 So how Chinese use to sound the challenges of an opaque writing system in the study of historical phonology and Just to give you an overview of the talk first. I will make a very brief methodological orientation Then I will touch on Sumerian as a sort of comporandum then look at Chinese and When it comes to Chinese discuss the nature of the script Phonetic information contained within the structure of the script phonetic information from external sources and That will be it so turning to the method There's this sort of theory of social science methodology from Karl Marx to ascend from the abstract to the concrete and I think that's that's kind of The the organizational principle. I'm going to be using here tonight the role Of Sumerian is to provide me with some of the necessary abstractions and then apply these two Chinese so I Should emphasize that I really know nothing at all about Sumerian zero which I see as and a decided advantage So, yeah, so Sumerian Begins the 31st century BCE very long time ago But the the older the texts are the more morphemes are omitted in the writing and Sumerian has let's say is typologically exciting yeah, has lots of prefixes and suffixes and agreement and all sorts of exciting stuff going on of the type that that we have in a lot of sign of Tibetan languages like Quiranti languages and Guaranic languages But generally the kind of stuff we don't think of Chinese as having but it's only In the early second millennium when Sumerian was probably already extinct and only spoken in Schools that all of the affixes are fully expressed Which is to say if you look at the early Sumerian when Sumerian was a robust living language from the point of view of An investigator today it might well look like a highly isolating language with very little morphology And that's just because the Sumerians didn't feel the need to write it down because they all knew it whereas in particular in this bilingual context of Acadians having to learn Sumerian as a second language It was they felt the need to spell everything out. So I think that's a useful Story to keep in mind which means that apparently isolating Graphic Information that does not necessarily say anything about what the morphological profile of the language was So let's look at some Sumerian concretely. This is The sign for the word sag which means head And then later it was written like this So you can see that it goes from more iconic to more abstract Partly just for reasons of the convenience of writing and then here is the sign for mouth So you see it's it's like the sign for head But with some indication that you should pay attention to this particular part and then that also gets stylized In a way that has a certain graphic relationship to the stylized version of head But no phonetic relationship, right? Okay, so Now you can make phonosematic compound graphs where you have Some indication of the Sound and some indication of the meaning so here I'm giving the example That you see at the bottom of the slide of MA which means tongue so this Sign is formed by putting meh, which means a hundred It doesn't have anything to do with tongue semantically, but has a certain phonetic similarity you put that inside the sign for mouth and And that's how you form the sign and may which means tongue. Okay Now just another example of these phonosematic compound graphs, so we start again with car But this time we put noon inside of it. We then get lip Noon doom We get the sign for lip So so lip has something semantically to do with mouth and something phonetically to do with noon So it's so that's how we make this sign noon. Yeah, so now look let's look at how you actually write a word I'm going to look at Raven and fish So you you find yourself wanting to write the word for Raven which was pronounced ooga and I guess I mean, this seems a little bit far-fetched to me But it's it's what the secondary literature says so so there you go you say well, what's the most similar sounding sign I already know of and apparently it's Naga But Clearly Naga and Ooga don't really sound that similar So I look for a sign ooh, and I look for a sign God and I put the ooh in front of Naga and the God afterwards and then I get this structure of of I guess the way to understand it is This sign Naga that you know, but you also know has the reading Ooga Well in this context read it Ooga because I put an ooh in front of it and a God after it so it's quite um Redundant in terms of information. Yeah, you could just use the ooh and you could just use the God And then say ooh God like a syllabary, but it's it's more redundant than that for historical reasons And I should say of course, there's a lot of spare spelling variation This is just one way of writing words that are pronounced about and then if you want to distinguish Raven from fish You can put this sign Mushen which means bird at the end of this Ooga or you can put the the could which means fish at the end of the Ooga and I think anyone who knows Japanese will also recognize that the purpose of These semantic determiners. That's what we would call them in Chinese Is not simply to give an indication of the meaning but also of morphine boundaries So that as you're reading rather than just seeing a string of meaningless syllables You see okay, I should bunch these signs together and try and read them as a morpheme I should bunch these signs together and try and read them as a morpheme. Let's now develop some abstractions about how Sumerian works. So a particular Grapheme is associated With at least because it can it usually often will be more than one, but at least one semantic value and one phonetic value So using a notation that William Bolz Puts forward. That's quite straightforward. You can think of the sign as kind of plus s and plus p So if you ignore the semantics of a particular Grapheme will call it g1. It can be added to another grapheme We'll call it g2 to disambiguate the pronunciation of the latter. So we're so we're so we have a sign with both with let's say with more than one But at least one p and s and we're adding another sign to it that we're ignoring the semantics of in order to disambiguate the second sign so the Sign that's being added. We've sort of turned off the s, but we're leaving the p, okay? and this We can call the rebus principle and then on the other hand if we ignore the phonetics of Grapheme g1 we can add it to a grapheme g2 to disambiguate the meaning so that's what we've seen in the in the raven versus fish example and Here we can say we've sort of turned off the p and we've left the s And we call this second one semantic extension now I called semantic extension because Because the the meaning is always a change of meaning you might be using the first of the Process the rebus principle from going from one pronunciation to exactly the same pronunciation in a different Morphing, but the the semantic determiners always will be associated with a change in meaning because two different morphemes In a language mean two different things by definition So that's a semantic extension So I think that's you know a nice abstract characterization of what's going on in Sumerian So now just some some observations about Sumerian the development of the writing system took a very long time complex graphemes are Not Spatially equivalent to simplex graphemes now this may seem like a strange observation to someone who works in Kinéa form studies, but it's important. It's one of the most striking things as a sinologist and I'll even back up and just look at it You know This this Uga meaning fish is a different size then just one Sign Whereas in Chinese we when we make compound signs we make everything small so that so that the size of a sign Stays always the same size. Yeah Another way of putting the same observation is that there's no notion of a character in the Chinese sense Another thing that I think is intriguing is that there's no strict standard of phonetic similarity underpinning the application of the rebus principle if we just look at the cases that I've discussed You have meh serving as the phonetic for and me Where if you want to have an abstract principle it'd be something like The application of this phonetic in a new environment means that the the new the second the second morpheme ends with the same CV as the one we're using to indicate it and then there's this noon used for noon doom Which Instead the principle would be starts with the same CVC and then using Naga for Uga I find a little particularly strange because it it's some kind of principle like ends with the same CV and Has some vowel before that CV, but not necessarily the same vowel. It looks to me like the Sumerians did not have a Phonological theory underpinning their application of the rebus principle Not that you would expect them to And this lack of consistency in principles is made up for with redundancy Because you know if I found myself thinking how does this possibly work in practice and then I think well It's because there's a lot of in-built redundancy. You say, you know, ooh, ooh go go then you've You've made sure that Although there are many dimensions of possible confusion on the part of the reader with the correct training in a given Circumstance, they will always know how to read it The other thing I've asked myself is how on earth did people you know after the death of this language Thousands of years later figure out any of this and I I don't know but but I assume it's because we have Basically the whole school curriculum of kind of classical Sumerian learning Preserved so that you know in a sense. Well if kids could do it, you know 3,000 years ago then scholars can do it today Despite the fact that there are so few kind of rigorous principles underlined the scripts organization So Chinese emerges into history with a fully developed the writing system in 1250 BCE so way way after Sumerian and Complex graphs are spatially equivalent to simplex graphs. So now we're going to apply our abstractions to the analysis of Chinese You start with an iconic Grapheme in this case It looks like a field. It's how you you know, they how you lay out a field So it's a picture of a field and it means field and it's pronounced the link and Then you can apply semantic extension to it This is to the same character, right? Reading it Rui Later with a different writing the meaning is is like a burr. It's like this kind of raised piece of earth between two fields And you can also use the rebus principle. So say, okay, I'm going to use it for another word pronounced Ling here hunt It's also later written with a different character Okay, now just to go through two more examples. You can draw a picture of fire to mean fire Then you can use that Same character semantically extended to to refer to this word conflagration That has a different pronunciation You can also take a picture of a nose that's pronounced bits and Using the rebus principle you can use it for this word subits Which means self but now let's look at the phonetic information inside the script So there are these things we call a shi shang series which are Different characters that are built using the same phonetic determiner and we're going to look at this one be a Where the pronunciations are coming out from Middle Chinese So there are tested pronunciations from many centuries later. Yeah, and you see okay, you have be a be a in a different tone the capital X refers to tone and Then you have Pierre and you have Pierre So looking at evidence like this you can come up with what we call the shi shang hypothesis So Lee feng kuei, it's not the first diversion of the shi shang hypothesis But we'll say it's the the one that is most used nowadays Proposes that characters with the same phonetic component share the same rhyme and Have a initial that is homo organic with each other So we saw this with the examples that I gave they have the same run But you can allow interchange between a be a P and an aspirated P. So this Hypothesis is the cornerstone of of of all progress. I would dare say in Chinese historical phonology and Let's say one way of summing up my observations on on Sumerian or at least some of them is there is nothing like the shi shang hypothesis It seems in the study of Sumerian So here's some examples We see that We have two characters that have the same rhyme Have the same phonetic But one begins with a gna and one begins with a ha So we want to sort of fix it, you know make it so that in all Chinese. These are Middle Chinese readings in all Chinese They would have had home or organic initials so The proposal that I think most people believe is that you had voiceless residents. So in this case your voice your voiceless vealer nasal changes to a voiceless vealer fricative And then in a similar case just to show you that it you know works In other cases too. We do the same thing for a label. So we have this we have Characters that have the same rhyme or characters which in Middle Chinese readings have the same rhyme but in their Middle Chinese readings one starts with an N and one again starts with a voiceless vealer fricative. So we propose. Oh, maybe there were voiceless Labial resonance And they those change, you know, maybe with some conditioning environment that I'm not going to get into here into Voiceless vealer fricatives. Okay, so now you see an example of the Sheshang principle in action how we how we use this this assumption although well-founded assumption to Let's say internally reconstruct Middle Chinese to fit the Sheshang hypothesis So there's a guy Shushan who in the very early Hong dynasty wrote a book called the Shouwen Jidze where he Says among other things he says of about 9,000 Chinese characters What the phonetics are and what the semantics are but he actually had he classified Chinese characters into six types and only one of them is this phonosematic compound type but We will not worry about most of the types What we do want to look at is the Hui Yi type So this according to him is when you take two characters and you combine them without reference to their sound at all but only from their meaning and Here are some examples and I one reason I've decided to talk about this is these examples are Like oftentimes when when someone is first told about the Chinese script these are examples that they're presented with so you know woman plus child equals good and Woman plus roof equals peace because you know having your wife at home. It makes you feel so peaceful. Yeah and The Sun and the moon together are bright because you know the Sun's bright and the moons are bright So somehow the Sun and the moon together are bright. So these are kind of just so stories that suggest You know, the Chinese wrote ideas and then they combined their ideas to create new ideas Writing the ideas directly with graph themes rather than via the mediation of language So you can probably guess that I have a certain skepticism about this Let's look at Sumerian does this does Sumerian have Huey characters, I don't know what the Sumerianists say they probably don't call them way characters in any case But here is one example. I have seen discussed in the secondary literature which is a Saga plus Ninda so head plus bread Can be read goo, which means eat and I just want to point out that it is not necessary to look at this picture and Analyze it as Saga plus Ninda. You can just say it's a person putting bread in their mouth. It's a it's it's the iconic Relationship is still there and it's it becomes a Philological question, you know, maybe this character is attested before Ninda as bread, maybe Ninda as a graphing was actually extracted from this character, right? There's there's nothing a priori That tells you that, you know, the relationship between iconic things if the iconicity is still active in any case Moving on to Chinese. I would say the same thing about this example of woman plus child equals good in Oracle bone inscriptions the character exists and it's still iconic Let's say it's a child sitting on a woman's lap That's not an uncontroversial interpretation, but it's iconic. That's the point I'm making. It's not necessary because it's iconic already to See it as a combination of the words woman and child William Bolz who is who I'm following for most of this Believes there are no way characters and that the reason that Shuxiang thought there were was because of changes both sound changes and meaning changes and And and Billy graphic changes between the origin of script and early Honda asked you So let's look at some of these quay II characters Okay, peace peace is a woman under a roof. Well, not necessarily Because in fact, there are two shesheng series that use this phonetic one with nah like readings and one with on like readings So it's perfectly fine to say Just that the semantic in peace is roof and the phonetic is woman To give a another example. We have this picture of the moon So it it means moon. Yeah, and there's one Series built on it in the meaning moon where it's read not and you can see the series there and There's one series built on it in the meaning night where it was found something like this is gut So that part everyone agrees with but now let's move to a third one Bolz would say there's a third Reading of this character in the meaning like bright Where there's a series that bright itself is built from and that The character for name is built from so bright would have been pronounced Marang and if you look at it, it's not the sun that the moon is next to it's the word for window In fact, it's Garza construction the word for window was pronounce them like come wrong So you can actually see it exactly like in the Sumerian case that we Where you know this picture moon had three readings So we're we're applying either Further phonetic specifiers or further semantic specifiers to say which of those three readings are intended Okay, so that's all I'm going to say about the Hui hypothesis But basically the the the lesson I want you to go away with is if anyone ever says, oh, you know in Chinese child plus woman is is good and woman plus roof is Is peaceful that's nonsense Woman plus child as good is not Shushan's example that I was that was a little bit of a sleight of hand Shushan only I I think he gives something like only seven Hui Yi and I thought about presenting those but they were less sort of Obvious So I didn't which is to say the the stupid Hui Yi examples should not be credited to Shushan Uh, they've they've come up, you know, uh later So now just a a few words about phonetic Source external phonetic sources. We have loans from Chinese and other languages and loans into Chinese and then Some explicit comments philological comments about pronunciation. So loans from Chinese into other languages Chinese has This word for 10,000, which in middle Chinese is something like a man Uh And backs from cigar actually in 2014 reconstructed for reasons that I'm not going to go into with an unspecified consonant prefix But if you look at old to our kick and to carrion, which are words that, you know, we're in contact with Chinese at the right time Uh, they both point to a t prefix And this t prefix also helps explain the occurrence of this word A scorpion that has a t initial and middle Chinese in the same shesheng series. So here's an example where Uh, you know backs from scar didn't know what was it a t prefix or p prefix or k prefix And then the borrowing of this chinese word into other languages shows us it must have been a t prefix moving on Indo-Aryan loans into old chinese Uh, so, uh, I think that the word for horse In old chinese was something like rama And that it comes from arwant, uh, in Indo-Aryan, and uh, this is based on Work that's still unpublished with hones felner and dito gunkal Maybe a less controversial example is the word chariot, which is borrowed from the Indo-Aryan word for wheel and archaeological evidence I think is totally unambiguous That Indo-Aryan speakers introduced the horse and chariot to china around 1250 bc Which incidentally is also when the chinese invented their riding system. Although that's you know, that's a surely a coincidence, right? Okay, now explicit comments on pronunciation So I don't want to go into this in too much detail, but you have comments like this one Where in the seventh century someone says how to pronounce Transliterations from uh sanskrit into chinese he says that Those characters used to transliterate short vowels in sanskrit should all be read in the rising tone So this is something that maizulin Points to as evidence that there was a glottal stop final in the rising tone Which would correlate with uh short vowels Uh, and then here's a discussion from the shouenjie to about dialect differences. So they say, okay, there's this word brush and in chu it's called murut and in wu it's called purut And in yang it's called putt and in chin it's called putt So this is you know good evidence for figuring out how all chinese was pronounced are Explicit discussion of dialect variation And here's another example where I will sort of skip over the details, but basically a guy travels to a certain area and um, and they pronounce Uh, a word That he pronounces as lip. They pronounce it like pick So this is maybe good evidence for a constant cluster something like prick in old chinese Uh, and then uh last but not least in these kind of explicit philological, uh discussions One of the best I think examples of dialect variation in old chinese Discussed by baxuan cigar is that that r changes to n in the west and r changes to ya in the east And here's an example nowadays the people of yanzhou all pronounce the family name yin As yi so they pronounce, you know ur as ayi And transcriptions of loan words. I'm not going to go through this uh in detail, but I'll just point out these are these are Middle chinese han chinese sanskrit and you see that you get in middle chinese capital h's Where you have an s in sanskrit and that's quite good evidence I think that these were actually pronounced as s's at at at this point. Yeah And then I don't know just a fun uh case alexandria So if you say this in, you know, even in middle chinese, you get It's not a great transcription of alexandria But in old chinese, it's alek sran rei. Okay. There are still some details that are confusing to me like Where did the d go? But uh, you know, I think it's pretty good evidence That for instance, this ya comes from l and was still l in 36 dc Okay, and then the very last thing I will talk about is rhyme patterns where if you look at old poetry And this is an ode and you read it in modern chinese you can tell it Should rhyme because you get things like door and law rhyming in the second stanza And you get things like jie and xie rhyming in the third stanza But in the first stanza, it sure doesn't seem like zai and yo rhyme So, um, these are these kinds of uh things are invitations to make proposals that fix the rhyme scheme And I'm going to just give one example. Although it's not from that poem Which is there's this idea that develop as opposed to r in middle chinese comes from ra in old chinese and if you look at some rhymes you get uh mac hua and Hyak rhyming so if you change the hyak into Into, you know, sorry the crack Hyak into crack To crack rather it would rhyme better And then similarly in another ode you get kang huang kwang Xiang so this kwang change into kwang or kruang Then it would rhyme better. So that's just an example of the you know one of the motivations We have either for coming up with hypotheses or for testing hypotheses about How old chinese was pronounced even though Chinese really doesn't give you much information about how it was pronounced The methods are the same as you would use in figure out how latin was pronounced or how greek was pronounced Which is you look at loan words in and out of the language you look at the the inherent structure of the script You look at philological comments like explicit, you know meta discursive things you look at uh poetic Devices that give you information about phonetics The tone we mark with capital h in baxter's Transcription of middle chinese comes from an s and I could point to evidence like sanskrit for that And then the tone that Baxter marks with an x It comes from a glottal stop. I think the evidence for that is much less convincing actually the um the the fact that You know this one comment that I pointed to of of uh use this tone when when trans or when pronouncing Gryptions of sanskrit short vowels, uh is one of the Considered one of the best pieces of evidence, but otherwise it's uh dialect pronunciations particularly in southern china where it's still articulated with a final glott stop One thing to to mention is that middle chinese has four tones So I just told you about two of them one of them We don't need to talk about because you can think of it as the the unmarked tone And then the fourth one is is stop finals So in in let's say in chinese internal analysis a syllable like pang and a syllable like pak Are the same but in different tones. So in in the west we would You know tend to say no, they're not the same one ends with a nasal and one ends with a stop, but um already, you know In let's say uh 602 When we get an explicit analysis of these things uh the chinese Phonological tradition treated those as the same modulo differences of tone And at that time there were four tones There are chinese dialects with way more than four tones. Those are secondary splits because of things like manner onset this gets into Sort of articulatory phonetics that it's not really my area, but for instance um Like if you look at the first formant of a syllable pronounced with uh a voiced onset And the same syllable pronounced with a voiceless onset The formant of the syllable pronounced with the voice onset will be lower That's just a sort of fact of human biology, right? So bak is going to have a lower first formant than pak so the Theory is and this of course is i'm giving a general presentation not actually what happened in in old chinese to middle chinese uh is that um speakers re phonologize the contrast From one of manner onset to one of pitch so then let's say in in old chinese something similar which is that um glottal stops tend to correlate with a decline in pitch and s's tend to correlate with a rise in pitch And what makes people do that? I don't know the probability space is constrained by uh human biology and human reason if you like, uh, but when and where these changes actually take places is I don't know is just what history is I reassure you that it is real. So, uh, let's take it this way. Uh, we start in 602. That's when we had middle chinese uh, and let's say as a as a As an as an abstraction as a kind of point of analysis We can pretend that at that moment although this is of course not true There was no dialect variation and then who you get dialect variations spreading out Um by the tong dynasty you had this northwest middle chinese Which has some very clear features associated with it like final t has changed into final r And nasal initials have become pre-nasalized stops. So, uh, just to give one example a syllable like not would have changed into Yeah, um, and you can tell that like There's loads of evidence Transcriptions into in and out of to bed and in and out of we gore. Just there's loads of evidence I think in and out of tong good And it's also clear that at a certain point that dialect Stopped being spoken got sort of engulfed in The mass of other chinese dialects around it. Zhong wei shan Has written about this in his new, uh, book, uh from kamesh university press and there are some features of uh, old northwest That that have been sort of let's say I don't quite like this way of talking but that surface as a substrate influence on the varieties of standard Mandarin spoken in northwest china, but old northwest, uh Chinese has sadly disappeared Well, i'll just talk about one example. It's three women in that case Bolts says Let's say that at a minimum one of those women is the phonetic And one of them is a semantic. Yeah, and I think that um Let's say and some people have said that's absurd right with people look at it and they see three women Uh, but then I would say, uh, you know, it it There's two questions that need to be sort of taken separately One is our attempt to model the psychology of pre-modern people and the other one is the correct methodology for Doing our own research and uh, I think it's a reasonable hypothesis to suggest that At the margin people preferred to have writings that were not wildly uneconomical in terms of their their sort of phonological analysis, which is to say, you know, look if if you can analyze Woman has a phonetic radical in that situation. Why not? now Then if you push that sort of methodology too far, then you come up with sort of absurd questions like Well, which woman is the phonetic and which one is the semantic right which clearly At that point you need to sort of take the perspective of the of the the internal psychology of the pre-modern people and say like, well That kind of question wouldn't arise