Mayan Spelling Patterns
[Revision of 4 March, 2008; 11 March 2008 addition of paragraph on CV-CV spelling logograms; 18 June 2008 slight editing for clarity and addition of note that Förstemann had discovered synharmonic spellings; 24 November 2008 added references to Josserand and Hopkins 1991, Lacadena and Wichmann 2008, Tanaka 2008, a separate page stating the connections between consonant-vowel correlations, syllable structure, and spelling patterns; and bits of clarification throughout]
This page is devoted to the so-called "unpronounced" vowels in the last syllabary sign used to spell words ending in a pronounced consonant. At the end is a short review of a recent paper on this topic in Research Reports on Ancient Maya Writing., and a listing of papers on this subject.
The beginnings of this topic were already in the discoveries by Förstemann and Knorozov of sequences of two syllabary signs with the same vowel, used to spell a word ending in a consonant, like ku-tzu used to spell kutz 'turkey' or tzu-lu used to spell tzul 'dog'. From the long version of the film "Breaking the Maya Code" we learn that Förstemann already had figured this out, but was discouraged when a critic said that the Maya word for 'dog' was pek not tzul (he had not found tzul in modern vocabulary lists, did not search hard enough to find pek.). This was taken as a counterexample and delayed development of the field. Knorozov was able to take the next step.
Still within the framework of synharmonic spellings, it was possible to consider a minor theme, the combination of syllabary signs to make either more complex syllables or else to function as new logograms. Lloyd Anderson noted to George Stuart at a AAA meeting in Chicago (1984 or 1987?) that a lot of infixation seemed to be signaling syllable-final consonants, so typically chu[ku] spelling chuc, the glyphically fossilized month name mo[lo] spelling mol, etc. The more common pattern is for the containing element to represent the syllable-initial, the infixed element to add the syllable-final consonant, despite exceptions such as ja[la] spelling either jal or laj, perhaps because these particular graphic shapes support easily only one kind of visual combination. Nick Hopkins proposed long ago (date?) that CV-CV were functioning as CVC logograms. Mark Zender proposed that pairs of syllabary signs in a long string were to be read as CVC, counting odds and evens, and that seems usually to be the case with a slight emendation, that we start counting from the root (or stem?) of a word, excluding the prefixes. Has anyone tested whether this reflects a phonetic dropping of alternate vowels, or precisely which examples would be consistent with such a formulation, and which would not be accounted for?
At a later point, exceptions were noted to synharmony. Josserand and Hopkins pointed at first to a rather common occurrence of a final syllabary sign <ti> with vowel /i/ when the first syllable had some other vowel. They later generalized this to final alveolar or apical consonants ( /t, t', tz, tz', ch, ch', n, s, x/), in their 1991 grant report to National Endowment for the Humanities. The most relevant part of that text is extracted and available here. It turns out that this generalization is too broad, but it does contain a large proportion of the exceptions to the pattern of "unpronounced" vowels being identical to preceding vowels.
Anderson early noted that the vowel /i/ is the most "closed" vowel, phonetically appropriate either to a phonetically reduced vowel or to an unpronounced vowel which might historically have once been pronounced. A comparison seemed apt especially with Japanese /u/ (high back unrounded) which especially after voiceless consonants devoices or even vanishes, finally or as the middle vowel in a sequence CVCVCV(C). Yucatec words can also delete medial vowels in this last context.
This range of questions took on forms in which they are now most discussed after a presentation in Austin, Texas in 1999 (?) by David Stuart, Stephen Houston, and John Robertson, where they suggested that in some cases a distinction of vowel quantity in roots was reflected by a difference in how the last consonant of a root word is spelled, by the choice of which syllabary sign would normally be used for it -- whether the vowel in that syllabary sign was ("synharmonic") the same as or ("disharmonic") different from the vowel pronounced in the final syllable of the root or stem, before the last consonant ending the word. This was a bit better match for the data than Josserand and Hopkins's simple rule about syllabary signs with /i/ if the consonant was palatal or alveolar. But this explanation by length of root vowel was proposed as an abstract rule with no phonetic motivation, and as an alternative to rather than an addition to the approach of Josserand and Hopkins..
There has been substantial work on this since, with further collection of examples by Robertson, Houston, Stuart, and others, and contributions by Anderson, by Alfonso Lacadena and Soeren Wichmann, by Terrence Kaufmann assisted by John Justeson, and by Tanaka..
The data which was being discussed a few years ago was integrated by Anderson into a single chart with explanations. The chart version dated 8 October, 2004 is available as a PDF file here. The legend at the bottom of the two pages of that chart summarizes which parts of the data on spelling patterns is accounted for by each explanation or "rule".
The explanatory framework, the logic linking consonant-vowel correlations, syllable-structure, and spelling patterns, proposed by Anderson in 1999, is here. The idea that the "unpronounced" final vowels might once have been pronounced, thus providing one possible basis for the phonetically reasonable parts of the observed spelling patterns, came earlier than that. (There might be other origins as well, that is just one plausible one.)
The chart was distributed at the Austin meetings in 2004 or earlier. There have been only a couple of changes of wording in the chart legend to make it simpler to understand. (The data itself agrees almost entirely with what Stuart, Houston, and Robertson had collected in the 1999 Austin workbook and what they had added later, up to Fall 2004, with the exception that some examples which do not fit their rules are included, rather than being excluded. However, the reasoning is very different, and the chart with its explanations is explanatory rather than purely descriptive.)
There seems to be no reason to change the basic conclusions presented along with that chart, though a few example words might surely now be added, deleted, or modified.
Those conclusions were that at least three hypotheses appear to be needed. There are exceptions to each of the three, but each of the three also explains parts of the pattern which the others cannot. The first two hypotheses seem to have phonetic motivations (at least conditioning contexts which make sense phonetically), while the third does not. This is so despite the fact that the last vowel is regarded as "unpronounced". That would remain a bit of a paradox unless unless we discover a chain of historical linguistic events giving rise to the spelling pattern. Already in 1991, Josserand and Hopkins clearly stated the existence of "echo vowels", which may provide part of such an explanation. Phonetically motivated explanations can be distinguished from a hypothetical mostly arbitrary rule invented by some scribe in order to make a distinction felt as needed, which is more the way some have looked at this. A fourth hypothesis has not been fully integrated with the first three because the data set has not yet been made fully accessible to the public. This fourth hypothesis may have a motivation from morphological analogy. Here are the four hypotheses:
(a) Consonant and Vowel of the Same Point of Articulation
A single hypothesis (by this author, reflecting thinking similar to that of Josserand and Hopkins 1991) is available here. It was formulated immediately after the first description of the data in Austin by Stuart, Houston and Robertson 1999, except that the role of glottalized vowels and consonants was added after Lacadena and Wichmann (2004). This explains a large part of the data, that the spelled "unpronounced" vowels closely reflect the points of articulation of the root-final or stem-final consonants in the "disharmonic" cases, that is when the root contains a long vowel. This hypothesis regards the observed patterns as being a result of the cumulation of results of phonetic influences from some earlier stage.
Very direct evidence showing which vowels are most similar in articulation to which consonants is available in an article in the journal Language vol.46 no.1 (1970) pp.77-88, by Sarah C. Gudschinsky and Harold and Frances Popovich of the Summer Institute of Linguistics: "Native Reaction and Phonetic Similarity in Maxakalí Phonology". The abstract states that "some [phon]etically disyllabic sequences of the shape [CVCV] are interpreted as [phon]emic monosyllables of the shape /CVC/". The relations are these (from pages 83-84 of the article), for vowels which are the manifestations of consonants at the four points of articulation. (If the consonant is nasal, then its vocalic allophone is nasalized.)
i with c ñ palatals (high front)
ï with k ng velars (high back unrounded)
ë with p m labials (mid back unrounded; labials are not a
"rounded" but a flat labial articulation)
[shwa] with t n apicals, alveolars (central vowels but with
considerable variation, low to high)
Transferring this to the Mayan context, where there are no distinctive back unrounded vowels and no distinctive shwa vowel, we would have the following:
/i/ with "ch"
/i/ or /u/ with "k" (velar consonants often manifest splits into front and back correlated with vowels)
/a/ with "p" (there is some typological evidence for a statistical correlation of /a/ with labials)
/e/ or /i/ with "t"
Very similar correlations, with high frequency, do account for much but not all of the choices of Mayan "unpronounced" but spelled vowels. Details for the Mayan case are in the chart here.
(b) Influence of Glottal Stop or Glottalization of Vowels
This second hypothesis remains not very different from the way Lacadena and Wichmann (2004) presented it. Since this one is also phonetically plausible, showing the kinds of effects phoneticians might expect from a glottal stop, and most evident in contexts where they would be expected, its explanation may also go back historically to phonetic origins. Commonly, a disharmonic sequence like ch'e-na or cha-nu ("unpronounced" /a/ after root vowels /e, o, u/; "unpronounced" /u/ after root vowel /a/) signals a glottal before a final resonant (l n), so these are read ch'e'n and cha'n. Disharmony can also reflect a glottalized stop, as in to-k'a for tok'. Conditioning contexts discussed here reflect a focus more on articulatory and acoustic causes. Particularly this applies to the favoring of sonorants through which glottalized vowels can extend their effects ( /n/ and /l/ are sonorants). By contrast, the rule as formulated by Lacadena and Wichmann is stated more generally and formally.
(c) Disharmonic Spellings Only with Long Root Vowels
The seemingly arbitrary rule of "synharmonic" vs. "disharmonic" spellings (same vowel vs. different vowel, disharmonic only possible if the root vowel is long or heavy, but varying with synharmonic even there) explains a part of the data which neither of the first two rules do. It has no obvious phonetic motivation. It is not a bidirectional implication, but a unidirectional one. When restricted to some subsets of data which rules (a) and (b) do not explain, it becomes more bidirectional "if-and-only-if". This applies to examples (using "C" for any initial consonant) Ca-ba read Cab vs. Ca-bi read Caab, and similarly for final b, t, k, j, h.
Addition in 2008: there is a possibility of predicting (rather than merely describing) that variation in final "echo" vowels would survive only with long root vowels, see here. But it appears so far not to be possible to predict precisely which final consonants show more numerous examples of the contrast CaaC-ci vs. CaC-ca. Those are simply listed in the notes to the table here.
(d) Correlation of Bare-Root Spellings with Suffix Spellings
A rule that the spelling of derived forms with a -V:l suffix predicts the spelling of the root when the final vowel is "unpronounced" (namely the same syllabary sign as if the vowel were pronounced) may be valid for many cases. This has been proposed by Terrence Kaufman (2003:29-34). The form of the statement of this rule is much like what linguists call "morphological analogy", in which one form of a word influences another form of the same word. At least earlier, Terry Kaufman was not comfortable treating it as analogy, and he appeared to regard it as a psychological rule rather than as a cumulative result of phonetic causes. Since the -V:l suffix forms are not as common as the bare root or stem forms (for nouns), linguists would normally expect the analogy to go in the other direction, that is to say, the spelling of the more common root word would cause the spelling of the -V:l suffixed form to use the same vowel. The problem with this last approach (not Terry's) is again that the vowel of the last syllabary sign is "unpronounced" (the one which is used to spell the final consonant of the bare root or stem). Even if one can predict "unpronounced" vowels in consonant-final forms of a word from the vowels used in -V:l suffixes, the problem still remains of explaining which vowels are used in the -V:l or other suffixes. This partly just pushes the search for an explanation back one step, although it does add the interesting observation of the correlation.
(e) "Echo Vowels"
The 1991 Josserand and Hopkins grant report here points to actually pronounced "echo vowels", which add a phonetically natural basis for predicting from root to spelling with a syllabary sign. "Echo vowels" are actually pronounced, so echo vowels could possibly have guided scribes in writing the forms. This general presentation can be supplemented with the Mopan data in Tanaka (2008). The 1970 paper by Gudschinsky and Popovich, mentioned above, also predicts vowel allophones of consonants. It is a typological parallel, it does not contain data from a Mayan language.
Sketch of a Review of RRAMW 62
A recent paper on this topic is one by John Robertson, Stephen Houston, Mark Zender, and David Stuart, published electronically as Research Reports on Ancient Maya Writing No.62, available at http://www.utmesoamerica.org/pdf_meso/RRAMW62.pdf. The title is "Universals and the Logic of the Material Implication: A Case Study from Maya Hieroglyphic Writing"
This is a sketch of a review.
This paper has a very important collection of examples which anyone interested in this problem can benefit from. However, in other respects it is an unfortunate paper. It is primarily John Robertson's writing. Other co-authors should have been more wary before signing on. Except for the valuable collection of examples, this paper is a disservice to the field.
This paper is needlessly abstract and offputting, and seems to be focused less on the content than on the logical but really very minor point that the pattern being discussed (hypothesis (c) of the four above) is not a bidirectional implication, because three rather than only two possibilities occur. Since that is true of rather many phenomena in the real world, it is not particularly remarkable, not enough so to justify the fixation on it which appears in this paper. Here are the three which do occur, in this view:
Root with long vowel, final consonant spelled with a "disharmonic" syllabary sign
Root with long vowel, final consonant spelled with a "synharmonic" syllabary sign
Root with short vowel, final consonant spelled with a "synharmonic" syllabary sign
(In other words, a "disharmonic" syllabary sign is not used to spell the final consonant if the root or stem has a simple short vowel in the last syllable.)
The data appears to be not much different from that in extended internet discussions in 2004. Publication of that data would have been more useful without the heavy overlay of discussion of logic which contributes virtually nothing, which makes the facts less transparent.
But there is one disturbing exception. Some spellings which were clearly exceptions to the generalizations made in those earlier years, and which were at that time admitted as exceptions, are now simply excluded from the main body of the discussion and relegated to the end (pp.54-55). This looks suspiciously like misrepresenting the data set in order to make one's hypothesis look better. The question has to be whether the primary author or the authors tried especially hard to discredit these items precisely because they didn't fit the theory they were advancing. It is of course entirely legitimate to exclude items if one applies the exclusion process in an even-handed manner, and does not select criteria for exclusion simply because one wants to discredit particular parts of the data. The items in question are /utz/ 'good', /tz'utz'/ 'coati', /k'uch/ 'vulture', /ikatz/ 'bundle'. It is certainly interesting that the items excluded have final consonants "tz" and once "ch" and all have "unpronounced" final /i/ in the spelling of their last syllabary sign. Those consonants are phonetically most conducive to an epenthetic vowel of the quality of /i/ (or the phonetic central vowel barred-i, but that is by virtually everyone in the field of Mayan epigraphy and linguistics assumed to not be a phonemic vowel and to not be available as one of the possible choices).
Although the paper does discuss hypothesis (b) above, it omits entirely to mention (a) or the integrated chart which is available here in .pdf form and was distributed in Austin a few years ago, including a copy handed to Robertson in person. It dismisses the work of Kaufman on the topic (hypothesis (d)) in a rather condescending manner. One can add here that Robertson some years ago said he simply did not believe their claim that there is a 90% correspondence between the vowels in -V:l suffixed forms and the unpronounced vowels in final syllabary signs of unsuffixed forms. Given the subject of this paper, a presentation of extensive data to argue for or against Kaurman's hypothesis would have been appropriate. This author's partial attempts to check Kaufman's data convinced him that there is a very substantial correlation, just as Kaufman claimed. It is unclear why Robertson claims otherwise.
There is an even bigger problem. The claimed "hypothesis" in RRAMW62 is not a unified hypothesis at all, but is a list of groups of cases. The paper gives a description of the partial patterns in the data, but does not have a hypothesis other than a list of those patterns. There are no typological parallels offered for such sets of cases, which of course leaves the advocate free to tailor the description to just the data being considered. When one does that, one can no longer claim to have "predicted" that data, although just such a claim is presented. Such reasoning is completely circular.
By contrast, all of the hypotheses (a, b, d) are motivated either phonetically or morphologically by processes known to occur in living languages, and each of those three is reasonably unitary or at least potentially so (the partial exception is Lacadena and Wichmann, who did not extensively address the issue of unifying the cases they observed, but a phonetician can do that plausibly to a substantial degree).
This is obviously not to deny that synharmonic vs. disharmonic spellings did occur with some systematic patterning in Maya writing. They did. But that is only one of at least three parts of the observed patterning, and the one with the least understandable motivation! Languages do have crazy rules, usually resulting from a complex combination of circumstances and then historical changes which remove the clues to the original motivations. Synharmonic vs. disharmonic spellings may be in part a "crazy rule" of this kind. Or there may be some other explanation how and why that factor correlates with a part of the data patterns.
There may also be some other oddities in the account of the history of discovery in the field. Please see the references at the end in chronological order for this author's best understanding of that history.
The only reference to Kaufman and Justeson's hypothesis ((d) above) is this footnote 10: [sic]
"Terrence Kaufman (2003:29-34) has a third proposal for the so-called silent vowels, which we will treat in another paper. Very briefly, he suggests that silent vowel come from the "-V:l suffix that was characteristically suffixed to that root in the Epigraphic Mayan language, which should be best reflected in Greater Tzeltalan languages today" (2003:ibid., p.31). We have carefully considered this proposal on its own merits, and have come to believe that the vocalic nature of the -Vl suffix are so hard to pin down diachronically and synchronically that we flatly reject the suggestion. A discussion of the proposal goes beyond the bounds of this paper, requiring us to show our reasoning in a later paper."
The above seems a rather cavalier dismissal for a reasonable unifying hypothesis, given that RRAMW62 is giving a descriptive list of small patterns without itself having any unifying principle. A "flat reject[ion]" is also not appropriate if an author has so far found it "hard to pin down" the data.
A problem in this paper, as in some previous writing by Robertson, is that he often phrases things to make it appear that he has predicted exactly what is found, whereas in reality it is postdiction not prediction. I could not state it better than was done by an eminent epigrapher in August of 2004:
"why disharmony is marked while synharmony is not [i.e. synharmony can occur with either long or short vowels in the root]. As written, your paper seems to imply that before ever looking at the evidence this was already your hypothesis. You said in the introduction: "But we do not expect all complex vowels to be written disharmonically. In fact, many will have a synharmonic spelling. From the same hypothesis we deduce that all reconstructible simple vowels must have a synharmonic spelling. Yet we do not expect all synharmonic spellings to reference simple (short) vowels -- to the contrary, the data would show many instances of synharmonic spellings with complex vowels." Why should the data show this? I agree that they do show this, but I don't see why they should. Your hypothesis appears to have been crafted only after examination of the data and was designed as broadly as possible to cover the curious fact that synharmony is unmarked and that we have no examples of simple vowels being spelled disharmonically. This is why I see your proposal as strong on description but not necessarily on explanation. Your paper may not have been designed to explain the spelling rules, instead just defining them, but when you use such assertive statements as "(a)nd finally, in accord with the hypothesis, there should not be a single instance of a disharmonic spelling with a simple vowel", you give the impression of attempting an explanation rather than a simple description."
Put more strongly, less in the form of a question, the paper does not have a hypothesis, and cannot claim to have predicted the distributional patterns of spellings. It rather describes those distributions. That can be useful in itself, and will aid others who are looking for explanations.
References limited to the approximate order of discovery and presentation:
Need to add here the earliest approaches to the phenomenon, by Förstemann and by Knorozov;
Explanation (precursor to (a))
Josserand, J, Kathryn and Nicholas A. Hopkins 1991 July. A Handbook of Classic Maya Inscriptions Part I: The Western Lowlands. Final Performance Report, National Endowment for the Humanities, Grant RT:21090-89, pp.123-126
Description (c):
Stuart, David, Stephen Houston and John Robertson 1999. "Recovering the Past. Classic Maya Language and Classic Maya Gods" In Notebook for the XXIIIrd Maya Hieroglyphic Forum at Texas, part II. Austin, Texas: Maya Workshop Foundation [specify the 3? pages]
[and an earlier publication in 1998]
Explanation (a):
Anderson, Lloyd 1999 [circulated privately: first statement of the hypothesis that point-of-articulation phonetics pairs consonants with their most similar vowels, those which preferentially occur in Mayan writing in disharmonic spellings. Also, not as often mentioned since then, a possible reason why disharmony might be restricted to roots with heavy first syllables, namely that they are slightly closer to being borderline between monosyllables and disyllables. (This will seem most reasonable with root-final consonant clusters such as /hk/, /jk/, /sk/, /xk/, which were present at some times in some languages. This also is closer to the "echo syllables" discussed by Josserand and Hopkins 1991.) So consider the remote possibility that the "unpronounced" vowels were once really there in a second syllable. This writer is well aware that Mayan proto-languages as currently reconstructed do not have pronounced vowels recorded as separate segmental phonemes in these positions at ends of roots, but this kind of a pattern could easily have arisen if they had once been there. Josserand and Hopkins 1991 discuss "echo syllables" by which /pat/ is pronounced as [pat.t(i)].
Description (c):
Houston, Stephen D., John S. Robertson, and David Stuart. 2001. "Quality and Quantity in Glyphic Nouns and Adjectives" Research Reports on Ancient Maya Writing 47:1-56. Washington, D.C.: Center for Maya Research
Statement of (d):
Kaufman, Terrence. 2003. A Preliminary Mayan Etymological Dictionary. FAMSI report
http://www.famsi.org/reports/01051/
Description (b):
Lacadena, A. and S. Wichmann 2004 "On the representation of the glottal stop in Maya writing." In Wichmann, Soeren (ed.) The Linguistics of Maya Writing, pp.103-162. Salt Lake City, Utah: University of Utah Press. [developed some years earlier?]
Description (c):
Robertson and numerous other epigraphers:: email discussions in 2004.
Explanation (a):
Anderson, Lloyd 2004 Chart of examples integrating the three hypotheses (a,b,c).
Click here for a .pdf download.]
Description (c): the paper here reviewed.
John Robertson, John, Stephen Houston, Mark Zender, and David Stuart 2007
"Universals and the Logic of the Material Implication: A Case Study from Maya Hieroglyphic Writing" published electronically as Research Reports on Ancient Maya Writing No.62, available at http://www.utmesoamerica.org/pdf_meso/RRAMW62.pdf.
Description (b)
Lacadena, Alfonso and Søren Wichmann 2008 ms. "Harmony Rules and the Suffix Domain: A Study of Maya Scribal Conventions" 50pp. [This paper argues that the same rules applicable to the spelling of final consonants of roots also work well for the spelling of final consonants of suffixes. They distinguish types of vowels in suffixes, for example -ul, -uul, -u'l.]
Explanation (a)
Tanaka, Yuki 2008 "A Comparative Study of Maya Hieroglyphic Writing and Japanese Orthography in Quiriguá Hieroglyphic Corpus". Paper at American Anthropological Association meeting, San Francisco, California.
This document is copyright © 2007, 2008 Lloyd Anderson. All Rights Reserved with the following exception: It may be copied (printed) for use by a single individual as long as it is properly cited and this copyright notice is retained unchanged. A copy shall be made only from the version on this web site, to ensure that only the latest version is copied.
|