphology at the beginning of words and languages like the giao rong languages show you a lot of morphology at the beginning of words so if we look really at a case by case basis we might be able to come up with morphological explanations for the irregular onset patterns um that are reconstructible to prophesior which is to say uh in short I think you know the giao rong languages can can kind of serve the role that Sanskrit served in Indo-European of um you know a language with quite transparent morphology that's quite systematic that they can serve as a sort of um template or model for how the prologue language you know might have been and um you know inevitably uh the discipline should move beyond that uh in the way that Indo-European you know went through a kind of 50-year phase of uh treat Latin and Greek as if they're just sort of funny Sanskrit and then eventually move beyond that and realize that that Sanskrit is not a perfect replica of the Indo-European morphologically speaking but I think in in um Sino-Tibetan it would do us very well to kind of go through that phase and pretend that all Sino-Tibetan languages are secretly giao rong and once then we've understood the morphology of Sino-Tibetan better we can discover those ways in which all Sino-Tibetan languages are different than giao rong or to put it differently those ways that giao rong itself innovated uh giao rong languages have a distinction between an alienable between alienable nouns and inalienable nouns and inalienable nouns have to have a possession prefix so in citation form we use the third person singular possessive prefix like so uh so which is to say I think these words for name actually mean his name but you cannot use the lexim without some kind of possession prefix but there would be a different word for my name or your name or something it would be overplaying our hands to kind of say wherever giao rong has one of these tuz or tuz we're going to throw it right back into proto sido tibetan right that would clearly be unjustified but it is probably worth keeping them and similar kinds of stuff in other languages in mind when we have weird correspondences so I'll just mention one that will come up I think in a forthcoming article of lion funds which is door for elbow where old Chinese seems to have the crew and old Tibetan has grew so elbow is one of the words unsurprisingly that is regarded as an inalienable possession in giao rong languages so you could say proto sino-tibetan had a system of inalienable possession prefixes the his elbow word was something like tukru and then in uh chinese they they lost the system but they kept the the word his elbow as the word elbow which is perfectly understandable that's an analogy right an analogical extension whereas in tibetan they sort of turned the bound morpheme into a free morpheme when the system of inalienable prefixes went defunct everything I've discussed until now are places where we think probably old chinese as actually spoken had one thing or another thing you know so either had the oval or the oval either had an r or didn't have an r but it's very hard to tell because of sound changes that happened later that that merged those origins so we are sort of looking to comparative data just to guide us in the interpretation of the chinese data but the the changes that we're about to look at can more easily be thought of as mergers on the way from sino-tibetan to old chinese so the first one is old chinese k has two correspondences let's put it that way has two correspondences with tibetan bermes so in the first one all three languages have a k and in the second one uh tibetan sometimes has this kind of velar final velar fricte final sometimes has nothing uh and bermes always has nothing so in my book I reconstruct that as k followed by a schwa I make a phonetic argument for that the reviews have been quite skeptical for reasons that I think are foreseeable which is you know this is kind of the only place that we reconstruct I don't know what to say a certain kind of disyllabic word so we have a schwa after k but we don't have a schwa after n or p or whatnot and fair enough it's a very mechanical reconstruction on my part but you know the correspondence is real so here are examples of k so you have night k in all three languages one k in all three languages uh torch or flame k in all three languages poison and poison has to be one of the clearest sino-tibetan cognates there is it's it's really dok in all three languages yeah pretty straightforward but here are examples where uh chinese has a k and the other languages don't so we have something like to pass arrow actually and the semantics is small semantic shift here where tibetan and bermes have arrow whereas chinese has hit with a bow and arrow um let's look at 100 that's a nice one you really expect the word for 100 to probably be cognate and it goes back to um prak in in chinese but uh something like bria in tibetan and bria in bermes so uh so where did this k go in in the other two languages have to somehow reconstruct a different correspondence so that's one merger you know maybe I don't have a good solution for it but it's a real merger uh now another merger which is kind of uh kind of the friend the kind of sister to the one we've just looked at which is a glottal stop final in chinese can either correspond to velars in the other two languages or to nothing in the other two languages and to index this correspondence and this I think is an idea of of cigars I reconstruct a q final would make sense right we have to have we've already reconstructed loads of velars sorry u velar initials why not use a u velar final and a u velar final changing into a glottal stop and a velar in different languages is perfectly normal uh and uh uh we just leave a glottal stop because you know the two languages lose glottal stop anyhow I think this is a less problematic uh proposal and in favor of it um cigar points out that there's an austronesian word of punuk uh brain and now you know he thinks austronesian is related to to chinese let's not worry about that now but I do think that punuk looks an awful lot like the word for brain in uh chinese and bermes so it's nuke in um in bermes and it's nu in chinese so it's punuk in austronesian so I think that's yeah that helps us say well maybe there is a u velar final in this word now we can extend it to the correspondence in general I won't go through all these but let's look at friend yeah so we have something like um yeah I I want to look at friend because I think it shows you that we're starting to achieve something akin to progress because if you look at the middle chinese for example you you look at you in chinese and grokks in tibetan they don't look anything alike you would never compare them but um we can get back to things that look quite similar using proposals that we already had to make for other reasons um yeah so I think that's a sign that oh look we're making progress I think I mean one of the struggles with any language family where uh where where we're at a fairly early stage of research is that you have you know um the let's call them the conservatives who just want to talk about again and again the really clear cognates like you know oh look tibetan has the word lock for hand the bermes has the word lock for hand yeah fine they're cognate but it doesn't help you go anywhere right uh and then you have other people who who are happy to make really wild comparisons that uh the phonology doesn't work the semantics doesn't work and it it's about finding that productive space between the two where you can make kind of bold conjectures and then you can pin them down with um with uh rigorous historical phonology and I think we're starting to see that you know we're getting in that direction with this example of friend so now moving on here are examples that I reconstruct with a glale stop so these ones are pretty straightforward you know let's look at five yeah so you have nga in all three languages there's a glale stop in chinese it's not there in tibetan and bermes so we can reconstruct the glale stop why not okay there are some problematic examples uh so let's look at them and actually i'll start with uh skin because it's the easiest the tibetan correspond the cognate would make you expect a glale stop final in chinese and it's not there so that's a problem you know I don't know how to deal with it if we look at uh you we have a velar nasal final in bermes and a glale stop in chinese so this is an interesting uh correspondence right because it's uh it's the kind of let's say it's the kind of word we expect irregularities in terms of very frequent very old word so I think it's promising for example for figuring out lost morphology and then last we look at tiger which I think is is uh worrisome probably a vondervoort right like I I don't think it goes without saying that there were tigers in the sign of tibetan homeland chris beckwith points out that the I would put it this way that the tibetan and the japanese are unlikely to be coincidences I mean you might not agree tora and stock are pretty different but you know they both have a t and having some kind of you know it makes perfect sense the final you don't need to worry about because japanese didn't didn't permit the cvc structure very easily and uh he says that's uh beckwith says look if if if tibetan and japanese have similar looking words like in the numbers for instance basically the only uh proposal that can be valid is that uh chinese was an intermediary you know either chinese was the donor language to both or it's a sign of tibetan cognate that then ended up in in japanese as a loan but the reconstruction that baxter and cigar have for tiger does not you know look anything like uh the the tibetan or japanese word um but there are t initials elsewhere in the shesham series so for details you can you can look at my book to basically say tiger is a mess and and is an interesting problem that i think more work needs to be done on but actually it's a um a fine example just in terms of sound laws if you wanted to reconstruct a q final right because you have a velar final in tibetan and a glow stop in chinese okay so we're moving now from the from the velar final questions to uh to resonance so it looks like chinese final yaw has two sources uh one we can reconstruct yaw so bermes also has a final yaw and it's lost in tibetan you have um yeah so slanting and be descended excellent and overdue slanting being inclined you see i'm i'm favoring the bermes cognates because it also preserves the final y so here cases you know look bermes has y chinese has y so we reconstruct y but there's also a correspondence uh of chinese y to tibetan l and generally speaking bermes has no final uh in these cases but it it keeps the final in question as a yaw after the vowel ooh that's a kind of bermes problem don't worry about it but we have to reconstruct something in in sign of tibetan and there's a merger going to all chinese so why not l because we need it for tibetan yeah so uh let's look at uh some examples so chinese uh to carry gay and uh tibetan let me load or or be loaded something like uh gal river in chinese guy and cross a river or gal in in tibetan and then maybe skip to the end of the slide sing or song in chinese cai and dance in bermes so that's just to prove that bermes loses the the l so let's reconstruct this corresponds l and then we have a a tricky one where again uh people aren't very happy with my solution i'm not very happy with my solution but it is another um correspondence so it has to be dealt with which is all chinese has an r tibetan has an r fine we can reconstruct r but all chinese has an r tibetan has an l what to do with that so i just reconstructed r l mechanically and you know we do have like like it's a possible sequence of sounds and human language so i'm not too bothered and just a point to remember are in sorry and in brackets in vexer and cigars uh system allows for r as an interpretation so you'll see some of those in the next uh couple slides so here are examples to be reconstructed r uh actually kind of uncannily good uh correspondence between chinese and tibetan score to wrap in chinese and score to go around in tibetan uh let's look at this next one ser fresh in chinese sar new in tibetan car polar rod in chinese car stick or staff in tibetan so fine nice correspondences right nice correspondence nice cognates uh they can be reconstructed r but here are the ones where it's not so nice maybe um you have ken for send away and it could be care right uh in chinese and then scowl in in tibetan so rl correspondence move uh guar in chinese gull in tibetan and so on so it's a different correspondence it needs a different reconstruction i propose rl it's not very attractive because we don't have those we don't have rn clusters we don't have those other kind of final clusters in general but it needs needs to be dealt with right okay and then just to point out that there are cases where um we can't tell whether it's r or rl in sign of tibetan um to look at neck so neck in chinese cun which could be cutter right uh but tibetan has mogul and mogul i don't know why mogul has ended up there twice that's some kind of mistake and then um uh and the other case to look at is uh sar bright and white uh which compares nicely in tibetan with matar which means fair or beautiful or bright but it also uh compares pretty nicely with sal why you can reconstruct sts al as the root in tibetan i'm not going to get into here but you can and that means clear or bright uh so you see it's just unclear which which tibetan correspondence uh or which tibetan cognate to pick so then it's unclear which um correspondence pattern uh exists and maybe these suggest that actually you know there's some kind of morphological thing uh that would explain these doublets in tibetan but these are the only two so um you know it's pretty marginal if if that is the answer okay um yeah so now we're moving on to the vowels so we looked at um mergers that all chinese did in the finals now we're looking at mergers all chinese did in the bow in the vowels so uh particularly all chinese oh corresponding to tibetan you well this is a little bit of a problem because we have an oh or correspondence and if you know uh sort of textbook case in historical linguistics right oh oh how do you reconstruct it how about oh oh we have oh how do you reconstruct it how about oh but then we have oh corresponding to oh how to reconstruct it i go with schwa followed by a w so oh yeah don't feel particularly committed to this solution uh but here are its advantages it fills a gap in all chinese so all chinese has a w but it doesn't have schwa w it has e w but it doesn't have schwa w and it doesn't have u w either but you wouldn't expect it to because they're both rounded yeah so it fills the gap in all chinese and it's only called for like which is a the the cognates where the correspondence occurs uh all are either open syllable or have velar coders which are exactly the environments where all chinese has a w so it has a w nothing and it has a w k so i think that's also an advantage but there are problems with it as well as a proposal one is that the the number of cognates that you would reconstruct schwa w is higher than the number of cognates where you could reconstruct oh you know kind of maybe it's not a problem because maybe we just need to find more cognates that are sitting around waiting to be discovered uh that have the oh oh correspondence but it it bothers me right like i would expect there to be more o's in uh pro-design or Tibetan than schwa w's and a w and e w merge has o in Tibetan with a lot of other stuff there there o has a lot of points of origin uh in Tibetan or that is to say Tibetan o has lots of points of origin in Sino-Tibet so as a kind of parallelism we would maybe expect schwa w to become o in Tibetan but it doesn't it becomes u by definition right so that's maybe an inelegance of this proposal as well so you you tell me those are the advantages those are the disadvantages uh is it a good reconstruction so here are examples where both languages have ooh this first one i think it's quite quite nice so pru womb in chinese at pruma after birth in Tibetan gu we've discussed before we don't need the r there right gu maternal uncle and uh ku paternal uncle ku nine gu nine yeah very nice comparisons and then here's the o examples so we have something like uh cut or break off be sharp pluck all chop basically zot chop something like that peel or or release lot in all three languages so o and o and o and then here is the problematic correspondence so we have steel is uh maybe kos or cross or maybe even rkos as we discussed in all chinese and it's orku in uh Tibetan we have uh kok for for bent in chinese and guk in in Tibetan so o and o corresponding right seems like enough examples in good enough you know that are overall good enough that we can't ignore them okay so um so that was uh it that was my survey of let's say ambiguities in old chinese uh that can be resolved with uh comparisons to other languages as well as those mergers that look like uh they took place on the way from sino-tibetan to um old chinese and now i will just review them in something like uh chronological order but you know as i said we don't know because they're all mergers and they don't interact uh so schwa w changes into o r and whatever this beast is that i reconstruct rl merge as r in chinese yu and l merge as yu glottal stop and final q merge as glottal stop k and this thing whatever it was that i reconstruct as k followed by schwa merge as final k and then we have pulley blanks conjecture which is that let's say r pre-initial and other pre-initials are one of the sources of um let's say rotic-like phenomena in middle chinese and we can write that as cr although it may never have actually passed through that stage and koblin's conjecture uh which i similarly write as a sound change but it's actually maybe more accurately put as a as a phonotactic theory which is that um these rotic phenomena in middle chinese uh never should be reconstructed as a medial when the initial is a dental instead they should be reconstructed as a pre-initial that's koblin's conjecture and then getting down to the the ambiguities that are left we're sort of we've already arrived at old chinese if you like but it just so happens that later changes obscure uh what old chinese looked like at that moment we're unable to recognize r in certain uh kinds of uh syllables it's difficult to distinguish ooh and uh before acutes and also uh uh so let's say rounded vowels merge with unrounded vowels before labials and i think that's it no there's one more uh yeah which is that uh in and in merge and this happened let's say just on the cusp of history is maybe the right way to put it so i think some shi sheng series distinguish them but they they the change seems to have already happened in the odes so um yeah so so i feel okay if you like presenting this as a chronological story even though um as most of them all of them are mergers that don't interact we can't really tell i want to just draw your attention and this is something that those of you looked at my book i do all the time is like you know you you find a story to tell and then focus in on all of the the the misbehaving words so um this this word for you uh it has a really strange correspondence but i think it's unlikely to be uh unrelated uh so that is something that needs work also the word for skin is irregularly missing on the last stop in chinese and this word for sand is an exception to polyblank's uh conjecture uh but in this case i you know i'm not totally i don't know just my intuition says maybe the word for sand in chinese and the word for earth in tibetan are just not cognates um so those are the exceptions and that's the end of this presentation yeah i think so um the the the issue is that we're going to have a lot of stuff it's not cvc right all these pre-initials and the the final glottal in some cases many cases of final s um so the question is kind of can we let's say redo uh cigars 1999 book uh with um updated old chinese and updated uh comparative evidence and arrive at a a root theory that would explain all of these pre-initials and let's call them post-finals i'm not sure that's a that's that's a word either uh as morphological your gloss as differently i i think pardon so you you glossed the final s with a different marker so i guess you mean it's like a post-finals yeah i i mean um there's actually a very nice and this is something that maybe i should actually go into in detail but um those of you who who actually know some chinese and some classical chinese will know that um uh let's say apart from the the the voicing alternation between transitive and intransitive verbs in middle chinese that we saw yesterday the clearest example of morphology in chinese is suffixation with s which from a kind of chinese perspective is understood as a a changing a character's reading from uh not chuxiang to chuxiang and the difficulty is that it seems like it can have you know a million different meanings uh and actually let's say i don't know i i i feel like it's almost rude uh to to um to say he said this but he did say in print uh harps mire in uh his long review of baxter and cigar's book uh says he thinks that um the the function of chuxiang derivation is just to say this character is not being used in its normal usage and to indicate that we are asking you to read it in a different tone so i i think this is the kind of proposal that can really only be made by a a sinologist who has not studied enough linguistics right because like how how is information like that passed on from mother to child you know like sonia that is just to say that um you know uh the the s suffix has clearly has many morphological functions in chinese so much so that uh you know uh faced with that confusion you end up with quite um let's say uh laughable suggestions like that one of harps mire's it's not that he who came up with it it's a kind of idea that's been sort of circulating in sinological circles um geom zhuck recently like two years ago has a nice article where he sort of looks at the different functions of uh chuxiang in middle chinese and compares them with different uh suffixes in gyarong and in some cases different segmentally like oh this one's actually just s this one's si that kind of thing so it um i think it's perfectly possible possible we have a good theory now if you like which is that uh a whole bunch of different suffix sole phenomena uh kind of all collapsed into final s kind of right around the time that uh chinese writing became widespread so that's why we have this you know morphological like it like it looks like an s can be used for 10 different things it's not that it can be you know from a historical perspective that's not the case and that situation only lasted very briefly but long enough to to be the evidence that we have that's the argument that geom zhuck makes and it will be followed up soon by a similar paper by jung shuyang so i think that's a very promising direction returning back to your question that's to say i think we can deal with a lot of these uh final s's and you know maybe we can deal with these pre-initials which is to say i think a a root theory for sino-tibetan of cvc uh where everything else is uh explained as affix sole uh is a good thing to try uh but uh it would be you know a huge undertaking i'm just going to paraphrase the question to make sure everyone understands it you've been looking at middle chinese and seeing these capital h's and these capital x's right and then you probably already noticed that in old chinese the the capital h goes back to an s and the capital x goes back to a global stop well i just explained that that s suffixation is a very widespread um morphological phenomenon in um in chinese so asking about the glottal stop which is to say the capital x and if it only occurred after what would otherwise be open syllables it you kind of no problem right you would just say like oh you know there's k finals and n finals and glottal stop finals but there are many uh well many it may be overstating it there are limited cases where you reconstruct things like m glottal stop and n glottal stop as finals in old chinese and so the question is does the that glottal stop suffixation have a morphological role i don't know and i think it's one of the real kind of uh holy grails uh i'll tell you that personally i find uh you know i know the the the the phonetics people tell me ah there's nothing wrong with you know ending a word m glottal stop but in terms of the i don't know the phonotactics of old chinese in terms of what's normal in in in east asia it bugs me and you know i i realize that's you know just my problem and i need to put my finger on exactly what bugs me if i'm going to complain about it so that's why i've never complained about it um uh but it's uh there let's say this is something that a little bit of work has been done on i think in in sagars 1999 book but it's it's uh tricky yeah and i also think this is something that beckwith really emphasizes is there's something a little fishy about this little glottal stop like one thing is it it seems to have stayed put exactly as is for like 2000 years which is just something glottal stops rarely do yeah glottal stops 10 10 the final glottal stops tend to be quite you know unstable um and like suffocation with a glottal stop to a word that ends with a nasal it just seems a little weird but you know again sort of saying oh that seems a little weird that's not science so so i feel like i should shut my mouth about it what i would say is and you know maybe i should have emphasized this more than i did at the very beginning is i'm a neogramarian so i think all sound change is only conditioned by phonetic position as a method like as a basic methodological principle which is uh to say what's her name the frequency person um by eb yeah so joanne by eb and let's say the sociolinguists talk a lot about oh high frequency words change different than low frequency words i don't want to say that you know all of that empirical work is nonsense i don't think that uh but as a methodological commitment historical linguistics do not allow themselves that move you know and i i like to think of think of it as like you know like when you do geometry as as an ancient greek you have two things an unmarked straight edge and a compass and you do all of your geometry with an unmarked straight edge and a compass and yeah you could do geometry with a let's say for instance a ruler where you could measure things but we don't allow it we don't call it geometry if you do that so i feel the same way about historical linguistics in historical linguistics we have totally exceptionalist regular phonology it as as one tool and and analogy as the other tool and anything you try and do without these two tools i don't call historical linguistics so i don't accept any kind of arguments about frequency