 So this talk is called Why Does Tibetan Stack Its Letters? And I think most of you know some Tibetans, so we'll be familiar with this. I'll have some preliminaries, then the Tibetan script, then look at prescripts, then superscripts. Oops, some of it is still in German. Then some intermediate conclusions, then look at classical Tibetans. Most of it is about all Tibetan, and then we'll end. Okay, so here's a Tibetan syllable. This is Drup which means accomplished, and I tried very hard to write it in a way that it might make sense even if you don't know Tibetan, although I think most people here probably do. So you see that the S goes on the top, so basically, let's start the vowel. You have constants that come after the vowel or on the right. The vowel and most constants that come before the vowel are in the middle, but then for some reason this B is on the left. And as a convention, in order to sort of make it legible throughout, I put the prescript in blue and the superscript in orange. So now we're going to look at the behavior of the prescripts and ultimate compounds, the behavior of superscripts and ultimate compounds, and then the pattern in classical Tibetan. Okay, so let's look at the prescript, and I'm going to look at three morphemes. So let's say words, how they behave in compound. They are pun, which means official, tsun, which means queen, and zik, which means leopard or sea. So when you have word initial before a vowel, then you keep the prescript. So here we have tri dupun, where dupun means official, and tri means 10,000. So it means myriarch, and you keep the d because it's after a vowel. Whereas when you have chiliarch, so official who runs a thousand unit in the army, tung pun, then you delete the d because it's after a consonant. And we can see both forms of the morphemes in a single sentence. Here you have tuktsam, tung pun, pun yokki, mun lam du sol wa. So prayers for the benefit of superiors and subordinates on behalf of the chiliarch, tuktsam. Where you see you have the word pun yok, which means the officials and subordinates. At the beginning of a word, you keep the d, but inside a word, after a consonant, you lose the d. Okay, now looking at queen, word initial and after a vowel, you keep the b, but word internal after a consonant, you lose the b. I gave this talk in Germany. I thought I had cleaned it all up and turned it into English, but it would appear not entirely successfully. So I didn't have a hurry, I guess. So sorry about that. Anyhow, now looking at zik, leperd or si. So here we just have two names. Snia gziks, where you keep the g after a vowel word internally, but you delete the g word internally after a consonant. So you get the idea, I think. So looking at these morphemes all together throughout, let's say, the documents in the Old Tibetan documents online and a couple more, you get sort of this pattern of keep the prescript after vowels and word initial, delete the prescript after a consonant word internal. I list here the regular and the exceptional examples and you see that there are exceptions, yeah. It needs to be looked at a little bit more, but the evidence is quite strong for this pattern. So now that was it for the prescript. Now let's look at superscripts. And this I'll do primarily with one morpheme as my example, which is bracen. Or, you know, sorry, I should have said bracen if I'm saying it in Old Tibetan and cen if I'm saying it in modern Tibetan. In modern Tibetan, you don't pronounce any of these things anyhow. So our custom of pronouncing things in modern Tibetan is not necessarily helpful for giving this presentation. But anyhow, here you see the word mighty. It also occurs in the word emperor and other things, where we have the b prescript and the r superscript. Now, if you just look at the forms that occur in texts, Old Tibetan texts naturally sort of cleave into two groups. One group that has two forms, brtsan and btsan. And one group that has four forms, brtsan, btsan, rtsan and tsan. And I call these group B and group A amazing counterintuitive that I don't call them group A and group B. But rather group B and group A. That's because as you will see anon that I think group A is older and group B is younger. But for the purposes of exposition, I think it's easier to discuss the simpler pattern first and the more complicated pattern second. And that's what I have just explained here. But we're going to look at, so text group B first that has two forms and we'll look at word initial and word internal. So looking at word initial, the word initial form is always btsan. That's to say with the b but without the r. You get things like tsempo emperor, tsenmo empress and tsentore which is a person's name. Now looking at word internally, the word internal form is usually btsan with the b and with the r. So you get things like dtsen which is a title and motsen which means mighty helmet and gelsen which means mighty victor. So that was it for texts in group B because it's a relatively simple pattern with only two forms. Now we will turn to texts in group A where we have four cases to consider. And these are word initial where we get btsan, same as in group B. Word internal after vowels btsan. Word internal after the consonants nga, gaba and ma. You get rtsan. And word internal after the consonants rla, na and sa. You get tsan with no b and no r. Now I just want to sort of, the linguist in me can't help but say these consonants do form natural classes to use Jacobson's terminology. The first ones are grave, the gga, gaba, ma and the second ones are acute, rla, na, sa. So it's not a totally random fact that, you know, oh some letters do it this way and some letters do it this way. These sounds form natural sort of classes in terms of their articulation in the mouth. Okay, so looking at word initial, you know, this slide will look familiar. But now we're looking at text group B. Tsenpo, tsenmo and centore. You get the b prescript but not the r superscript. And looking at word internal, you get both the b prescript and the r superscript. You've seen these examples before, at least some. Detsen, tritsen and platzen, divine emperor. Okay, now we get to the ways that text group A is different. So word internal after nga, gaba and ma, we usually have rtsan. So we have things like moqtsan, mighty helmet, tongtsan, a name and gimtsan. Who's all the name? And then after the acute finals, so rla, na and sa, we usually get tsan, T-S-A-N. In examples like gyaltsan, kartsan, mighty castle and muntsan, which is a name. Okay. And we do have nice cases where two forms of the same word are contrasted in a single text, which let me just explain methodologically why I see that as important. You can't say, oh, it's just noise. So some texts have these forms, some texts have these forms. If you actually get in one single passage, both forms following the rule, then it means the language actually behaved this rule as a systematic thing. So just let's look at the example. So it's nice because their names are extremely similar, one tsansong, one tsongtsan. But you get the rtsan form and the btsan form depending on the context that I explained previously. And then for the second one, we have zhang, tritsen, dang, zhang, gyaltsen, uncle tritsen and uncle gyaltsen, where uncle tritsen gets the b because tree ends in a vowel, and uncle gyaltsen doesn't get the b or the r because gyal ends in an l. Then lun, gyaltsen, tang, lun, tsansong, minister gyaltsen and minister tsansong. So gyaltsen again without the b or the r and then tsansong because it's word initial, the b but not the r. And here so I've contrasted the form sort of pairwise using textual passages. So now let's look at exceptions. I've sort of given you, in order to kind of, well, sort of experimenting with different presentation styles in part, I've decided this kind of top down version where I just tell you the answer and then convince you it's true is better than sort of more empirically saying, well, let's just see. Have we noticed a pattern? Have we noticed a pattern that's more time consuming and harder to follow? So I've just given you the pattern, but now let's look at the exceptions. So there are many exceptions in group A texts. I mean, partly, you know, group A has a more complicated pattern, so you would expect more exceptions, whereas group B has a simpler pattern, so there are fewer exceptions. So, but the exceptions predominantly occur in three texts. Those are the royal genealogy, the chronicle, and the prayers for the foundation of Degu Jutsu Monastery. One thing that's fun about all Tibetan studies is we all read about the same six texts over and over and over again. So, you know, the prayers for the foundation of Degu Jutsu Monastery, for instance, came up a lot in Dr. Jue's talk, as did actually, for that matter, the chronicle, but I think less so the genealogy anyhow. So my proposal for explaining the exceptions is that group A texts are earlier and group B texts are later, but the three exceptional group A texts properly belong to group B, which is to say, I think those three texts are later texts. But then why have I put them in group A? Well, I put them in group A because they have all four forms. Well, why do they have all four forms if they're group B texts? My explanation is that they have retained some type A examples as archaeasms, and that's of course possible, right? Older things can be transmitted to younger things, younger things can't be transmitted to older things. So if I'm right that group B texts are later, you wouldn't necessarily mind seeing a few type A cases in group B. So that's the proposal I'm going to make, and I'm going to try and argue for it a little bit. So first of all, well-behaved type A texts are early. Let's try it on texts of known date, if you like. So the Zhole inscription from Syrca 763 is a type A text, and the annals, and kind of when the annals are from is a very hard question to answer in a sense. But if anything can claim to be the oldest text we still have in Tibetan, it's probably the annals. So, well, let's just say it's a type A text as well. Whereas type B texts and poorly-behaved type A texts, and here I'm just listing the poorly-behaved type A texts, are later, like the chronicles, and I forget exactly when Brandon and Agnieszka said the chronicle is from, but it's late. Or, I mean, it's still old Tibetan, we're talking about pretty small time spans here, but it's late-ish. And the prayers for the foundation of the Yuzhou monastery we know are from around 822, so they're late. Okay. And now, just so, I think maybe I've already convinced you that there's these two patterns, and there's type A and type B, and type A is older and type B is younger, but some type A texts are actually type B texts that just have retained some type A examples. But now let's look at some actual sort of philological evidence. And I'm going to discuss a principle of historical linguistics here, which is the fourth law of analogy of Kuru-Wovitz, which specifies that when a language presents a doublet, the more transparent form is older, and the more transparent form is younger, and the more opaque form is older, and that both means formally and semantically. So I give you two examples. Lost in English is new, whereas forlorn is old, and you get that both in the R, which reflects this Germanic process called gamata-shevexo that has been analogically leveled in English. It's still preserved in the alternation is R and was were, where you get the R in the R and the were, right? So you get this SR alternation. So it used to be in English you had something like loose with an S and lorn with an R, and you keep that in forlorn with the specialized meaning and the archaic form. And then also melted versus molten, where you have the verb melt and the productive, both formally productive and transparent past tense is melted, that also has the more broad meaning, whereas the more specific meaning molten is only really used in metal and chocolate, has the formally archaic look, has this different oblau grade, so it has the O rather than the E, and a more specific meaning. So that's Kurolo's law, where if you see two things that are sort of similar in form and similar in meaning, the one that has the more specific meaning and the more wacky form is the inherited one and the other one's new. So now let's look at a case where I'm going to apply this this law in Ulta Baden. So in the prayers for the foundation to go to monastery, we have type A Karzen, which is a place name, and you have type B Mokzen, which means mighty helmet. So Karzen should mean mighty castle, but it's a place name, whereas in the text Mokzen actually means mighty helmet. So my explanation is the one with the clear meaning, the transparent meaning, the type B example is new, whereas the one with the specialized, you know, not no longer active specific meaning is old. So yeah, so I think that helps, you know, confirm my hypothesis that prayers, for instance, is a type B text, but it's kept type A features where you would expect it to keep them as archaeisms in things like the conventionalized spellings of place names. Is that the right form of Karzen? Ah, let's come back to that in the questions, yeah. I've given this talk several times and no one's spotted that so far. Okay, so now moving along, now we have a linguistic criterion to separate Ulta Baden documents and Leite Baden documents. So first I say, well, we have this mess, we're trying to figure out the pattern of the spelling alternation. Oh look, it correlates with old texts and new texts. Well now we can actually, for texts of unknown date, we can use this criterion to assign them either to the earlier or the later period. So let's do that and I actually have to give Brandon full credit for making the talk interesting by suggesting this case study. So here it goes, the Pari Bell, which was first published in 2011, so it's a bell from Pari, and it mentions Emperor Tzuk-sen, who lived from 704 to 755, and the fellow who published this article thinks that the bell is from the life of this emperor, so thus is older than the Zhol inscription, which would make this bell the oldest, you know, datable extant Tibetan text, older than the Zhol inscription, if we believe, I think his name is Lachok, the guy who published this article. But let's look at its spelling, so the Emperor's name, Tzuk-sen, is Type B. So it's not that old, it's my suggestion. If the bell is in fact older than the Zhol inscription, it should have been Tzuk-sen without the B, Tzuk-r-sen. Now I just want to make clear that I'm not saying that the bell was made yesterday, or that it's fake, or that it's not even pretty early old Tibetan. I'm just saying it at least belongs to Type B, old Tibetan texts, which means that it's not older than the Zhol inscription, and it is from after the time of the Emperor who's mentioned in it. Okay, now, so there's this lovely pattern in old Tibetan texts. Is there any sign of it in classical Tibetan? And the answer is a little bit, so particularly if we look at the behavior of numerals, you'll notice that the word for ten in the decades, so in words like ten, twenty, thirty, forty, fifty, the word for ten keeps its B prescript when the preceding syllable ends in a vowel, but it loses the B prescript when the preceding syllable ends in a consonant. So it behaves, so let's call it the prescript deletion rule that you saw as active in the words like pun, which was official, and sun queen in old Tibetan, is still, that pattern is still followed by the word ten in these numerals, in classical Tibetan. And now I'll just look at a few nouns in classical Tibetan where I think that these patterns elucidate at least the etymology or the spelling of classical Tibetan words. So if we look at sonam, which means punya, which means merit, it's also a common man's name, just we know because of the Mahadyupati that it is a compound of se, which means to foster, and nam, which means to gain, but hitherto we would just say well why when they made this compound did they delete the BS? Well it's perfectly regular, like in old Tibetan when you make a compound the second element, if the second element begins with a prescript and a superscript and follows an element that ends in a D, you should delete the prescript and the superscript. We saw that pattern already, which is to say writing sonam and deleting the BS from the second morpheme is good behavior for compound internal sundae in old Tibetan. Now the word you've all been waiting for, tsuklak, which Dr. Zhu talked about. So whatever it means, according to the Mahadyupati it means archa or vihara or chastra. As far as its etymology there's been lots of ink spilled on this question, but let's go with the most recent contribution, which is that of Micheal Khan in 2003 and he said it's a compound of the two verb stems, future verb stems, penetrate, tsuk and read, black. And then he just sort of, I mean in I think a quite convincing way, he draws parallels. He says, well and here are some other examples where a future stem deletes the B in the second element of a compound. But here we can say it's not just that oh some compounds delete the B in the second element in a future stem. That's a rule. It's a phonotactic rule that we saw operational in old Tibetan. So I think let's say this is a reason to maybe think maybe his etymology is right or maybe in any case my explanation of how prescript and superscript deletion works in old Tibetan compounds is consistent with Han's etymology. And there's a mean question lurking in this word too. We'll see whether Brandon figures it out for the question answers session. Okay, and then last but not least, ten chö which means shastra. It's the same ten is in ten chö, right? Ten chö is something like ten chö bei gyurwa or something, yeah, I don't know. It looks like it violates the pattern because you say oh ten chö there shouldn't be a b there. But in old Tibetan texts or even in some classical Tibetan texts ten chö is a frequently met with spelling. So actually just already I think from kind of the principle of lectio difficile or in text editing you would say the one without the b prescript is probably older. But here I can have also demonstrated that the one without the b prescript is, you know, correct if you're interested in following the rules that I've been articulating. And what I think is exciting is the bun, you know, the bun pos they don't like the word chö. So wherever they find the word chö they change it into bun and they changed it into bun in this word. Which means when a bunpo, you know, whoever did it the first time, saw the word ten chö he was thinking without the b prefix. Because then it looks like the word chö that means dharma and it should be changed into bun. Now when I gave this talk previously someone objected and said but the chö in ten chö has nothing to do with the chö that means dharma. And I completely agree with that as a sort of a buddologist or a linguist or whatnot. My point is just at least one bunpo once saw ten chö and thought he saw the word dharma there because otherwise it wouldn't have been changed into bun in bun terminology. So let's say these two pieces of evidence, the fact that you actually get orthographic variation on it and the fact that bun pos somehow see the correct form as without the b prefix indicates to me that in old Tibetan it probably shouldn't have had the b prefix. And the spelling with the b is a kind of morpho, what's it called, morphophonemic spelling if you like to reveal its etymology. Anyhow, that's the end of my presentation. And just sort of in passing I'll point out that because prescripts and superscripts behave differently in terms of their phonotactics inside of compounds, it makes sense to write them different places in the syllable. You can say if you want to answer the question why did Tumi Sambota, if such a man existed, which he didn't, why did he decide to write some things to the left and some things on top. It's because he had noticed this quite complicated pattern of compound internal sundae and thought well why don't I index it somehow in my spelling system. Which is to say I think that the Tibetan, far from being some sort of arbitrary arcane frippery of the nefarious llamas, this idiosyncrasy of the Tibetan spelling system actually shows that it's a very well designed orthographic system that encodes a lot of linguistic information. So that's the end of my talk. Okay, shun gong. Oh, let me say and make sure the cameras are rolling. I'm very indebted to my collaboration with Abel Zadox on this paper. In fact, the initial insights to look at sen and pun in their distribution in compounds. That was his suggestion. I did the kind of philological legwork but I want to say to the extent that you see a good idea here is his good idea.