Cracking the Indus Script.

Mallory Locklear has a piece at the Verge on an old and probably unsolvable problem, the Indus Valley script. She writes that “new work from researchers using sophisticated algorithms, machine learning, and even cognitive science are finally helping push us to the edge of cracking the Indus script,” but that’s your basic science-journalism hype — that edge is a long way from the crack, and the crack is purely hypothetical. Be that as it may, if you’re interested in the problem, this is a useful summary of the current situation, with descriptions of techniques like conditional entropy and Markov models, and even some juicy academic brawling:

“You would be better off getting medical advice from your garbage man than you would getting ideas about the Indus script from listening to Steve Farmer,” says Wells. “None of the three authors have a degree in archaeology, epigraphy, or anything to do with ancient writing. Their underlying subtext is, ‘We’re all so brilliant and we can’t decipher it so it can’t be writing.’ It’s ludicrous.”

Thanks, Trevor!

Comments

  1. “I am authorized to say that the action we are now reporting may well bring the [translation] within measurable distance of its end.”

  2. -cultural neurobiologist and comparative historian Steve Farmer

    I have not a faintest idea what a cultural neurobiologist does.

    But at least he doesn’t claim to be a cultural neurosurgeon. That’s reassuring.

  3. David Marjanović says:

    Interesting article!

    Long, long ago (possibly around the time Star Wars came out, actually) there was an article in Scientific American (which I read much later) that briefly presented a decipherment as Proto-Dravidian which exploited the fact that certain words are homonyms in Dravidian and apparently share a sign in the Indus script. Does anyone know why this hasn’t caught on?

    To me it’s obvious that it’s writing, BTW – the symbols are highly stylized and stereotyped, too much so to be emoji, and heraldic signs would be unnecessary on the seals that already have all these animal depictions on them.

    and even some juicy academic brawling:

    The next sentence is even better!

    I have not a faintest idea what a cultural neurobiologist does.

    Yeah, me neither. I’m surprised Witzel the IEist goes along with this.

  4. Richard Hershberger says:

    You left out the best part: ‘Wells compares fact-checking Farmer to fact-checking Donald Trump. “You have to fact-check every single thing he says because it’s mostly wrong.”’

    I have no basis for judging whether he is right about this, but I can sympathize. I am currently reading a book, published by the leading academic press within the field, whose first half is very much within my area of expertise, but the second half less so. I have just finished the part within my expertise. It is rife with errors and omissions both large and small. The author simply doesn’t understand the topic. I will probably not read the second half, for fear of internalizing errors. The second half is the meat of the topic, with the first half essentially setting the context. It is possible that the author skimped on research time for the earlier material. But I’m guessing that the second half is pretty dreadful, too.

  5. ə de vivre says:

    There’s a video of a lecture from the University of Chicago Oriental Institute that touches on the Indus script. The claim there is that the writing comes from pre-Harappan potter’s marks. But, given the extent of the Indus Valley Civilization (from the Wakhan Corridor in Afghanistan to Gujarat in India), there’s a good chance it wasn’t used for just one language (just like early cuneiform was used for three unrelated languages, or the early Semitic script for a fairly wide continuum of dialects). Seals from South Asia are square, but there are also round seals in a Persian Gulf style that have a distinct distribution of sign-frequencies. The speaker also claims that the different animals featured on the seals represent different communities. Only the bull is found on seals outside of India, perhaps the symbol of the merchant community. It’s pretty a pretty interesting lecture overall if you’ve got an hour-ish to spare.

    Just call me uncle Toby, because I’m going to get on my hobby-horse again. Ancient Mesopotamia (groans from the audience) had fairly close contact with the Indus Valley Civilization. It’s a topic that fascinates me because it’s a case where the early textual sources and the archaeology line up in straight-forward ways. Early on Assyriologists found textual references to ‘Meluhha’ starting with Sargon of Akkad down to the beginning of the 2nd millennium BC—coincidentally when the IVC disappears. There are also references to ‘Meluhha’ translators, and even a community of ‘children of Meluhha’ settled in Mesopotamia in the 21st century BC. On the archaeological side, they’ve found remains of water-buffalo, cylinder seals made of shell, lapis lazuli, and bead-making technology that could only come from South Asia.

  6. It is rife with errors and omissions both large and small. The author simply doesn’t understand the topic. I will probably not read the second half, for fear of internalizing errors.

    Sad!

  7. The Verge article doesn’t seem to be reporting on anything more recent than the “conditional entropy” papers by Rao et al., which were discussed in several Language Log posts a few years ago (including references to comments and an earlier paper by Farmer, Sproat, and Witzel), e.g.

    http://languagelog.ldc.upenn.edu/nll/?p=1374

    and, more recently, in a guest post by Sproat:

    http://languagelog.ldc.upenn.edu/nll/?p=4652

    [edit to add:] Ah, and the debate was also discussed here back in 2009:
    http://languagehat.com/indus-script-squabble/

  8. From the Verge article, quoting Bryan Wells: “None of the three authors have a degree in archaeology, epigraphy, or anything to do with ancient writing …”

    Of course, this is also (or even more) true of the team of Rao et al. (Rao is “director of the National Science Foundation’s Center for Sensorimotor Neural Engineering and a professor in the Computer Science and Engineering Department at the University of Washington”; Yadav is “a researcher in the Department of Astronomy and Astrophysics at the Tata Institute of Fundamental Research”, as is the third author on the 2009 paper, Vahia; and Adhikari is “a physics professor”. Only the very last author on their paper is someone who’s studied the ancient Indian scripts, and is apparently a longtime proponent of the Dravidian hypothesis for the Indus script.)

    At least Sproat and Witzel have some background in linguistics, which seems a bit relevant.

  9. Ah, and the debate was also discussed here back in 2009

    And I’d forgotten all about that post. Lots of comments, too. Sad!

  10. No one has responded to my Nineteen Eighty-Four reference, so I feel like I should elaborate. The decipherment of the Indus Valley script is a long-running problem, with very little progress having been made. It’s something that I’ve been following for the last thirty years (since my third-grade social studies book had a section entitled “Mohenjo-daro: Where is There?”), and we are still a long, long way from translating the seal inscriptions. Any suggestion otherwise as this point are pure meaningless hype.

    Calculation of conditional entropy is something that can be done using off-the-shelf statistics packages, with little original though required. If the results of Rao’s analysis are correct (and I have no reason to doubt them), all that indicates is that the results are probably a real language. That is millions of miles from there being a decipherment! (Compare the Beale cipher, which is almost certain to be a forgery, but which satisfies many statistical tests for encoding a real message. Statistical tests can tell you only the most minimal information in high-entropy systems like natural languages.)

  11. I find Farmer, Sproat, and Witzel 2004 absolutely convincing.

    First of all, Harappan culture as a whole doesn’t look literate: there are no graffiti or written-on potsherds, and if the longer texts existed in manuscript form (and so have been lost), where are the ink-pots, desks, and other paraphernalia of literacy? Next, how is it that there are so many hapax legomena and low-frequency glyphs, but on the other hand some glyphs are enormously popular? What is more, very few inscriptions have more than one instance of a given glyph, unless indeed they have multiple consecutive instances. Lastly, the Harappan civilization lasted two millennia, but there is no discernible difference between one age and another, unlike Chinese or Egyptian.

  12. David Eddyshaw says:

    “You would be better off getting medical advice from your garbage man than you would getting ideas about the Indus script from listening to Steve Farmer,” says Wells. “None of the three authors have a degree in archaeology, epigraphy, or anything to do with ancient writing. Their underlying subtext is, ‘We’re all so brilliant and we can’t decipher it so it can’t be writing.’ It’s ludicrous.”

    This reminds me of what Paul Krugman calls “pulling rank.” It’s argument ad hominem, so far as it can be dignified with the term “argument” at all.

    Michael Ventris was an architect.

    The paper JC links to is indeed convincing.

    Wells might learn something if he listened to his garbage man, instead of imagining him as an epitome of ignorance. (Full disclosure – I have been a sewage worker in my time.)

  13. David Eddyshaw says:

    The least convincing bit of the FSW paper for me is the part that speculates on what on earth the inscriptions are if they aren’t a script. There doesn’t seem to be anything comparable elsewhere in terms of sheer elaboration. As Rao is quoted as saying in the original article, that in itself is a very interesting question.

  14. Wells might learn something if he listened to his garbage man, instead of imagining him as an epitome of ignorance.

    Tolkien wrote a letter to his Aunt Jane talking about Welsh postmen:

    Sir John Morris Jones, a famous Welsh scholar […] said, commenting on the work of a learned French scholar (Loth) on Welsh metres: ‘I get more learning and sense on the topic out of my postman.’

    “Which did not mean, of course, that Loth was as ignorant as a mere postman ‘passing the time of day’; but that the postman was better read and more learned than a French professor. It may have been true – in Welsh matters. For as a ‘poor country’ even yet Wales has not learnt to associate art or knowledge solely with certain classes.

    There doesn’t seem to be anything comparable elsewhere in terms of sheer elaboration.

    What about the comparison to heraldry, which is very complex but entirely non-linguistic (except that there is a way to translate it into text)?

  15. David Eddyshaw says:

    Good point; I put it sloppily. But if the Harappans had a system comparable in complexity to heraldry which they used to decorate their artefacts like this, that is if anything even more interesting than if they’d had a script.

    It does occur to me that there may be some hidden assumptions lurking about what scripts are actually *for.* We owe our systems to Mesopotamian bean-counters, only later hijacked to big up the big men with accounts of Alternative Facts, and later yet by the poets. If the Harappans’ ideas of the purpose of writing were quite different ab initio, you might perhaps end up with something that doesn’t match our data from bean-counter/royal flatterer usages.

  16. David Marjanović says:

    Next, how is it that there are so many hapax legomena and low-frequency glyphs, but on the other hand some glyphs are enormously popular?

    What if – being found on seals and such – the inscriptions are mostly names, and we’re looking at a culture that was more creative with names than the Romans? Then the enormously popular glyphs could be genitive markers or something.

    Close to heraldry, but not quite…

    It does occur to me that there may be some hidden assumptions lurking about what scripts are actually *for.* We owe our systems to Mesopotamian bean-counters, only later hijacked to big up the big men with accounts of Alternative Facts, and later yet by the poets. If the Harappans’ ideas of the purpose of writing were quite different ab initio, you might perhaps end up with something that doesn’t match our data from bean-counter/royal flatterer usages.

    Quoted for sheer beauty.

  17. David Marjanović says:

    Imagine deciphering Chinese from nothing but names and the occasional de or zhī.

    Or indeed just names, if the enormously popular glyphs were like Chinese family names…

  18. See also this 2010 post by Rao as well as the cited article by Vidale, which is on JSTOR.

  19. It does occur to me that there may be some hidden assumptions lurking about what scripts are actually *for.*

    That’s an extremely important point, and exactly the kind of thing that doesn’t spring readily to mind. Our blind spots are many and, well, blind.

  20. David Eddyshaw says:

    It seems clear from its non-arbitrary complexity that the system is at least meaningful, in the sense that both a script and heraldry are meaningful, while an abstract arabesque (however beautiful and aesthetically worthy of contemplation) is not. There are two distinct questions about this meaning: does it map into a particular language? and does it map into Language at all? The latter would make it a script, but does not entail the former: think of early cuneiform texts which could equally well be read as Sumerian or Akkadian. The political obfuscations which seem to bedevil this matter seem to owe a lot to not keeping these two questions separate.

  21. I think the name idea falls down on the repeating glyphs, unless you suppose that someone is named Gggggg Ff Rrrrrrrr.

  22. Suggestion that Harrapan symbols are actual deity signs runs into a big problem – according to Parpola, 398 distinct signs and no less than 1839 variants were identified to date.

    Having 1839 gods seems a bit excessive even for India.

  23. ə de vivre says:

    It does occur to me that there may be some hidden assumptions lurking about what scripts are actually *for.*

    Indeed. I’m also not sure the hard line they’re drawing between ‘real writing’, ‘proto-writing’, and ‘symbols’ is as analytically useful as they’re making it out to be. Just calling it ‘proto-writing’ assumes that the natural course of logographic symbols is to become a system capable of capturing any given utterance.

    It’s ironic that the non-writing ‘deity signs’ on page 41 contain many symbols that were also used as logographic and eventually syllabographic signs in cuneiform writing. In fact, archaic cuneiform (ie before it was actually cunei-form) has considerable overlap between stylized representational images and writing proper.

    I think the name idea falls down on the repeating glyphs, unless you suppose that someone is named Gggggg Ff Rrrrrrrr

    Only if it’s alphabetic. They might have perfectly reasonable names, like Huwawa, Inana, Anunanki, or Igiigi.

    We owe our systems to Mesopotamian bean-counters, only later hijacked to big up the big men with accounts of Alternative Facts, and later yet by the poets.

    Incidentally, (one of the, if not the) oldest surviving text (rather than 4th millennium Excel spreadsheet), has perhaps my favourite line in the Sumerian corpus, “Nothing is of value, and yet life is so sweet.” Like most expressions of this kind, there’s a politically quietist interpretation (“you can’t buy happiness, so just accept your lot in life as a serf”) and a more troublesome interpretation (“hey, the king’s just a person who shits and farts like me!”). There’s no way to know what the “accepted” interpretation was, but about 700 years later, the line shows up as the first line in a group of compositions, one of which suggests that making burnt offerings to the gods doesn’t actually do anything. All that to say, there’s at least some textual evidence that the “no gods, no masters” tradition is just as old as the “the king is infinitely just and wise” tradition.

  24. –might have perfectly reasonable names, like Huwawa, Inana, Anunanki, or Igiigi.

    Harappan language could be related to


    Proto-Euphratean was considered by some Assyriologists (for example Samuel Noah Kramer), to be the substratum language of the people that introduced farming into Southern Iraq in the Early Ubaid period (5300-4700 BC).
    Benno Landsberger and other Assyriologists argued that by examining the structure of Sumerian names of occupations, as well as toponyms and hydronyms, one can suggest that there was once an earlier group of people in the region who spoke an entirely different language, often referred to as Proto-Euphratean. Terms for “farmer”, “smith”, “carpenter”, and “date” (as in the fruit), also do not appear to have a Sumerian or Semitic origin.
    Linguists coined a different term, “banana languages,” proposed by Igor Dyakonov and Vladislav Ardzinba, based on a characteristic feature of multiple personal names attested in Sumerian texts, namely reduplication of syllables (like in the word banana): Inanna, Zababa, Chuwawa, Bunene etc.

  25. Even if the super-common signs are assigned the meanings “Mac” and “Von”, the idea the script encodes names still doesn’t really satisfy. A system for encoding language, with hundreds of signs, maintained for thousands of years, but only ever used to label pots? Despite contact with Mesopotamia and so on? It just doesn’t seem plausible.

  26. David Eddyshaw says:

    ə’s point about the “hard line they’re drawing between ‘real writing’, ‘proto-writing’, and ‘symbols’” is very pertinent, thinking about it. The dividing line between a ‘script’ which represents Language but not necessarily just one particular language, like the oldest cuneiform, and more abstract systems (like the conventional signs used in traffic signals or international airport signs) is going to be fuzzy. It’s not a clearcut binary script/non-script thing. How much structure does there have to be for it to count as a representation of language?

  27. ✯O□∆

    Is it a real writing? Some Russians find deep meaning in it

  28. David Marjanović says:

    Our blind spots are many and, well, blind.

    Unknown unknowns.

    one of which suggests that making burnt offerings to the gods doesn’t actually do anything.

    Wow, that beats Cārvāka by, what, a thousand years?

    A system for encoding language, with hundreds of signs, maintained for thousands of years, but only ever used to label pots?

    Well, if they had invented papyrus or something, or wrote on textiles or leather, we’re not going to find out anytime soon.

    Also, it seems that, for a long time, runes were used by very few people for very few purposes and were otherwise as secret as their name says. (German raunen = to pass on a rumor in a very low voice.)

  29. I would argue that a script _is used_ to represent a given language from the moment when the rebus principle is applied, letting signs that have a concrete signification stand in for other similar-sounding words that don’t (or ultimately when signs have lost any concrete signification they once had, like alphabets for instance).

    In theory even a script where every single sign has a ‘basic’ concrete meaning can be used to represent different languages, ordering signs according to their different grammars and applying different sets of rebus equivalents.

    I don’t know if any cases are known where a symbol system is used to communicate complex meaning without depending on a specific spoken language for structure. (Potentially begging the question of what complex is, of course — but the preamble to the declaration of human rights seems to be a popular thing to try).

  30. Well, if they had invented papyrus or something, or wrote on textiles or leather, we’re not going to find out anytime soon.

    True, but the same could be said for any civilization with no evidence either way. I can see the argument for multiplying the ol’ entities in this particular case, because the Indus script is certainly something, but personally I’m not convinced. If archaeologists dig up some earth-shaking new evidence, I’ll revise my opinion. (Because obviously I would love to be proved wrong here — the more ancient scripts, the better, as far as I’m concerned.)

  31. perhaps my favourite line in the Sumerian corpus, “Nothing is of value, and yet life is so sweet.”

    I like that; what is it in Sumerian?

  32. ə de vivre says:

    Re: Banana languages. I wouldn’t put too much stock on there being any actual linguistic connections between “Proto-Euphratean” and the language(s) of the IVC. Some of those (C)VCVCV names (Inana, for example) probably have Sumerian etymologies, especially since newer research indicates that the language had 3 different types of more or less productive reduplication.

    The owners of IVC seals might also have names like ‘Eight Deer’, who’s named after the day he was born. There’s also no reason the IVC seals have to bear personal names. They might belong to the ‘Ten Tribes’ like the Onoghuz Turks, or to ’50 House’, the largest temple in Girsu.

    A good deal of the FSW paper seems likely true. I just think the analysis really struggles when they try to bring their statistical data (writing as an autonomous system) to bear on writing as a culturally bound phenomenon.¯\_(ツ)_/¯

    Wow, that beats Cārvāka by, what, a thousand years?

    There was a strong tradition of ritual lamentation and the fickleness of gods in Ancient Mesopotamia. It’s maybe not a huge leap from “sometimes the gods ignore us” to “do the gods ever hear us?”. I imagine that even back then the fact that bad things happened to ritually observant people made people question things, after all ancient people weren’t any stupider than we are now. I’d bet it’s more an issue of those kinds of counter-narratives finding expression in writing. Most of the literary Sumerian tablets come from school exercises, so it might be that scribes had their own private jokes about egotistical kings and holier than thou priests (who weren’t always identical with the most literate classes), but that’s just my own speculation.

    I like that; what is it in Sumerian?

    “niŋ₂-nam nu-kal zi ku₇-ku₇-dam”, 𒃻𒉆 𒉡𒆗 𒍣 𒆯𒆯𒁮 if you’ve got the fonts. Which was probably pronounced something more like [niŋnam nukʰal tsi kʰukkʰudã].

  33. Thanks! As for the fonts, would you recommend I download CuneiformNA, CuneiformOB, or Cuneiform Composite?

  34. ə de vivre says:

    Cuneiform Composite is the best for general use. It covers all the unicode points and maintains distinctions where CuneiformNA, which is based on a later style, has collapsed them. Personally, I think CuneiformOB is the prettiest, but it’s sadly lacking a few important signs.

  35. I don’t know if any cases are known where a symbol system is used to communicate complex meaning without depending on a specific spoken language for structure.

    Blissymbolics certainly fits that description, but that is what Peter Daniels (of Daniels and Bright) calls “sophisticated grammatogeny”: it’s not clear that it could have been invented from scratch by someone who did not know ordinary writing.

  36. Well, I thought I downloaded and extracted it to my Fonts folder, but it doesn’t appear to be in there. Do I have to restart my computer? As you see, my ignorance about these things is massive.

  37. If you didn’t see a pop-up telling you that new fonts were being registered, try extracting to a temp folder and then dragging and dropping into Fonts.

  38. Blissymbolics — and that sort proves my point because there are lots of little squiggles that don’t represent concrete objects. True that they aren’t derived from spoken words by rebus substitution, but as mentioned that is an idea that might be easier to come up with now than 5000 years ago.

    And even though Blissymbolics claims that it’s not an encoding of English (which I’ll tentatively agree with) it’s still anchored in Western culture. I’m sure that HEART+FIRE+VERB = ‘want’ is not a universal of human experience. (As an example found on the Wikipedia page). It ould just as well mean hate.

  39. David Eddyshaw says:

    @Lars:

    You’re surely right about that. Not exactly the same, but close: in Kusaal “My heart has gone white” means “I am angry”; “My heart is cool” means “I’m happy.” There isn’t an idiom “My heart is hot” AFAIK. The word for “common sense” is “gall” (as in “gall bladder.”)

    In Seri “I am angry” is “My spirit stinks” and “I am happy” is “My spirit lands.”
    In Fongbe, “joy” is the compound “belly-open” and to ignore or neglect someone is to “hideous-reptile” them.

    Talking of Western cultural presuppositions reminds me that in rural Turkey some decades ago I discovered by awkward experience that the little figures of men and women on the doors of public lavatories in fact featured a lady in modest Muslim trousers and a man in an evzone-style skirt/kilt.

  40. Some say the world will end in fire,
    Some say in ice.
    From what I’ve tasted of desire
    I hold with those who favor fire.
    But if it had to perish twice,
    I think I know enough of hate
    To say that for destruction ice
    Is also great
    And would suffice.

         —Robert Frost, “Fire and Ice”

  41. David Marjanović says:
  42. LH:
    ———————————————
    perhaps my favourite line in the Sumerian corpus, “Nothing is of value, and yet life is so sweet.”

    I like that; what is it in Sumerian?
    ———————————————–

    tu’lu’bogh lo’laHghach ‘ej ‘ach yIn vaj quv.

  43. David Marjanović says:

    One of many things that sound better in the original Klingon.

  44. Except that it is, as usual, gibberish: ‘ej ‘ach ‘and but’.

  45. Owlmirror says:

    [2nd attempt — I tried correcting a few errors that I spotted before the time elapsed, then clicked “Save”, and the comment disappeared. Delete if duplicate.]

    @ David Marjanović:

    Long, long ago (possibly around the time Star Wars came out, actually) there was an article in Scientific American (which I read much later) that briefly presented a decipherment as Proto-Dravidian which exploited the fact that certain words are homonyms in Dravidian and apparently share a sign in the Indus script. Does anyone know why this hasn’t caught on?

    A bit of archive searching on Nature’s website (SciAm now appears to be a Nature property) found several articles on the Indus Valley civilization, and two that specifically mentioned the Dravidian notion.

    The first was a brief reference (in the “Science and the Citizen” column) to a reference (The Indus Script Deciphered? Clauson, Gerard;Chadwick, John — Antiquity; Sep 1, 1969) to the actual work in question: Asko Parpola’s 1969 Decipherment of the Proto-Dravidian Inscriptions of the Indus Civilization, in The Scandinavian Institute of Asian Studies, Special Publications, No. 1) with co-authors Seppo Koskenniemi, Simo Parpola and Pentti Aalto.

    The second was a 1983 article by Walter Fairservis: “The Script of the Indus Valley Civilization”, with photographs and illustrations (which does not even mention Parpola by name, oddly).

    The latter is slightly closer in time to when Star Wars came out; the former, though, is better described as being brief.

    As to why it didn’t catch on — well, the paper that John Cowan links to @January 30, 2017 at 5:56 pm ( Farmer, Sproat, and Witzel 2004 ) discusses the work of Parpola and Fairservis.

    The publication in Antiquity I mentioned looks a bit skeptical; there are more reviews, and another one that I found (Indo-Iranian Journal Vol. 12, No. 2 (1969-70), pp. 126-134, by Arlene R. K. Zide and Kamil Zvelebil) was very critical indeed.

    Some comments extracted from the above review:


    Further, as far as the concrete readings for Dravidian contained in the “First Announcement” are concerned, fifty percent or more of the Dravidian equations are incorrect. They unfortunately reveal – or at least a good many of them do – a not too firm grasp on the nature of comparative linguistic methodology, particularly as applied to Dravidian. Though there are many instances, we shall quote only a few that will serve as flagrant illustrations.


    * Vide p. 21, where Parpola says, quite mistakenly, the order in Proto-Dravidian was stem + case + plural. All twenty-two Dravidian languages known to date including a language (Tamil) whose records go back as far as the third century B.C., have the order of morphemes : stem + plural + case. How does Parpola reconstruct the reverse order for the proto-language?


    Thus, one must note that the monograph is characterized by a great many inconsistencies, and careless errors. It is most unfortunate that this should be so, since considering the potentialities for significant accomplishment which are contained in the varied abilities of the authors, their access to computers, and the general courage required to tackle an undertaking of this nature, one would have hoped for a more accurate and fruitful product. The reviewers hope that in their forthcoming publication(s), Parpola et al. will be able to incorporate more rigorous and promising procedures and methodology. […]

    The review also encompasses a follow-up work by Parpola:

    Progress in the Decipherment of the Proto-Dravidian Indus Script, by
    Asko Parpola, Seppo Koskenniemi, Simo Parpola and Pentti Aalto
    (= The Scandinavian Institute of Asian Studies, Special Publications,
    No. 2). Copenhagen, 1969. 47 pp. Notes, Tables, Bibliography. (Paper.)

    More extracts from the review:


    So that, as regards specific Dravidian examples, though there is some improvement in this respect in the second publication (e.g., the plausible etymologies suggested on p. 17), which show that the authors are aware of some basic Dravidological literature like Burrow’s papers […] they have committed several errors which are unpardonable.


    From the point of view of comparative Dravidian linguistics, the derivation of Śiva from PDr. *ceva, civa, etc., “red” is totally impossible since there is no such PDr. reconstruction as *ceva, civa “red”. It is surprising (and somewhat staggering) that this team of scholars is capable of such an error; they say, “PDr. *ceva, civa, ‘red’ (DED 1607, found in all Dravidian languages)”. Anybody who knows something about the methodology of comparative and historical linguistics sees at once, when consulting the entry DED 1607, that the PDr. reconstruction of this root is not *ce- but *ke-, since in all the languages except Ta.Ma. (and sporadically Te.), the item begins with *k- or, in NDr, by its regular correspondence, [x]; and though *k- > c- is simply accounted for in terms of palatalization, the reverse *c- > k- would be unexplainable in the respective environment(s). It is now exactly twenty-six years since Burrow in his “Dravidian Studies III”, BSO(A)S, 11 (1943) discussed the sound-change in question. Hence, the derivation of Śiva from a PDr. form beginning with the palatal and meaning “red” has been invalid for more than a quarter of a century now.


    The rest of this second publication abounds in quasi-cultural and religous speculations that, aside from their relative unimportance for the decipherment and explanation of the mechanics of the script are, to say the least, somewhat premature. […] Further, the text abounds in miscellaneous loosely-supported statements, such as that on p. 42, where they say, “we may occasionally find in the inscriptions Munda words borrowed into Proto-Dravidian”. The evidence of such (not-yet-conclusively-reconstructed) “Proto-Munda” words has yet to be produced for Proto-Dravidian (itself hardly that solid), or even for the extant Dravidian languages. One also wonders on what (or whom) they base their various assertions about Munda.


    And yet it should be noted that the reviewers would hope that the Parpolas might continue their work, for with the availability of the use of the computer as a tool, the knowledge of Assyriologists and Egyptologists as well as the background Asko Parpola himself brings to the work in terms of Ancient Indian history and language, there still exists promise for a measure of success, given more rigorous and scholarly methodology and patience. One awaits their next publication eagerly, in the hope that it will incorporate the information and competent linguistic methodology available to them, as befits a scholarly publication, rather than the careless, if highly exuberant, manner more characteristic of sensational journalism.

    (Sorry this was so long, but I thought it worthwhile to show how they demonstrated Someone! Is! Wrong! In! The! Literature! syndrome)

    I find myself wondering something else: Another paper on the Indus civilization describes finding bone assemblages of domestic animals, and human grave sites. Might it be possible to do genetic assays of those remains, and try and figure out where the descendants of the civilization, and/or their animals, have ended up currently? That might at least point to what languages their descendants were/are speaking, and therefore provide the possibility of some clues to firmer ground for linguistic reconstruction.

  46. David Marjanović says:

    Wow, thanks!

    Might it be possible to do genetic assays of those remains, and try and figure out where the descendants of the civilization, and/or their animals, have ended up currently?

    Sure – assuming any DNA is left, which isn’t exactly guaranteed by the climate of the area.

  47. Earlier this year it was announced that DNA has been extracted from sketelons found at Rakhigarhi, the site of an IVC city. Sequencing results have not yet been published.

  48. David Marjanović says:

    Awesome.

  49. David Marjanović says:

    The second was a 1983 article by Walter Fairservis: “The Script of the Indus Valley Civilization”, with photographs and illustrations (which does not even mention Parpola by name, oddly).

    The article I was thinking of is indeed by Walter A. Fairservis, Jr., who has also written a book I haven’t seen:

    The Harappan Civilization and its Writing. A Model for the Decipherment of the Indus Script. Brill (Leiden) 1992.

    I’ll have to read the Farmer, Sproat & Witzel paper, then. Fairservis’s article relies heavily on signs being used for two different words that are homonyms or near-homonyms in an unspecified reconstruction of Proto-Dravidian. Oddly, it suggests a hierarchical society (lots of “chiefs” and “politeness syllables”), while IVC sites look really remarkably egalitarian from the archeological point of view.

  50. David Marjanović says:

    I’ve read the paper. It’s good, but:

    1) Writing systems capable of use for a wide range of purposes won’t necessarily be used for a wide range of purposes. There are books in runes, but they’re from Christian England, apart from a few law codes from Christian Sweden; there are casual graffiti in runes, but they come from a time when every Viking had become a cheap version of a runemaster and the religious associations were apparently gone.

    2) Page 46: “It should finally be noted that recognition that the Indus civilization was not literate helps explain a number of well-known features of the society that distinguish it sharply from third-millennium literate civilizations (D.P. Agrawal 2001, personal communication). A few of these include the lack of monumental architecture, large temples, massive standing armies, and clear evidence (besides the tantalizing suggestions in the inscriptions) of large-scale bureaucratic organization.” That’s shortly before the lack of writing in the Aztec and Mixtec civilizations is mentioned, and those had all of these things except, I think, the large-scale bureaucracy.

  51. @ DM: Lack of writing for the Aztecs? Am I missing something?

  52. David Marjanović says:

    Lack of writing for the Aztecs?

    They were just beginning to experiment with that “picture-writing” system for representing names (incidentally the very purpose most often suspected for Indus inscriptions). By the definition of Farmer et al. (2004), that’s not closely enough linked to language to count as “writing”.

  53. Also there are ways and ways to be literate: khipus certainly count as writing as far as I am concerned, though they aren’t used to represent a specific language.

  54. Doug Henning Jr says:

    The Sumerian line referenced above is the Instructions of Šuruppag line 242:
    niĝ2-nam nu-kal zi ku7-ku7-da
    niĝnam nu=kal zi kudkud=’a
    anything neg=rare/valuable life sweet(redup.imprf)=rel. clause(?)
    “Nothing is of value, life being sweet.”

    The next line is:
    niĝ nam-kal-kal-en niĝ-e me-kal-kal
    niĝ na=m=kal[kal]=en niĝ=e m=e=kal[kal]
    thing prohibitive=ventive=valuable[imprf]=2SAgent thing=ERG VENTIVE=2SPatient=valuable[imprf]
    “Do not be of value to things. Things are of value to you.”

  55. Thanks!

  56. David Marjanović says:

    I just read the Fairservis article again. It’s actually very short, so without the 1993 book I can’t provide a fair assessment, but it does provide arguments that the Farmer et al. paper doesn’t address. First, overlapping strokes show that the signs were drawn right-to-left (unspecified work by Lal and the abovementioned Mahadevan is mentioned). Assuming a general right-to-left direction of writing, it turns out there’s some kind of syntax: some signs only appear at the beginnings of inscriptions, others at the ends, and several of the common ones are always in the same order whenever two of them appear in the same inscription. This alone doesn’t disprove a symbol system as assumed by Farmer et al., but it’s certainly a more parsimonious assumption that we’re looking at the syntax of a language. Next, of course there’s no question that Dravidian is a good candidate; better than that, however, there are signs that consist of 1 through 7 vertical strokes (6 is 3 on top of 3, 7 is 4 on top of 3), never more. If these are number signs, they suggest that 8 had a symbol of its own. (Fairservis did propose signs for 8, 9 and 10 in a figure, but didn’t explain why.) In the Dravidian languages, “8”, “9” and “10” are literally “number”, “many minus one” and “many” (citing the Dravidianist Kamil V. Zvelebil). This is evidence that the Indus script is at least slightly more tightly bound to language than Farmer et al. thought.

    Assuming Dravidian explains, wrote Fairservis, why a sign interpreted as “a plant – presumably a cereal stalk”* occurs in connection with numerals so often: “Which language has a word for cereal that also means month or moon, and also uses a base-eight number system? Concerning the first point, nel means rice in five Dravidian languages, while nilā or nela means moon in three of those and in five other languages of this family. In some of these languages it also serves as the designation for month.”* The rebus principle, the use of pictograms for homonyms and near-homonyms found very early in China, Mesopotamia and Egypt, is exactly what Farmer et al. claim was absent. Fairservis goes on to provide several more examples, some too briefly for evaluation, but here’s an interesting one: “Next, let’s look at the sign that dominates so strongly in column 5 [of the grid used to show some of the evidence for syntax]. A few years ago Mahadevan concluded that this sign, which ends most seal texts, was the pictogram of a pitcher with handles. He further pointed out that several Dravidian words for such a pitcher are homonyms of terms that mean ‘male’. One of them is the suffix an̠, which has been appended to the names of male persons as an expression of respect at least since the beginning of the Christian era. The fact that such a sign ends seal texts which are thought to contain personal names corroborates Mahadevan’s conclusion.”* Further, homonymy may explain what numbers are doing on seals: “3” is mu(n), “most noble” is mun̠; “4” is nāl, “good” is nal.

    The end mentions “perhaps to prove even more conclusively that the Indus culture, being an ancestor of Indian village culture, has never disappeared”* as part of a goal. Frankly, that’s bad.

    There is no sharp boundary between language-independent symbols as described by Farmer et al. and logograms; I’m sure many Indus seals can be interpreted either way, the difference being just that Fairservis assigns sound values where Farmer et al. think that’s missing the point. But if there’s evidence for the rebus principle, it’s also evidence for a script by the definition of Farmer et al., even though this script wasn’t used for Mesopotamian-style bean-counting and may not even have been suitable for such tasks.

    * My retranslations from German.

  57. Very interesting!

  58. SFReader says:

    There was a post-conquest phase in Nahuatl writing which lasted a few decades, before final shift to Latin script.

    In that phase, the Aztecs quite ingeniously were using their characters to write quite a few Spanish names.

    For example, a sign “water” (pronounced in Nahuatl as “A”) and a sign bird” (“toto” in Nahuatl) on top of it were read as A:to (ie, Anton – Nahuatl version of Spanish name Antonio).

    Any Spanish name or borrowed Spanish word could be represented by this system.

    to-min-co (Domingo), cax-ton (Gaston), ton a-lon-xo (Don Alonso), lo-ix (Luis), men-toza (Mendoza), i-xi-le (for Spanish visorrey – viceroy), pala (Sp. fray – friar) and my favourite – xan pala-ci(ci)-co (San Francisco).

    If it’s not writing, then what is writing?

  59. SFReader says:

    This has to be seen.
    https://ibb.co/h0zygQ

  60. Concerning the first point, nel means rice in five Dravidian languages, while nilā or nela means moon in three of those and in five other languages of this family. In some of these languages it also serves as the designation for month…

    One has to be careful with such arguments, impressive as they might look at first blush. The Dravidian languages, understandably, have rich vocabularies connected with the lexical field of rice farming (which facilitates pseudo-etymological cherry picking). It doesn’t mean that all modern Dravidial forms can we projected back several thousand years. It would be like listing English rice, Hungarian rizs, Basque arroz etc. and announcing that we have reconstructed an ancient European word for ‘rice’.

    As a matter of fact, English rice is ultimately Dravidian. It comes (via Greek and Romance) from an Iranian word related to Skt. vrīhi, which in turn comes from the actual Proto-Dravidian ‘rice’ word, *war-inci-. It’s true that PDr. *nel-a-nc(c)- meant ‘moon, moonlight’ (hence Tamil nilā, etc.), and that the ‘month’ word (*nel-V-) was related to it, but the argument based on its alleged similarity to a ‘rice’ word is totally spurious. The same goes for other such anachronistic “lookalike equations”. There was a masculinising suffix (not a noun), *-anṯu, and there were several words with the approximate meaning ‘man, male, husband, lord’ to choose from, but any similarity of any of them to a modern ‘pitcher’ word is just as probative as the homophony of Modern English ewer and you’re. Surely any “proto-writing” system in which a pitcher-like symbol occurs frequently must be English, since in no other language can a word for ewer stand for something as common as that.

  61. By the way, did you know that ewer = aquarium (etymologically)?

  62. Wells might learn something if he listened to his garbage man, instead of imagining him as an epitome of ignorance. (Full disclosure – I have been a sewage worker in my time.)

    Petty credential games, apart from being a form of ad hominem, are also a weapon that cuts both ways. I won’t say more coz the guy got himself an AK-47.

    https://ca.linkedin.com/in/bryan-wells-86b82269

  63. minus273 says:

    ewer = aquarium

    Yeah, even French people pretty non-deficient in Latin can be tripped by évier.

  64. David Marjanović says:

    Hm. It didn’t occur to me that Fairservis wouldn’t use an actual reconstruction of Proto-Dravidian. I thought he was just illustrating what the reconstructions were based on… That’s bad, then.

    By the way, did you know that ewer = aquarium (etymologically)?

    No, but I didn’t know the word ewer in the first place.

  65. The word for “common sense” is “gall” (as in “gall bladder.”)

    One of my favorite lines is from Oliver Cromwell (a moderate Puritan) addressing some extreme Puritans: “Bethink ye, bethink ye, in the bowels of Christ, that ye might be mistaken.”

    It wasn’t quite as funny in Early Modern English, where bowels was a standard biblical (Greek? Hebrew?) metaphor for ‘compassion’.

  66. One of my favourite quotes too, though the form of the quote I know is “think it possible you may be mistaken”, as quoted by Jacob Bronowski in a very powerful moment in The Ascent of Man. There’s also the background to it in an interview with Michael Parkinson, one of Parkinson’s favourites.

  67. One of mine as well.

  68. The quotation is from Cromwell’s letter To the Generall Assembly of the Kirke of Scotland; or, in case of their not Sitting, To the Commissioners of the Kirk of Scotland (1650). He wrote:

    I beseech you in the bowels of Christ, think it possible you may be mistaken.

  69. The question is, of course, whether Cromwell thought it possible that he might be mistaken. It’s always easy to press doubt on the other guy!

  70. I haven’t had time to read through all the comments here … too many of them, but I just thought I should add a few cents’ worth to a couple of points made here.

    Locklear’s piece is indeed, as has been suggested, largely a regurgitation of old stuff. Really the only reason I agreed to be interviewed for it was that if I didn’t, I feared that the “other side” would get all the coverage.

    My points that she discusses (she also cites a 2014 paper in Language, IIRC), are really quite simple: the entropic measures that Rao and colleagues propose are simply not useful for telling us anything about *what* the Indus Valley symbols represented. To the extent that they are informative about structure in the system (and even that is not clear) the presence of structure in a system does not tell us that it is writing, and therefore represents spoken language in some way. So one can conclude nothing from their results, other than (maybe) it was a system that had some structure to it. But we’ve known that since at least the 1930’s.

    As Peter Erwin says, I do have a background in linguistics — my PhD was in it, and I’ve worked for 30 years as a computational linguist in industry and academia, and I have also worked a lot on writing systems, so I do understand something about this area. But as has also been pointed out, professional background isn’t determinative here: ultimately the argument I make is very simple and shouldn’t require any specialized background to understand.

    So Brett is wrong about that point: the entropy statistics tell you nothing about whether this was a real language.

    The 1984 reference is, however, apposite: one of the characteristics of true decipherments is that once one has had a breakthrough, it leads to fairly rapid cracking of much or all of the rest of the code. Witness Rawlinson’s decipherment of Babylonian, or Ventris’s decipherment of Linear B, where once the key was found, the rest followed pretty quickly, well within the lifetime of a single individual. Egyptian took maybe longer, but still things progressed very smoothly after Champollion found the key.

    Contrast that with the situation for the Indus system: Parpola and his team claimed a breakthrough in the late 1960’s, but almost nothing has happened since. Oh, sure, Parpola has occasionally found an interpretation of one or another symbol, but nothing that has allowed him to read a full text. Supposing that the Indus symbols are a full writing system, that suggests that the key has not been found. If the language was indeed some early form of Dravidian, then once that initial key had been found, others should have followed. Of course the extreme brevity of the inscriptions (one of several reasons that Farmer, Witzel and I were suspicious) makes it all too easy to fit interpretations to the texts (witness the 100’s of “decipherments” that have been claimed), so even if one had come up with readings of whole texts, one would have to be a bit skeptical. And if the symbols don’t represent language, then naturally the attempts at decipherment are pointless.

    I am not naive enough to think that this issue is going to go away: people will continue to try crack this code. But I do think that we provided a valid alternative to the standard view that has been with us for over a century, and that my own work on the statistics of symbol systems has called out the dangers of overly facile interpretations of statistical results.

  71. Very interesting; thanks for taking the time to comment!

  72. One of the characteristics of true decipherments is that once one has had a breakthrough, it leads to fairly rapid cracking of much or all of the rest of the code.

    And vice versa: Gordon’s Linear A, Fischer’s Rongorongo, and many other claims start with decipherments for a handful of signs or words, and that’s the last we hear of them.

Trackbacks

  1. […] Hat links to an article concerned with the question of how to try cracking the Indus Valley […]

Speak Your Mind

*