Editor's note:
In this issue of Science Online, we present an experiment in web publishing. The print version of Science contains a research report by Justeson and Kaufman on decipherment of epi-Olmec inscriptions on a stone monument. The online version contains both the version which appears in print and a longer version (below) presenting the results of the decipherment in much greater detail. Most readers should find the print version useful for learning about the results in general, whereas specialists in mesoamerican studies will benefit from the expanded description of how the decipherment was carried out. By keeping the print version concise and using the web to present the in-depth analysis for experts, we hope to capture the best of both worlds.

[Reprint (PDF) Version of this article]

A Newly Discovered Column in the Hieroglyphic Text on La Mojarra Stela 1: A Test of the Epi-Olmec Decipherment

John S. Justeson, Terrence Kaufman

A badly weathered column of hieroglyphs was discovered in November 1995 on the side of Stela 1 from La Mojarra in southernVeracruz, Mexico. Most of the signs in this column have now beenidentified by nighttime examination under artificial lighting,making possible a nearly complete transcription and translationof this column. This data expands the modest corpus of epi-Olmechieroglyphic texts and confirms various aspects of the deciphermentof the epi-Olmec script.

J. S. Justeson, Department of Anthropology, State University of New York, Albany, NY 12222, USA. E-mail: jsjusteson{at}aol.com
T. Kaufman, Departments of Anthropology and Linguistics, University of Pittsburgh, Pittsburgh, PA 15260, USA. E-mail: topkat+@pitt.edu

Authorship is equal; correspondence may be addressed to either author.

The remains of a distinctive script tradition are found on monuments from southern Mexico, ranging from Cerro de las Mesasin the north to Chiapa de Corzo in the south. Because the associatedart style and artifacts descend from the Olmec tradition, we referto this script as epi-Olmec.

Only four legible texts currently attest the epi-Olmec writing system (1): three are virtually complete, whereas one fragmentarytext preserves portions of just a few words in parts of two shortsequences. Epi-Olmec signs are visible on a few other, badly weatheredmonuments, but these signs are few and isolated, not forming legibleconnected text. There was too little text data on which to basea decipherment of this script until the discovery and publicationof the fourth legible text, that of La Mojarra Stela 1 (2).This stela bears a lengthy text, one of the longest known fromancient Mesoamerica.

It was this fourth text that made it possible for us to decipher a substantial portion of this writing system (3, 4).The fact that it was a single continuous text was important tothe decipherment. Topic continuity in texts of this length leadsto repetition of lexical items and of larger grammatical units;with a model for the grammar of the text's language, pre-proto-Sokean,this repetition was exploited to determine the boundaries of manyof the words in the text (boundaries are not explicitly markedin the epi-Olmec script) and thereby some aspects of their grammaticalstructure. The text on the Tuxtla Statuette provided importantadditional clues to the decipherment. The state of the deciphermentas of December 1992 is summarized in (3), and its state asof January 1994 is partially described in (4).

In 1994, we were able to make a drawing of the other two short texts. These texts have helped to confirm the essentials ofthe decipherment, to determine the readings of a few additionalsigns, and to refine some of the uncertainties we had had earlier.Additional progress in the decipherment has been achieved throughconstant reanalysis, and by applying insights from several yearsof descriptive and comparative work that we have undertaken onthe extant languages of the Mije-Sokean family [work that wasidentified in (3) as one of the two main avenues for advancingthe decipherment]. Few of our readings or interpretations havehad to be revised: for example, only two syllabic signs in (3)were changed in (4), and only two in (4) have been changedsubsequently. There has, however, been a fair amount of revisionin our knowledge of reconstructed proto-Sokean and proto-Mije-Sokean,based on Wichmann's work (5) and on Kaufman's comparative analysisof data from our own Mije-Sokean Language Documentation Project.

Further evidence concerning the epi-Olmec script now depends chiefly on the recovery of more textual data written in it, butno more texts have since come to light. For now, it remains apoorly attested writing system. A magnetometer survey of the siteof La Mojarra was undertaken in an attempt to locate additionalinscribed monuments there, but it did not prove to be successful(6).

La Mojarra Stela 1 was brought to scientific attention after it arrived at the Museo de Antropología in Xalapa, Veracruz,Mexico, in 1986. The importance of the text was recognized assoon as it was examined at the museum. George Stuart made a drawingof the monument that was produced on the basis of photographsby Logan Wagner and a rubbing by John Keshishian, both made whileit was lying on its back in the basement of the museum; Stuartthen revised the drawing by direct examination with a flashlight.Under such conditions, it is difficult to discern all of the detailsof signs that are at a distance from the edges of the monument;it is remarkable how accurate Stuart's drawing has proven to be,erring mostly in the absence of some lines that are difficultto see. This drawing has been almost the sole basis for researchon this text up to the present time.

For reasons that epigraphers have never found compelling (7), doubt was cast on the authenticity of the monument, and recentdirectors of the museum left it in the basement. Access to themonument was restricted, and even for those of us who were ableto see it first hand, viewing conditions were so poor that itwas difficult to evaluate many details of the text, particularlyon the more badly damaged glyphs.

In September of 1995, Sara Ladrón de Guevara became the director of the museum. Soon thereafter, she arranged to put themonument on public display, which was done in connection witha public ceremony on 24 November 1995. It is now possible to inspectthe details of this monument at first hand. Our own detailed examinationof the whole text has led to improvements in the identificationof several signs, especially those in damaged areas.

A Previously Undetected Column of Glyphs

Early in November of 1995, while the stela was being readied for display, geologist Fernando Muñiz and archaeologist SergioVásquez discovered what appeared to be the remains of a columnof glyphs on the left side (viewer's right) of the monument (8).On 23 November, the day before the public ceremony, Vásquez satisfiedhimself that it was a series of eroded glyphs, a little more than20 by his estimate.

Before going to Xalapa to investigate the matter for ourselves, our knowledge was based on (both first- and second-hand) verbalreports from the scene that it was in poor condition; it was clearthat hieroglyphs had once been present, but they were so badlyweathered that little could be recovered of the original text.It was not even clear if any of the glyphs could be identified,or how long the text had been.

The main reason for the poor condition of the side text is that the stone is built up in thin layers. Most of the side isdeeply scored by weathering, evidently between the layers of stone,and it is badly pitted as well, which makes it difficult to recognizewhat signs are inscribed. It was mainly the surviving horizontallines that made it possible to recognize at the outset that acolumn of text had once been present.

Justeson examined the monument from 10 to 12 June 1996, including a nighttime session on 11 June. He was able to identifya few signs and became convinced that more could be identifiedif the surviving details were highlighted at night by artificiallighting. We returned to the museum in October to attempt this.Justeson conducted such an examination and made and revised preliminarydrawings of portions of the text from 9 to 12 October 1996. Togetherwe continued this examination and produced further revised drawingson 13 to 16 October, and Justeson rechecked the most badly weatheredsigns on 20 January 1997, for a total of approximately 42 man-hoursof work over 10 nights. Justeson made the final drawing.

The result was an almost complete recovery of the final column of text on the monument (Fig. 1, A and B). The column is atthe far left of the surface of the side, ranging 1 to 2 cm fromthe face of the monument. In this position, it is likely to bethe final (22nd) column of text, which we label column V (9).There appear to be 30 sign groups, which we number 1 through 30;four of the legible sign groups contain two signs each (denoteda and b).

Fig. 1. The text on the side of La Mojarra Stela 1. (A) Slanted hatching indicates lines that have been deepened and widenedby erosion; dashed lines are plausible but open to doubt. Signgroups are designated sequentially as V1 through V30; in signgroups consisting of two vertically juxtaposed signs, the individualsigns are labeled with the position number of the group, followedby a or b. (B) Reconstruction of original form of text.Each sign is identified by its previously determined values. Afreestanding ? means that the sign is illegible; a ? after a sign'svalue, as in ni?, reflects some uncertainty over the identificationof the incised form, with the reconstructed form transcribed bythat value. We reconstruct SHAPESHIFTER2 rather than jaat V25 and ni rather than DRUM/YEAR at V28 because ourprovisional reading of the text uses these identifications, butthe alternative reconstructions are not precluded. (C)Analysis of text. T = transcription of sign values; R = readingin the epi-Olmec language; MG = morpheme-by-morpheme gloss ofR; LT = literal translation; FT = free translation. Conventions:In transcriptions, a hyphen joins transcriptions of signs thatare part of the spelling of a single word but are visually inseparate groups, in that they are vertically in sequence separatedby space; a + joins transcriptions of signs that are part of thespelling of a single word but that visually occupy a single signgroup. In readings, +X means X is enclitic; X+, proclitic; -X,suffix; X-, prefix; =X, postpound; and X=, prepound. In morphemeglosses, a hyphen marks any morpheme boundary, and grammaticalmorpheme functions are described by the following abbreviations:A, absolutive-person marker; E, ergative-person marker; X, exclusive(first) person; RES, resultative nominalizer; NUM, numeral-formingsuffix; INC, incompletive-aspect suffix; CMP, completive aspectsuffix; REL, relativizer enclitic; and MAN, manner-adverbial-formingsuffix. Dots join English words or abbreviations that togethertranslate a single epi-Olmec linguistic unit. A freestanding ?transcribes a morpheme or word whose identity is unknown; parenthesized,it means that such a unit may or may not have been present. A? preceding a form reflects uncertainty over the meaning or phonologicalvalue of the sign whose use is reflected by that form.
[View Larger Version of this Image (0K GIF file)]

In three of the sign groups (at V5, V9, and V16), we detect no surviving features diagnostic of the identities of the signs,and one other sign (in a two-sign group, at V6a) is distinct fromall previously known epi-Olmec signs. However, the content ofthe surrounding contexts of each of these signs provides enoughconstraint that we can parse, and determine the general senseof, the phrases in which they occur.

There is essentially no doubt about the identity of 18 of the remaining 26 sign groups, containing 22 of the remaining 30signs. Of these 22 signs, 20 preserve so much detail that theyare almost completely recovered; the two remaining signs forma single sign group (at V7ab), found on the front of the monument,whose features, together, are also quite distinctive.

Most of the remaining eight signs have much less surviving detail or have detail that is less definitely discernible. In threecases (at V6b, V13, and V23), the clear details are consistentwith just one sign and are rather eccentric, so that there islittle chance that these details are the remains of some otherwiseunknown sign. In one sign, at V25, the surviving features area circle far to the right, a central vertical line, and a horizontalline extending leftward from the vertical about halfway acrossthe left half of the sign. Each of these details is simple enough,but the circle alone is enough to identify this sign with jaor with SHAPESHIFTER2/JAMA (which contains ja) (10),and the additional features seem adequate to secure that it isindeed one of the two.

Identifications that are open to doubt are those for the signs at V22, V24, V26, and V28, a region where the text has sustainedthe most damage from weathering and breakage. In three of thesecases, the surviving features are consistent with just one or(at V28) two known signs each; in the present state of our knowledge,these are the only viable identifications for these signs. Thesedetails, however, involve simple enough visual features that theymight plausibly be the remains of previously unattested signs.Most doubtful is the identification of the sign at V26. Its survivingdiagnostic features are a pattern of vertical lines, but somemight be the result of erosion; many vertical lines in this columnhave been subjected to heavy erosion, one proof of the authenticityand antiquity of the text.

In our previous decipherment work (11), we had already read all of the recognizable signs in column V. This text thereforeprovides not only more evidence concerning the epi-Olmec script,but also a serious test of the decipherment. We began workingon an analysis and translation of the text in October 1996 afterthe drawings of the individual glyphs had been completed. It hasproven possible to provide a complete, coherent, and grammaticalanalysis and translation for the entire column in terms of previouslyreconstructed Sokean vocabulary and grammatical structures andof previously established representational principles of the script.Nevertheless, the translation provided here is not definitivein all details. Given the present state of our knowledge of theepi-Olmec language, there are alternatives for some clauses, andthe few signs that are unidentifiable or that have only a fewsimple diagnostic features bring some uncertainty to the identificationof linguistic elements, especially with respect to the identityof the verb at V9.

Analysis and Translation

The text in pre-proto-Sokean:

V1-5: 7is mak=metz-a 7ame7 [TITLE]

V6-8: AND.THEN tuku7 ?paks-pa

V9-10: [UTTER]-wu

V11-18: 7i+ne7w-e je7-tzu ?ki7ps-i ? TITLE3+wu7

V19-24: na+tzetz-e nip7-i wu=tuk-i

V25-30: jama masa=ni7-APPEAR-wu

Running translation:

V1-5: Behold, there/he was a twelve-year [title].

or Behold he was for 12 years a [title].

V6-8: And then a garment got folded.

V9-10: He [utter]ed:

V11-18: -- The stones that he (had) set in order were thus symbols, ?kingly ones--

V19-24: "What I chopped has been planted and harvested well."

or "What I chopped is a planting and a good harvest."

or "What I chopped has been planted; the latter was well harvested."

V25-30: (A) shape-shifter appeared divinely in his body.

Comments on sign identifications and interpretations.

V1-5: The glyph at V5 is partly traceable, but not identifiable, and is possibly otherwise unattested. No verb occurs in thisclause, and V5 must be the final sign in the clause, because V6represents a conjunction that separates clauses. The clause musttherefore be equational, giving a title or status that the protagonist,Harvester Mountain Lord, held for the mentioned 12-year period.Parallel clauses with this type of interpretation, 7i-si 1YEAR-me TITLE4 and 7i-si213 YEAR BUNDLE-ti,are found at R26-27 and I1-4. The numeral at V3 could be 13rather than 12.

V9-10: The (large) illegible glyph group at V9 may be two conjoined glyphs, 9a and 9b. It likely spells an intransitive verbwith a third-person subject, because it is followed by wu(= {-wu} "completive aspect") and has no person marking proclitic.It likely spells a verb of speaking, because a probable exclusiveergative marker with a probable noun (possibly a dependent incompletiveverb) follows at V19-20. Alternatively, V9-10 might spell a nounor adjective followed by the relativizer {+wu7}, but such aninterpretation would not yield a parseable string.

V11-18: The glyph at V16 is illegible and might be either a phonetic complement for the preceding or following word, or anadditional word logographically or phonetically. The last possibilityis perhaps the least likely. In the given context, if V16 standsfor a word, it may represent either a noun or an adjective. Ofknown signs, ti, 7a, ke, we, ja,mi, and GOD are all possible; ma, ko, andku are not.

V25: The surviving details are consistent with two signs, ja and SHAPESHIFTER2/JAMA; the latter contains theright half of the former as an infix. Visually, either is equallyviable. We reconstruct SHAPESHIFTER2 in the drawing because ourreading of the text uses this identification, but ja isnot precluded.

Comments on the analysis.

V6-8: At V6a is a previously unknown glyph that occurs conjoined to AND.THEN at V6b. It might be a phonetic complement toAND.THEN, which is known from two other contexts, MOJ I5 and TUXG13. Otherwise, V6a may represent an additional (probably adverbial)morpheme.

Some clauses were previously identified as verbless predications and were therefore interpreted as equational clauses; equationalclauses, and no other clauses, are verbless predications in Mije-Sokeanlanguages, and in Mesoamerican languages generally. This interpretationis confirmed by the clause that begins column V. No sign for averb suffix, one of which must occur on any verb, occurs in V1-5.The only verbs that sometimes lack orthographic evidence of sucha suffix, which is nevertheless present, are incompletive dependentverbs, which begin with ergative prominal markers; for the epi-Olmectexts we have, they begin with 7i or na. The sign7i occurs at V1, but it is part of the spelling of an adverb,{7is} "behold!", as the parallel passages beginning with the sequences7i-si2 NUMERAL YEAR show. Thus, no verb occurs in V1-5.The sign pair at V6 spells a conjunction, which elsewhere separatesclauses, so V1-5 must constitute a complete, verbless clause.The structure, with an adverbial phrase followed by a seeminglyuninflected noun, is in fact one of the possible structures ofan equational clause with third-person subject.

Another verbless clause (or two of them) appears to be represented by the sign sequence at V19-24, but the argumentation requiredto show this is too complex to pursue here.

Comments on the translation.

V1-5: The phrase "Behold there/he was a 12-year [title]" most plausibly refers to a status held by the protagonist for 12but not for 13 full years, after the last date referred to inthe previous part of the text (that on which the protagonist'sbrother-in-law was executed; narrated at S7-T46). This bringsus to, or somewhat past, 9 August 169 A.D. Gregorian ( This date fell 58 days before the end of an epi-Olmecdecade at 6 October 169 A.D. (, a type of date on whichmonuments were dedicated among neighboring Lowland Mayans, andperhaps the most likely date for the erection of La Mojarra Stela1. It is possible, however, that what is read as 12 here shouldbe 13. The effaced title borne by the protagonist may refer toa rank he achieved after the defeat of his brother-in-law (narratedin R31-40), maybe something like "regional overlord."

V6-8: The phrase "... a garment got folded" may refer implicitly to a bloodletting event, because folding garments has thisassociation at O*32-33 and Q6-8. An alternative parsing is "...he garment-folded."

V11-18: "The stones that he set in order ..." are most likely the same stones that are referred to at R28-30 ("when he placedstones in order ...") and T24-30 ("the symbol[-stone]s got replacedupright"). In "... were thus symbols, ?kingly-type ones," "thus"translates/je7-tzu/, literally "in yon way," referring to a relativelydistant past event, rather than a relatively recent one; the eventis, presumably, the of setting stones in order referred to onthe face of the monument.

The sign at V17 appears as a title of Harvester Mountain Lord at R24, at P31, and, although presumably for a different person,on the Tuxtla Statuette at G4. Visually, at least, this titleis of Olmec vintage; whether it refers back to Olmec times orinstitutions has not yet been determined. In the context at V17,where it is combined with the relativizer {+wu7}, it serves asa qualifier, TITLE3-wu, meaning something like "kingly"or "royal."

According to our reading of this text segment, V11-18 constitutes an aside or parenthetical remark by the narrators (13)of the text about the situation of the protagonist when he madethe remark found at V19-24. Groups V25-30 might be a continuationof the quote at V19-24 or, more likely, is information providedby the narrative voice of the text.

V19-24: "What I chopped is a planting and a good harvest." No coordinating conjunction like "and" has been found in epi-Olmectexts, although several passages seem to offer lists of nouns,adjectives, or verbs that would require the insertion of "and"in English translation. The words /tzetz-e/, /nip7-i/, and /tuk-i/are resultative verbal nouns or non-active participles that canbe rendered both "having been VERBen," and "VERBen thing," thusthe rendering as "chopped thing," "planted thing" or "planting,"and "harvested thing" or "harvest." These three nouns presumablystand for three actions that are linked in some logical order.The chopping may refer to the beheading of a prisoner at L4-7("when I chopped [off his head] ...") and/or the execution of theprotagonist's brother-in-law at S44-T6. The heads of his enemiesor their blood, or both, may be the buried things referred to("to plant" and "to bury" are the same word, {nip7}). This interpretationmay have the prosaic interpretation that the "harvest" is thefruitfulness of the land with respect to some crop or crops atthe point of and/or as a result of the burial of the heads orthe blood. Human sacrifice was believed to promote good harvests.Among Mayans, at least, the ruler was expected to carry out orsponsor the rituals that would ensure good harvests at the endsof years, especially at the ends of 5-, 10- (as perhaps here),and 20-year stations in their calendar. "Harvester Mountain Lord"(found at L2-3, O10-11, and Q16-17) may be an epithet of the protagonist,rather than a name, and may refer to his success over time inensuring good harvests for his people.

In this column of text, the protagonist who speaks at V19-24 is not named, although a title is given him at V5. He is presumablythe same person as Harvester Mountain Lord, who is named and depictedon the face of the monument, and given various titles.

V25-30: "(A) shape-shifter appeared divinely in his body." The text spanning Q48-T23 refer to ritual acts that resulted inthe protagonist (or him and his supporters) taking 23 jaguarsover a 23-day period [confirming the chronology of (4)]. Thus,V25-30 may refer to this set of events; to the garment foldingevent at V6-8, when a public bloodletting and attendant visionprobably took place, perhaps at the end of a 10-year period; orto a public event after one of the harvests (there could havebeen 24 each of corn and beans in 12 years) that may be referredto at V24.

Support for the decipherment of epi-Olmec writing.

At a general level, the grammatical model for the decipherment is supported by the results reported here; it is this modelthat makes it possible to set the parameters for the syntacticparsing, and these parameters constrain the more detailed featuresof the analysis and translation. Certain specific features ofthe decipherment are also supported, along with the grammaticalframework and much of the specific phonetic and lexical deciphermentson which these features were based.

New contexts supporting grammatical analyses.

In any text as long as that on La Mojarra Stela 1 (about 535 glyphs), there is a substantial amount of lexical repetition.This feature, along with a model for Mije-Sokean word structure,was a key to the grammatical analysis that was the primary basisfor our ability to decipher the epi-Olmec writing system.

In the case of sign sequences that were not repeated, the determination of word boundaries in the original decipherment workhad to be based on a systematic grammatical analysis of the entiretext, along with a uniformly applicable set of sign readings andvocabulary identifications. Three of the sign sequences that hadoccurred only once in the previously known columns of text--butthat were analyzed as belonging to single linguistic units onthe basis of the way they fit into the overall structure of thedecipherment--are now known to occur also in column V. Their recurrencein this column supports the view that the decipherment had builta correct parsing of all of these units and, in part at least,of their immediate surrounding contexts.

1) The combination je-tzu, which occurs at M2-3 and is interpreted as /je7-tzu/"thus," occurs also at V13-14. Thisrecurrence confirms the analysis of this sequence as a word, andthe meaning of the word fits both contexts.

2) At V26-27, ma-sa spells an incorporated preposed variant of proto-Mije-Sokean *masan (14) "god" that is foundalso at D2-3; /maas=/, from *masa=, is attested as a variant of*masan used as a preposed incorporee in present-day Soteapan andAyapa.

3) At V20, tze-tze spells the word /tzetz-e/, consisting of the verb root {tzetz} "to chop" followed by the resultativenominalizing suffix {-e}. The same sign sequence is found at L5,where it spells the root {tzetz} only, given that L4-L7 spells/na+tzetz-ji/"when I chopped it." The repetition shows that tze-tzeat L5 was correctly isolated from its context as spelling a fullverb root.

4) At V21-22, PLANT-7i spells the word /nip7-i/, with 7i representing the final /7/ of the root and the resultativenominalizing suffix {-i}. This recurrence confirms the interpretationof the sign PLANT as representing the verb root {nip7} in itsother two instances, on the Tuxtla Statuette and the O'Boyle Mask.It was correctly isolated as a logogram, not a syllabogram spellingan affix, nor partially spelling a root in combination with othersigns in its context.

Support for the identification of grammatical patterns.

The occurrences of FOLD+pa2tu+CLOTH at O*32-33 and FOLD+pa2CLOTH at Q6-8 already suggested that thesesign sequences go together in a single clause. This interpretationis supported by the combination tu+CLOTH FOLD+pa2atV7-8. In all three cases, the meaning is "cloth(ing) gets folded,"but the contrary order of this third instance requires recognizingthat different ordering of words can occur under some conditions.This change in word order involving the same subject and verbis quite interesting, because it conforms to a pattern of word-ordervariation that emerged from the overall decipherment describedin (3). In fact, we found two grammatical patterns that canaccount for it: mediopassive usage and noun incorporation.

In every Mije-Sokean language there are many transitive verb roots that can be used intransitively without any intransitivizingderivational affixes and that have a mediopassive meaning in suchusages; that is, the logical subject of the transitive verb doesnot appear, and its logical object appears as its grammaticalsubject. Whether a particular verb has this property in a particularlanguage is a lexical fact about the verb; not all transitiveverbs do this, and this property cannot be predicted from otherfacts. When different Mije-Sokean languages have descendants ofthe same transitive verb root, these descendants do not alwaysagree with each other as to whether that root can be so used.This difference shows that this lexical feature has been subjectto change through time in individual languages. Because everyMije-Sokean language has many transitive roots that have thisproperty, the pattern itself is reconstructible for pre-proto-Sokean.Our recognition of this pattern in several straightforward casesis part of the body of evidence confirming that the epi-Olmeclanguage was Mije-Sokean.

Whether or not the proto-Sokean verb *paks "to fold (cloth, among other things)" had this property in pre-proto-Sokean cannotbe determined from the evidence of the modern Mije-Sokean languagesalone: There are languages in both subgroups of Sokean that haveit, and languages in both subgroups of Sokean that do not. OnLa Mojarra Stela 1, however, this is the only possible grammaticalinterpretation of the first two instances of the cloth-foldingclauses, in which the word /tuku7/ "cloth" follows the verb: Becausethere is no ergative marker preceding the verb, it must be intransitive,even though "to fold" is semantically transitive.

This grammatical interpretation is consistent with the subject-verb word order found at V7-8. We previously found that nonactivesubjects of intransitive verbs usually follow but often precedethese verbs. This variable order is not the case for active subjects,which invariably precede the verb (transitive or intransitive).

The other possible explanation is object incorporation. Noun incorporation of various sorts is quite common in Mije-Sokeanlanguages and many instances are found in the epi-Olmec texts.More specifically, in these languages, a transitive verb can beintransitivized by incorporating its direct object, and in suchcases the object immediately precedes the verb stem. In the caseat V7-8, this would yield the intransitive verb form /tuku7=paks-pa/"he garment=folded." Object incorporation is possible only whenthe object is unpossessed, as it is here, and not definite; themajority of nouns in the text appear to be definite, but no overtmarker of definiteness or indefiniteness normally appears. Wehave recognized several other instances of object incorporationin epi-Olmec texts.

New contexts supporting sign readings.

Several signs are used to spell the same morphemes as in previously known instances, and these examples (for example, thesign wu for the verb suffix {-wu}) support the grammaticalmodel for the uses of these signs. The values of three signs aresupported by their use in new contexts:

1) All previously known instances of the syllabogram wu are used to spell either the completive suffix {-wu} or therelativizer enclitic {+wu7}. At V23, wu seems to spellthe root {wu} "good." This example confirms the reading of thissign as the syllable /wu/.

2) The probable ni sign seems to spell the prefix /ni7-/ on/ in the body" at V25-30. This prefix is known from bothSoke and Mije, as well as Oluta (as {ni:-}), and so must haveexisted in pre-proto-Sokean. This is the first instance of theprefix {ni7-} recognized in epi-Olmec texts. It confirms the phoneticreading of the sign that represents it, which only occurs in oneother instance, as the syllable /ni/.

3) The string 7i-si 12 YEAR/7AME7 TITLEx at V1-5 is mirrored at H3-I4 (7i-si213 YEAR/7AME7BUNDLE/PIT-ti) and T7-10 (7i-si2 ONE YEAR/7AME7-meTITLE4); in all three cases, the reading is "behold he/there wasa ROLE for n years." The occurrence of 7i-si at V1-2 alongside7i-si2 at H3 and T7 also demonstrates the previously postulatedequivalence in value of si and si2.

Additional evidence concerning certain grammatical constructions.

The five instances of verbs of speaking that occur on the face (columns A through U) of La Mojarra Stela 1 all have the incompletivesuffix {-pa}. Sign groups V9-10 apparently spelled a verb of speaking(because it is followed by a first-person ergative pronominalagreement marker, which should occur only in a direct quote) thatoccurs with the independent completive suffix {-wu}. In our documentationof present-day Mije-Sokean languages, incompletive {-pa} is typicalon verbs of speaking, even with past time reference, althoughthe completive {-wu} is not proscribed.

Sign groups V6-8 and V9-10 spell two clauses in sequence linked through what we have labeled the pa-conversive; in this construction(i) the two clauses are adjacent, (ii) the events referred toare in close temporal succession, (iii) and one verb is markedwith incompletive {-pa}, whereas the other is marked with completive{-wu}, and both verbs are understood as being in the completive.Usually, as here, it is the first verb that takes {-pa} and thesecond that takes {-wu}. It is plausible that V6-10 is an instanceof the pa-conversive.

Additional evidence for a phonological structure.

The spelling PLANT-7i "plant-ing/plant-ed" at V21-22 shows that the epi-Olmec descendant of proto-Mije-Sokean *ni:p7"to plant," with the nominalizer {-i} suffixed, was pronounced/nip7i/, with a postconsonantal /7/, in pre-proto-Sokean. Thispronunciation is consistent with epi-Olmec postconsonantal /7/in /RULER ko7=mon7a/ "ruler's head-wrap," spelled KNOT+GOVERNOR-7aat Q41-42, and /poy7a/ "moon, month," spelled po-7a atJ3-4. In proto-Sokean, as it would be reconstructed from survivinglanguages, postconsonantal /7/ had been lost from these words.Data from Mijean languages show that postconsonantal /7/ appearedin all such items in proto-Mije-Sokean (5).

Additional evidence for a spelling convention.

A distinctive spelling convention is also supported. In quite a few cases, a syllabogram's iconic origin is apparent, andin these cases the syllabic value is based on the consonant andvowel that begin the word whose depiction the sign reflects; somesuch signs can be used as a logogram for that word and as a syllabogram.For example, the icon for "earth," which was *na:s in proto-Mije-Sokean,is almost always used for the syllable /na/, but it occurs atO26 as a logogram for "earth" in spelling /nu.tzat7.e=nas/ "groundjointly measured by handspans." Similarly, the sign for the numeral2 is used at R43 to spell the syllable /wu/ (15), a syllabographicuse of what is otherwise known only as a logogram; "two" in Mije-Sokeanhas two suppletive allomorphs, *metz and *wustuk.

A confirming example is now provided by the sign ne, which shows a hand setting down a stone; its logographic originis pre-proto-Sokean {ne7w}, "to set stones in order." At V12,this glyph is used to spell the logogram ORDER.STONES, not thesyllable /ne/. As a logogram for the verb {ne7w}, it may alsospell the corresponding participle or verbal noun /ne7w-e/ "havingbeen set in order, of stones," and that is what it does at V12.The example also confirms the convention by which logograms forverbs could be used to spell their nominalizations.

New evidence concerning uses of titles.

The use of TITLE3-wu at V17 shows that word for titles, offices, or statuses were applied not only to rulers but alsoto artifacts and activities that are associated with those statuses.


This study shows that a previously unknown segment of text can be read and understood in terms of the same model for languagestructure, sign values, and spelling conventions that were developedin the previously achieved decipherment of the epi-Olmec script,and shows that the segment's content is well integrated with thepreviously read portion of the same text. Conversely, there areno phenomena in this stretch of text that challenge the modelin any way. It is difficult to imagine that this model would yielda complete, coherent, and grammatical text if these portions ofthe decipherment--language structure, sign values, and spellingconventions--were not essentially correct. In our view, the dataconfirm the results obtained in the first two of our by now sixyears of our work on the decipherment of epi-Olmec writing.


  1. These texts are, in order of their discovery, the Tuxtla Statuette (TUX), the Chiapa de Corzo Sherd, the O'Boyle Mask, and La Mojarra Stela 1 (MOJ). When not otherwise mentioned, the text is from La Mojarra.
  2. F. Winfield Capitaine, La Estela 1 de La Mojarra, Veracruz (Publ. 16, Research Reports on Ancient Maya Writing, Washington, DC, 1988).
  3. J. S. Justeson and T. Kaufman, Science259, 1703 (1993).
  4. ___, Arqueología 8, 1992 (1996).
  5. S. Wichmann, The Relationship Among the Mixe-Zoquean Languages of Mexico (Univ. of Utah Press, Salt Lake City, UT, 1995). Wichmann had made versions of this work available to us in 1991.
  6. R. Diehl, A. Vargas, S. Vásquez, in Memoria del Coloquio Arqueología del Centro y sur de Veracruz, S. Ladrón de Guevara and S. Vásquez, Eds. (Universidad Veracruzana, Veracruz, Mexico, 1997), pp. 197-205.
  7. A. M. H. Schuster, Archaeology 47 (no. 5), 51 (1994).
  8. In a note to the director, Vásquez wrote that "una revisión cuidadoso de la estela me permitió identificar poco más de 20 glifos, muy erosionados en la costado de la estela [a careful examination of the stela allowed me to identify a little over 20 glyphs, very eroded, on the side of the stela]."
  9. Stuart had labeled four signs represented as tattooed on the body of the standing figure, with column labels V through Y. However, these are one-word designators of some status of the individual who bears them, that is, they are captions; they are not texts or text segments, they have no "reading order" relative to one another or to the remainder of the text, and they would not ordinarily have column designations at all.
  10. We write Mije-Sokean forms in a practical, Spanish-based orthography. Most letters have their usual Spanish pronunciations, but j represents [h]. We use u to represent a high, central-back unrounded vowel, like the u of put and bush as pronounced by many Southerners and Westerners, and of just as in just now. The symbol 7 represents a glottal stop. Phonologically explicit representations of Mije-Sokean words are between slashes, and transcriptions of morphemes are between curly braces; phonetic transcriptions of epi-Olmec signs are in bold italics. We indicate logograms in transcription by rendering them in capital letters. These transcriptions are in bold italics when they specify a Sokean word they are known or thought to represent, and are in roman type when they specify its (usually basic) meaning only. For example, the sign representing Sokean /tuku7/ "cloth, clothing" may be transcribed either TUKU7 or CLOTH. In phonologically explicit representations of Mije-Sokean words, grammatical affixes are joined to roots or to one another by a hyphen; elements of compound words are joined by =; and clitics are joined to adjacent words by +.
  11. Readings for almost all these signs were reported in the original announcement of the decipherment (2). Only three readings have been subsequently revised. One, appearing at V14, was revised from tzi to tzu in the spring of 1996, when it was realized that the final vowels of words we believed to be spelled with these signs would conform exactly to existing and reconstructible words if the sign values were exchanged. This change does not impinge on any semantic interpretation. Another, appearing at V29, was identified as a logogram for an intransitive verb, initially identified as referring to the performance of some kind of ritual, but later revised semantically to "to appear." The reason for this change is that the verb refers to something done by or happening to both a throne (inanimate) and to human beings, to jaguars, probably to a god, and to a constellation, all in a ritual context. The constellation helps to narrow the semantics fairly tightly; becoming manifested (appearing, being revealed) in some way seems to be the only feasible category. The last change was made on 1 May 1997: Although the reading of the sign at V25 as /jama/ is unchanged, we have discovered that the meaning of this term in Mije-Sokean languages is "shape-shifter" rather than "animal spirit counterpart".
  12. Gregorian equivalents of these epi-Olmec dates, expressed in the so-called "long count" calendar, are based on a 584,265 correlation constant, making them 18 to 20 days earlier than the corresponding Mayan dates would be under the two variants of the almost universally accepted correlation of that system with European chronology. A summary of the evidence for this correlation is given in (4).
  13. The narrators were apparently elite supporters of Harvester Mountain Lord. We know they were a group, not an individual, because they twice referred to their own actions on the face of the monument using the (plural) ergative inclusive pronominal {tun+}, at S7-12 and T42-46.
  14. Reconstructed words, which are labeled by an asterisk, are pre-proto-Sokean unless otherwise stated.
  15. The identification of this sign as that for the number 2 is another result of direct examination made possible by the erection of the stela. The drawing in (2) suggested that the numerical dots were ovals, presumably because the photographs on which it was based had to be done at an angle, and they showed a spurious internal line, presumably an effect of lighting.
  16. Our greatest debt for the work reported here is to S. Ladrón de Guevara, Director of the Museo de Antropología in Xalapa, Veracruz, Mexico. It was she who made the discovery of the text possible, and who, for the first time, provided full public access to the monument for all who wish to examine it. She facilitated our work in particular by granting us access to the museum when it was closed to the public, especially at night, and by helping us with the physical arrangements to make accurate tracing of the text possible. I. Graham photographed the monument for us. In 1993 and 1994, our linguistic work documenting the modern Mije-Sokean languages was supported by funding from the National Geographic Society (grants 4910-92 and 5319-94) and the Faculty Research Awards Program grant 320-9753P of the State University of New York at Albany; since 1994, it has been supported by the National Science Foundation (grants BNS-9411247 and SBR-9511713). This funding included support for the use of these materials in comparative reconstruction of earlier stages of these languages and direct support for our continuing work on epi-Olmec decipherment, including the funding for the work reported here. Travel support for our collaboration since 1993 has been provided under these grants and by the Texas Workshops on Mesoamerican Writing and Iconography, under the direction of P. Keeler.

26 December 1996; accepted 17 April 1997

Volume 277, Number 5323, Issue of 11 July 1997,
©1997 by The American Association for the Advancement of Science.