As the authors point out (Kaufman and Justeson Reference Kaufman and Justeson2007:232) in reference to the work of Dakin and Wichmann (Reference Dakin and Wichmann2000), one word cannot be used to establish the speech of Teotihuacan. The linguistic analysis by Kaufman and Justeson is comprehensive and, as far as I am able to evaluate its technical aspects, convincing that the term kakaw has its origins in the Mije-Zokean family. They conclude this based on empirical critiques, as well as distinct approaches to analysis. Just as processual or symbolic archaeologists will approach the same data differently stemming from their commitment to one set of theoretical assumptions or another, something similar seems to be happening here. I for one would like to see a better explanation of the differing principles that lead Dakin and Wichmann to focus on phonemic bases for comparison while Kaufman and Justeson focus on phonetic rules (pointed out in Kaufman and Justeson Reference Kaufman and Justeson2007:196). Surely these are problem-driven choices, and this is arguably the level at which archaeologists require tutelage rather than the nitty-gritty of a phonetic analysis itself.

I assume that I am not alone among archaeologists in hoping for some alternative approaches to synthesis that better convey linguistic conclusions to those less accustomed to dealing with those data. I appreciated the authors' spatial characterization of identifiable loan words from Mije-Zokean into other Mesoamerican language families (2007:Figure 6), and further visualizations of this sort would go far in presenting specialized data to an audience of non-linguists. I similarly found myself wishing for charts arranged vertically (by relative time), with horizontal indicators of loans from one level of proto-language reconstruction into another, so as to better tease out the linguistic stratigraphy of Mije-Zokean loans now variously attributed to Gulf Coast Olmec or Teotihuacan interregional interaction. This could encapsulate some conclusions more succinctly. I am also convinced that Mesoamerican linguists could make use of Christopher Ehret's (Reference Ehret1976 and later publications) categorization of borrowing events between African languages, in which he associates the kinds of words and scale of borrowing with the nature and intensity of contact. With such a deductive approach, the significance of a single term like kakaw (and apparently, sacatl “grass”?) could be better contextualized without reference to an elaborate culture-historical edifice.

My perspective on the issue stems largely from an interest in the Nawa migrations. There are two basic models for the appearance of Nawatl in Mesoamerican prehistory, with sub-variants for each. The first sees Nawatl speakers as having participated in agriculture and been in contact with Mesoamerican peoples for a very long time (Hill Reference Hill2001). A second model would see Nawatl as an introduction from outside Mesoamerica, citing the known linguistic ties to Southern Yuta-Nawan languages in western (Wichol, Kora) and northwestern Mexico (Taraumara, Kájita, etc.), and models would then vary on the timing, nature of the introduction, and so on (Fournier and Bolaños Reference Fournier, Bolaños, Kowalski and Kristan-Graham2007 vs. Smith Reference Smith1984).

Although the models seem distinct enough, there is a tangle of underlying assumptions that perpetuate miscommunications between scholars. Many publications continue to associate Nawatl with “Chichimecs”, which is then assumed to indicate that Nawas were nomadic, and further assumed that they were therefore outside of the orbit of Mesoamerican practices. The sixteenth-century sources make it clear that the term Chichimec included a range of adaptations from nomadic hunter and gatherer to agriculturalist, and encompassed speakers of Otomi, Nawatl, Pame and other languages (Viramontes Reference Viramontes Anzures2000:33–48), many of whom shared in Mesoamerican practices to varying degrees. On the other hand, Hill's (Reference Hill2001) argument that Southern Yuta-Nawan speakers were agriculturalists from northwestern Mesoamerica does not automatically default to central Mexico, much less Teotihuacan. Mexico west of the valley of Toluca continues to be excised from discussions of Mesoamerican dynamics despite years of productive research (Beekman Reference Beekman2010).

Contact period linguistic data place Korachol languages, of which Nawatl should have been a neighbor, in northern Jalisco and Nayarit. Kaufman and Justeson leave a blank hole on their Figure 1 between these Korachol languages and Pame, but this is where other maps of the sixteenth-century situation place the Zacatecos, Guachichiles, and Guamares, all usually assigned to Korachol languages, and the Cazcanes, who are often considered a variant of classical Nawatl (Barragan and Yáñez Rosales Reference Barragán Trejo and Rosales2001; Moctezuma Zamarrón Reference Moctezuma Zamarrón, Zamarrón and Hill2001:381–382). This blank spot corresponds to Los Altos of northeastern Jalisco, western Guanajuato and the southeastern edge of the Sierra Madre Occidental. This places Korachol and even a variant of Nawatl in an area that had participated in agricultural and Mesoamerican practices during the Classic period.

Published biological data support the contention that migration, and not other mechanisms, brought Nawatl into central Mexico. Non-metric and metric biological data document a significant population influx into central Mexico from northwestern Mexico sometime during the Classic to Early Postclassic (Beekman and Christensen Reference Beekman and Christensen2003:127–133) or Epiclassic to Early Postclassic (González-José et al. 2006). DNA analyses have supported an Early Postclassic migration (Fournier and Vargas Sanders Reference Fournier and Sanders2002). Complex multiethnic relations at sites such as Tula suggest that substantial biological variation may lurk beneath similar material culture however (Beekman and Christensen Reference Beekman, Christensen, Cabana and Clark2011), and the small samples of undisclosed context that have been reported to date do not test adequately for that variability.

Bracketing the problem in this way, the widespread disruptions of the Epiclassic period (with traditional dates of a.d. 600–900) take center stage. This is certainly a period of political and economic transformation across Mesoamerica (re-evaluated recently in Demarest et al. Reference Demarest, Rice and Rice2004; Kowalski and Kristan-Graham Reference Kowalski and Kristan-Graham2007; Manzanilla Reference Manzanilla2005; Solar Valverde Reference Solar Valverde2006; several in Faugère Reference Faugère2007) and some movement of population, whether as refugees or opportunists, seems increasingly hard to deny. Alexander Christensen and I carried out an analysis of the Nawa migrations that coordinated existing linguistic, biological, ethnohistoric, and archaeological data across the northern rim of Mesoamerica (Beekman and Christensen Reference Beekman and Christensen2003). We feel that it resolves many of the faulty assumptions of other models. Our basic conclusion associates Nawatl with sedentary agricultural communities from the small Mesoamerican polities of southwestern Guanajuato, and proposes that they migrated through central Mexico during the tumultuous Epiclassic. A discontinuous string of multiethnic communities resulted throughout central Mexico and beyond to the Gulf Coast, but the Nawatl language only spread laterally with the rise of its symbolic capital in the Postclassic under a succession of powerful polities—Tula, Azcapotzalco, and Tenochtitlan. In places like Jalisco state (Yáñez Rosales Reference Yáñez Rosales, Ávila, Emphoux, Gastélum, Ramírez, Schöndube and Valdez1998) or El Salvador (Fowler Reference Fowler1989:50–65), Nawatl and its variants typically remained a single component of multiethnic communities, because no particular prestige came to be attached to its use. Note that this model is entirely consistent with the evidence for rebus writing implying Nawatl speech at Epiclassic Cacaxtla and Xochicalco (cited by Dakin and Wichmann Reference Dakin and Wichmann2000:68; Wichmann Reference Wichmann, Blench and Spriggs1998:302) and proposed Nawatl loan words in Mayan inscriptions from a.d. 650–750 (Macri and Looper 2005).

Our more detailed recent research in the Mezquital region of southwestern Hidalgo (Beekman and Christensen Reference Beekman, Christensen, Cabana and Clark2011) found a spatial disjunction between the use of Nawatl and Otomi in Early Colonial period parish registers (Christensen Reference Christensen2002) that corresponds closely to an archaeological distinction in the Epiclassic separating those sites sharing in the Prado-Corral-Terminal Corral phase materials found around Tula from sites participating in the Xajay Regional Development slightly further west. We propose that a leapfrog migration probably following the Lerma River brought Nawatl speakers into the eastern Mezquital and the nature of the ensuing interethnic interaction led to varying expressions in material culture.

If this is all true, how does it relate to the role ascribed by Kaufman and Justeson to Nawatl prior to the Epiclassic period? They see its speakers as engaging in linguistic interaction with Mesoamerican communities since the Late Formative, precisely when complex polities begin to emerge in the Bajío out of earlier Chupicuaro communities (Darras and Faugère Reference Darras, Faugère, Williams, Weigand, López-Mestas and Grove2005). Scholars have long observed architectural and ceramic similarities between Chupicuaro and Cuicuilco in the Basin of Mexico (critically evaluated by Darras Reference Darras2006), but the ties to central Mexico take an interesting turn with the emergence of Teotihuacan. The widespread emergence of enclosed patios in public architecture (Cárdenas Reference Cárdenas García1999; Ramos and Crespo Reference Ramos de la Vega, Crespo, Williams, Weigand, Mestas and Grove2005) may suggest Teotihuacan inspiration, and specific finds of Tzacualli, Tlamimilolpa, and Xolalpan phase ceramics in a handful of sites in Querétaro and Guanajuato (Saint-Charles Reference Saint-Charles Zetina, Crespo and Viramontes1996) indicate high level interregional relationships.

Nawatl speakers in the Bajío would have been in a position of ongoing interaction with central Mexican polities from the Late Formative through the Epiclassic, with various opportunities for borrowed words to enter the linguistic repertoire of other language families. The Nawa migrations are best seen as a migration stream of cohesive agricultural communities from one area of Mesoamerica into another, and not the late invasion of nomadic gatherers and hunters across cultural boundaries extrapolated from the ethnohistoric record alone. The Epiclassic Bajío is reconstructed as suffering abandonment in the early part of that period (Filini and Cárdenas Reference Filini, García and Faugère2007) or as experiencing population nucleation and political centralization just prior to an Early Postclassic abandonment (Castañeda et al. Reference Castañeda López, Cervantes, Crespo and Flores1989). There is surely a connection to the widespread intensification of arid conditions in the Epiclassic and Early Postclassic described by some paleoclimatologists as “probably the driest of the Holocene” (Metcalfe and Davies Reference Metcalfe and Davies2007:169).

For purposes of illustration, I will briefly consider the opposing position in relation to this reconstruction of Nawatl's role in northern Mesoamerica. Dakin and Wichmann's (Reference Dakin and Wichmann2000) alternative interpretation argues for the occurrence of kakawa as a Nawa loan in dated Maya hieroglyphic texts by a.d. 400. As already pointed out, it is difficult to interpret the significance of such an isolated loan without a more comprehensive consideration of other potential loans in the same direction. Was the interaction represented by the one word actually that important? Or was it a fluke, as Dakin and Wichmann would apparently have to argue for sacatl “grass” or solotl “cork tree” (Dakin and Wichmann Reference Dakin and Wichmann2000:68)? Did Nawatl speakers in the Bajío already have some experience with cacao to the west, along the western Mexican coast? Did they carry out long distance trade in the service of central Mexican powers? Or, are all of us partly right and was Nawatl spoken in the Basin of Mexico at Cuicuilco? This center has long been cited as sharing links to Guanajuato during the Late Formative, and is often described in opposition to the emergent Teotihuacan on the opposite side of the Basin of Mexico. Was there a linguistic frontier running through the Basin, one that broke down with Cuicuilco's volcanic destruction and the absorption of part of its population at Teotihuacan? Once again, the exceedingly narrow range of culture-historical interpretations considered by Dakin and Wichmann is evident.

On one hand I am impressed by the detail potentially accessible through linguistic analysis, but it is a very different data set from that with which archaeologists work, and the potential explanations for loans can comfortably co-exist within a range of historical explanations. We need to seek broad accumulations of data integrated with other datasets rather than smoking gun arguments that seek to explain major questions at a single stroke.

The twin issues of Teotihuacan's languages and of Nawatl's presence in Mesoamerica are far from resolved. I am puzzled over the seeming irrelevance of Otomi to the debate. I also wish the authors had addressed the claims of Macri (Reference Macri2005) and Looper (Macri and Looper Reference Macri and Looper2003) for the presence of Nawatl words in Maya inscriptions. All of the dated examples are from the seventh and eighth centuries, but Macri (Reference Macri2005) makes reference to two Maya vessels with three potentially Nawatl words modifying the term kakaw. This should pique our interest and demand close examination. Kaufman and Justeson (Reference Kaufman and Justeson2007:Table 7, Section A.1.e) list the same phrases in their discussion of cacao vessels but consider the terms untranslated. Perhaps they could turn their critical eye to these cases.

My comments on Kaufman and Justeson's paper are as an archaeologist, and I have little to say about linguistic aspects. In the hope of reducing confusion, I adopt their spelling conventions for languages. Full assessment of Kaufman and Justeson's wide-ranging article would require not only competence in linguistics, archaeology, ethnohistory, and biological anthropology, but also control of the entire span of Mesoamerican regional studies, from west Mexico to the Maya Lowlands. No single person today is such a polymath, and I concentrate on what I know best—the archaeology of highland central Mexico, especially Teotihuacan and the “Epiclassic” period immediately following collapse of the Teotihuacan state. I start from what seems most clear to me, and move on to less certain matters. I focus on the question of whether there was a sizable migration of Nawa speakers into central Mexico (i.e., the 6000 square km Basin of Mexico and its close neighbors) around the time of the collapse of the Teotihuacan state. To a lesser degree, I comment on the case for the prominence of Mije-Sokean speakers in that state.

Chronological precision remains low in Mesoamerica, but recent opinion favors a date somewhere around a.d. 550 to 650 for the burning of major civic-ceremonial structures in the city of Teotihuacan and the collapse of the Teotihuacan state. Ceramics of the ensuing “Epiclassic” period retain some features of the Teotihuacan tradition, but show many drastic changes. That fact is not controversial, but there is much controversy about the explanation for these changes. Some archaeologists, notably the late William Sanders, have argued that they were local innovations. Others, including myself, while acknowledging that styles can spread without the spread of people, think it more likely that they represent substantial migrations from regions west and northwest of the Basin of Mexico, somewhere around the state of Guanajuato and its neighbors. This broad region seems to have been the source from which many of the ceramic innovations were derived. Important recent publications on this topic include Beekman and Christensen (Reference Beekman and Christensen2003), who draw on archaeological, linguistic, ethnohistoric, and biological data, and many of the chapters in the volumes edited by Manzanilla (Reference Manzanilla2005), Solar (Reference Solar Valverde2006), and Faugère (Reference Faugère2007).

There is a tolerable fit between the general region that seems to have been the source of many of the Epiclassic ceramic innovations that appear in and near the Basin of Mexico, and what looks to me to have been a plausible homeland region for Pre-Nawa and/or Proto-Nawa. On the surface, there seems to be some disagreement among linguists. Hill (Reference Hill2001) argues that the homeland of Proto- Uto-Aztecan (PUA) was far to the south of what many had thought, but Kaufman and Justeson (Reference Kaufman and Justeson2007) argue for a “northern” homeland for Proto-Nawa. However, if I understand correctly, Hill actually argues against a far northern homeland for PUA, somewhere in what is now the state of Arizona, and argues that it was instead far enough south for its speakers to have participated in maize agriculture from a very early date; while Kaufman and Justeson argue that early Nawa speakers were in contact with speakers of Koran and Wichol. But current thinking is that maize was probably first domesticated around the Balsas river valley, in the modern states of Guerrero and Michoacan. Since Koran speakers lived as far south as the state of Nayarit and the northern parts of Jalisco around a.d. 1500 (Kaufman and Justeson Reference Kaufman and Justeson2007:194, Figure 1), it seems plausible that during the time Teotihuacan flourished (roughly 100 b.c. to a.d. 600) Proto-Nawa speakers may have been living somewhat east and south of this Koran-speaking region; that is, centered around the state of Guanajuato and the eastern part of Michoacan. This would have put them far enough north to be in contact with Koran speakers (if those speakers occupied roughly the same region then as in a.d. 1500) but far enough south for their ancestors to have participated early on in maize agriculture. There may well have also been speakers of unrelated languages in this general region. Nevertheless, it is the same broad region likely to have been the source of many of the ceramic innovations that appeared in central Mexico after the collapse of the Teotihuacan state.

None of the current archaeological evidence proves beyond reasonable doubt that Nawa-speaking migrants first became politically prominent in the Basin of Mexico around the time of the Teotihuacan collapse. Full proof or refutation of substantial migration at that time is likely to come from bioarchaeological investigations, including ratios of stable isotopes of strontium and oxygen in teeth and bones, hopefully DNA, and perhaps skeletal morphology. Nevertheless, the archaeological evidence combines neatly with linguistic interpretations to suggest a sizable migration of Nawa speakers around a.d. 600, as proposed by Beekman and Christensen (Reference Beekman and Christensen2003). It is tempting to speculate that these migrants may have contributed to the collapse of the Teotihuacan state, although that state may well have already become vulnerable due to internal problems. Perhaps the migrants took advantage of a polity already in disarray.

A second issue concerns the dominant language of Teotihuacan. Kaufman and Justeson present lengthy evidence that, during Teotihuacan's heyday, languages of the Mije-Sokean family were the major sources of borrowings into languages belonging to other families, while there are few, if any, borrowings from Nawa until later. I am unable to evaluate most of their linguistic arguments, although their methods look sound. Certainly some of their criticisms of alternative methods are well-taken, such as the mistake of operating as if phonemic inventories rather than phonetic interpretations were a reliable basis for establishing the direction of borrowing.

Kaufman and Justeson further argue that the powerful influence of Teotihuacan throughout Mesoamerica implies that whatever language was spoken by its elites is likely to have been the source of many loan words (and possibly some other linguistic features) in other languages and that, conversely, any language not the source of many loans during Teotihuacan's heyday is unlikely to have been widely spoken by Teotihuacan's leaders. This argument is plausible, but I do not find it airtight. It seems conceivable, if not very likely, that Mije-Sokean speakers living in the Gulf Lowlands could have been middlemen responsible for much of the spread of Mije-Sokean vocabulary. If they were traders, possibly they were agents of linguistic change even for speakers of Tarasko, presumably always located in western Mexico.

Kaufman and Justeson correctly note that Tolstoy (Reference Tolstoy, Sharer and Grove1989a:98) finds it significant that Olmec-related features of iconography, ritual, and utilitarian objects all appear suddenly in the Basin of Mexico, early in the Early Preclassic, around 1200 b.c. But Tolstoy (Reference Tolstoy, Sharer and Grove1989a) also states that their fairly rapid fading is already quite evident later in the Early Preclassic. In another chapter in the same volume he says that “…the obvious candidates to occupy the Central Highlands prior to the appearance there of Nahua speakers … are the ancestors of the Pame and Otomí…” (Tolstoy Reference Tolstoy, Sharer and Grove1989b:276). Later he notes that “… the Basin of Mexico provides little assurance of cultural and linguistic continuity” across the first millennium b.c. (Tolstoy Reference Tolstoy, Sharer and Grove1989b). I agree with Tolstoy; the archaeological evidence does little to support a strong Mije-Sokean linguistic presence after Early Preclassic times in the Basin of Mexico. Some early Preclassic sites, such as Tlatilco, do have some ceramics related to those of the Olmec heartland in the southern Gulf Lowlands. But, after that, during the Middle and Late Preclassic, ceramic resemblances to the Gulf disappeared, and relations were closer to western Mexico, although resemblances to specific sites, such as Chupicuaro, look general rather than specific. Only around a.d. 100–200 were there again important connections with the Gulf in material culture, apparent in such features as interlocking scroll motifs and direct-rim tripod cylinder vases. These features are additions to the continuing local ceramic tradition and seem too limited to suggest any significant migration.

It is of some interest to note that by at least a.d. 200–300 there were significant interactions between Teotihuacan and societies in eastern Michoacan. Sergio Gómez has found ceramics from that region in a Teotihuacan apartment compound a short distance east of the well-known enclave of people with Oaxacan ties (Gómez Reference Gómez Chávez and Gallut2002; Gómez and Gazzola Reference Gómez Chávez, Gazzola and Faugère2007), and Filini and Cárdenas (Reference Filini, García and Faugère2007) illustrate Teotihuacan-related objects from the Cuitzeo Basin of northeastern Michoacan, just south of the Guanajuato border. They illustrate an everted-rim cylinder vase so typical of the Miccaotli-Early Tlamimilolpa phases of Teotihuacan that it may well be an import (Filini and Cárdenas Reference Filini, García and Faugère2007:148, Figure 6, vessel on the right). This suggestion could be tested by instrumental neutron activation analysis. They also illustrate a slate disk that has a raptorial bird motif so typical of Teotihuacan that it may well have been made there. The archaeological evidence suggests interactions between Teotihuacanos and people in this region strong enough to make linguistic influences from Teotihuacan plausible. If a Mije-Sokean language was dominant at Teotihuacan, might this explain Mije-Sokean loans into Nawa and Tarasko?

However, I am troubled by the immense difference between the corpus of 150 or so signs in Teotihuacan's standardized notation (Langley Reference Langley1986) and any of the signs in the “Epi-Olmec” script found on the La Mojarra stela and elsewhere. It seems to me that if Teotihuacan had been any part of an Epi-Olmec sphere, there would have been less disparity in their signs.

In sum, the case against Nawa as a dominant language of the Teotihuacan state looks rather strong. The case that it was, instead, a Mije-Sokean language looks less strong to me. I think languages belonging to other Mesoamerican families, including Oto-Pamean and Totonakan, are still in the running.

IntroductionFootnote ¹

Although in their far-ranging article, Kaufman and Justeson discuss many terms they claim to be loanwords into NahuatlFootnote ² and other languages, these comments will center on Kaufman and Justeson's criticisms of Dakin and Wichman's (Reference Dakin and Wichmann2000) proposal that the words kakawa-tl, chi[h]kola:tl/chokola:tl and xi:kal-li Footnote ³ are of Uto-Aztecan, specifically Nahuatl or pre-Nahuatl origin.

Given that Ancient Mesoamerica is directed at an audience from a number of disciplines, I should note that historical linguists often have differences of opinion regarding the proper reconstruction for particular forms in a language family. These differences derive in part from the particular analyses of the rules of sound change that each considers to offer the best explanation for the data available for the language family, whether based on existing work or their own developed in the course of their analysis. During the past several years Kaufman, Justeson and some other linguists have worked to establish their hypothesis, based in large part on shared loanwords found in Mesoamerican languages, that the Olmecs spoke an early Mixe-Zoquean language and that that language has served as the source, along with some other languages, for example Totonac and proto-Zapotecan, for the loanwords. They argue that the only loans from Nahuatl entered other languages at a relatively late date. Kaufman and Justeson's decisions about the direction of the loans have come from their understanding of the historical development of the families in the area. Uto-Aztecan, and in particular Nahuatl, have been important in their considerations and Kaufman (Reference Kaufman1981) proposed a number of hypotheses about the development of that family and the southern-most language in it. In more recent work, including Kaufman and Justeson (Reference Kaufman and Justeson2007), Kaufman and others have used his perspective in analyzing forms they believe to be loanwords into Nahuatl. They contrast these hypotheses with their understanding of the evolution of the same forms in the Mixe-Zoquean language family, and in Zapotecan for the case of xi:kalli. In Dakin and Wichman (Reference Dakin and Wichmann2000) we described evidence and cited other work that made us question details of the analyses that led to their conclusions. I will expand aspects of our analysis here. I think it is important to remember that in linguistic reconstruction, we try to create frameworks that are as systematic as possible, but one is always dealing with hypotheses based on the empirical linguistic data that have survived for us to consider both in historical sources and as modern languages, and that the most probable situation is that we will never be able to have all the evidence we need, for example, from key languages that have disappeared, some without leaving any trace.Footnote ⁴ For that reason, all proposals for reconstructions, one would think, would best be considered as hypotheses, and not as absolutes.

Independent hypotheses based on linguistic arguments can introduce doubts into scenarios proposed on the basis of hypotheses developed from archeological and other non-linguistic evidence. That was the case for Dakin and Wichman's 2000 paper. Both authors had wondered about the origin of the word cacao in particular. For Dakin, finding the word ka'wa-/kawa in Sonoran Uto-Aztecan languages led to consideration that it might be a reduplicated word on the pattern of other Nahuatl terms such as na-naka-tl ‘mushroom’ from naka-tl ‘flesh, meat’ and ko-kone:-tl ‘doll’ from kone:-tl ‘woman's child’. It was Wichmann who brought in the philological evidence in Classic Mayan inscriptions that leads to the possibility for dating the presence of the word, which if it is of Nahuatl origin, points to an important participation of Nahua speakers at a much earlier point in Mesoamerica. In turn, archeological hypotheses would need to accommodate such participation. For that reason the different interpretations of the linguistic evidence need to be as precise as possible. Kaufman and Justeson have pointed out some difficulties they see in our identification of cacao and other words as of Nahuatl origins, and while I will try not to repeat points that Wichmann makes in his comment in this issue, I will attend to the questions from the point of a Uto-Aztecanist, beginning with the three terms that we called a kind of ‘chocolate complex’. These words are cacao, Nahuatl kakawa-tl, a beater-stick or chi[h]kolli, our hypothesized source for the Nahuatl word chi[h]kola:-tl ‘chocolate’, and the jícara or gourd cup used for serving, Nahuatl xi:kalli.

Cacao / kakawa-tl

Kaufman and Justeson are critical of our arguments against kakawa coming from MZ, but Wichmann discusses those in this forum. I will consider their comments about our proposed Uto-Aztecan etymology. Kaufman and Justeson make two basic criticisms against a Uto-Aztecan etymology.

The first is that according to their analysis all reflexes of proto-Uto-Aztecan intervocalic *p's are lost in processes of vowel fusion or changed into an h in Nahuatl. They hypothesize that if there were a pre-Nahua stem *kakawa-, it would have become *kaka:-tl. Their second principal argument is that the forms of the word for cacao in most languages do not have the final absolutive suffix, yet other loans from Nahuatl do, a point which Wichmann treats, and I will add brief comments.

Their criticism about intervocalic pUA *p in the protoform we reconstruct for cacao, takes into account the southern forms basically: Kaufman and Justeson state that they would accept the reconstruction of a proto-Sonoran form *kava on the basis of the words for ‘egg’ in Sonora languages cited in Dakin and Wichman (Reference Dakin and Wichmann2000), and derived from a pUA *kapa. They comment (2007:214) “These data are sufficient to reconstruct proto-Sonoran *kava.” But then they continue, “Contrary to Dakin and Wichmann's assumption, however, a Southern Yuta-Nawan form *kava (which would yield Sonoran *kava) cannot be the source of Nawa kakawa-tl.” Kaufman and Justeson (Reference Kaufman and Justeson2007:215) hypothesize that “postvocalic Yuta-Nawan single *p (like initial *p) changes to [v] in Southern Yuta-Nawan, which shifts to [h] in Koran and Nawa, and this is subsequently lost in Nawa.”

Although I am in agreement that pUA *w and *p can go to h or when part of a VCV sequence be fused into a long vowel in Nahuatl under certain conditions, those conditions are quite specific and do not affect a stem final syllable –wa. It is true that reflexes of both intervocalic pUA *w and *p fuse into long vowels in Nahuatl (Kaufman and Justeson Reference Kaufman and Justeson2007:214; Dakin Reference Dakin1996a). The specific conditions, however, depend on the position the syllable with the w/p occupies within the word. When followed by a derivational morpheme, the sequence VC ^[R]V can fuse into long vowel, for example *awo > o: in their cited form *navo-tsi-ABS. ‘prickly pear fruit’. Here *navo- goes to no:- but it is important to note that the change occurred because a second derivational morpheme *-tsi followed.

The nature of the second vowel has also been key in determining whether a change would take effect. A mid-vowel *i after *w/v is lost before the absolutive suffix, but when the vowel following *w/v is *a, it is not lost. The following examples show these contrasts.

In forms such as pUA *kwaLa-wi- ‘eagle’Footnote ⁵, we do get a long a: from an *aLa in Nahuatl because the *aLa sequences is followed by augmentative *–wi, giving k ^wa:-w-tli. Note that the–i of the –wi suffix is lost here, but the w does not go to h because it follows a long vowel, a general rule in Nahuatl (Dakin Reference Dakin, Fernández and Aguilar2000). The *wi after a short vowel other than i also reduces to w and but then does goes to h, as in *tawi- ‘hawk’ > Nahuatl tloh-tli ~ toh-tli Examples from Hill/Miller ta-51 bird sp. include Cahuilla távix; Cupeño távix; Guarijío tahi'iwe, Cahita tá:we and Classical Nah tlo'tli ‘sparrow hawk’.

Kaufman and Justeson's key example in their argument that intervocalic *p and *w always goes to h or is lost in Nahuatl, is se-tl ‘cold’, found transcribed with a short vowel for some dialects, but they have perceived it to have a long e, and give “se:tl” in their data. They claim that this long e: developed from a “proto-Sonoran” *seve. Given that the reflexes of pUA *i contrast with those of *a in other positions, the example is not relevant, since one would expect that sequences with eve/ewe would behave differently from those with Vwa/Vva in this position as well. The primary point is that when the *-wa or *–pa of a proto-Uto-Aztecan root is not followed by a derivational morpheme, the sequence does not fuse into a long vowel or lose the final –a permitting a w to go to h. Contrary to their analysis, Kaufman and Justeson's proto-Sonoran *RED-kava- would be expected to go to Nahuatl kakawa-tl.

Nahuatl Nouns in –wa-tl

There are a number of UA cognate sets for Nahuatl words ending in –wa like kakawa-tl:

1. ewa-tl ‘skin’, (Hill/Miller pi-11) Eudeve vewát (gen. véhte, ac. véhta) ‘skin’; véve ‘one who has skin’; Cahita bé:wa ‘skin, peel’. Also Huichol (Grimes et al. Reference Grimes, de la Cruz Avila, Vicente, Díaz, Díaz, de la Rosa and Rentería1981:59) has heu- ‘on the surface of the body’, and Tubar –pe-. The northern Uto-Aztecan languages have a –hi second syllable, for example, Mono pihi, ‘hide’, Western Shoshone pihi ‘fur, outer covering’, which may or may not correspond to the –wa in Nahuatl and the Sonoran languages, and Hopi has piik ^ya ‘hide, skin’

2. owa-tl ‘green corn stalk’: Eudeve sún-oma ‘corn stalk’; perhaps Hopi öövi('at) ‘cob heel, butt end of the corn cob; Stubbs Reference Stubbs2008:289: #2109, adds Cupeño öá’ya ‘trunk, stem’ and Comanche owóora ‘tree trunk’.

3. pawa-tl ‘a kind of big tree; a species of avocado tree’, cf. Stubbs(Reference Stubbs2008:223) #1504. *pawa ‘oak’; Cupeño páwi-x ‘blue oak’; Tubar amwá-t ‘oak’, CN a:wa-tl ‘holm oak’, (Carochi Reference Carochi1983 [1645]:116: ‘encina’); Guarijío awé ‘a kind of oak’. A possible northern cognate is Ute pawá-pu, ‘cedar tree’ (Southern Ute Tribe, 1979). This word, which Kaufman and Justeson consider a loan into Nahuatl from Totonac l-paw is of special interest. A more detailed examination of one of the first sound changes recognized in the development of Nahuatl from Uto-Aztecan, that initial pUA *p > Ø, is not really so straight-forward. Kaufman and Justeson (Reference Kaufman and Justeson2007:196), however, are less critical and state that “Any non-verb in Nawa with initial p is a strong candidate for a borrowing, because p regularly disappears in word-initial position, except in some words that habitually occur with prefixes so that the p is not usually initial.” However, a number of intransitive verbs begin with *p in Nahuatl which have no subject prefix in the third person, perhaps the most common form in discourse. In addition there are a number of other Nahuatl nouns that begin with p that have cognates in other UA languages and that are not commonly possessed, such as pawa-tl. This word has a paradigmatic relationship with a:wa-ka-tl, since a number of Nahuatl (C)VCV-ka- nouns are derived from (C)VCV- roots (see Dakin Reference Dakin2004), and pawa-tl also has a doublet a:wa-tl ‘oak’, so the term may refer more to a generic kind of large tree. In any case, rather than explain it as a borrowing from Totonac we need to find an explanation for the doublets. Several words show the same alternating initial p/Ø for pUA *p. There are enough exceptions and variations to see that the rule regarding initial pUA *p and its Nahuatl reflexes as p, Ø, and h is a complicated one. One would expect that there was some element that protected the initial *p, and that has since disappeared, which Ken Hill (personal communication 2008) has suggested might be some kind of Uto-Aztecan laryngeal. He points out there are similar variants providing doublets in Serrano for h/s and other alternations

4. Kaufman and Justeson (Reference Kaufman and Justeson2007:214) note *kopa with the round vowel o before the *pa, which drops, so that *kopa develops to Nahuatl *k ^wa-, ‘head’. It is in that way a clear contrast to *kawa/*kava with its sequence of two unrounded *a's and also to kowa:-tl ‘snake’ discussed below.

Nouns with a Word-final -wa:-

There are some long final vowels in Nahuatl, but they do not derive from proto-forms that included –wa as the last syllable of the root. Siwa:-tl and kowa:-tl are examples of this exception. The words in which –wa- with a short vowel comes before the absolutive contrast with the ones that end in –wa:- with a long a: in that position. All or almost all of the latter refer to animate nouns only, and examples include siwa:-tl ‘woman’ (PUA-KD #suwaLa, Stubbs [2008:336] #2508a-d), ko(:)wa:-tl ‘snake’, awa:-tl ‘furry caterpillar’, wa:wa:-tl ‘azotador (a kind of furry stinging caterpillar)’, (probable cognate with Cora háhari, pl. hahari-te ‘worm’). Where does the –wa:- come from? Evidence from the Tarahumara and Guarijío sequences –wa-ra or –wa-la on nouns of various kinds (Medina 2008) would lead one to suggest that Nahuatl *-wa:- comes from a UA *wa-La or *pa-La sequence incorporated as part of the nouns siwa:-tl, etc., and perhaps also cognate with the pUA possessed-noun or incorporating noun (Dakin 1993) markers *wa'a / *pai/pan.

In their analysis of –wa:-tl nouns, Kaufman and Justeson contrast the single intervocalic *p, which they propose is always lost or becomes h in Nahuatl, with their reconstruction of Uto-Aztecan*-nw- clusters to explain the presence of intervocalic w's in Nahuatl that reconstruct to pUA *w. Their forms *konwa ‘snake’ and *sunwa ‘woman’ with nw sequences are at variance with the reconstructions of other Uto-Aztecanists (Munro Reference Munro1973:136). Even if some cognates in northern languages have –nw-, as Kaufman and Justeson's hypothesize, that sequence probably developed from *-wa-La, which gave *w-L and then the lateral l become a nasal by regular sound change *pUA l > NUA *n, and assimilated to the velar w, –NGw-. In Nahuatl the *-wa-La sequence did become –wa:- with a long a:.

Kaufman and Justeson's (Reference Kaufman and Justeson2007:204) second criticism of our cacao etymology is that forms of the word cacao that appear in other languages in general do not end in the absolutive suffix; they claim that most Nahuatl borrowings into other languages include the absolutive suffix, so that cacao with no absolutive suffix must not come from Nahuatl. However, Kaufman and Justeson (Reference Kaufman and Justeson2007:205, Table 3) state that the Nahuatl words borrowed with the absolutive, entered the Soteapan Zoque after the arrival of the Spanish. This would have been a much later stage of language contact with quite different sociolinguistic characteristics from those for the borrowing of cacao. Wichmann points out that actually a fairly large number of Nahuatl words have been borrowed into Mayan and other languages without the absolutive suffix. The borrowed forms fit into four fairly regular general patterns that may reflect on the Nahuatl dialect that was their source: (1) Words ending in resonants that take an epenthetic i in some but not all Nahuatl dialects have neither the i nor the absolutive suffix: xa:m-i-tl > xan ‘adobe’; tena:m-i-tl/tena:n-tli > tenam/tinam/ tinamit ‘stone wall/town’ in contrast with similar chinamit, a which has the i- and the absolutive –t, but refers both to a fence and to a post-classic political division, and must be later; (2) Also when borrowed, the –in after –ch- is dropped in Nahuatl nouns that are found with –in and sometimes with -tli: michin > -mich/-mech ‘fish’, to:chin/to:ch-tli > Kanjobal tox ‘rabbit’;(3) Words in –chi-tl drop the –i-tl: xo:chi-tl > Mam suxh (?), (xo:chi-tl is also found as xo:chin in old texts); (4) Nahuatl nouns in –l have no –li. (5) Papantla Totonac has l-paw for the pagua tree. If my analysis (Section B) that it is Nahuatl is correct, paw would be another Nahuatl word ending in wa- like kakawa-tl borrowed without the absolutive suffix into Totonac.

Additional Notes on the Semantics of ‘Cacao’

Kaufman and Justeson (Reference Kaufman and Justeson2007:195) state that “Throughout Meso-America, the peanut …is named after ‘cacao’. Usually it is called ‘earth-cacao.” This translation seems more likely a folk etymology by borrowing languages, with the original formation based on the resemblance of the brittle-shelled underground peanut to an eggshell, as ‘earth-egg’. A compound with es- ‘blood’ in the colonial Molina dictionary <ezcacauatl> ‘scab or crust of mange’ also reflects the sense of a brittle shell (Campbell Reference Campbell1985:36).

The article by Terrence S. Kaufman and John Justeson (Reference Kaufman and Justeson2007), “The History of the Word for Cacao in Ancient Mesoamerica,” was written in response to an earlier article by Karen Dakin and Søren Wichmann (Reference Dakin and Wichmann2000) that argued for a Yuta-Nawan (Uto-Aztecan) origin for the widespread Mesoamerican terms for cacao and chocolate. Kaufman and Justeson counter this argument in a lengthy, detailed, and often near-vitriolic response, presenting evidence that the terms for cacao are of Mije-Sokean (Mixe-Zoquean) origin. Polemics aside, their thesis is convincing, and more recent research can in fact provide further support for the association of Mije-Sokeans and Theobroma cacao.

Kaufman and Justeson note (Reference Kaufman and Justeson2007:193–195) that contemporary archaeological evidence demonstrates the presence of cacao in Mesoamerica as early as 600 b.c., in Belize (Hurst et al. Reference Hurst, Tarka, Powis, Valdez and Hester2002), and that vessels associated with the preparation and consumption of beverages made from cacao are found earlier, 700–1000 b.c., in the Ulua Valley of Honduras (Henderson and Joyce Reference Henderson, Joyce and McNeil2006). A majority of indigenous Mesoamerican languages have terms for cacao of the form kakaw(a), and the authors' goals are, first, to demonstrate that these terms derive from Mije-Sokean antecedents, and second, that a Yuta-Nawan origin for the terms is highly unlikely, even impossible.

In order to make their detailed treatment of the linguistic evidence more understandable, the authors discuss their methodology for the analysis of suspected language borrowings in detail (2007:195–199), with special attention to the identification of loan words from Yuta-Nawan languages. The point of this discussion is to show that many apparent loans from Yuta-Nawan to other Mesoamerican languages, especially those proposed by Dakin and Wichmann, can be shown to have diffused in the opposite direction, if the intimate details of the languages involved are taken into account. Having prepared the groundwork, Kaufman and Justeson (Reference Kaufman and Justeson2007:199–203) proceed to propose their “Mije-Sokean hypothesis”, namely that *kakawa ‘cacao' was a proto-Mije-Sokean word, and that it was widely borrowed into other languages of Mesoamerica, including Nawatl, the Yuta-Nawan language to which the origin of the term is attributed by Dakin and Wichmann. They repeat their position (stated in Justeson et al. Reference Justeson, Norman, Campbell and Kaufman1985) that the diffusion of this term took place in post-Olmec times.

In the case of Mayan languages, Kaufman and Justeson date the borrowing to between 200 b.c. and a.d. 400. The latter date is established on the basis of the appearance of the term kakaw in Classic period Maya hieroglyphic inscriptions, presumably with the meaning ‘cacao', Theobroma cacao (as evidenced by chemical analysis of the contents of vessels labelled as ‘drinking vessels'), and they present evidence that a second hieroglyphic term, probably read b'ajläm kakaw (‘jaguar cacao'), referred to T. bicolor (Spanish pataxte). The earlier date is based on data from Greater Tzeltalan, a branch of Western Mayan that includes Ch'olan and Tzeltalan languages. Before diversifying into its component parts around 200–100 b.c., Greater Tzeltalan underwent a sound change that transformed k into ch, and had the term kakaw been borrowed before this sound change. it would manifest itself in these languages as *chachaw, which it does not. Therefore the borrowing of Mije-Sokean kakaw(a) into Mayan—at least this branch of Mayan—must have taken place after about 200 b.c., and the loan word is attested in (Ch'olan) hieroglyphic texts by a.d. 400. These dates correlate well with the archaeological evidence that the breakup of Greater Tzeltalan took place about the time its speakers began to move into the Chiapas Highlands, where they would have been in intimate contact with the Mije-Sokeans who occupied the Grijalva Valley at that time.

Noting that the physical evidence for the presence of cacao in Mayan archaeological context predates these happenings by a considerable amount of time (in Belize, as early as 600 b.c.), the authors note that there is a native term for cacao in some Mayan languages, derived from proto-Mayan *pe:q. Based on its meaning in languages where the term has survived and not been replaced by the borrowed kakaw, this term may have referred to uncultivated cacao, that is, Theobroma bicolor. A number of languages have replaced this term with a phrase like b'ahläm te' (literally, ‘jaguar tree') which Kaufman and Justeson believe may be derived from the hieroglyphic term b'ahläm kakaw cited above; this replacement can be dated to no later than the end of the Classic period.

Having stated and supported their Mije-Sokean hypothesis, the authors turn to the offensive (in both senses of the word), and attack the proposal of Dakin and Wichmann (Reference Dakin and Wichmann2000) that kakaw terms are of Yuta-Nawan origin (Kaufman and Justeson Reference Kaufman and Justeson2007:203–215). This entails a discussion of the general features of borrowing into and out of Nawa (or Nahua, a language grouping that is composed of a large number of variants, including those known as Nahuatl, Nahuat, and Pochutec) and a lengthy attack on the treatment of Mije-Sokean data by Dakin and Wichmann. The gist of their argument is summarized in six points (Kaufman and Justeson Reference Kaufman and Justeson2007:204–206): (1) Nawa in general is influenced by Mije-Sokean (and Totonakan, and to a lesser extent by Wastekan), but not vice-versa. (2) Proto-Nawan borrows heavily from Mije-Sokean, both in vocabulary and in grammar. (3) Proto-Mije-Sokean has no loan words from (Yuta) Nawan, although individual Mije-Sokean languages do. (4) No Nawa loans in Mesoamerica can be dated to before the Late Classic (ca. a.d. 1000). (5) The form of 'cacao' terms does not conform to the pattern of Nawa loans, and in any case, (6) cacao doesn't grow near the Basin of Mexico nor is it found in the northern areas from and through which the Nawa migrated into Mesoamerica. In other words, the Nawa migrated into Mesoamerica from an area where cacao was unknown, and were takers of influence from Mesoamerican languages and cultures, not givers, until about a.d. 1000, by which time the Mije-Sokean term for cacao was widely diffused, including into the language of the Nawa.

To the south of Mesoamerica, the term of Mije-Sokean origin was further diffused to many languages of Central America (Kaufman and Justeson Reference Kaufman and Justeson2007:215–217). The shortened forms of the word that occur in Chibchan languages, among others, suggest a route of diffusion from Mayan to Honduras Lenka (Lenca), and from there southward. There are no South American Chibchan cognates that would support a hypothesis of Chibchan origin.

A second question regarding cacao is the origin of the names for the drink 'chocolate', a subject also addressed by Dakin and Wichmann, who propose an origin in Nawa. Other authors have proposed an origin in Mayan languages. Kaufman and Justeson (Reference Kaufman and Justeson2007:217–231) present linguistic, ethnohistorical and even epigraphic evidence to the contrary, and suggest (p. 234) that since terms like chocola:tl are unattested in early Colonial period dictionaries of Nawa, they may not have existed in pre-Columbian times.

Kaufman and Justeson (Reference Kaufman and Justeson2007:232–234) conclude their article with a discussion of the culture historical inferences of the distribution of terms derived from Mije-Sokean *kakaw(a). They agree with Dakin and Wichmann that Teotihuacan may have played a central role in this diffusion, although they accept none of Dakin and Wichmann's reasoning on this point. The basic disagreement here is that Dakin and Wichmann take Teotihuacán to have been Nawa speaking, while Kaufman and Justeson do not. In a series of recent papers, some as yet unpublished but widely circulated, Kaufman (Reference Kaufman2000–2007, Reference Kaufman2001) and Kaufman and Justeson (Reference Kaufman and Justeson2007) point out that the large number of recently identified Mije-Sokean loans into languages of central Mexico can best be explained by the presence of a culturally-influential Mije-Sokean elite at Teotihuacan. Their linguistic influence is seen at its strongest in Totonakan, which has more Mije-Sokean loans than any other language group in Mesoamerica (about 50). Of the languages of northern Mesoamerica, Tarasko is in second place (with 17 Mije-Sokean loans). Nawa (18) and Wastekan (12) are next. The authors interpret these loans—and the associated archaeological record—as the result of a migration to the Basin of Mexico of a previously unsuspected branch of Mije-Sokean, the Northern Mije-Sokeans, about 1200–1000 b.c., more or less the time when Mijean and Sokean were diversifying from their common source. The huge impact on Totonakan suggests that the Northern Mije-Sokeans coexisted for centuries with Totonakans, and while they do not pursue this point here, it is notable that Juan de Torquemada (Reference Torquemada1969 [1615]:bk. 3, ch. 18, p. 278) was told by the Totonakos that they had lived in Teotihuacan before moving to the Sierra de Puebla, where he recorded their oral history.

A Teotihuacano center of diffusion for the Mije-Sokean term for cacao would be consistent with the dates for Mayan borrowing of the term, sometime between 200 b.c. and a.d. 400, and there is ample evidence for Teotihuacano presence in the Maya area in this time period. Under just what circumstances the Mayas replaced their native term for Theobroma with a borrowed term is unclear, but Kaufman and Justeson (Reference Kaufman and Justeson2007:234)suggest that the borrowing took place under conditions of culture change; either the introduction of the cultivation of cacao (leaving the native term to describe wild Theobroma bicolor), or the introduction of new methods for processing cacao, or, “more likely, a rising importance of its use—perhaps in a ritual context or perhaps through an economic importance, for example, as money.”

New Evidence in Support of the Hypothesis

Having summarized the argument put forth by Kaufman and Justeson in 2007, I would now like to call attention to recent research that gives further support to their hypothesis by suggesting an even stronger—and earlier—relationship between cacao and the Mije-Sokeans.

In a series of papers, John Clark and his coauthors (e.g., Clark and Pye Reference Clark and Pye2003, Reference Clark and Pye2005) have noted that around 1900 b.c., in the Soconusco region (the Pacific coastal plain of Chiapas and parts of Guatemala), the earlier Chantuto hunters and gatherers (described by Voorhies Reference Voorhies1976) were suddenly replaced by agriculturalists. This sudden culture change is unlikely to represent a local development; rather, it implies the entrance of a new population to the area. Besides agriculture, this new population introduced elite ceramics to the region, ceramics that had no local antecedents. Clark has named these people the Mokaya, a name derived from the Mije-Soke term for maize, since he can trace their development forward in time to the Mije-Sokean peoples represented by Izapan cultures on the Pacific coast and Olmecs on the Gulf coast. That is, the Mokaya people must have been speakers of pre-proto-Mije-Sokean, the ancestor of the Mije-Sokean family of languages.

Proto-Mije-Sokean, the immediate ancestor of Mijean, Sokean, and Northern Mije-Sokean, dates to a somewhat later period, around 1500–1000 b.c., the earliest phase for which we can use the comparative method of historical linguistics to gain insights into the vocabulary (and therefore the culture) of the speakers of an ancestral language. Reconstructed proto-Mije-Sokean (Kaufman Reference Kaufman2001) has an agricultural vocabulary that includes nouns for corn (maize), corncob, cornfield, leached corn (Spanish nixtamal) and dried ground corn (pinole), as well as beans, chile, edible tubers, and the verbs for plant, harvest, and grind corn, a complex that fits well with the Mokaya archaeological culture.

The Mokaya ceramic tradition (Clark and Pye Reference Clark and Pye2003) begins with the Barra and Locona phases (ca. 1900–1700 b.c. and 1700–1500 b.c., respectively), which would correspond to the period of pre-proto-Mije-Sokean. Proto-Mije-Sokean dates to about 1500–1000 b.c., which fits well with the Ocós phase (ca. 1500–1400 b.c.), and Ocós ceramics are associated with agricultural villages growing maize and manioc. Around this time, proto-Mije-Sokean began to diversify into two (and perhaps three) distinct branches, and it appears to be the case that this was brought about by the migration to the Gulf side of southern Mexico of one segment of the population. Note that the Ojochi phase at San Lorenzo, on the Gulf side, can be described as “the local manifestation of the Ocos farming-village pattern” (Adams Reference Adams1991:49). The population on the Gulf side became the Sokean branch of the family, and the bearers of Olmec culture. The population that remained on the Pacific side became the Mijeans. Izapa, in the Mije-Sokean homeland, became the dominant force of the Pacific coast, and the cultural developments along the coastal plain from Soconusco to El Salvador (Izapa, Takalik Abaj, Bilbao, etc.) probably represent an expanding Mijean population. Later developments resulted in a more complex situation as Sokeans with Olmec culture came back to the Pacific coast and some Mijeans spread across the Isthmus of Tehuantepec to settle in the Sokean territory. At some point in this development, around 1200–1000 b.c., a third group migrated to central Mexico and became the Northern Mije-Sokeans at Teotihuacan. In the Early Classic period, the influence of these Mije-Sokeans stretched into the Maya area, at Kaminaljuyu, Tikal, Copan, and elsewhere (Kidder et al. Reference Kidder, Jennings and Shook1946; Martin and Grube Reference Martin and Grube2000:29–31, 193).

The Mokaya culture, then, is the antecedent of Olmec culture, the putative “mother culture” of Mesoamerica, along with the other early centers of cultural influence, Izapa and Teotihuacan. But recall that the Mokaya make a sudden appearance on the southern Mexican coast, bearing culture that has no local antecedents. Where, then, did they come from? Even asking this question runs the risk of accusations of heresy: How could the mother culture of Mesoamerica not be Mesoamerican in origin? Nonetheless, the evidence seems strongly to suggest that the Mokaya came from somewhere outside Mesoamerica. (Note that Clark has steadfastly declined to pursue the implications of his discoveries.) We cannot resolve this question with current knowledge, but it deserves to be asked, with the goal of guiding research towards a resolution, a new and more informed view of cultural development in Mesoamerica.

The Ecuadorian Hypothesis

In his landmark definition of Mesoamerica, Paul Kirchhoff (Reference Kirchhoff and Tax1952:24) noted that along the coast of Ecuador and northern Peru there was “a grouping of traits so typically Mesoamerican that it allows of no other interpretation but that of cultural diffusion.” In 1960, Michael Coe pointed out the similarities between Ocós ceramics and contemporary ceramics of western Ecuador (Coe Reference Coe1960). At the time this seemed highly unlikely, and this article has been generally ignored. But more recently, a good case has been made that cultures further up the Pacific coast have strong connections to western Ecuador. Patricia Anawalt (Reference Annawalt1992, Reference Annawalt and Townsend1998) has pointed out that west Mexican cultures share with western Ecuador such varied traits as shaft tombs, stirrup pots, fabric and clothing designs, whale tusk and Spondylus shell jewelry, and even biological species that must have been carried one way or the other, the hairless dog and the painted jay.

West Mexico has long been recognized as a distinct subarea within Mesoamerica, as it differs from the norm in having circular site layouts and circular mounds (the Tarascan yácatas) as well as the features just mentioned and a distinct ceramic tradition of large hollow figurines (albeit these are not dissimilar to some Olmec figurines). The archaeological record reveals a situation not unlike that of the Mokaya, the sudden appearance of a highly developed culture that replaces an earlier sparse population of hunters and gatherers, around 1500 b.c. The derivative cultures, the Teuchitlan Tradition, flourished in west Mexico until the Classic period when, except for the Tarascan area, they disappeared and were replaced by the Yuta-Nawan cultures that occupied the area in historic times. Their extinction took place about the same time as the fall of Teotihuacan, and probably for the same reasons, having to do with the incoming Yuta-Nawans.

Authors who have dealt with the similarities in west Mexican and Ecuadorian cultures have usually ascribed the situation to the result of contact and cultural diffusion rather than the intrusion of a new population. There is ample evidence of a long-distance trade route stretching from Peru to west Mexico (Martin Reference Martin2001, Shimada Reference Shimada, Salomon and Schwartz1999:350–517). From as early as Ecuador's Valdivia phase (2800–1100 b.c.) one of the items of local trade was Spondylus princeps (spiny oyster) shells, gathered offshore from the Valdivia area northward to Panama, and trade in Spondylus eventually expanded throughout Peru (Paulsen Reference Paulsen1968, cited in Martin Reference Martin2001). It is a reasonable speculation that the search for other areas of this scarce resource could have brought South American traders to the shores of west Mexico, where a second area of S. princeps stretches from Jalisco to Baja California (Lamprell and Whitehead Reference Lamprell and Whitehead1987:60). The round trip from Ecuador to Mexico requires a long stopover to await the annual shifting of ocean currents, and traders would have established permanent settlements along the route and at its terminals (Marcos Reference Marcos1977–1978). It is my hypothesis that the Ecuadorian features found in west Mexico are not the result of sporadic contact, but result from the establishment of permanent colonies that ultimately came to dominate the area culturally. Note that in the area of west Mexico that continued to develop the Teuchitlan Tradition, Michoacan, the indigenous population speaks a language unrelated to anything else in Mesoamerica, Tarasko (Tarascan, Purepecha, or Phorhepecha).

Other language isolates scattered along the Pacific coast may also represent the remnants of this trade route: Seri (to the north), Kwitlatek (in Guerrero), Tekistlatek (or Chontal of Oaxaca), Wavi (on the Isthmus of Tehuantepec), Xinka (on the Guatemalan coast) and Lenka (in El Salvador). In at least some of these areas there is physical evidence of long-distance contacts, such as the four-legged jaguar effigy grinding stones that are characteristic of Kwitlatek sites (Hendrichs Pérez Reference Hendrichs Pérez1946:69–77) and are clearly related to artifacts of the Chiriqui region of Panama.

In this context, we can see Mije-Sokean as just another language isolate in origin, but one which grew and diversified. There is linguistic evidence that Mije-Sokean once had a distinct grammatical structure that has now assimilated to greater Mesoamerican patterns, basic word order (the typical order of Subject, Verb, and Object). Campbell, Kaufman, and Smith-Stark note that Mije-Sokean “has certain constructions typical of verb-final languages, e.g., Noun + Postpositions, Adjective + Noun order, and Possessor + Possessed word order” (Campbell et al. Reference Campbell, Kaufman and Smith-Stark1985:547), but while that may represent an earlier SOV stage, Mije-Sokean now conforms to Mesoamerican “non-verb-final” order. Note that the other Mesoamerican languages that have SOV basic word order are Seri and Tol (Jicaque or Xikake, a language isolate of northern Honduras); Muisca, the prototypical Chibchan language of Colombia, also has SOV order. It is worth noting that the five Mesoamerican languages that have SVO word order are Mazatec (surely a special case because the rest of Otomanguean doesn't have SVO), and Tarasko, Kwitlatek, Tekistlatek, and Wavi, all Pacific coast language isolates. Is this simply a coincidence?

A Return to Cacao

Returning to the central topic of the paper under review, what does all this have to do with cacao? Kaufman and Justeson cite the earliest date for archaeological evidence of cacao in Mesoamerica as 600 b.c., in Belize (Hurst et al. Reference Hurst, Tarka, Powis, Valdez and Hester2002), but we can now push that date much further back and associate cacao directly with Mije-Sokean languages. The examination of a Mokaya bowl from the Barra phase at Paso de la Amada, Chiapas, and a vessel from El Manatí, in Veracruz, both in Mije-Sokean territory, “shows that chocolate (Theobroma cacao) was consumed by the Mokaya as early as 1900 b.c. and by pre-Olmec peoples as early as 1750 b.c., pushing back the chemical evidence of cacao use by some 700 years” (Powis et al. Reference Powis, Hurst, Rodríguez, Blake, Cheetham, Coe and Hodgson2007).

Thus, the earliest known cacao in Mesoamerica comes from the pre-proto-Mije-Sokean area, and Soconusco, the Pacific coast of Chiapas and western Guatemala, becomes one of the most productive areas of cacao cultivation, as does the Gulf coastal region occupied by Mije-Sokean peoples. Since T. cacao is apparently native to South America (Bletter and Daly Reference Bletter, Daly and McNeil2006:31), it is more than likely that this evidence represents the first importation of the species to Mesoamerica—all the more reason for the word for cacao to be of Mije-Sokean origin.

The colonization of Soconusco by the ancestors of the Mije-Sokeans may be a parallel phenomenon to that of west Mexico. Cacao had become a major trade item in South America, and the search for fertile grounds for cacao plantations may well have been the initial attraction of the area for the incoming settlers. Whatever the case, the Mije-Sokeans were spectacularly successful immigrants, and virtually all of the early centers of cultural development can be attributed to them. The glaring exception is Oaxaca, and it is notable that there is virtually no borrowing of the Mije-Sokean term for cacao in Otomangean languages, the possible exceptions being Chinateko and Amusgo (Kaufman and Justeson Reference Kaufman and Justeson2007:217).

In their comprehensive study, Kaufman and Justeson (Reference Kaufman and Justeson2007:193) trace “the origin and spread of the widely diffused form *kakawa (and variants) as a word for Theobroma cacao in Mesoamerican languages.” They conclude that their “study has shown that *kakaw(a) has an unimpeachable pedigree as a native Mije-Sokean word” (Kaufman and Justeson Reference Kaufman and Justeson2007:232) that originated before the two languages separated between 1200 and 800 b.c. (Kaufman and Justeson Reference Kaufman and Justeson2007:206). Kaufman and Justeson (Reference Kaufman and Justeson2007:203–204) further argue that the inhabitants of Teotihuacan spoke a Mije-Sokean language and that the word for cacao spread across Mesoamerica with Teotihuacan's regional influence at some time between a.d. 100 and 500 (Kaufman and Justeson Reference Kaufman and Justeson2007:232). For their part, Dakin and Wichmann (Reference Dakin and Wichmann2000) propose that Teotihuacanos spoke an early version of Nawa and that the word for cacao is derived from a Nawa word for egg. Dakin and Wichmann are in agreement, however, with Kaufman and Justeson that the word for cacao spread across Mesoamerica during the Early Classic period as Teotihuacanos expanded their influence. Instead of evaluating arguments for the linguistic source of the Teotihuancos' word for cacao, in the comment that follows I expand the discussion temporally and discuss the origins of this word into Mesoamerica. To do so, I review the implications of the recent evidence from Soconusco that Theobroma cacao arrived in Mesoamerica as early as 1900 b.c. (Powis et al. Reference Powis, Hurst, Rodríguez, Blake, Cheetham, Coe and Hodgson2007), and so, considerably early than the separation of Mije and Sokean.

Implications of Early Cacao in Soconusco

The origins of the Mesoamerican word for cacao is a different issue from where the plant Theobroma cacao originated? Or, when it was first used to make a caffeinated drink? Evidence indicates this tree originated in the South American Amazon (see Ogata et al. Reference Ogata, Gómez-Pompa, Taube and McNeil2006). A drink made from its seeds was then put in ceramic vessels dating to the Barra phase (1900–1700 b.c.) in the Soconusco region of Mesoamerica (Powis et al. Reference Powis, Hurst, Rodríguez, Blake, Cheetham, Coe and Hodgson2007). The Barra phase ceramic assemblage is the earliest known in Soconusco with nearly 90% of the vessels consisting of elaborately decorated tecomates likely used to serve liquids (Clark and Blake Reference Clark, Blake, Brumfiel and Fox1994:23–27). Cacao therefore must have spread north from the Amazon and up through Central America at some point prior to its earliest documented use in Soconusco. This date for cacao use is approximately a millennium older than that which was available to, and cited by, Kaufman and Justeson (Reference Kaufman and Justeson2007:194). Therefore, rather than arriving in Mesoamerica at the time of proto-Mije-Sokean, this tree arrived while some ancestral form of the language was spoken.

Given a date as early as 1900–1700 b.c. for cacao use in Soconusco, and the need for it to have spread up from South America prior to that point, a very early, Early Formative arrival is necessary and an Archaic-period origin of cacao in Mesoamerica is a distinct possibility. While there is no direct evidence of Archaic period cacao use, I propose that is likely enough the case to be worth exploring the implications.

The Soconusco region is a likely place for a South American crop like cacao to have first arrived and been cultivated. The region is centrally located on a transportation route with great antiquity that runs along the Pacific coasts of Oaxaca, Chiapas and all the way down through Central America (Navarrete Reference Navarrete, Lee and Navarrete1978). Soconusco is also where the confluence of rainfall, constantly warm temperatures and well drained soils provide some of the very best cacao production anywhere in Mesoamerica (Lowe et al. Reference Lowe, Lee and Espinoza1982: 43–44). If Theobroma cacao spread from the Amazon up through Central America at some point around 4000 years ago, Soconusco would have been a likely place for it to have first arrived and been cultivated.

As the inhabitants of Soconusco spoke a Mije-Sokean language, Kaufman and Justeson's historical reconstruction of the earliest Mesoamerican word for cacao in this language family is consistent with the early evidence for cacao. This raises the question, however, where did the ancestral Mije-Sokean speakers get the word from? The Archaic period foragers or the very early, Early Formative villagers in the Soconusco may have themselves spoken a Mije-Sokean language, but this would be extremely hard to know. If they did, they may have invented a new Mije-Sokean word for the cacao tree and its fruit. Alternatively, the early inhabitants of Soconusco might have spoken some other language that supplied the loanword and the only surviving traces of this language that went extinct are a handful of loanwords borrowed by the proto-Mije-Sokean peoples who replaced them. A third possibility is that the Archaic period foragers or very early, Early Formative peoples in Soconusco (whatever language they spoke) adopted the word for this tree that their neighbors to the south employed. Given the current lack of evidence, these three possibilities seem equally viable hypotheses.

As a non-linguist I will not evaluate the detailed points of disagreement that Kaufman and Justeson have with Dakin and Wichmann's claim of the non-Mije-Sokean nature of the word (presented in Kaufman and Justeson Reference Kaufman and Justeson2007:206–212). It is worth noting though that of the six words Kaufman and Justeson (Reference Kaufman and Justeson2007:Figure 4) find morphologically unanalyzable, five (cacao, mamey/plantain, guava, Guiana chestnut and peccary) are foods. These words would therefore be good candidates for a systematic set of loanwords, perhaps similar to French gastronomic loanwords into English, as Kaufman and Justeson (Reference Kaufman and Justeson2007:227) mention. As South American or lower Central American peoples introduced cacao into Mesoamerica, it is not unreasonable that the names for this crop travelled in the same direction. Coe (Reference Coe1962) long-ago proposed that the earliest ceramics in the Soconusco (including the Barra phase types Powis et al. [2007] have documented cacao residue in) may have been spread from South or Central America. Clark and Gosser's (Reference Clark, Gosser, Barnett and Hoopes1995:Figure 17.1) more recent review of the earliest known ceramic complexes is consistent with the northward spread of ceramics (or at least the spread of how to make them) from South America. In fact, the earliest ceramics are documented in the Amazon (Roosevelt Reference Roosevelt, Barnett and Hoopes1995) where cacao also originates. Therefore, crops, words for foods and food preparation technology could have all spread together in a manner similar to that proposed by Bellwood (Reference Bellwood2005). But, rather than spreading as a package of subsistence crops that transformed the economic structure of society, specialty items such as cacao and technology to make fancy vessels to serve it in could have spread together.

The extent of interaction between peoples in distant areas of the New World during the Archaic period is generally not incorporated into reconstructions of society during this time. What glimpses we do have indicate that plants and raw materials were moved over considerable distances by these peoples. We know that obsidian was transported hundreds of kilometers from the Guatemalan highlands to the Soconusco during the Late Archaic (Nelson and Voorhies Reference Nelson and Voorhies1980). At approximately the same time, maize was transported from the Balsas region in the central Mexican highlands south with the earliest dates in South America between 2200 and 2700 b.c. (Blake Reference Blake, Staller, Tykot and Benz2006:60). An Archaic period movement of cacao northward would therefore make it roughly contemporaneous with the spread of maize in the opposite direction. Presumably, the early inhabitants of the Amazon and the Balsas Valley each had their own words for cacao and for maize. But, of course, such names are not necessarily the same as those that have persisted into modern times. Kaufman and Justeson (Reference Kaufman and Justeson2007:201) provide an example of this as the early Maya word *pe:q was replaced by *kakaw(a) with the former surviving only in restricted or semantically associated meanings. The existence of other extinct such terms, but ones that lack evidence is also quite possible.

Ogata et al. (Reference Ogata, Gómez-Pompa, Taube and McNeil2006:84) posit that undomesticated strains spread north into Mesoamerica from Amazonia and were then bred to become the cacao we know today, and a separate strain was domesticated in South America. This scenario is possible but there still must have been a desirable use of this tree species for it to have been spread so far north. To explain the early spread of maize south long before cob size made it the food source it was later to become, Blake (Reference Blake, Staller, Tykot and Benz2006:68–69) proposes that there were numerous uses for it, such as for making an alcoholic drink. Neither maize nor cacao would have moved from its natural habitat without purposeful human intervention. The onus is on Ogata et al. (or others who propose a similar scenario) to explain why the cacao tree was spread so far from its natural habitat in undomesticated form. If the spread of cacao was not originally for its caffeinated beans perhaps it was for fermenting the pulp, as later peoples did, to make an alcoholic drink. Maize beer and cacao wine might thus have been the original impetus for the spread of these crops.

Concluding Thoughts

A number of questions are raised by a very early, Early Formative period, and the possibility of an Archaic period arrival of cacao into Mesoamerica. Precisely when, and following what route, did cacao spread north? And, what sort of social mechanisms were responsible for its spread over the considerable distance from the Amazon up to Mesoamerica? Linguistic data from Central America seem to indicate that the Mije-Sokean word for cacao (especially present in Chibchan languages) spread from north to south (Kaufman and Justeson Reference Kaufman and Justeson2007:215–217). Such evidence is not very helpful in tracking Theobroma cacao's known spread northward. Further, “there is no ‘cacao’-like form in any South American language that is not the result of Colonial period diffusion” (Kaufman and Justeson Reference Kaufman and Justeson2007:216). Archaeological data might therefore be better suited to provide answers. However, a serious impediment on this front is that macrobotanical and microbotanical cacao remains do not preserve well. Furthermore, the recent detection of caffeine residue left in ceramic vessels in the course of cacao processing and/or consumption (Hurst et al. Reference Hurst, Tarka, Powis, Valdez and Hester2002; Powis et al. Reference Powis, Hurst, Rodríguez, Blake, Cheetham, Coe and Hodgson2007) is obviously not possible for Archaic period peoples that did not use ceramics.

Without the concerted effort to find relevant data, the origins of cacao in Mesoamerica will remain a mystery. With the limited Archaic period remains known from Mesoamerica, we will need to be very lucky with botanical preservation to find earlier evidence of cacao use than we currently possess. However, large programs of flotation from Archaic period sites would increase our chances of recovery. In addition, residue analysis of Archaic period stone tools used for grinding cacao might be another way to recover traces of theobromine. Detailed analysis of Amazonian languages for words that could serve as sources for the word cacao or other terms such as the Maya *pe:q might also prove useful. In addition, if viable uses for non-domesticated cacao trees can be proposed (such as alcohol from fermented pulp) then such words could also be sought in Mesoamerican, lower Central American and South American languages. Histories of such terms might provide more clues to the early movement of cacao into Mesoamerica.

The arguments presented by Kaufman and Justeson (Reference Kaufman and Justeson2007) were not meant to address the Archaic period Mesoamerican world. However, by tracing the lexical diffusion of the word for cacao their study is relevant “…to work on the intercultural interaction that probably dates to Preclassic times” (Kaufman and Justeson Reference Kaufman and Justeson2007:193). Here I have proposed that tracing the history of both the plant and the word for cacao can potentially provide insights into very early, Early Formative or even Archaic period intercultural interaction. Kaufman and Justeson's detailed analysis of the Mije-Sokean root of the word for cacao therefore outlines a substantial portion of the history of this word. Further evidence must be employed to finish the job by tracing this history back to document cacao's origins in Mesoamerica. Cacao was one of a host of plants and ideas that spread through the New World during the second or third millennium b.c. (or earlier). Unfortunately, few Mesoamerican archaeologists think about (let alone investigate) the forms of intercultural interaction that must have occurred during the Archaic period to have allowed maize, obsidian and quite possibly cacao to have spread over long distances. Linguistic evidence such as that presented by Kaufman and Justeson (Reference Kaufman and Justeson2007) have a role to play in helping us to do so.

Kaufman and Justeson (Reference Kaufman and Justeson2007) make many intriguing suggestions that I will not attempt to address. I will limit my comments to their discussion of the origin of the word ‘chocolate’ because recent archaeological and ethnohistoric research in southern Mesoamerica appears to provide an answer to the conundrum they present of differing origins and ages for the terms ‘cacao’ and ‘chocolate’. Linking archaeological and linguistic histories of cultures is fraught with problems but the task is irresistible nonetheless. As anthropologists we believe that material culture and language are part of the same nexus of social action and if given enough data, we can ferret out connections. Archaeological understanding of zones occupied by the Pipil at the time of Spanish conquest is currently expanding, allowing us to address several of the points made by Kaufman and Justeson regarding chronology and the relationship of the Pipil to cacao and its circulation. Three points discussed by the authors are particularly interesting from a southern Mesoamerican archaeological perspective: (1) that the diffusion of the words for ‘chocolate’ and ‘cacao’ was not related and did not occur at the same time (2007:219); (2) that the words chikola:tl and chokola:tl may have originated in a peripheral form of Nahua and then spread to Central Nahua (2007:221–222); and, (3) that chukula:t (Campbell Reference Campbell1985:200) is attested in Pipil (2007:216). From an archaeological point of view, all three points make sense. The Pipilteca were relative latecomers to southern Mesoamerica but appear to have played a part in transforming the consumption of cacao. I think Kaufman and Justeson are right that ‘chocolate’ is a Nahua term, and I suggest that the term spread from Pipil zones to other parts of Mesoamerica as part of changes in the production, consumption, and distribution of cacao during the Late Postclassic period.

That cacao has been consumed in Mesoamerica for a very long time is attested by the archaeological evidence of Formative period spouted vessels (Henderson and Joyce Reference Henderson, Joyce and McNeil2006; Powis et al. Reference Powis, Valdez, Hester, Hurst and Tarka2002). Spouted vessels are found throughout much of southern Mesoamerica and are a consistent part of Late Preclassic period assemblages in El Salvador (e.g., Andrews Reference Andrews1976; Sharer Reference Sharer1978). The authors' linguistic reconstructions suggest a similar antiquity for the word ‘cacao,’ which referenced both unprepared and prepared states of the beans or kernels. In contrast, ‘chocolate’ appears to refer only to prepared cacao and seems to be a more recent term. Kaufman and Justeson's (Reference Kaufman and Justeson2007:217) suggestion that chokola:tl could be a borrowing from the Central Nahua speakers (“mejicanos”) brought into Guatemala by the Spanish after 1525 is undermined by their own data, which provide little evidence for the word in Central Nahua until well into the Colonial period. Lockhart (Reference Lockhart1992) also noted that the use of the term chocolatl is extremely rare, yet is certainly Nahua. If ‘chocolate’ originated in Pipil, then the sporadic occurrence of the word in central Mexican texts would be expected. The earliest Spanish reference to ‘chocolate’ should instead appear in documents from the Audiencia of Guatemala, which encompassed much of the Pipil zones. Unfortunately, I know of no such early reference, but researchers have only begun to explore this large corpus of documents.

Archaeological evidence for chocolate is circumstantial; a special preparation of cacao that is clearly chocolate is not yet archaeologically demonstrated and may very well have involved multipurpose tools and vessels or artifacts already associated with cacao. My suggestion is that a new name for a preparation of cacao would have most likely occurred in an area devoted to cacao production, distribution, and consumption. Language is more likely to be marked concerning an element of particular abundance, and new uses might well emerge in these same areas. New forms of consumption are often linked to broad societal changes, including significant changes in political economy (Mintz and DuBois Reference Mintz and DuBois2002). Both documentary and archaeological evidence from the early Spanish contact indicate that the Pipil zone of the Izalcos in western El Salvador was just such a region of intensive production and that by the time of Spanish contact cacao was commodified. Cacao cultivation was widespread in Mesoamerica, but intensive cacao production zones were limited due to the fickle nature of the domesticate (Bergmann Reference Bergmann1969). Xoconochco (Soconusco), the Izalcos of western El Salvador, and Suchitepequez far outshone all other production areas if Spanish tribute quotas are a reliable index of productivity (Ciudad Real 1973; Fowler Reference Fowler1987, Reference Fowler1989, Reference Fowler and McNeil2006; García de Palacio 1985 [1576–1587]). Of these three, the Izalcos was the smallest zone, yet tribute quotas were comparable to the other areas, so per unit production was exceptional. Fowler (Reference Fowler1987, Reference Fowler1989) has argued that high productivity occurred because of Spanish involvement, but the timing of peak production, the time required for new plantings to mature, and multiple references to Spanish ineptitude all suggest that the spectacular intensification was pre-Columbian and the Spanish managed only to mess it up.

Reorientation toward intensive cacao cultivation in western El Salvador appears to correlate with the arrival of the Pipil. Kaufman and Justeson (Reference Kaufman and Justeson2007:216–217) argue that the Pipil arrived in western El Salvador around a.d. 900, but to push Pipil arrival before a.d. 1000 would not square with well-dated archaeological remains. Archaeological patterning in the Izalcos region shows a comprehensive, radical change in the Postclassic period. Late Classic and Postclassic settlements very rarely shared the same locale, and settlement structure was entirely different. Late Classic sites were usually nucleated and organized into plaza groups; Postclassic settlement was dispersed, which accommodated the need for regularly spaced labor to tend cacao orchards and associated irrigation networks (Hirth Reference Hirth, Marcus and Sabloff2008; Sampeck Reference Sampeck2007). This shift in the archaeological signature of the region then laid the foundation for patterns observed in well-dated sites and contexts of the early Colonial period. Other elements of material culture such as ceramic forms and decoration exhibit equally distinct changes from the Late Classic to the Postclassic as well as similar continuities from the Postclassic to the early Colonial period. This patterning suggests that the Pipil arrived in western El Salvador in the Postclassic, and the strong Postclassic- to Contact-period continuities indicate that the Pipil reoriented the economy towards intensive cacao production when they arrived. The beginning of the Postclassic period is not securely dated in this region. It is possible a hiatus occurs between Terminal Classic depopulation and the dominance of the Pipil. A hiatus might have been very long and the Postclassic assemblage then changed quickly in the face of political and economic developments in central Mexico and coastal Veracruz from the fourteenth century to the fifteenth century, including the rise of the Aztec state. The earliest possible date for the inception of the Postclassic period in western El Salvador would have to postdate Plumbate pottery manufacture, about a.d. 1100, because Plumbate is associated with the last of the Late Classic patterning in material culture. Stylistic similarities of Izalcos region Postclassic ceramics with central Mexican wares also support this date. The arrival of the Pipil in El Salvador, though, is another matter, and several lines of evidence suggest that central El Salvador may have been occupied by Pipil earlier, during the Terminal Classic. Kaufman and Justeson (Reference Kaufman and Justeson2007:217–218) state that linguistic separation of Pipil from Gulf Nahua occurred about a.d. 900, which is too early for Pipil arrival in western El Salvador, but not central El Salvador. The linguistic and archaeological chronologies are not in conflict for Pipil expansion in general, but the Pipil did not arrive in every locale at the same time.

Kaufman and Justeson (Reference Kaufman and Justeson2007:194) are right to argue that all pre-Columbian societies tried to control the production and distribution of cacao. The nature of this control is not clear and undoubtedly changed over time, but Contact period sumptuary laws, such as noted in Garcia de Palacio 1985 [1576–1587], give us an inkling of forms of control over some kinds of consumption. Countering these restraints on consumption was cacao's Mesoamerica-wide distribution at contact. The extensive distribution of cacao could indicate merely that sumptuary laws were ineffective, but a reasonable alternative is that some kinds of consumption were limited and others permitted or promoted. In fact, Kaufman and Justeson provide evidence of the widespread linguistic linkage of ‘cacao’ with ‘money.’ Particularly provocative is that Xinca /tuwa/ means both ‘cacao’ and ‘money’ because Xinca speakers were in close proximity to the Pipil at the time of Alvarado's entrada (Barón Castro Reference Barón Castro1950). Kaufman and Justeson cite multiple examples (in Zoquean, Nahua, Zapotec, Yucatec, etc.) of the linkage of ‘cacao’, ‘8000’, ‘bag’ or ‘sack’ and ‘skirt.’ All of these terms are related to the ways cacao was measured as a commodity and the ways commodities were measured in terms of monetarized cacao. Even the linkage of ‘cacao’ with ‘skirt’ may not be chance, as cloth also functioned in some instances as a kind of money. I suspect the monetarization of cacao and cacao hyperproduction in the Izalcos region were related, mutually reinforcing processes that occurred relatively late in the history of pre-Columbian cacao use.

Two process of commodity consumption identified by Mintz (Reference Mintz1985:154) are extensification, the use of a commodity on a regular basis and in new ways by increasingly larger numbers of people, and intensification, a kind of emulation of past elite commodity usage by people of lower class and status, which had the effect of freeing the commodity of some social and political content it formerly carried. This emulation is not a perfect re-enactment of elite usage, but an appropriation of those aspects of elite usage that pertain to the new social environment. Intensification depends upon established patterns of elite use, so Kaufman and Justeson's (Reference Kaufman and Justeson2007:215) suggestion that cacao was not initially a commodity is likely. Intensification (sensu Mintz) of cacao use probably occurred when it became a commodity. Dramatic extensification of cacao probably corresponded with its use as money. The Izalcos Pipil appear to have been key actors in making extensification possible. The location of Pipil settlements at key nodes in southern Mesoamerican exchange routes also provided the necessary infrastructure to encourage extensification (Sampeck Reference Sampeck2007).

The Pipil were part of the Late Postclassic synergy in trade, production, and consumption of cacao, and their unusual emphasis on cacao makes them a likely source for the origin of ‘chocolate.’ The Pipil term may have traveled along with commerce in their cacao. Dakin and Wichmann (Reference Dakin and Wichmann2000) suggest that Nahuas controlled the Mesoamerican trade in cacao, but Nahua involvement probably happened later, interdigitating a southern commodity complex with northern empires. The spotty references to ‘chocolate’ in the historical record postdate the apex of Izalcos cacao trade in the Colonial period and may be reflecting the region's involvement in the colonial exchange system. The differing origins and ages of the words ‘cacao’ and ‘chocolate’ underscore that the meaning and uses of commodities change over time, and this process continued with the entrance of cacao into even wider spheres of trade in the colonial period.

Cacao and Chocolate: A View from the Mixteca and BeyondFootnote ⁷

Dakin and Wichmann (Reference Dakin and Wichmann2000, henceforth DW) and Kaufman and Justeson (Reference Kaufman and Justeson2007, henceforth KJ) present differing analyses regarding the origin and distribution of the word ‘cacao’ in Mesoamerica. In their discussions, DW and KJ make little use of data from Otomanguean languages, arguably the oldest and most diversified language family of Mesoamerica yet identified. This is understandable since Otomanguean languages generally do not have words of a shape that resembles [kakaw] (cf. KJ 2007:217). Nevertheless, important, early sources exist for some Otomanguean languages that can broaden and inform the debate about ‘cacao’ as well as the ancillary discussion of these authors about the origin of the word ‘chocolate’. In this brief note we will examine sixteenth-century documentation relating to cacao and chocolate in Mixtec, a highly diversified language group of the Mixtecan branch of the Otomanguean family.Footnote ⁸ The final section is a brief mention of some of the earliest references in Spanish of the word ‘chocolate’, one of which is a 1593 Spanish-Mixtec vocabulary.

Since the Postclassic period, Mixtec has been spoken across an extensive area including most of the western part of what is today the state of Oaxaca and adjacent regions in the states of Guerrero and Puebla. This area is known historically as the Mixteca. Today cacao is grown for local consumption in some Mixtec communities along the coast of Oaxaca and adjacent parts of Guerrero. However, cacao is imported in most of the highlands and interior lowlands of the Mixteca. In the sixteenth century, cacao was an important trade good for tribute payment and ritual use during matrimony. Cacao is depicted on both polychrome ceramics from the region (Figure 1) and in the spectacular Mixtec pictographic screenfolds known as codices (Figure 2).

Figure 1.

Drawing of a Postclassic period (a.d. 900–1521) polychrome ceramic sherd with representations of cacao, Pueblo Viejo, Teposcolula. (Courtesy of the Archivo técnico del Proyecto Pueblo Viejo de Teposcolula Yucundaa).

Figure 2.

Drawing of the representation of a ceramic vessel containing a cacao beverage from the Codex Vindobonensis, folio 18 (drawing by Enrique Montes).

Figure 3.

Map of Mixtec communities mentioned in the text (drawing by Enrique Montes).

An etymon of the shape *sɨʔwa may be posited for proto-Mixtec based on cognates for ‘cacao’ in different forms of Mixtec today (see Table 1). In some forms of Mixtec the cognates no longer refer to the cacao seed, but rather to ‘chocolate’, both as a drink and occasionally even as chocolate bars. These Mixtec forms are not cognates with the words for ‘cacao’ in Cuicatec (du ⁴ndu ³ča ³; Anderson and Roque Reference Anderson and Roque1983) or ‘chocolate’ in Triqui (ni ³kee ¹; Erickson de Hollenbach Reference Erickson de Hollenbach2005:43), the two other language groups pertaining to the Mixtecan branch of Otomanguean.

Table 1.

Reflexes of the proto-Mixtec *sɨʔwa.

Note: The column of Mixtec forms includes a practical orthography in italics which is followed by a proposed phonological form, often based only on an interpretation of the practical orthography, which usually does not include the representation of tone. When data on tone is available it is noted in the phonological form: a grave accent refers to a low tone, a macron refers to a mid tone (or more correctly, the absence of tone), and the superscript H refers to an unassociated high tone.

In several forms of Mixtec, especially those spoken along the Sierra Sur of the Mixteca, reflexes of the abovementioned etymon refer to a broader category of ‘seeds’ or ‘nuts’, such as squash seeds, chilacayote (Cucurbita ficifolia) seeds, peanuts. In these forms of Mixtec, ‘cacao/chocolate’ is generally either the unmodified form or specified by being the ‘good’ kind (see Table 2).

Table 2.

Reflexes of the proto-Mixtec *sɨʔwa which refer to different kinds of seeds or nuts besides cacao.

Since most of the Mixteca is not a cacao-producing region and since the most probable candidate for a proto-Mixtec homeland is the dry highlands of western Oaxaca, it may be that the etymon *sɨʔwa did indeed originally refer to another kind of seed or nut. The seeds of Jatropha spp., edible when toasted and called “nuez” and “piñón” in local Spanish, are one possible candidate. Three or more species of Jatropha occur at lower altitudes in the upper Balsas, Papaloapan and Verde drainages in the Mixteca; the seeds of J. neodioica appear to have been one of the staple foods of hunting and gathering bands during the archaic period in the Valley of Oaxaca, just east of the Mixteca (Flannery Reference Flannery1986). Pine nuts of Pinus cembroides appear to be a less likely prototype for *sɨʔwa, given that the southernmost contemporary occurrence of this species has been recorded in the mountains above the Tehuacan Valley, well beyond the historical area of Mixtec speech.

Cacao in the 1593 Vocabulario en lengua misteca

The most important source available for sixteenth-century Mixtec vocabulary regarding cacao is the Vocabulario en lengua misteca, printed in 1593 under the auspices of the Dominican Order (Alvarado Reference Alvarado1962 [1593]; Jansen and Pérez Jiménez Reference Jansen and Jiménez2003). The Vocabulario is written in the now extinct Teposcolula variant of Mixtec. Teposcolula Mixtec was also spoken in adjacent regions, among which the Tamazulapan-Teotongo Valley (Reyes Reference Reyes1593: prologo; Swanton Reference Swanton, Jansen and van Broekhoven2008). The ecological zone in which this Mixtec variant was spoken is characterized by pine-oak forest, chaparral and xerophytic scrub. Cacao was not produced in this region. Nevertheless, the Vocabulario includes 22 Mixtec-language entries related to cacao.

Nineteen entries in the Vocabulario include the Teposcolula Mixtec reflex of the aforementioned proto-Mixtec etymon for ‘cacao’, alternatively written dzehua, dzéhua or dzèhua. In the Dominican orthography, <dz> perhaps represented a voiced dental fricative /ð/or a voice affricate /dz/, <e> frequently represents a high central vowel /ɨ/, and a vowel bearing an acute or grave accent before a consonant a /Vʔ/. The Teposcolula Mixtec form for cacao might then be rendered as *ðɨʔwa or *dzɨʔwa.

The 1593 vocabulary provides six verb constructions that describe activities involving cacao:

• yocachiñeendi dzehua : Alimpiar cacao quitandole la corteza

• yodzacaandi [n]duta dzéhua : Echar cacao de vna xicara en otra para hazer espuma

• yodzacayendi : Cacao tostar

• yodzacaindi [n]duta dzehua : Echar cacao de vna xicara en otra para hazer espuma

• yodzacuijndi : Alimpiar cacao quitandole la corteza; Cacao mondar

• yodzayatnandi : Cacao tostar

These constructions are glossed in terms of three actions: (1) removing the cacao seed skin, (2) toasting the cacao seed and, (3) pouring a cacao beverage (literally, ‘cacao water’) from one jícara (gourd bowl) to another in order to make it frothy. These actions describe the preparation of cacao for consumption; there are no actions that involve the cultivation or harvesting of cacao. Furthermore, words such as the name of the cacao tree do not appear in the Vocabulario. These facts reinforce the historical record which documents cacao as a trade and tribute good for the highland Mixtecs and not an agricultural product.Footnote ⁹

The rest of the Vocabulario entries that refer to cacao include the aforementioned word dzéhua. These entries provide expressions for types of cacao or cacao-like seeds (including a neologism based on this word), the cacao pod, drinks prepared from cacao, and the froth of cacao beverages (see Annex).

Among these entries appears: “Cacao hancho y blanco. dzehua ñaña” (f. 40). The Mixtec expression (dzehua ñaña) literally means ‘jaguar cacao’ and undoubtedly refers to pataxte (Theobroma bicolor). The pataxte seed is indeed wide and white as indicated in the Vocabulario description. Furthermore, as KJ (2007:201–202) note, several Mixe-Zoquean and Mayan languages refer to pataxte with a term that contains a word for ‘jaguar’. Teposcolula Mixtec can therefore be added to their list of Mesoamerican languages with such a term.Footnote ¹⁰

The Dominican vocabulary also provides names for five types of cacao-based drinks:

• [n]duta dzehua : Beuida de cacao (f. 34v); Cacao beuida (f. 40)

• [n]duta dzehua nina : Beuida de cacao (f. 34v)

• [n]duta dzehua ñuhu : Beuida de cacao (f. 34v); Chocolate beuida de cacao (f. 64v)

• dzehua nina ndico : Beuida de cacao (f. 34v)

• dzehua ticoo : Beuida de cacao (f. 34v)

These five entries all employ either nduta dzehua, ‘cacao water’, or just dzehua ‘cacao’ to refer to the cacao-based beverages. This usage supports KJ's argument (2007:219–222) that drinks make from cacao “may have been called by the name of its principal ingredient”.

Sixteenth Century References to ‘Chocolate’

In the 1593 Dominican vocabulary one of the two Spanish translations of nduta dzehua ñuhu (literally ‘earth cacao water’Footnote ¹¹) is ‘Chocolate beuida de cacao’. This is the earliest printed reference to the word chocolate in Spanish we know of from the province of Oaxaca.Footnote ¹² KJ point out that the word chocolate also appears in Joseph de Acosta's 1590 Historia natural y moral de las Indias as well as in the so-called Motul dictionary of Yucatec Maya around this same time. Nevertheless, they conclude:

However, the term chocolate is attested in several sources prior to and contemporaneously with Acosta and seems to have been of common usage in the last two decades of the sixteenth century. The earliest reference we know of occurs in a letter written by the Jesuit Pedro de Morales in 1579 regarding the festivities organized to celebrate the shipping of important relics by Pope Gregory XIII to New Spain one year earlier. An allegoric arch was erected in which Guatemala was personified by a woman holding a branch of the cacao tree and a drinking cup. The related poem went: “De lo ques mi propio dote/ Le traygo en un tecomate;/ Que es cacao y achiote/ Para hacer chocolate” (‘From my very dowry, I bring you in a gourd, that is cacao and achiote to make chocolate’) (Mariscal Haz 2000:57). Morales stated explicitly that this drink “was invented in Guatemala”. In fact, in the Relación Geográfica de Zapotitlán, written in the same year in Guatemala (Acuña Reference Acuña1982:45), chocolate is again mentioned as a drink made from cacao and achiote.

Shortly thereafter, during their 1586 trip through Xoconusco, the Franciscan Alonso Ponce and his secretary observed the use of cacao and wrote: “they make of it a great variety of very good drinks, some of these they drink cold, other hot, and among these [hot ones] there is a very common one that they call chocolate, made from the said ground cacao and from honey and hot water, wherein they add other mixtures and materials of hot things”Footnote ¹³ (De Ciudad Real Reference De Ciudad Real1993:183). The Dominican Agustín Dávila Padilla (Reference Padilla, Agustín1955:626–27), who finished the manuscript of his book in 1592, wrote about the use of chocolate in Oaxaca (which again is said to be a warm drink), the use of which started in Chiapas according to this author. He cites Juan de Cárdenas' book Problemas y Secretos maravillosos de las Indias from 1591, in which doctor Cárdenas wrote extensively about chocolate. Among many observations about the ways of preparing cacao drinks the doctor observed: “although it is true that every lady boasts of making her new invention and way of chocolate, with all this, generally the most common in all of the Indies is the one formed in tablets, which took its origin with the Guatemalan ladies, and this same one is the one that is dissolved with hot water and a touch of sweet, which gives it great charm”Footnote ¹⁴ (f. 105). These descriptions suggest that the cold drinks made from freshly ground toasted cacao were used since pre-Hispanic times, but that the current Oaxacan use of producing hot chocolate from prefabricated tablets already including the spices, was a sixteenth-century fashion introduced from Guatemala. This early colonial innovation may have spread along the trading routes, explaining the spread of the word chocolate during the 1590s over most of Mesoamerica.

Taking into consideration the Mixtec Vocabulario and other texts, the Spanish word ‘chocolate’ is attested in several sixteenth-century sources from different regions of Mesoamerica: Guatemala, central Mexico, the Yucatan Peninsula, and the Mixteca Alta. This suggests that Acosta's usage of the word in Spanish cannot be dismissed as “simply different.” On the contrary, the diversity of these sources and the nature of the sources themselves (which include an expensive printed dictionary) suggest an established use of this word in the Spanish of New Spain. Furthermore, it is worth noting that none of these early sources attest to a word like chicolate (from chicolatl), which Dakin and Wichman (Reference Dakin and Wichmann2000) propose as a possible original form.

The purpose of this comment is to provide an assessment of the state of the evidence concerning the etymologies of the words for cacao and chocolate (concentrating almost exclusively on the former). I will discuss only directly relevant matters and avoid quibbles over details that do not affect the larger picture. Towards the end of my comment I assess the evidence for the idea that the Teotihuacano elite was comprised by speakers of Mixe-Zoquean.Footnote ¹⁷

Cacao and Chocolate

Dakin and Wichmann (Reference Dakin and Wichmann2000) argued that the widespread word for cacao in Mesoamerica (kakawa and similar forms) could have a Nahuatl rather than a Mixe-ZoqueanFootnote ¹⁸ origin based on the following arguments: (A) the word does not behave as a typical Mixe-Zoquean word; and, (B) a plausible derivation within Nahuatl can be posited, something which would not be possible if the word had entered Nahuatl from some other language. These two arguments have different components, all of which have been criticized by Kaufman and Justeson (Reference Kaufman and Justeson2007). First I will treat the three components of argument A.

A1 states that the word is of untypical shape for Mixe-Zoquean, A2 states that if it were inherited from proto-Mixe-Zoquean into the modern languages there would be irregularities with respect to the placement of stress, and A3 states that there would also be irregularities with respect to the phonemes contained or not contained in the inherited forms.

Regarding A1 we argued that lexical stems of three syllables are very rare in proto-Mixe-Zoquean. Due to data gathered under the auspices of Kaufman and Justeson's language documentation project they have been able to add to the putative proto-Mixe-Zoquean *kakawa one other proto-Mixe-Zoquean trisyllable: *wakata ‘apompo’ (there are a few more trisyllables in proto-Mixean and proto-Zoquean, but not in proto-Mixe-Zoquean). A form meaning ‘cockroach’, which we had cited as another proto-Mixe-Zoquean trisyllable, is analyzed by Kaufman and Justeson as composed of a *ma, said to be “some kind of optional modifier” (Kaufman and Justeson Reference Kaufman and Justeson2007:210) and *kokoʔ. If we concede that the word for ‘cockroach’ could be a complex form even if the meanings of each of its elements are unknown, the argument A1 still remains unchanged: a putative proto-Mixe-Zoquean *kakawa would be one out of just two trisyllabic proto-Mixe-Zoquean (pMZ) lexical stems. As mentioned there are some more trisyllables at later stages in the development of the language family, but even taking those forms into account the word for ‘cacao’ would be the only stem where the two first syllables are identical; as such, it has a distinctly non-Mixe-Zoquean appearance.

Argument A2 relates to stress. The principal new observation on the part of Kaufman and Justeson (Reference Kaufman and Justeson2007) is that the loss in Gulf Zoquean of the second vowel, which was seen as an irregularity by Dakin and Wichmann (Reference Dakin and Wichmann2000) is, in fact, regular. This observation is largely based on data which have become available in the meantime. Other elements of their analysis are less clear. According to the authors' suggestions (2007:208) a trisyllabic word such as pMZ *kakawa would have had stress both on the first and the second syllable, something which is typologically unusual; a similar phenomenon is said to be found in Santa María Chimalapa Zoque, but this is hard to verify since the data are unpublished (the reference does not even appear in the bibliography). It is also not clear why North Highland Mixe and the rest of the Oaxaca Mixean languages behave differently from one another. The former shows the effects of stress on the initial syllable, while the latter have it on the last syllable. Forms similar to kaká:w in non-North Highland Mixe are said to derive their long vowel from an h inserted before the w. Such an insertion of an h took place in proto-Mixean, but there is no evidence for it in proto-Oaxaca Mixean other than before consonant clusters (Wichmann Reference Wichmann1995:164). Finally, it is not clear whether proto-Zoquean had *kákawa or *kakáwa. The first is preferred by the authors, but both are possible, the choice depending on whether one sees the Gulf Zoquean form with its initial stress or the other Zoquean languages with their penultimate stress as having innovated. Be that as it may, I concede that the Gulf Zoquean reflexes are regular, that the Oaxaca Mixean reflexes may be regular although the proof is still lacking, and that the reflex kakwa in the northern dialect of Chiapas Zoque may be explained as a borrowing from a Gulf Zoquean language. These new interpretations do nothing to destroy the loanword scenario, however. It is perfectly possible for a loanword to be treated as a normal inherited form provided that it is borrowed early enough; and, as Dakin and Wichmann (Reference Dakin and Wichmann2000:57) argued, the word seems to have entered the Mixe-Zoquean family after its break-up into the two major branches, Mixean and Zoquean, but before further differentiation.

Component A3 is based on the observation that the forms that can be reconstructed for proto-Mixean (pM) and proto-Zoquean (pZ) are different in a way which is not regular. Based on the evidence from the current languages we may reconstruct pM *kakaw and pZ *kakawa. These are discrepant in that there is no regular loss of a final vowel in pM. Kaufman and Justeson do not identify a regularity that would account for this discrepancy either. Instead they “explain” the irregularity by showing that there are other irregularities. Wichmann (Reference Wichmann1995) contains 699 reconstructed pMZ forms; of these around ten percent exhibit uncertainties in one way or the other because Mixean and Zoquean forms are too similar to not be related but nevertheless not entirely accounted for by regular diachronic developments. Kaufman and Justeson (Reference Kaufman and Justeson2007:212) list such instances in their Table 6. What mainly emerges from this list is that none of the Mixean and Zoquean irregular matches are irregular in the way the pM *kakaw vs. pZ *kakawa are. Specifically, in none of the cases does a Mixean form lack a final vowel that a Zoquean form has. Instead, the opposite is found in three cases, e.g., the word for ‘meat’, where Mixean has *sisi and Zoquean *sis.

Let me recapitulate and summarize the state of the A argument. We argued that the word for ‘cacao’ has a highly peculiar structure for a Mixe-Zoquean stem and that it had an irregular phonological trajectory within the history of the family. I will maintain that its structure indeed is peculiar and that its reflexes are irregular, but I concede that the reflexes are irregular only with respect to the presence/absence of the final vowel, not necessarily with respect to stress. What the A argument in its present slightly weakened form amounts to is still a strong indication that this could be a borrowing, and this should motivate a search elsewhere for a the origin of this word.

This leads us to the B argument, which is the Uto-Aztecan etymology. The expertise concerning this argument rests with Karen Dakin, and I refer the reader to her comments, in particular on the issue of the lenition of p in her proposed reconstruction *ka^Npa^N ‘hard pod, shell’ (which may or may not be further decomposed) to w. This is essential for the derivation of the reduplicated form *kakawa from *ka^Npa^N. The morphological analysis is important for determining the direction of a borrowing event. If a given borrowed word is analyzable in language X and not in language Y, then it is safe to conclude that the word passed from language X to language Y and not the other way around.Footnote ¹⁹ If a word is not morphologically analyzable in any of the two languages, then either direction is equally probable, all else being equal. The latter is the situation which would arise if we consider the possibility that Dakin's derivation, which after all is only a hypothesis, is not correct for Nahuatl. We would then be left with a form which is unanalyzable in both Nahuatl and Mixe-Zoquean, and would have to establish the direction of borrowing on other grounds. Kaufman and Justeson (Reference Kaufman and Justeson2007:214) provide some important evidence in this regard which is in conflict with their own view when they cite the Huasteca Nahuatl forms i-kakawa-yo ‘its thick bark’ and kakawa.ti.k ‘hollow’. These have meanings far removed from ‘cacao’, and yet contain the very stem that also means ‘cacao’. These, as well as other forms with meaning broader than or simply different from ‘cacao’ cited by Dakin and Wichmann (Reference Dakin and Wichmann2000:60 [Table 3]), indicate that ‘cacao’ is a special instance of a broader range of meanings of Nahuatl kakawa-tl, and that kakawa-tl in the sense of ‘cacao’, whether or not it is morphologically derived, certainly is semantically derivative. It is easy to imagine, along the lines of Kaufman and Justeson, that Nahuatl speakers could have borrowed a word meaning ‘cacao’ from a language whose speakers resided in cacao-producing areas and would have had more familiarity with the item, but it is very hard to imagine that they would begin to use the word to refer to ‘bark’, ‘husk’ or ‘shell’, since they would have their own words for such objects.

An issue raised by Kaufman and Justeson (Reference Kaufman and Justeson2007:204–5) is that the –tl suffix marking non-possessed forms in Nahuatl is missing from the borrowed forms. One might expect the word to turn up in borrowing languages as something like kakawat. They cite Nahuatl loanwords in the Gulf Zoquean variant of Soteapan (a.k.a. Sierra Popoluca) in favor of the view that something corresponding to –tl or its allomorphs should show up on borrowings of Nahuatl nouns. But it is important to realize that we cannot expect the early Nahuatl language that we are positing as the source of words for ‘cacao’ in Mesoamerica to behave exactly like modern dialects spoken around one and a half millennia later. Rather, one should look at possible early borrowings in order to establish the typical behavior. In my referee report on the manuscript for Kaufman and Justeson (Reference Kaufman and Justeson2007) I therefore offered the case of the word for ‘town’, which takes tenam or similar shapes in Mamean, Q'anjob'alan, Chujean and Huastec (all Mayan languages or subgroups). This is not borrowed with the absolute suffix, and contrasts with tinamit and similar forms found in K'ichean, which also mean ‘town’. While Kaufman and Justeson (Reference Kaufman and Justeson2003:76–77) posit a Nahuatl origin for both forms, the same authors now think that “the form #tena:m is only possibly of Nawa origin, not definitely so” (Kaufman and Justeson Reference Kaufman and Justeson2007:2004). I do not see any reason to doubt that they are both from Nahuatl, and that they in fact are etymologically one and the same word.

Another example may be a word for ‘owl’ in Mayan languages which is found throughout the family, with the exception of Huastecan. Kaufman and Justeson (Reference Kaufman and Justeson2003) reconstruct Common Mayan *tuhkur(uu7) and note that the word has diffused internally. This form is quite similar to Classical Nahuatl tekolo:-tl. If an early Nahuatl form of a similar shape is the origin of the rather irregularly behaving Mayan forms it would again have been a form without the absolute suffix.

The Mayan epigraphic record offers two more intriguing examples. On Tikal Stela 31 (left side, at L2) we find the spelling ko-sa-ka. Although there is no contextual control to verify this, it may well be that the word represented there is a cognate of Classical Nahuatl ko:ska-tl ‘jewel, ornament, necklace’ (this was suggested by Alfonso Lacadena to me in personal communication in 1999 and, unknown to both of us, had already then been considered by David StuartFootnote ²⁰). The spelling is conspicuously followed by a glyph representing a mat throne very similar to representations of such thrones in later Aztec iconography, and this glyph does not otherwise appear in the Maya hieroglyphic corpus. Moreover, as is well known, the entire stela is full of Teotihuacan iconography and its text refers to close political ties between the Tikal elite and Teotihuacan (Stuart Reference Stuart, Carrasco, Jones and Sessions1999). If the reading and identification of the Nahuatl word is correct we have another candidate for an early (a.d. 445 to be precise) Nahuatl borrowing, and one which is lacking the –tl suffix.

Finally, there is an interesting spelling on Seibal Stela 13, which dates to the Terminal Classic. This spelling consists of the string of syllabic signs e-je-ke and seems to spell the name of the central Mexican wind god, which was pronounced e ʔ e:ka-tl in Classical Nahuatl. The identification was made by Lacadena (Reference Lacadena, Vail and Hernández2010), who notes, as others before him, that the sculptural program to which the stela belongs is full of central Mexican calendrical and iconographical parallels, and who clarifies the details of his identification as follows:

Thus, in addition to kakawa it is possible to muster four other possible early Nahuatl loanwords which lack the absolute suffix in the borrowing languages: ‘town’, ‘owl’, ‘precious ornament’, and the name of the wind god. It therefore seems that the missing –tl in the word for cacao as found in borrowing languages is simply the expected pattern, and that it is not necessary to take recourse to the hypothesis that the word was most commonly heard in its possessed form, as originally suggested by Dakin and Wichmann (Reference Dakin and Wichmann2000:67–68).

Kaufman and Justeson offer other, more circumstantial points of critique of the hypothesis that kakawa is from Nahuatl. One argument (Kaufman and Justeson Reference Kaufman and Justeson2007:205) is that Mixe-Zoqueans are geographically associated with cacao-growing districts while Nahuas are not. Dakin and Wichmann (Reference Dakin and Wichmann2000:67) anticipated this sort of reaction by noting that the word would have diffused along trade routes, and that it would be the word of those who controlled the trade routes (rather than that of the producers) which became the adopted word throughout Mesoamerica. There is no reason to belabor this simple point further, but just to give a perhaps parallel example it may be noted that several words in Spanish which refer to items that were known in the Iberian Peninsula before the Moorish invasion were later borrowed from Arabic, including zanahoria ‘carrot’ (Dworkin Reference Dworkin2006).

At this point I will add a few words about the relationship between čikola:-tl and kakawa. Dakin and I wrote in 1990 that “[i]t is reasonable to suppose that these two words share their center of dispersal”. The wording signals a pure speculation. We do not have any direct evidence for identifying a certain geographical region as being the center of origin of either of the two words, and we certainly did not wish to imply that the words must necessarily have spread at the same time. The spread of čikola:-tl could well have been relatively late, but going so far as to entertain “reasonable doubt as to whether the Nawa forms chikola: tl and/or chokola:tl even existed in pre-Columbian times” (Justeson and Kaufman Reference Kaufman and Justeson2007:219) seems like going too far. I can, however, fully agree with their more limited statement that “[t]he words chikola:tl and chokola:tl may have arisen in a peripheral type of Nawa, at an undeterminable date, and only later spread to Central Nawa” (Justeson and Kaufman Reference Kaufman and Justeson2007:220). My conjecture is that it did not spread from central Mexico at the time of the Triple Alliance but from a Nahuatl-speaking group which was once dominant and later became marginalized. A closer identification, such as our suggestion regarding Pipiles, is a matter of speculation, but speculations may stimulate research (as ours have clearly done) and are therefore occasionally relevant.

To summarize my assessment of the state of the evidence for a Nahuatl origin of the word for ‘cacao’ I consider the Mixe-Zoquean evidence to point a likely origin of the word outside of this family, and a Nahuatl origin is a strong hypothesis because of the wide range of meanings associated with the lexical stem kakawa and the possibility of a morphological analysis. Finally, there is no circumstantial evidence that would speak firmly against a Nahuatl origin of the word.

Of main interest, perhaps, to the readers of this journal is the issue of a connection between Nahuatl and Teotihuacan. Kaufman and Justeson (Reference Kaufman and Justeson2007) argue that a single loanword is not enough to identify a language of Teotihuacan. Against this, I would maintain that it is a mistake to expect a directly proportional relationship between the amount of loanwords donated by a given culture to other cultures and the political importance of the donating culture. Political dominance need not imply language contact at all levels of society. In contrast, a more or less equal political status of two groups can lead to a great number of loanwords simply because of the mingling of populations. Kaufman and Justeson cite loanwords from Nahuatl in Soteapan Zoque (Sierra Popoluca). In spite of their relatively great number they do not reveal anything about political dominance, but only indicate that there has been close contact between Nahuas and Popolucas in Veracruz. Regarding the language of Teotihuacan, some important pieces of circumstantial evidence are the geographical association of Teotihuacan and Nahuatl speech as well as the graphic continuity between the writing system used at Teotihuacan and the one used by Nahuas at the time of the conquest. These facts cannot settle the discussion, but they can serve to render a connection of Teotihuacan to Nahuatl vastly more probable than the proposed connection to Mixe-Zoquean, which I comment on in the next section.

Teotihuacan and Mixe-Zoquean?

In continuation of the preceding remarks I will expend a few words on the new proposal regarding a form of Mixe-Zoquean as being the language of the elite at Teotihuacan. The proposal rests entirely on putative Mixe-Zoquean loanwords in languages surrounding the central Mexican region. The actual data are neither given in Kaufman and Justeson (Reference Kaufman and Justeson2007), nor in Kaufman and Justeson (Reference Kaufman, Justeson, Arnold and Pool2008), which reiterates the same proposal, and are not published elsewhere, so I will be referring to Kaufman (Reference Kaufman2006).

The most massive borrower, in Kaufman's scenario, is Totonacan, for which 52 forms with semantic and phonological similarities to Mixe-Zoquean forms are listed. Independently of Kaufman's work and in collaboration with David Beck, Greg Kondrak, and myself, Cecil Brown has also prepared a list of words that are similar in form and meaning in Totonacan and Mixe-Zoquean. There is a large overlap among the lists, with some forms from Kaufman's list not appearing in the list of Brown et al., and 23 forms in the list of Brown et al. not being on Kaufman's list. It is highly striking that in the total set of more than 70 Mixe-Zoquean-Totonacan agreements or apparent agreements all but one or two pertain to basic vocabulary. They include some kinship terms, words for common animals or generic animal terms like ‘louse’, ‘bird’, ‘rabbit’, and ‘snail’; generic plant terms or words for common plants like ‘tree’, ‘fruit’, ‘sweet potato’, and ‘manioc’; the basic color terms ‘white’ and ‘red’; the numerals for 1-4 and (perhaps) 20; bodypart terms like ‘nose’, ‘tooth’, ‘feather’, ‘bone’; and finally verbs for basic actions such as ‘to come’, ‘to go up’, ‘to hear’, ‘to eat’, ‘to sleep’, ‘to sing’, ‘to spit’, ‘to suck’, ‘to defecate’, ‘to hit’, ‘to blow’, ‘to wash’, ‘to cut’, etc. Holman et al. (Reference Holman, Wichmann, Brown, Velupillai, Müller and Bakker2008) single out the 40 universally most stable items on the one hundred-item list developed by Morris Swadesh for lexicostatistical and glottochronological purposes (e.g., Swadesh Reference Swadesh1955). Of these, there are 11 items for which agreements in form and meanings between Mixe-Zoquean and Totonacan are found. The overall agreement in basic vocabulary between the two families is comparable to what is found within well-established families such as Indo-European, Hmong-Mien, Arawakan or Chibchan and is therefore suggestive of common inheritance, which would imply a genetic relationship between Mixe-Zoquean and Totonacan. The lexical similarities are most definitely not suggestive of language interference emanating from a minority Teotihuacano elite. The kinds of words that the two language families share are simply not the kinds of words that one would expect to be shared under cultural dominance; rather, they are the sorts of words that would be expected to diffuse among peacefully coexisting and intermarrying groups if not, as just mentioned, the sorts of words typically inherited from a shared ancestor.

In the case of agreements between Huastec and Mixe-Zoquean words we find a quite different situation. Here the meanings are centered in the domains of plants and animals. It is expected from loanwords that they should be concentrated in certain semantic domains, and these domains will tell us something about the nature of the events that gave rise to the loanwords. The Huastec borrowings are indicative of a shift of these Mayan immigrants to a new natural environment in the vicinity of Mixe-Zoquean speakers and contribute nothing to a hypothesis regarding the language(s) of Teotihuacan.

The borrowings listed for Tarascan are also quite numerous. Some may really just be accidental similarities and the majority do not show close semantic matches with Mixe-Zoquean forms. In any event, givin that Tarascan is an isolate for which previous stages cannot be reconstructed we do not have any way to chronologize these putative loanwords, so they cannot sustain a hypothesis concerning a specific association with the Classic period or with Teotihuacan.

For the rest of the language groups that have borrowed or appear to have borrowed from Mixe-Zoquean the data similarly show a mixture of what may be accidental similarities and loanwords from horizons that need not be associated specifically with Teotihuan.

In conlusion, the impressive data assembled in Kaufman (Reference Kaufman2006) will be valuable for the further exploration of the linguistic prehistory of Mesoamerica, but they certainly do not, in my view, support the hypothesis of a Mixe-Zoquean presence at Teotihuacan. I expect that, eventually, new epigraphic data will provide us with the kind of evidence needed to finally settle the question of the linguistic affiliation of this important city.