Summation

The widespread and early diffusion of the word *kakaw(a) into a large number of Mesoamerican languages and language families is consistent with what is known about the diffusion of words for other cultigens from Mije-Sokean languages. In contrast, no words for cultigens are known to have diffused so widely at such an early date from any other language family. In addition, one of the prime areas of cacao cultivation, in the lowlands of Tabasco, was part of the (Mije-Sokean-speaking) Olmec heartland, where Gulf Sokean languages are still spoken today. These facts provide strong circumstantial support for the linguistically motivated hypothesis that this word is native to the Mije-Sokean family and diffused into other Mesoamerican languages, ultimately, from speakers of one or more Mije-Sokean languages. In fact, setting aside the recently contested case of *kakawa, no currently reconstructed proto-Mijean, proto-Sokean, or proto-Mije-Sokean term has a demonstrated foreign origin, so any alternative to a Mije-Sokean source requires substantial independent evidence.

General Features of Borrowings into and out of Nawa

Before evaluating Dakin and Wichmann's specific arguments, they can be framed in the context of what is known about lexical diffusion in Mesoamerica more generally. Our view that Nawa kakawa-tl is a borrowing from Mije-Sokean, and not into it, is concordant with a number of facts about linguistic diffusion in Mesoamerica that are set forth in the previous section, while the contrary view is not consistent with what is otherwise known about diffusion from Nawa. Setting aside the contested case of the word for cacao, four empirical observations can be made about cleanly established borrowings between Nawa and other Mesoamerican languages:

(1) Nawa in general is heavily influenced by Mije-Sokean and Totonakan, and to a lesser degree by Wastekan in its pre-Nawa stage (see “Demonstrating Borrowing”). In contrast, neither Mije-Sokean in general nor Totonakan in general is much influenced by Nawa. Even Wasteko (a single language rather than a family of languages), which has had Mesoamericanized Nawa as a neighbor since at least a.d. 900, has relatively few lexical borrowings from Nawa. This makes Nawa look like a relative newcomer to Mesoamerica, not in a position to provide very early loans into languages throughout Mesoamerica.

(2) More specifically, proto-Nawan—the common ancestor of all forms of Nawa—shows substantial borrowing from Mije-Sokean languages in both vocabulary and grammar.

(3) Setting aside the contested case of the word for cacao, proto-Mije-Sokean does not show a single plausible instance of a lexical borrowing from Nawa or Yuta-Nawan, and neither does proto-Mijean or proto-Sokean or any other genetic subgroup of Mijean or Sokean. (Individual Mije-Sokean languages have borrowed some Nawa lexical material.)

(4) Setting aside the contested case of *kakawa, there is no demonstrated instance of a Nawa loan into any Mesoamerican language that clearly predates the Late Classic period. In particular, no loans from Nawa have undergone sound changes characteristic of any genetic group of languages. Rather, Nawa loans in Mesoamerican languages reflect Nawa phonology as we know it from the sixteenth century and therefore cannot go back earlier than about a.d. 1000.

There are proposed counterexamples to these claims, mostly by Dakin, but none is cogent. (See “Demonstrating Borrowing”, earlier, for refutations of a sample of such proposals.)

Given points (2)–(4), compelling evidence is required to make a case for a word of Nawa origin having diffused into Mesoamerican languages at a substantially earlier time period, especially one that was borrowed as widely as *kakawa.

There are two other general problems with the claim of widespread early borrowing of *kakawa from Nawa.

(5) Nawa nouns are almost always borrowed in their unpossessed form and reflect the absolute suffix -tl~-tli~-li when this suffix is present in the Nawa model (Kaufman Reference Kaufman2000–2007). In Mayan and Mije-Sokean languages in particular, then, one would expect to find something like kakawat*, rather than the attested kakaw, if this word were indeed a borrowing from Nawa. By way of illustration, we first discuss a set of borrowings into a single language, and then we discuss borrowings that are widespread in a particular language family.

Table 3 presents all of the Nawa loans in Kaufman's data on Soteapan Gulf Sokean (Kaufman and Himes Reference Kaufman and Himes1993–2005), which can serve as a typical instance of how the Nawa absolute suffixes—in particular, -tl(i)—appear in loans into other languages. In all 15 cases where the Nawa source has a word-final -tl or -tli, Soteapan Gulf Sokean has a straightforward reflex of it. In five cases, Soteapan Gulf Sokean has a reflex of -tl, -tli, or -li where present-day Gulf Nawa lacks the suffix (taanajti, kukujti, lupujti, manteeka7t, xik7ipiili). These borrowings into Soteapan Gulf Sokean presumably reflect an earlier stage of Gulf Nawa, though one that postdates the arrival of Spanish speakers.

Table 3. Nawa loans into Soteapan Gulf Sokean and the fates of their absolutive suffixes

Note: SOT = Soteapan Gulf Sokean (Kaufman and Himes Reference Kaufman and Himes1993–2005); MEC = Mecayapan Gulf Nawa (Wohlgemuth 2000); COX = Coxcatlan Gulf Nawa (Kaufman 1969/1984–1993/2006); PAJ = Pajapan Gulf Nawa (Peralta Reference Peralta2002–2007).

^aThis Nawa form occurs both with and without -tl.

^bFrom Spanish via Nawa. The source of these Nawa loans is Gulf Nawa as spoken in Mecayapan (Wolgemuth, et al. Reference Walters, Carl, de Wolgemuth, Hernández Pérez, Ramírez and Hurst Upton2000) and Pajapan (Peralta 2004), Veracruz. The Nawa forms are cited, though, as they occur in the more conservative tl dialects.

There are 13 nouns of Nawa origin that have a fairly wide spread in Mayan languages. In each case, some languages are likely to have adopted the word directly from Nawa, while others probably received it from another Mayan language.

Three of these—#mis ~ #mistuun ‘cat’, #masa:t ‘deer’, and #xunakat ‘onion’—date from after the arrival of the Spanish.

Eight are found in the Guatemala highlands and bear the traits of Gulf Nawa or Pipil: #koht ‘eagle’, #to:ch(in) ‘armadillo’, #karat ‘frog’, #chikiwit ‘basket’, #no:chti7 ‘prickly pear’, #matzahti7 ‘pineapple’, #nakatamal ‘meat tamale’, #nawal ‘shape-shifter’. Whatever absolute suffix was present in Gulf Nawa is also present in the Mayan borrowing. Gulf Nawa /kohti/ ‘eagle’ has been shortened to make it look more Mayan.

Two Nawa loans into Mayan present interesting problems:

Nawa tena:mitl ‘fortified place’ (possessed theme -tena:n) appears in Mayan languages of Huehuetenango and in Wasteko as #tena:m ‘fenced-off area, town’. These forms seem to lack the Nawa absolute suffix. However, the possessed form of tena:mitl is Poss-tena:n, which ends in /n/, not /m/, so the possessed form was not the basis of the borrowing. It is plausible that, if the form #tena:m is borrowed directly from Nawa tena:mitl, it has been shortened by Mayans, and that is why the /m/ was kept. K'ichee7an languages borrow tena:mitl as #tina:mit ‘town’, and the absolute suffix is preserved. However, there is room for doubt as to whether the word is of ultimate Nawa origin, since it is found in only some varieties of Nawa and is not analyzable. While one may start by hypothesizing that the segment /te-/ corresponds to the root {te} ‘stone’, the sequence /-na:m(i)/ is not a known Nawa morpheme. Thus, the word may have originated outside Nawa. While there can be no doubt that #tina:mit comes from Nawa, the form #tena:m is only possibly of Nawa origin, not definitely so.

The last item constitutes an exception to the norm that nouns of Nawa origin are borrowed with a reflex of the Nawa absolute suffix, when one is in fact present in the Nawa original. The word #xa:n ‘adobe brick, wall’ occurs so widely in Mayan that a form *xa:n would be reconstructed for proto-Mayan if the Nawa word xa:mitl, possessed-xa:n ‘adobe’ were not attested. The Mayan word definitely seems to reflect the possessed form of the Nawa noun. It is irrelevant that Nawa xa:mitl is in turn borrowed—from Mije-Sokean. The motive for borrowing ‘adobe’ in its possessed form is not obvious.

Two other proposed borrowings, from Nawa into Mayan, are not demonstrably early loans from Nawa. One is not demonstrably a loan, and one is demonstrably not early.

One word, *ku:m ‘pot’, is reconstructible for proto-Yukatekan, which broke up circa a.d. 1000. This word is likely to share a common history with Nawa ko:mi-tl ‘pot, water jar’. This could conceivably be by borrowing from Nawa, but the word has no Sonoran etymology, and it is equally plausible that it entered both languages from another source. It may be entertained that the word entered Nawa from an ancestor of proto-Yukatekan, but it is not clear where the contact could have occurred that would make this a realistic option.

On the basis of Ch'ol chikib' and Ch'orti7 chiki7, Kaufman and Norman (Reference Kaufman, Norman, Justeson and Campbell1984:118) reconstructed *chiki7 to proto-Ch'olan because it was formally possible to do so. As a historical statement, however, this was misleading in that the form chiki7 reflects a contraction from Nawa chikiwi-tl that is otherwise known only in Gulf Nawa, which probably did not become distinct from other forms of Nawa before a.d. 800, whereas epigraphic evidence and glottochronology both put the breakup of Ch'olan before that date. The Ch'ol and Ch'orti7 forms therefore represent independent borrowings from some kind of Gulf Nawa.

In summary, out of ten pre-Columbian borrowings from Nawa in Mayan languages, only one was borrowed without the absolutive suffix. These data bear on generalizations about tendencies in the process of borrowing Nawa nouns (see “Evidence that *kakaw(a) was not Borrowed from Nawa …”, below).

(6) Cacao does not grow anywhere near the Basin of Mexico, nor farther north in areas from which Nawas entered Mesoamerica. It is completely plausible that Nawa would have borrowed its word for cacao from a language localized in the area in which it grows. This is in fact the norm when newcomers encounter unfamiliar plants and animals, whenever they do not devise neologisms using native resources. Nawa was strongly affected by one or more Mije-Sokean languages in its vocabulary, its morphology, and its syntax. Apart from Mije-Sokean, early forms of Nawa show substantial influence only from Wasteko and Totonakan, and the word kakaw of Wasteko and Totonakan, if borrowed into Nawa, would have been expected to yield kakaw-tli rather than kakawa-tl. On the face of it, Mije-Sokean is the only plausible source for Nawa kakawatl.

Given these characteristics of lexical diffusion from Nawa, any proposal for a widespread lexical borrowing from Nawa in the Preclassic or Early Classic has to be approached with skepticism and requires compelling linguistic evidence to be accepted. We show below (“Refuting the arguments …”) that the evidence presented by Dakin and Wichmann for a Nawa origin of the word for cacao does not meet such a standard. Instead, their proposed Yuta-Nawan etymology is speculative, as is their attempt to deal with the absence of the Nawa absolute suffix—the only one of the linguistic issues that they address—and the supporting arguments for each of these proposals are invalid. In addition, nothing in the evidence provided by Dakin and Wichmann undermines the arguments that Mije-Sokean *kakaw(a) was the source of the word for cacao.

Dakin and Wichmann's proposed scenario for the diffusion of kakawa from Nawa involves the hypothesis that the word for cacao was diffused by Nawa-speaking Teotihuacanos who controlled the Mesoamerican trade in cacao. But Teotihuacan had a massive impact on other Mesoamerican societies. This level of impact would have been accompanied by a substantial impact on at least the vocabularies of many other Mesoamerican languages. Their historical scenario therefore requires a substantial body of early Nawa loans into Mayan, Xincan, Sapoteko, Wastekan, and probably Tarasko and Matlatzinkan languages. So substantial an early impact would be detectable in a large number of obvious and unproblematic candidates for early borrowings.

Dakin and Wichmann's proposal for the diffusion of kakawa from Nawa in association with Teotihuacan's interactions with other Mesoamerican societies raises a further chronological problem. They argue (Dakin and Wichmann Reference Dakin and Wichmann2000:57) that internal evidence within Mije-Sokean shows that “a word kakawa or one close to that in form was borrowed into the [Mije-Sokean] linguistic family, but at a time when it was still at an early stage of differentation.”

The breakup of Mije-Sokean probably took place between about 1200 and 800 b.c. (1) Elsewhere (Kaufman Reference Kaufman2000–2007; Kaufman and Justeson Reference Kaufman, Justeson, Arnold and Pool2007), we provide evidence that the source of Mije-Sokean loans in central Mexican languages was a northern branch of Mije-Sokean that was in the Basin of Mexico. The loans reflect a vocabulary that was not differentiated between Mijean and Sokean, so the speakers of this language probably arrived before or around the time of the breakup of Mije-Sokean proper. Archaeologically, their arrival can be dated to 1200 b.c. or earlier. (2) This constraint agrees with the glottochronological estimate of 1000 b.c. for the breakup of Mije-Sokean. (3) The previous section shows that a body of Sokean loans, including names of four animals associated with day names, entered pre-Sapoteko from Sokean, so Mijean and Sokean must have been well differentiated at that time. One of these borrowings can be dated to before 200 b.c. (4) Epi-Olmec data from the La Mojarra stela (a.d. 157) and the Tuxtla Statuette (a.d. 162) show only two lexical retentions from proto-Mije-Sokean that are now associated with Mijean but 37 words that are now found only in Sokean. This much differentiation is likely to require at least 1,000 years to develop. Certainly, it does not occur within as few as 500 years of differentiation, after which dialects are always close enough to be inter-intelligible. The Epi-Olmec data therefore put the early differentiation of Mije-Sokean well before 300 b.c., and probably back to 800 b.c. or earlier.

Even the latest of these estimates is far too early for the diffusion of anything from Teotihuacan, which was not established until around 150 b.c., and whose non-local impact did not begin until the second century a.d. It is even further out of line with the known timing of lexical diffusion from Nawa.

In the remainder of this paper we show that, contrary to Dakin and Wichmann, the Mije-Sokean data are consistent with a Mije-Sokean origin for kakawa and that their alternative Yuta-Nawan origin in a reduplicated form, descending from a word for egg, is impossible.

Consistency of the Mije-Sokean Evidence

Dakin and Wichmann (Reference Dakin and Wichmann2000:57) claim that their analysis of Mije-Sokean data pertaining to the word for cacao demonstrates that “it is not possible to continue to attribute a Mixe-Zoquean origin to it. Instead we argue that kakawa—most likely pronounced kàkawá by its donors—entered [Mije-Sokean] from the outside …”. They make three claims in favor of the view that kakawa, or something similar, was borrowed into Mije-Sokean languages rather than having descended from a native Mije-Sokean lexical item. One is the claim that “morphemes consisting of three open syllables (CV.CV.CV)” are virtually unknown in Mije-Sokean, while they are common in Nawa.” (Verb roots are either monosyllabic ending in a consonant or disyllabic ending in a consonant, and most affixes are monosyllabic; these and other morpheme-structure constraints in Mije-Sokean languages mean that only noun roots can be at issue.) A second claim is that the stress pattern of proto-Sokean *kakawa was demonstrably in conflict with Sokean stress rules. Both claims are untrue, and the arguments based on them would be invalid even if they had been true. Their third claim is the correct observation that while Sokean languages point to an ancestral *kakawa , Mijean languages point to an ancestral *kakaw; these two forms cannot both descend normally from a single proto-Mije-Sokean form. Evidence cited later shows that there are numerous cases of discrepancies (though they are hardly prevalent) between proto-Sokean and proto-Mijean reconstructions for what are incontrovertibly single etymologies. All three of Dakin and Wichmann's claims are taken by them as evidence that kakawa was not native to the Mije-Sokean family.

This section addresses the reconstructibility of *CVCVCV(C) trisyllabic roots in Mije-Sokean and Gulf Sokean evidence for initial stress in these roots, establishing the regular descent of Sokean forms from proto-Sokean *kakawa. We address the evidence for the regular descent of all pre-Columbian Mijean forms for ‘cacao’ from proto-Mijean *kakaw. This shows that the form was in Sokean before the proto-Sokean era and in Mijean before the proto-Mijean era. Finally, it addresses the nature of the discrepancy between proto-Sokean *kakawa and proto-Mijean *kakaw.

Our discussion is based on Table 4, which presents our Mije-Sokean data on all of the trisyllabic roots that are reconstructible for proto-Mijean, proto-Sokean, or proto-Mije-Sokean. Each reconstructible proto-Sokean trisyllabic form survives as a trisyllable in Soke but is reduced to a disyllable in Gulf Sokean.

Table 4. Trisyllabic *CVCVCV roots reconstructible to proto-Mijean, proto-Sokean, or proto-Mije-Sokean

Notes: AYA = Ayapa Gulf Sokean; SOT = Soteapan Gulf Sokean. The data cited here were collected by linguists working on the PDLMA, except data from Rayón Soke (from Harrison and Harrison Reference Harrison and de Harrison1984). The trisyllabic variant of Wichmann's (1995:464) incorrectly reconstructed proto-Mije-Sokean *tjk(ay) ‘yesterday’ would also agree with our rules in having initial stress. However, the correct reconstruction, *tj 7k, is a disyllabic word followed (optionally, in just one descendant) by a clitic {7ay}; as such, stress rules operate on *tj7k alone. This root therefore does not provide data on trisyllabic roots in Mije-Sokean languages and is not included in the table.

Discussion of Mije-Sokean Forms

This section discusses the details of the Gulf Sokean data.

Cacao, mamey, fire, guava, and Pachira (apompo)

These words show that proto-Sokean had trisyllabic noun stems of shape CVCVCV(C) and that, in proto-Gulf Sokean, these were stressed on the initial syllable. Generally speaking, Soteapan Gulf Sokean and Ayapa Gulf Sokean stress the first syllable and lose the second syllable, while Texistepec Gulf Sokean stresses the first syllable and loses the last syllable. Since Texistepec Gulf Sokean disagrees with Soteapan Gulf Sokean and Ayapa Gulf Sokean on which syllable is lost, this syllable reduction must have taken place after the breakup of proto-Gulf Sokean. In Totontepec Highland Mije (see Suslak Reference Suslak1996–2002), /7i:tzm/ ‘peccary’, from proto-Mijean *7i:tzm, also has initial stress. In Soke proper, all trisyllabic words are stressed on the penult syllable (and, as stated earlier, in Santa María Chimalapa Soke the first syllable is stressed, as well).

In some of these cases, Soteapan Gulf Sokean and Ayapa Gulf Sokean lengthen the stressed syllable, while in others they do not. There is too little data to provide a secure account for the presence or absence of vowel length in these forms. Possibly related is the fact that, in Soteapan Gulf Sokean, disyllabic roots serving as prepounds may lose their second vowel and may lengthen the first, stressed vowel.

Reduction of the second syllable is obvious in the Gulf Sokean words for cacao, mamey, fire, and Pachira, but the case of *patajaC requires more extended explanation. The attested Soteapan Gulf Sokean form, patanh, has a short vowel in the first syllable, which can only result from an underlying synchronic tt. The only possible source for this tt that is consistent with the forms of cognates in Soke would be a cluster tj, resulting from a reduction of the middle syllable of *pataja(C), yielding pre-Soteapan Gulf Sokean *patja(C). In Ayapa Gulf Sokean, roots of the shape CV:7CV regularly arise from proto-Sokean and proto-Gulf Sokean *CVCV7; Ayapa Gulf Sokean [pa:7da] therefore suggests a pre-Ayapa Gulf Sokean *pata7, again presumably from *patja7. Both Soke and Gulf Sokean therefore support reconstruction of word-final 7 in this form, and in particular require the reconstruction of *pataja7 for ‘guava’. Soteapan Gulf Sokean patanh in place of expected pata7* is consistent with a sporadic but fairly common phenomenon in individual Sokean languages whereby word-final 7 or V in one language corresponds to nh or n in another language. If the Texistepec Gulf Sokean form patanh is descended from an antecedent form *patajanh, which may well be the case, we do not know the rules that would yield this form. Alternatively, the form may be a simple borrowing from Soteapan Gulf Sokean.

Ayapa Gulf Sokean has one word, pa7tanh=kuy ‘guava tree’, in which the word for guava is the first element in a compound. This compounding form has nh at the end of the root. In Soteapan Gulf Sokean, patanh occurs both in compounds and as an independent word. Since the internal 7 of Ayapa Gulf Sokean pa7tanh= must originally have been root-final, the current nh at the end of the root must have developed in pa7tanh=kuy after the collapse of the unstressed penult syllable and, therefore, after the breakup of Gulf Sokean. In waktanh= kuy, Ayapa Gulf Sokean has nh at the end of the root in the compounding form of wakta < proto-Sokean *wakata ‘Guiana chestnut’.

The form kakwa of Rayón Soke—a dialect of Chiapas Soke, which generally has kakawa(7)—seems to reflect the Gulf Sokean pronunciation and may be a borrowing from the direct antecedent of Ayapa Gulf Sokean [ka(:)gwa7], although perhaps not from the precise current location of Ayapa Gulf Sokean in Tabasco.

Table 5 provides forms that have the appearance of being trisyllabic roots but are in fact morphologically complex and, correspondingly, show different stress patterns.

Table 5. Morphologically complex forms that resemble trisyllabic roots in some Mije-Sokean languages

Notes: The data cited here were collected by linguists working on the PDLMA, except data from Francisco León Soke (from Engel and Allhiser Reference Engel and de Engel1987) and Rayón Soke (from Harrison and Harrison Reference Harrison and de Harrison1984).

^aThe form [púguyu7] reported by Wichmann (Reference Wichmann1995:428) differs from the form collected by Suslak (Suslak et al. 1997–2007), and, among other problems, has two violations of Ayapa Gulf Sokean phonotactic constraints: the first vowel, which is stressed, should be long, and underlying word-final glottal stops should jump to the penult syllable. We therefore reject this form as inauthentic. The Gulf Sokean form *pukuyyu is not directly cognate with proto-Mije-Sokean *pu7+juyu7, though it is obviously related to it or derived from it—by some irregular process. Because of its form, with a medial closed syllable, it cannot instantiate the Gulf Sokean initial stress on words of original shape *CVCVCV(7).

Pre-Columbian Mijean Forms were Inherited from Proto-Mijean

Wichmann (1995:343–344) argues that the proto-Oaxaca Mije form corresponding to proto-Sokean *kakawa is *kakaw. In fact, all Mijean (not just Oaxaca Mije) forms are consistent with a reconstruction of proto-Mijean *kakaw, except for one highland Mije form, Mixistlán [kaká:wa], which derives from regional Spanish cacahua. The highland and lowland Mije forms of approximate shape [kaká:w] cited by Wichmann (Reference Wichmann1995:343–344) are not Spanish borrowings, as he states, but follow regular Mije developments from *kakaw. Wichmann's evidence for borrowing from Spanish is that the final vowel in these forms is long and stressed. In Totontepec Highland Mije, it is true that descendants of proto-Mijean *CVCVC forms received initial stress, as in káku, which Dakin and Wichmann recognize as descending from proto-Mijean *kakaw. However, in other forms of Mije, *CVCVC forms receive final stress. In all Mije, stressed final syllables and monosyllabic stressable words insert [h] before the final C, and this inserted [h] goes to vowel length before resonants. Thus *kakaw ⇒ kakáw ⇒ kakáhw ⇒ kaká:w. Dakin and Wichmann (Reference Dakin and Wichmann2000:57a) go further than Wichmann (Reference Wichmann1995) in erroneously suggesting that the [kaká:w] forms do not develop from the same form as Totontepec Highland Mije /kaku/. Oluta Mijean kakaw [kaka7w] shows that this form also descends from one with a final consonant.

Wichmann (Reference Wichmann1995:343–344) cites the following data (language labels follow his terminology):

The Mijean form *kakaw has nothing to do with the fact that in Mije proper, in Sayula, and in Tapachula word-final proto-Mije-Sokean *V and *V7 are dropped in polysyllables. (The development in Sayula and Mije may be a shared change; that in Tapachula is not.) When these Mijean languages drop a final vowel, the phonological reflexes are different from those in words that never had a final vowel; it is shown here that the surviving Mije forms are consistent with the reconstruction of proto-Mijean *kakaw. Oluta does not drop word-final vowels, and the phonetic [7] before the final consonant of Oluta Mijean kakaw [kaka7w] is regular for Oluta Mijean roots of two or more syllables descending from proto-Mijean and proto-Mije-Sokean consonant-final forms but not for those descending from roots ending in V or V7.

Accordingly, except for one late loan from Spanish, the Mijean forms are consistent with direct inheritance from proto-Mijean *kakaw. They supply no evidence for any pre-Columbian borrowing that postdates the breakup of proto-Mijean.

The Discrepancy between Proto-Sokean *kakawa and Proto-Mijean *kakaw, and Its Source

The only remaining puzzle, then, is that proto-Mijean *kakaw and proto-Sokean *kakawa are not identical. Note that under the model we propose—that *kakaw~*kakawa is native to Mije-Sokean—borrowing between Mijean and Sokean does not account for this difference. Had Mijean *kakaw been borrowed into Sokean, it would be expected to have retained its pronunciation in Sokean, which has a large number of *CVCVC noun roots—far more of them than the number of its *CVCVCV roots. So such a borrowing is implausible. Similarly, although there were few *CVCVCV roots in proto-Mijean, it is unlikely that an ancestor of proto-Mijean would have reduced a trisyllabic noun root *kakawa to two syllables on being borrowed into a form of Mijean predating proto-Mijean. Not only are there at least two solid proto-Mijean *CVCVCV roots—*wakata ‘Guiana chestnut’ and *7i:tzm ‘peccary’—but CVCVCV is within the range of phonotactic shapes of complex words in proto-Mijean.

The remainder of this section explores a viable alternative to borrowing, for which there are numerous parallels: that the differences reflect an ancient heritage of this word within the Mije-Sokean family that dates back to proto-Mije-Sokean.

Very often for a given meaning, Mijean and Sokean have completely different morphemes or combinations of morphemes. Sometimes, though, the forms are clearly related phonologically, but show discrepancies that forestall the reconstruction of a single phonological form to proto-Mije-Sokean. Examples currently known to us are provided in Table 6. Note again that borrowing between an ancestor of proto-Mijean and an ancestor of proto-Sokean, in either direction, cannot account straightforwardly for any of these differences.

Table 6. Non-identical but phonologically related reconstructions for Sokean and Mijean

^a*tin is the proto-Mije-Sokean word for ‘shit,’ as can be seen in the proto-Mijean *tintzay ‘gut’ (i.e. “shit-vine”); proto-Mijean t:n7.i ‘shit’ is a nominalization of proto-Mije-Sokean *t:n7 ‘to shit’.

Refuting the Arguments for a Nawa Source of *kakawa in Mesoamerican Languages

[kakaw] outside Mijean

In Mayan, /kakaw/ is the typical form. It is borrowed, not native, because a proto-Mayan *kakaw would not have preserved [k] in all languages. Mayan tolerates disyllabic noun roots (though they are relatively few) but does not have native trisyllabic roots nor roots ending in vowels. The model for the diffused forms could have been either [kakaw] or [kakawa]. The borrowing reached Greater Tzeltalan after the change *k > ch had run its course, so sometime after 200 b.c. Mayan has a reconstructible root *pe:q, now referring to uncultivated cacao (Spanish pataxte); that Greater Lowland /pi:k/ expresses the numerical value of 8,000 (and proto-Yukatekan *pi:k ‘skirt’; cf. “The Mije-Sokean Hypothesis”, above) suggests that the Mayan term *pe:q also referred to cultivated cacao at one time.

Totonako, Tepewa, Salvador Lenka, and Paya all have forms like /kakaw/. The Paya and Salvador Lenka forms plausibly have spread from Mayan, but possibly from Mijean, or even from pre-Sokean if proto-Sokean *kakawa developed after the proto-Mije-Sokean stage. The Totonako and Tepewa forms more likely spread from Mije-Sokean.

• Totonako /kakaw/ ‘peanut’ (Kaufman field notes, Reference Kaufman and Justeson2003)

• Tepewa /kakaw/ ‘cacao’ (Susan Smythe, personal communication 2005)

• Salvador Lenka [k'á:gaw] = /k'akaw/ ‘cacao’ (Lehmann Reference Lehmann1920; the anomalous initial k' may well be an error on Lehmann's part)

• Paya [kaku] ‘cacao’ (Holt Reference Holt1999a)

Forms like kaw(a)

Most of the following forms, all from Lower Central America, resemble the Mije-Sokean *kakaw(a); some seem to reflect a model [kaw], which is not known in Mesoamerica proper. These forms are listed in the geographical order in which they occur (Figure 5):

Figure 5. Languages of lower Central America that have words for cacao derived, ultimately, from Mije-Sokean, showing their locations as of about a.d. 1500. To provide a sense of the linguistic geography when the word for cacao was diffusing, intrusive groups that reached their contact-period locations after a.d. 500 have been removed, and their territory has been divided among remaining adjacent groups. We have no data on words for cacao from Misumalpan languages, which divide Paya from the rest of the Chibchan family. Chibchan : BRK = Boruka; DRS = Doraske; MOB = Mobe; PAY = Paya; RAM = Rama; TRB = Terraba; WTS = Watuso. Other: LNKh = Honduras Lenka; LNKs = Salvador Lenka; TOL = Tol; XNK = Xinka. After Kaufman and Justeson (Reference Kaufman, Justeson, Arnold and Pool2006:Figure 6.2).

• West to East in El Salvador and Honduras

  • Mayan kakaw

  • Tol [khaw] (Holt Reference Holt1999b)

  • Honduras Lenka *[kaγaw] > /kaw/ (Lehmann Reference Lehmann1920)

  • Pipil kakawa-t (Campbell Reference Campbell1985; inherited from proto-Nawa, which had borrowed *kakawa from Mije-Sokean by about a.d. 500)

  • Salvador Lenka /k'akaw/ (or /kakaw/) (Lehmann Reference Lehmann1920)

• North to South on the Pacific Coast

  • (Chiapaneko nuusi [Fernández de Miranda and Weitlaner Reference Fernández de Miranda and Weitlaner1961:53])

  • (Chorotega nyuusi [Fernández de Miranda and Weitlaner Reference Fernández de Miranda and Weitlaner1961:53])

  • (Sutiaba uusi) (Lehmann Reference Lehmann1920)

• North to South on the Atlantic Coast (Paya and Rama are Chibchan)

  • Paya kaku (Holt Reference Holt1999a)

  • [Sumu-Ulwa] (forms not found)

  • [Mískitu] (forms not found)

  • Rama kuk (Colette Grinevald, personal communication circa 2000)

• West to East in Costa Rica (all of these are Chibchan languages)

  • Proto-Watuso-Boruka *'hú7 (Constenla Reference Constenla1981:373)

  • Watuso kaju: (Constenla Reference Constenla1981:373)

  • [Kabékar tsirú] (Dakin and Wichmann Reference Dakin and Wichmann2000:74)

  • Térraba kó (Dakin and Wichmann Reference Dakin and Wichmann2000:75)

  • [Bribri sirú] (Dakin and Wichmann Reference Dakin and Wichmann2000:74)

  • Boruka kaw7 (Constenla Reference Constenla1981:373)

  • Doraske koa (Dakin and Wichmann Reference Dakin and Wichmann2000:74)

  • Waymí ku (Dakin and Wichmann Reference Dakin and Wichmann2000:74)

Besides forms like [kaw], Chorotega and Sutiaba, whose speakers probably invaded the region circa a.d. 800-900 (Salgado Reference Salgado González1996:303; Steinbrenner Reference Steinbrenner and McNeil2006:257) and circa a.d. 1200, respectively, share a form like [(ny)uusi]. Tlapanekan (= Tlapaneko + Sutiaba) and Chorotegan form a node on the Oto-Mangean family tree, but since no form for cacao has been identified in Tlapaneko, the Sutiaba form could be a borrowing from Chorotega. Given the Chiapaneko form, *nuusi7 can be reconstructed for proto-Chorotegan and rolled back to central Mexico, where Chorotegan originated.

Kabékar and Bribri share a form like [(t)sirú]. Its further connections are unknown to us.

Salvador Lenka, Paya, and Rama reflect the typical Mayan pronunciation [kakaw], and this is plausibly their immediate source.

Going back to forms like [kaw], the following scenario may be envisioned:

First of all, the Chibchan data are limited geographically to only some of the Chibchan languages in Central America, and they are not found in any of the Chibchan languages of South America. The distribution does not correspond to a genetic subgrouping within Chibchan. Constenla's reconstruction *'hú7 may more properly be treated as a formula subsuming the phonological regularities between the Watuso and the Boruka forms as if they were cognate, even though they are due to diffusion. The rest of the Chibchan forms cited represent most of the branches of the stock, but there is no “cacao”-like form in any South American language that is not the result of colonial-period diffusion. Except for the Rama form, which cannot be fully accounted for, we can postulate that the phonological antecedent for all these forms is something like [kahaw]. Central American Chibchan languages show the development [káhaw] (with first syllable “stress”) > [*'hú7]  >  kaju:, káw7, kó, ku, kuk. Tol may have developed [kaháw] (with second syllable “stress”) to [khaw] or borrowed its word from one of the Chibchan languages with a (possibly intermediate) Boruka-like form [káw7]. Honduras Lenka may have done likewise.

The postulated Central American antecedent [kahaw] would have been borrowed from the general Mayan form /kakaw/. If the first intermediary into Central America from the Mayan area had been Honduras Lenka, there would be an explanation of the shift of medial /k/ to [h], because in Honduras Lenka, single intervocalic /k/ is pronounced [γ], which does not sound like [k] in a language having only [k] and [h] but not [γ] or [g]. If Honduras Lenka is the intermediary, why it should have simplified [kaγaw] to [kaw] is not crystal clear, but it may be observed that Spanish vacas yielded Honduras Lenka /waš/ ‘cattle’ through an intermediate form [waγaš], so this is a plausible internal development in Honduras Lenka.

Salvador Lenka /k'akaw/ (maybe /kakaw/) and Paya /kaku/ could be direct borrowings from Mayan with no Honduras Lenka intermediary, but Mayan languages are quite far away, and there are viable alternative explanations. The [g] of Salvador Lenca [k'á:gaw] could reflect the [γ] of an earlier Honduras Lenka [kaγaw] rather than Mayan [k]. Similarly, since Paya borrows proto-Mije-Sokean *pa:=ju7 ‘coyote’ as /paku/, it is more likely that it borrowed an antecedent Central American form [kahaw] as /kaku/ than that it made a far more distant borrowing from Mayan.

Altogether, the evidence from Central American languages does not clearly support an antecedent form like [kaw]; in fact it more strongly suggests an antecedent form like [kahaw], borrowed from Mayan /kakaw/, maybe specifically via Honduras Lenka.

There are two Oto-Mangean words that resemble [kakawa] or [kawa], both found in Oaxaca: (1) Proto-Chinanteko has *kwá:7 ‘case; peeling; pod; shell’ (Rensch Reference Rensch1989:50, no. 163); cf. also *kwé:7 ‘bark, peeling’ (Rensch Reference Rensch1989:50, no. 160). In no Chinanteko language does a form like [kwa] or [kwe] actually mean ‘cacao’. (2) Amusgo has /tεh šuah/ (literally, “bean cacao”) (tone pattern is low, mid) ‘cocoa bean’, plural /tεh nguah/ (Tapia 1999:216). There is a class of nouns in Amusgo that take the prefix {tz-} in the singular and {n-} in the plural; given this, and the fact that the underlying sequence //tz-k// is realized as š in Amusgo, the underlying form of the word for cacao can be seen to be //-kuah//, with the singular //tz-kuah// realized as /šuah/, and the plural //n-kuah// realized as /nguah/. Both the Chinanteko and the Amusgo forms can derive from an antecedent [kVwa]. The identity of the first vowel cannot be determined.

Dakin and Wichmann (Reference Dakin and Wichmann2000:74) cite Chocho /ka:kaú7/ (from Mock (Reference Mock1977), which could be borrowed from Spanish. No other Oto-Mangean words for cacao reflect [kakawa] or [kawa].

The Chinanteko and Amusgo forms seem as though they could reflect specifically [kawá]. Both of these forms lack the initial syllable [ka] of *kakawa, and an antecedent *kawa could yield both of them. The [kawá] form has stress or prominence on the second vowel. This may reflect habitual accentual patterns: in Oto-Mangean languages (apart from the Mije-Sokean-influenced Sapoteko, Misteko, and Kwikateko), polysyllabic words have highest prominence on the last syllable. In the earliest stages of Oto-Mangean languages, lexical stems had one or two syllables; any lexical material in antepenult position is a clitic or a classifier. Hence, if a form like proto-Sokean *kakawa were taken into an early stage of an Oto-Mangean language (at least 2,000 years ago), something would have to be done with the first /ka/. If it did not correspond in a meaningful way with an existing proclitic or classifier in the target language, it might be eliminated. We suggest that this is indeed what happened to *kakawa in Amusgo and what might have happened in Chinanteko if proto-Chinanteko *kwá:7 ~ *kwé:7 (dating before about 1,500 years ago) is a borrowed word (pre-proto-Chinanteko underwent sound changes whereby antecedent CVCV forms were reduced to CCV, and the initial cluster was subject to simplification in certain cases).

The kaw(a) forms discussed in this section might suggest that there was originally an “unreduplicated” form meaning ‘cacao’ drifting around that was reduplicated to produce the form *kakawa. The discussion here shows that this would be an unnecessary assumption and that these forms provide no viable evidence that *kakaw(a) was a reduplicated form.

Evidence Concerning the History of Nawa chikola:tl~chokola:tl

All suggestions so far offered for the origin of the first element are unsatisfactory. Nawa ch in native words should occur only before i (Campbell and Langacker Reference Campbell and Langacker1978). If the earliest Nawa form was chikola:tl, the form chokola:tl could have developed from it by assimilation of the first vowel to the second. If the earliest form was chokola:tl, {chokol=} is perhaps borrowed. Evidence for either of these two possibilities is lacking, though the first is more likely, and Dakin and Wichmann argue for it, adding the consideration of additional data that may not be relevant. This is not to say that we (Kaufman and Justeson) have a perfectly obvious derivation for chikola:tl~chokola:tl, because we do not. In any event, from the forms of the borrowings listed here, it appears highly likely that the languages that have it received the loan from Nawa.

Dakin and Wichmann (Reference Dakin and Wichmann2000:62b) cite chikola:tl for the Nawa towns of Ocotepec (Morelos), Ameyaltepec (Guerrero), Cuetzalan (Veracruz), and Rafael Delgado aka San Juan del Río (Veracruz). This form is also found in North Puebla Nawa (Una Canger, personal communication 2005), a type of Central Nawa. In the case of Cuetzalan Nawa, Dakin and Wichmann (Reference Dakin and Wichmann2000:62b) point out that chokola:t is said now, but the older form is said to have been chikola:t. While Dakin and Wichmann cite data from Rafael Delgado, attributed to uncited work by Tuggy, as having chikola:tl, the more recent PDLMA data from Rafael Delgado (Romero Reference Romero1999–2002) has both chokola:tl and xikola:tl (not chikola:tl*). Dakin and Wichmann argue plausibly that chikola:tl was the original pronunciation of this word. The pronunciation xikola:tl and its implications need investigation.

The form chokola:tl was documented by Kaufman in 1978 for Santa María Izhuatlan. Huasteca Nawa, both West and East (Kaufman Reference Kaufman1969/1984–1993), has chokola:tl ‘chocolate’; but in Chontla Huasteca Nawa the word means ‘caldo de tripa de puerco’ (soup/broth made from swine gut). This suggests that a Nawa morpheme {chokol=} (or {chikol=}) combined with {a:} ‘water’ has a generic meaning with at least two applications. Unfortunately, the more generic meaning of {chokol=} is not easy to divine. Pipil, which has been separate from other forms of Nawa since circa a.d. 900, has chukula:t (Campbell Reference Campbell1985:200). (Fowler [Reference Fowler and McNeil2006:310] cites Sampeck Reference Sampeck2005 for an archaeologically based date of circa a.d. 1200 for the arrival of Nawa speakers in the Izalcos region of El Salvador. This does not square with the linguistic data, if we assume that the Pipils arrived in Izalcos about the time when they became linguistically separate from other forms of Nawa—specifically, that of the Southern Gulf coast—which was probably around a.d. 900. If, however, they were in Chiapas or Escuintla for a while before a.d. 1200, the archaeological and linguistic chronologies would not be in conflict.) This raises the possibility that the form chokola:tl is at least 1,000 years old, but it could also be a borrowing from the Central Nawa speakers (“mejicanos”) brought into Guatemala by the Spanish after 1525. In light of the other data discussed in this study (both above and below), the latter possibility seems more likely.

Even if {chokol=} is assimilated from {chikol=}, the closest comparanda in Nawa are chiko ‘bent in a half-circle’ and chihkol-li ‘thing bent in a half-circle’, neither of which is in fact {chikol}. Dakin and Wichmann's (2000:63–66) hypothesis is that chihkol-li meant ‘cacao-beater’ in some kinds of Nawa, but chihkol-li has a preconsonantal /h/, and chikol=a:-tl does not. On the one hand, there is no straightforward evidence for a Nawa word chikol-li*; on the other hand, chiko and chihkol-li must be related, both reflecting the meaning ‘bent, hooked’, and this is not consistent with Dakin and Wichmann's hypothesis, which derives chihkol-li and their hypothetical chikol-li* from a supposed Yuta-Nawan *ci' ‘small/pointed stick’ (Dakin and Wichmann Reference Dakin and Wichmann2000:63–64) plus *ku- ‘tree, pole’ (Dakin and Wichmann Reference Dakin and Wichmann2000:64–65). These difficulties render their proposed etymology unconvincing.

Their proposed Yuta-Nawan etymology in particular is untenable. Dakin and Wichmann (Reference Dakin and Wichmann2000:63–66) argue for a Yuta-Nawan etymology for this word by attempting to analyze between 15 and 20 polysyllabic Yuta-Nawan words into monosyllabic roots, with the aim of providing evidence for putative proto-Yuta-Nawan elements *ci'- ‘pointed stick’, *ku- ‘tree, pole’; and *-ri ‘noun derivational suffix’. This section is of interest mainly as an illustration of Dakin's long-term research program of etymologizing Nawa lexical items of two or more syllables as compounds made up of two or more monosyllabic roots. This type of analysis is not employed by most other Yuta-Nawanists, who recognize a limited number of monosyllabic roots in each language and in proto-Yuta-Nawa itself—the majority of roots being disyllabic. This analysis into monosyllabic roots is not required for the morphological analysis of the lexicon into its constituent morphemes in the individual languages. As is to be expected if these items are not in fact composed of monosyllabic elements, the meanings associated with the parts of Dakin's proposed compound words do not often bear a compositional relationship to the meaning of the hypothesized compound, and the resulting monosyllabic elements typically lack finely focused semantic specificity.

The case at hand illustrates these problems. There is practically no identity of structure or gloss between any two Yuta-Nawan languages in the cited vocabulary, just partial overlap. There is one example only of an apparently plausible cognate—one with a close semantic matching—between two Yuta-Nawan languages:

• Kawaiisu či-ku-li ‘stirrer’ (cited from Dakin and Wichmann)

• Nawa chikol=a:-tl, which might say literally “stirring stick water”

However, this comparison is not valid, because Kawaiisu /u/ corresponds not to Nawa /o/ but to Nawa /i/. The proposed Yuta-Nawan etymology must be rejected.

If there were a Nawa word chikol-li* ‘stirring stick’, then Dakin and Wichmann's hypothesis that chikol=a:-tl meant ‘stirring stick water’ (their own term is ‘beater-drink’) would be plausible and would have no opposition from us. It would not, however, support their supposition that the diffusion of the word *kakawa and that of the name for chocolate were related in any way.

Further, Dakin and Wichmann's claim that Mayan forms like [chukul] that mean stirring stick are borrowings from Nawa is false. These forms descend from a noun *tuuk.ul ‘stirring stick’ that can be reconstructed for Greater Q'anjob'alan and K'ichee7, at least, and that is derived from a Mayan verb *tuk ‘to mix, stir’ that can be reconstructed from Eastern Mayan and Greater Q'anjob'alan languages (Kaufman with Justeson 2003:395).

In any case, why the above combination in Nawa should mean both ‘powdered roasted cacao whipped/shaken/stirred/frothed with water and seasonings’ and ‘chitterling soup’ is by no means clear. We may remind ourselves of how unexplainable some lexical formations are by considering the name of the storm god Tla:l=o.k, which literally translated is ‘one who lies on the ground’, or the dwelling of the blessed after death Tla:l=o.k.a:n, which literally is ‘place of lying on the ground’.

The Diffusion of *chikol=a:-tl and its Relationship to that of *kakaw(a)

Dakin and Wichmann (Reference Dakin and Wichmann2000:62) argue that the distribution of chikola:-tl and chokola:-tl across Nawa suggests that chikola:-tl originated within Eastern Nawa, and that chokola:-tl developed from it within Western Nawa based on a propensity of the latter for “vowel harmony” (by which they mean assimilation between vowels of adjacent syllables). As stated in the previous section, we agree with them that the form chikola:tl is likely to have been the original pronunciation of this word. Their further conclusions about its dialect history, however, are speculative, not secured by either linguistic or culture-historical data.

1. 1. The assimilation of vowels in adjacent syllables is not a regular process (or rule) in any Nawa dialect. Rather, it is basically sporadic and occurs not only in Western Nawa but also (contrary to Dakin and Wichmann) in Eastern Nawa. (Some forms of Western Nawa have rules for the assimilation of the short vowels in inflexional prefixes, but rules or outcomes of this type are not general in the lexicon.) In addition, chokola:tl is found in Eastern Nawa dialects, including Huasteca Nawa dialects, Santa María Izhuatlan Nawa, and Pipil. While we consider it likely (because of specific, known culture-historical data) that some of the forms with o probably spread into some forms of Nawa and, perhaps, even of Spanish, it is not at all certain that all of this diffusion was due to this influence. The origin chokola:-tl from chikola:-tl therefore cannot be reliably attributed to either the Eastern or the Western branch of Nawa.

2. 2. Even if the assimilated form did happen to originate in one of the Western dialects, this provides no evidence whatsoever that the earliest chikola:-tl originated in an Eastern dialect rather than in a Western dialect, nor that its diffusion into other Mesoamerican languages was from an Eastern dialect. Not only do some modern Western dialects have the chikola:-tl form, as Dakin and Wichmann observe, but since chokola:-tl is a later development, chikola:-tl could have developed in a Western dialect and spread from there to other Nawa dialects and/or to other Mesoamerican languages before changing their own pronunciation of the word to chokola:-tl.

While Dakin and Wichmann raise this Eastern-origin scenario quite tentatively—“Could it be that the čikola:tl form is an Eastern Nahuatl form …?” (Dakin and Wichmann Reference Dakin and Wichmann2000:67)—the status of this speculation is elevated to the status of a fact on its next mention:

They go on to address possible objections to this possibility that involve differences in the grammatical forms and regional distributions of the two words.

Note that the dispersal that they had argued for on page 62—albeit based on the incorrect claim of a restriction of assimilatory processes to Western dialects—had been for dispersal among Nawa dialects. Here, however, Dakin and Wichmann incorrectly present their conjecture as having been a conclusion that it was from Eastern Nawa that chikola:-tl was spread to other Mesoamerican languages. (The first sentence of the quoted passage is also misleading in that they in fact present no linguistic argumentation for the conclusion that Pipils were responsible for the claimed spread of kakawa-tl to other Mesoamerican languages.)

Dakin and Wichmann nowhere explain why they consider it “reasonable to suppose that these two words share their center of dispersal”—here clearly presented as an assumption, not a conclusion. The only answer that we have been able to divine, making full use of the context of this claim, is that they imagine these words to have diffused together—from the same cultural group as part of a single cultural process, if not at the same time.

• Nawa chikola:tl yields:

  • Sayula Mijean chikúla:t

  • Kora tzikura:

  • Chayuco Misteko sikula

  • San Mateo del Mar Wavi chikolt (perhaps via Spanish)

• From Spanish chicolate:

  • Huichol sikurá:ti

  • Mitla Sapoteko chikulahd~chigulahd

Besides the forms listed here, Dakin and Wichmann (Reference Dakin and Wichmann2000:62) cite i-forms for San Juan Colorado Misteko, Tlaxiaco Misteko, Warijiyo, Chamorro, Asturian Spanish, Catalan, and Dutch.

• Nawa chokola:-tl ‘chocolate’ yields:

  • Zinacantán Tzotzil chukul 7at ‘chocolate drink’

  • Oaxaca Chontal -tzugulalh

The incidence of other Mesoamerican languages borrowing the Nawa form chokola:tl is slim, indeed.

• From Spanish chocolate:

  • Juchitán Sapoteko dxuladi/chu+lati!7/ ‘chocolate’ < [chukulá:ti]

  • Pajapan Gulf Nawa (Peralta Reference Peralta2002–2007) chokola:teh.

  • Coxcatlan Huasteca Nawa (Kaufman Reference Kaufman1969/1984–1993) chokola:teh

The last two forms suggest that the version of chocolate made and drunk by Spanish speakers tends to get its name adopted even by the people who invented the word and the drink in the first place. It is in fact the case that, at the present time, Nawa speakers often think that a Spanish word of Nawa origin is actually the source of a Nawa word they use natively; following up on this false impression, they often adopt the Spanish pronunciation of a word and forsake the Nawa pronunciation.

It is quite clear that the pre-Columbian diffusion of this Nawa word, in either form, was extremely limited compared with that of kakaw(a) in pre-Columbian times. Certainly, they could not have diffused together. It is also extremely unlikely that the term for chocolate and the term for cacao diffused through Mesoamerica anywhere near the same time period: *kakaw(a) was borrowed into Lowland Mayan languages before a.d. 400 (and not from Nawa) and shows the results of sound changes affecting whole subgroups of Mayan languages. The Nawa word for chocolate was borrowed in diverse forms into a variety of Mesoamerican languages, with no internal evidence in any instance for substantial antiquity. No distributional or linguistic evidence directly suggests that the diffusion of these words was related in any way. It is in fact not reasonable to simply “suppose” that they were.

The Timing of the Origin of *chikol=a:-tl

A further wrinkle is that neither form of the Nawa word is even attested from the first decades of the Spanish colonization. It is not found in Alonso de Molina's Vocabulario en Lengua Castellana y Mexicana and Vocabulario en Lengua Mexicana y Castellana (neither in the 1551 edition nor in the expanded 1571 edition) or Bernardino de Sahagún's Historia general de las Cosas de Nueva España (1577). Given the lateness of its first citations, reasonable doubt may be entertained as to whether the Nawa forms chikola:tl and/or chokola:tl even existed in pre-Columbian times. Indeed, Corominas [1980–1983:2:385–386] and many other students of the history of Spanish do not believe that Nawa chokola:tl existed in the pre-contact period.

The drink made from cacao certainly existed in pre-Columbian times, but, among other things, it may simply have been called by the name of its principal ingredient. In Zinacantán Tzotzil, /kokow/ refers both to the seed and the drink (Laughlin Reference Laughlin1975:176). This is probably true in other forms of Tzotzil and in Tzeltal, in which /kokow/ (Tzotzil) or /kakaw/ (Tzeltal) is translated both as ‘cacao’ and ‘chocolate’. In Q'anjob'al of Santa Eulalia, /kakaw/ means both ‘cacao’ and ‘chocolate’ (Mateo-Toledo, personal communication 2005).

In Nawa itself, the word for cacao could also be used to name drinks that were made from it. In sixteenth-century Nawa, Sahagún (Reference Sahagún and Sullivan1558–1561, 1577) refers to the drink made from cacao as /kakawa-tl/ (“cacao”) (see Anderson and Dibble Reference Anderson and Dibble1951–Reference Anderson and Dibble1982:8.39–40, 11.119-120; Sullivan and Stiles 1988:202). Sahagún (Reference Sahagún and Sullivan1558–Reference Sahagún and Sullivan1561) has the following passage in a section dealing with feasting by nobles:

Sahagún Reference Sahagún, Anderson and Dibble1577 also refers to <quauhnecujo cacaoatl> /kwaw=nekw.yoh kakawa-tl/ ‘honeyed cacao’, <xochiocacaoatl> /xo:chi.yoh kakawa-tl/ ‘flowered cacao’, <chichiltic cacaoatl> /chi:.chi:l.ti.k kakawa-tl/ ‘red cacao’, <vitztecolcacaoatl> /witz=tekol=kakawa-tl/ ‘thorn-charcoal cacao’, <xochipalcacaoatl> /xo:chi=pa.l=kakawa-tl/ ‘flower-painted (i.e. pink or orange-colored) cacao’, <tiltic cacaoatl> [sic] /tli:l.ti.k kakawa-tl/ ‘black cacao’, and <itztac cacaoatl> [sic] /ista:-k kakawa-tl/ ‘white cacao’—all referring to kinds of chocolate (the drink), not kinds of cacao. Molina (Reference Molina1571:Nawa-to-Spanish section 10v, column b) has <cacauaatl> /kakawa=a:-tl/ (“cacao water”) ‘beuida de cacao’ (cited also in Siméon 1885:56b). Molina (Reference Molina1571:161a) lists <xochiayo cacauatl> /xo:chi=a:.yoh kakawa-tl/ (“flowerwater-having cacao”) ‘beuida de cacao con ciertas flores secas y molidas’ (drink made from cacao with certain dried ground up flowers). Molina (Reference Molina1571:Spanish-to-Nawa section 22r, column a) lists ‘cacao, beuida’, meaning by this “cacao—a drink”, “the drink called cacao”, showing that for Molina, chocolate was called simply cacao in his use of Spanish. On page 19r of the Spanish-to-Nawa section, Molina lists ‘beuida de cacao con mayz’ (drink made from cacao with maize): <cacaua atl> /kakawa=a:-tl/ (“cacao water”); ‘beuida de cacao con axi’ (drink made from cacao with chilli pepper): <chillo cacauatl, chilcacauatl> /chi:l.loh kakawa-tl/ (“peppery cacao”), /chi:l=kakawa-tl/ (“pepper cacao'); ‘beuida de cacao ſſolo’ (drink made from cacao alone): <atlanelollo cacauatl> /ah tla-nelo:.l.loh kakawa-tl/ (“unmixed cacao”). All these are cited as well in Molina's 1551 edition (right-hand column on page 34), which also cites ‘beuida de cacao compueſta con flores’ (drink made from cacao put together with flowers): <xochiayo cacauatl >, <xochayo cacauatl> /xo:ch(i)=a:.yoh kakawa-tl/ (“flower-watery cacao”)].

Bierhorst (Reference Bierhorst1985:54) cites <cacahuaoctli> /kakawa=ok-tli/ (“cacao pulque”) from the Cantares Mexicanos, which date from circa 1582 and lack the word <chocolatl> or <chicolatl > . The word /kakawa=oktli/ is probably the same as the /ok.yoh kakawa-tl/ mentioned in Sahagún (Reference Sahagún and Sullivan1558–Reference Sahagún and Sullivan1561).

In the Nawa poems Romances de los Señores de la Nueva España (from 1582) there are two instances of kakawatl naming the drink chocolate:

In the Spanish document Relación de Juan Bautista de Pomar (Tezcoco, 1582) (Garibay 1993:193), the following passage occurs:

Fray Diego Durán's Historia de las Indias de Nueva España and his Islas de la Tierra Firme, written before his death in 1586, do not use the word chocolate, but they do use the word cacao 16 times in reference to the drink and 27 times for the seeds of or for ground-up cacao (Durán Reference Durán and Kintana1965).

Hernando Ruiz de Alarcón's 1629 Treatise on the Heathen Superstitions that Today Live among the Indians Native to this New Spain (cf. Andrews and Hassig Reference Andrews and Hassig1984:132; Coe and Whittaker Reference Coe and Whittaker1982:188), which reports on magical practices by Nawa speakers in Guerrero, refers to the drink made from cacao by the Spanish word cacao. The kind of Nawa found in Ruiz de Alarcón is from the same general dialect group as the Nawa of the Basin of Mexico, which can be called Central Nawa.

The foregoing is evidence suggesting that, in the Nawa of the Basin of Mexico—and perhaps in Central Nawa generally—no word chikola:tl or chokola:tl existed in the sixteenth century or before and that drinks made from cacao were referred to as kakawatl or by expressions that included it. The words chikola:tl and chokola:tl may have arisen in a peripheral type of Nawa, at an undeterminable date, and only spread later to Central Nawa.

The first known use of the Nawa word chokola:tl is cited in Corominas (Reference Corominas1980–1983:2.385–386) from Francisco Hernández (Reference Hernández1959 [1577]) as <chocollatl>. In Chapter lxxxvii (“cacahoaquáhuitl árbol del cacao”), where Hernández discusses cacao and its varieties (Hernández Reference Hernández1959 [1577]:2:303–305), a good deal of information is presented. On page 304, he names four varieties of what he considers to be the same basic plant, which must be Theobroma cacao: <quauhcacahoatl> /kwaw=kakawa-tl/ “tree cacao”, <mecacacahoatl> /meka=kakawa-tl/ “vine cacao”, <xochicacahoatl> /xo:chi=kakawa-tl/ “flower cacao”, and <tlalcacahoatl> /tla:l=kakawa-tl/ “earth cacao” (not the homophonous ‘peanut’, discussed later). On pages 304–305, he considers adding <quauhpatlachtli> /kwaw=patlach-tli/ “broad/flat tree” to this group. Sahagún Reference Sahagún, Anderson and Dibble1577 distinguishes three types of cacao by the color of the fruits (Anderson and Dibble 1951–1982:8:39). Hernández (Reference Hernández1959 [1577]:2:304 says that “hacen tambien de ella una bebida” (they also make a drink from it [cacao]), but he goes on to describe four different drinks made with maize (maize dough, as Clavijero's description makes clear) and <cacahoatl> /kakawa-tl/, by which term he specifically refers to the kernel:

1. 1. <atextli> /a:=tex-tli/ (“water flour”) ‘pasta aguada’: ground <cacáhoatl> mixed with ‘grano indio’ (maize)—for refreshment and nourishment; also as an aphrodisiac;

2. 2. [unnamed]: made from kernels of <cacahoapatlachtli> /kakawa=patlach-tli/ (“broad/flat cacao”), <cacáhoatl> , and ‘grano indio’ (maize)—for nourishment and refreshment;

3. 3.  <chocóllatl> /chokol=a:-tl/: made from an equal number of kernels of <pochotl> /po:cho:-tl/ (Ceiba) and <cacahoatl> , with ‘grano indio’ (maize)—drunk lukewarm as a fattener and as a medicine for tuberculosis;

4. 4.  <tzone> /tzon.eh/ (“hairy”, “furry”): equal parts of roasted ‘grano indio’ (maize) and <cacáhoatl>—for refreshment and nourishment, not as medicine.

Before describing these four drinks, he mentions another use: a drink made from the cacao kernel alone that medicinally serves to reduce heat in the body (Hernández Reference Hernández1959 [1577]:2:305).

Other plants reported by Hernández that contain the element {kakawa} include <quauhcacáhoatl> /kwaw=kakawa-tl/ “tree cacao” (Hernández Reference Hernández1959 [1577]:2:305) and <tlalcacáhoatl> /tla:l= kakawa-tl/ (“earth cacao”) ‘peanut’ (Hernández Reference Hernández1959 [1577]:2:306–307; both are apparently different from the terms of identical form listed above); <iztactlalcacáhoatl> /ista:-k tla:l=kakawa-tl/ ‘white peanut’ (Hernández Reference Hernández1959 [1577]:2:307); <cacahoaxóchitl> /kakawa=xo:chi-tl/ “cacao flower” (Hernández Reference Hernández1959 [1577]:2:307–308; in Huasteca Nawa this is Hamelia patens, ‘scarletbush’], <cacahoapatli> /kakawa=pah-tli/ “cacao medicine” (Hernández Reference Hernández1959 [1577]:2:308). None of these is identified as a kind of <cacáhoatl>. (The acute accents on Nawa words cited by Hernández were probably added by the Latin-to-Spanish translator.)

The next known citation of the word chokola:tl in Nawa is from Clavijero (Reference Clavijero1780), who cites Nawa <chocolatl> with the gloss ‘alimento hecho con almendras de cacao y semilla del árbol llamado pochotl, en partes iguales [a food made from cacao kernels (“almonds”) and the seed of a tree called po:cho:tl (Ceiba, silk-cotton tree), in equal parts]’ (Siméon Reference Siméon1977:107).

Clavijero (cited in Santamaría Reference Santamaría1992:412), who was born in Veracruz, writes as follows of drinks made from cacao:

Clavijero's description (from 1780) matches that given by Hernández (from circa 1580) in a great many (not all) of the details. Though Hernández's descriptions of drinks made from cacao do not specify any sweeteners (and this has wrongheadedly been made much of by historians of chocolate), Sahagún (Reference Sahagún, Anderson and Dibble1577) does refer to “honeyed cacao”.

Apparently because it is not attested in the earliest sixteenth-century sources, etymologists of Spanish (see, Corominas Reference Corominas1980–1983:2:385–386) seem universally to agree that chokola:tl is not a genuine Nawa word. Rather, they think chocolate is a Spanish word created by Spanish speakers through the mangling of a Nawa word or expression. As a result, etymologists of Spanish have proposed a great number of hypothetical origins for this word. Not a single one of the etymologies suggested by members of this brotherhood seems to have a chance of being correct. By way of illustration, we discuss proposals by Corominas (Reference Corominas1980–1983:2:385–386). He discusses and discards several proposed etymologies of chokola:tl and chocolate that depend on Spanish speakers' mangling the pronunciation of some no longer extant Nawa expression, and then this mangled word being borrowed into Nawa in the guise of a Nawa word. Then he offers his own suggestion. According to Corominas, since chocolate was made from equal amounts of cacao (Nawa kakawa-tl) and ceiba (Nawa po:cho:-tl) kernels, the Nawas probably called it pocho-kakawá-atl “ceiba cacao drink”. This was contracted to cho(ca)cahuatl, and this, in turn, would have been mangled into chocolatl in the mouths of Spaniards. He also contemplates that a form something like xochayocacahuaatl* (which is a misspelling of Molina's <xochiayo cacauatl>) could be the source of chocolatl, again through mangling in the mouths of Spanish speakers.

This proposal is so fraught with speculation that it hardly merits serious discussion. In the interest of explicitness, we observe the following specific difficulties. First, the premise is incorrect. There were many types of cacao drinks, as discussed above, and not all were made with ceiba seeds. Furthermore, there is no evidence of any such word as po:cho:=kakawa-a:-tl* in Nawa, and deriving chokola:tl from it requires four separate manglings: the loss of two non-adjacent syllables, the initial syllable po and one of the ka syllables; the change of a in the remaining ka to o; and the change of w to l, a substitution not found in established Spanish borrowings from Nawa. Deriving chokola:tl from Molina's <xochiayo cacauatl> requires loss of three syllables (xo, chi, and ya) and the following segment y, along with the changes of a to o and w to l. These are wholly implausible hypotheses.

The fact is that the Nawa forms chokola:tl and chikola:tl are not plausibly the re-Nawatizations of Spanish chocolate and chicolate. There is no evidence that any Spanish word of the shape /..te#/ was ever borrowed into Nawa ending with /..V:tl/, and it is clear from the discussions of both Hernández and Clavijero that they thought that chocolatl was just as much a Nawa word as all the other Nawa terms they brought under discussion.

The first occurrence of the Spanish word chocolate is found in Book 4, Chapter 22, of Joseph (José) de Acosta's Historia Natural y Moral de las Indias (Reference Acosta1590): “El principal beneficio deſte cacao es, vn breuaje que hazen q <ue> llaman Chocolate, que es coſa loca lo que en aquella tierra le precian, y algunos q <ue> no eſtan hechos a el, les haze aſco [The main benefit of this cacao is a drink that they make that they call chocolate, which in that land they prize like mad: some who are not used to it are disgusted by it].” An English translation of Acosta's work from 1604 provides the first citation of chocolate in the Oxford English Dictionary. The Spanish word chocolate also occurs in the Motul dictionary of Yukateko as the gloss for <chacau haa> ; this work is thought to have been produced around 1590.

This evidence also shows that in the Spanish of central Mexico, chocolate was called cacao until well into the seventeenth century, and we know of no evidence of the word chocolate or chicolate being used there at that time, even though José de Acosta, who lived in both Mexico and Peru, was using the word chocolate by 1590. We must suppose that his usage in Spanish was simply different, for reasons that we are not at the moment able to determine, though his place of writing, his place of origin, and his social-group affiliation are all possibly relevant. Its use in the Motul dictionary suggests that chocolate may have been used by 1590 in Spanish in the Yucatan.

The use by Hernández of Nawa chokola:tl (spelled <chocollatl>) shows that the word existed in Nawa by 1577, but it does not show where in the Nawa-speaking world it was used. Hernández collected information in various parts of Mexico, as well as in Peru and the Philippines. The fact that the next known citation of this Nawa word is in Clavijero (Reference Clavijero1780) shows that there are serious gaps in our documentation of this Nawa word.

Epigraphic Mayan Evidence about Cacao, Pataxte, and Chocolate

As in the case of colonial Nawa, several Mayan languages use their word for cacao as a word for the drinks made of cacao. This is true in Epigraphic Mayan as well. This section explores the linguistic expressions used in Classic Mayan texts to refer to such drinks. Our conclusions are in general agreement with Stuart (Reference Stuart and McNeil2006), in part because both were completed after his 2005 workbook and in part because Justeson and Stuart discussed many of the issues before our respective papers were written. We differ on the interpretation of a few individual terms in the passages to be discussed, but mainly in our grammatical analyses and correlated issues involving relationships among ingredient terms. Our interpretation of elements of the phrase makes use of mostly the same data as that cited by Stuart (Reference Stuart2005) in his workbook, although we originally extracted it from Mora-Marín's (2003) database.Footnote ²

The key sign that represents ‘cacao’ is a depiction of a fish, sometimes rendered as the head of a fish with fins at the back. It has long been known that this sign represents the syllable /ka/ in some contexts, in which it may be transcribed as <ka₂> (<ka> is the transcription of a separate syllabogram). Stuart (Reference Stuart1988) demonstrated that this sign is regularly used in spellings of the word for cacao in the Primary Standard Sequence (PSS) on Classic Mayan vases. The attested spellings show the following pattern of variation, ordered by their approximate relative frequencies in Mora-Marín's (2003) database:

• 80% <ka₂-wa>

• 10% <ka-ka₂-wa>

• 5% <ka₂>

• 5% <²ka₂-wa>

The spellings <ka₂-wa> and <ka₂> show that this sign is generally treated as a logogram for cacao in these contexts; we transcribe it here as <CACAO>. The spelling <ka-ka₂-wa> can be treated as a fully phonetic spelling or as a logographic spelling with phonetic complements, <ka-CACAO-wa>. The spelling <²ka₂-wa> shows that Mayan scribes made use of the sign in two logically distinct values in this context. Similar variation is found in other logosyllabic writing systems.

These spellings of the word for cacao almost always occur in a noun phrase within the part of the PSS that refers to the drinking vessel of a prominent person, used for some kind of cacao. This phrase usually has the following formulaic structure:

which should be translated as ‘NAME's drinking vessel for (a drink made of) NOUN₁ (and NOUN₂ …)’.

The word kakaw is usually the head of the noun phrase specifying what the drink was made of. Because Epigraphic Mayan had no regular way to express “and”, this meaning and parsing mostly has to be supplied by the reader. Grammatically, a modifier of an expression parsed as a conjunction of two phrases can apply only to the first of those phrases.

About 10% of these phrases consist of a mere noun phrase rather than a prepositional phrase. About 5% consist of a string of two or more prepositional phrases; each of these prepositional phrases applies to y-uk'-ib' but cannot qualify one another. The word kakaw is to be understood as the head only of the prepositional phrase that it ends and not of any preceding or following prepositional phrase.

We have investigated the structure of references to drinks made from cacao in Mayan hieroglyphic texts, using as our corpus mostly the texts transcribed by Mora-Marín. To anticipate our conclusions, these drinks are referred to by their ingredients. The most commonly mentioned ingredients are (1) cacao (<CACAO-wa>, etc. /kakaw/); (2) “tree cacao” (e.g., <TREE-(7e-)le CACAO-wa> /te7-e:l kakaw/); and (3) maize (<MAIZE>, <7i-MAIZE> /7ixi:m/). In some cases, the drink itself may be referred to as kakaw.

If we restrict our attention to those texts that make explicit reference to cacao, the following observations summarize these occurrences:

(1) In one case (on K2777), the spelling of the word for cacao is used with a final syllabogram <la>, which spells a suffix {-a:l}, seemingly reflecting an adjectival use as a modifier of a subsequent word (Stuart Reference Stuart and McNeil2006:191–192).

(2) In all other PSS texts, the word for cacao, when present, is the final word in the phrase referring to the contents of the vessel. In about 10% of cases, it is not preceded by a modifier but immediately follows the word y-uk'.ib' ‘his drinking vessel’. Typically, it is preceded by one or two modifying nouns or adjectives. In about 8% of cases, three or more modifiers precede it.

(3) With one exception, there is no statistically significant tendency for any pair of modifiers of ‘cacao’ to occur together. The exception is that there is a strong tendency for a word spelled by a sign <MAIZE>, depicting the head of the young maize god, to occur in the same noun phrase as te:7 or te7-e:l (spelled <TREE > , <TREE-le>, or <TREE-7e-le>). In about 80% of the instances in which one of these occurs, the other occurs as well, and the order is always with ‘maize’ before te7-e:l. Often they occur together in the same glyph block. It seems that they are (or might be) part of a single expression.

Martin (Reference Martin and McNeil2006) provides an interpretation of this association in terms of a mythological complex that he identifies and analyzes through its iconography. One key piece of this iconographic complex is a set of depictions of a cacao tree bearing ears of corn. Martin suggests that this tree is named by the glyphic sequence <MAIZE TREE> in the noun phrases referring to the contents of drinking vessels in the PSS. Stuart (Reference Stuart and McNeil2006:197) proposes that <MAIZE> is to be read as ixim (transcribed by us 7ixi:m) based on two cases in which the sign is preceded by <7i> (K791, K8764). Both Martin and Stuart read <MAIZE TREE> literally as <7iximte'> ‘maize tree’. (In our representation, this would be 7ixi:m=te:7.)

In our view, Martin has provided substantial evidence for the existence of an iconographic and mythic complex associating maize with cacao trees. However, he does not provide detailed argumentation for the reading as iximte' of the glyphic phrase discussed above. We show here that this reading is inconsistent with the range of variation in spellings of this phrase, and we provide an alternative interpretation that is consistent with all of the data known to us. Before doing so, however, it is relevant to point out that the expression 7ixi:m=te:7 is widely found in Mayan languages of the highlands, sometimes referring to the breadnut (Spanish ramón)—a fruit that is eaten in times when the maize harvest is poor—and otherwise to a large, very tall tree that grows in the rain forest. (In most Mayan languages the breadnut is called by a form descending from proto-Mayan *7ojx.) Among Lowland languages, it is found only in late colonial Yukateko where it refers to a kind of bush; and in modern Ch'orti7, where it refers to a wild tree, Karwinskia calendroni, whose leaves are used medicinally (Stuart Reference Stuart and McNeil2006:198–199, citing a personal communication from Johanna Kufer). In existing Mayan languages, these terms do not name a mythological “maize tree” or define “maize tree cacao” as a variety of cacao. The most we can squeeze out of all this is that maize is a highly evocative plant and has spawned several plant names that are morphologically the same but do not otherwise refer to biologically related plants.

In spite of their strong statistical association and consistent word order, the two terms ‘maize’ and te7-e:l appear to have separate relevance to the chocolate drink. This is certainly the case in a text in which the two seem to occur separately as modifiers of ‘cacao’ <yu-k'i-b'i MAIZE CACAO-wa TREE-le CACAO-wa> (K5857)—that is, in the expression “his drinking vessel [for] maize (and) [plain/default] cacao (and) tree-type cacao”. Their separate relevance is further suggested by the fact that ‘maize’ and either te:7 or te7-e:l each occurs about 20% of the time without the other when ‘cacao’ follows. These examples do not support an interpretation of <MAIZE TREE> or <MAIZE TREE-le> as spelling a single (compound) lexical item.

(4) In 90% of the texts mentioning cacao, there are noun and/or adjective modifiers before that word. The norm in these cases is for the string of modifiers to begin with a preposition, indicating that the vessel is “for” a drink made with the stated type of cacao. In case there are no modifiers, the norm is for no preposition to occur (the word ‘cacao’ is directly preceded by the preposition in a few cases, although only in a minority of instances in which this word occurs without modifiers).

(5) In a few instances, more than one phrase or term is preceded by a preposition (the word <yu-ta-la> or <yu-ta> is preceded by one of these prepositions in most of these cases, and kakaw is involved in about half of them):

In these situations, the prepositional phrases are separate statements, and cannot be conjoined to modify the word ‘cacao’. The same thing occurs when the word ‘cacao’ does not appear in the text, including cases that are followed by nominal phrases and so do not involve truncation of the portion of the text that names the contents, as in:

Details of sign execution and text content suggest that these texts, along with K7459, are not independent cases.

(6) With this background, it is possible to arrive at a straightforward interpretation of the meanings of many of these expressions that is also consistent with the detailed accounts of cacao and cacao-based drinks from colonial Nawa sources.

The term kwaw=kakawa-tl “tree cacao” refers to a particular variety of the cacao plant mentioned by Hernández for colonial Nawas (see “The Timing of the Diffusion of chikol=a:-tl”, above). In fact, all of the Nawa terms in which a word for a general class of plants (not specific plants) is used as a modifier of kakawa-tl were used to distinguish varieties of cacao. “Tree cacao” is also known in Sokean: kuy kakawa7 in Tecpatan Soke refers to the cacaté “tipo de fruta como nuez silvestre que es amarga [a kind of fruit like a wild (wal)nut which is bitter]” (the fruit resembles an almond, and it is toasted, salted, and eaten), while in Soteapan Gulf Sokean, it refers to cacao. We take te:7 kakaw and te7-e:l kakaw to have been a variety of cacao plant, “tree cacao”, and we interpret almost all cases of te7-e:l in these texts as meaning “tree-type”, even if the head of the construction has been left out.

A viable alternative interpretation entertained by Mora-Marín (Reference Mora-Marín2003:Figure 1) is that te7-e:l kakaw refers to ‘forest (i.e., uncultivated) cacao’. The comparative support makes “tree cacao” seem to us a more attractive hypothesis, but in either case, the term would designate a kind of cacao plant, not a kind of drink.

Since half of the vessels in which kakaw takes a modifier use the term te7-e:l kakaw to refer to the contents of these vessels, and no frequent modifier of kakaw is in complementary distribution with it, this must have been a common type of cacao for use in these vessels. It could have been the type of cacao that was implicit when the word kakaw was used without te:7 or te7-e:l as a modifier, but it is also possible that kakaw and te7-e:l kakaw labelled contrasting varieties of Theobroma cacao. Further evidence for this interpretation is provided below, under point (7).

It should be noted that the word kakaw may have referred, and probably did refer, to a range of related varieties of T. cacao. It was no doubt possible to make these varieties explicit, but readers would have had the knowledge to infer what varieties were intended as default interpretations of these terms in particular contexts. We today lack this knowledge, so certain types of information will not be readily accessible to us.

Nawa sources have no word that literally translates as “maize cacao” or the like. However, in these sources, all of the terms in which words for foods or food ingredients were used as modifiers of kakawa-tl were used to refer to drinks made of cacao that used those ingredients. We suggest that the use of the word for maize was such a usage. A similar case is nal kakaw, attested in the phrase <yu-k'i-b'i TA-?? na-la CACAO-wa> on an Early Classic vessel (Stuart Reference Stuart and McNeil2006:192; Mora-Marín, personal communication 2005, interprets some of the MAIZE logograms as spelling nal).

As in the case of te7-e:l, half of the vessels in which kakaw takes a modifer have MAIZE in these phrases. This suggests that cacao mixed with maize was a very common type used in these vessels. Most of the drinks made from cacao by the proto-historical Nawas were prepared from ground-up roasted cacao kernels combined either with maize dough or with ground-up roasted maize kernels. This includes most of the cacao-based drinks distinguished terminologically by Nawas that were discussed in the previous section. This practice continues to the present day in indigenous communities in Mesoamerica, including among Mayans, Mije-Sokeans, and Nawas, and was doubtless general in proto-historical Mesoamerica.

It may be noted that 7ixi:m is the only one of the frequent terms in these phrases that precedes kakaw and that does not occur with a -V:l suffix. All of the other terms show variation (although in the case of tzi:h ~ tzih-i:l, the form with a -V:l suffix is rare).

Under the interpretations presented here, the phrase <yu-k'i-b'i ta MAIZE TREE(-le) CACAO(-wa)> can be read y-uk'.ib' ta 7ixi:m [and] te7-e:l kakaw and can be translated, ‘his drinking vessel for [a drink made from] maize [and] tree cacao’.

The order of these terms is consistent in always placing all ingredient terms before the only recognized plant-type modifiers.

Color terms were used among the Nawas in names of types of cacao and in names of other kinds of trees that include the root kakawa. Sahagún mentions that some types of cacao plants were distinguished by the colors of their fruit. On these analogies, <k'an-na CACAO-wa> seems more likely to be spelling k'an kakaw (‘yellow cacao’ or ‘ripe cacao’) and less likely k'a:n kakaw ‘prized cacao’) in <ta yu-ta k'an-na CACAO-wa> (K625), and the expression might refer either to a variety of cacao from which a drink is made, or it could be a general modifier of an ingredient, describing the state (ripe) of the cacao kernel.

(7) Two other words that occur in phrases referring to the contents of the drinking vessels are <tzi>~<tzi-hi>~<tzi-hi-li>, and <yu-ta_i(-la)>. MacLeod and Reents-Budet (Reference MacLeod and Reents-Budet1994:118) identify this word with pre-Cholan *tzi:h (which, however, means ‘unripe, uncooked/raw’ rather than ‘fresh’, as she translates it). This appears to be a feasible interpretation, and we know of no other root whose distribution in the Mayan family suggests that it could be cognate with a Lowland Mayan form consistent with the phonetic spellings. Unlike the other terms considered so far (with the exception of k'an if for ‘ripe’ rather than ‘yellow’), the word tzi:h refers neither to a variety of the cacao plant nor to an ingredient in the drink. In the frequent phrase tzi:h(-i:l) kakaw ‘raw cacao’, the word kakaw must refer to the cacao kernel or pulp—thus, to cacao as an ingredient and not as the name for a drink.

Stuart (Reference Stuart and McNeil2006:188) notes that <yu-ta_i(-la)> occurs in a possessed form <7u-yu-ta-la> on a carved vase illustrated by Dütting (Reference Dütting1992:Figure 17); another possible instance is on K8088, which seems to be <7u-yu-ta₂-la>. This shows that we are dealing here with a noun yut (presence of vowel length unverifiable). We know of no prior interpretation of this term that is consistent with Stuart's observation. We raise the possibility that <yu-ta_i(-la)> may imply the ability to confer good fortune. In colonial Yukateko, <yut> or <yutal> (whose possessed form is <u yutal>) names the bezoar, a stone found in the stomachs of some deer when they are butchered. This type of stone is believed to confer good luck on its owner even by present-day Mesoamerican Indians. If this is the correct connection for yut(-a:l), the word may imply either the power of the bezoar or the use of it in the preparation of the cacao drink. From descriptions provided to Kaufman by some present-day Indians, the bezoar is not really a stone but more like a hair ball and is made of organic matter. Others say it is indeed a stone.

Further progress in the interpretation of these words depends on distributional analysis in terms of other words that appear in the same phrases. Such analysis is complicated by the fact that there is a general absence of multiple terms within the phrases under discussion, apart from 7ixi:m and te7-e:l.

In 75–80% of the texts in which two or more terms precede the word for cacao, 7ixi:m and te7-e:l are among them. Only 10% of texts that mention cacao have two or more added terms when 7ixi:m and te7-e:l are not both among them. Since it is rare for more than two added terms to occur together, except in the case of 7ixi:m and te7-e:l with one another, the occurrence of any one term is usually negatively correlated with the occurrence of any other. As a result, most of these terms can tell us little or nothing about any uninterpreted terms in these phrases.

This statistical pattern, however, may explain one seeming anomaly in the data: 80% of references to te:7 kakaw and te7-e:l kakaw are preceded by 7ixi:m as a modifier, but only about 15% of references to kakaw are preceded by it. Apart from 7ixi:m and te:7 ~ te7-e:l, the presence of one modifier reduces the likelihood of the use of any other modifier, so just the opposite effect would be expected. Accordingly, te7-e:l kakaw is more strongly associated with a specification of maize as an ingredient than is kakaw. It is known that maize was typically mixed with kakaw, so this cannot be the reason for the difference; rather, it is presumably due to a far greater susceptibility of kakaw than of te7-e:l kakaw to the use of tzi:h and yut(-a:l). This supports the view that kakaw and te7-e:l kakaw are names for two different varieties of T. cacao.

A more involved set of inferences shows that neither yut(-a:l) kakaw nor tzi:h kakaw refers to a variety of the cacao plant.

In several texts, the word tzi:h alone refers to the contents of the drinking vessel. In many of these texts, a reference to the owner follows, so that these expressions do not involve truncation, which sometimes occurs in the PSS when space is lacking for completing the text.

When tzi:h and yut(-a:l) occur in the same phrase, they can occur in either order. Otherwise, each is usually the first of the modifiers in any phrases in which it occurs that specify the contents of the vessel. The only exception is in K2323, in which ta 7ixi:m is followed by ta yut-a:l kakaw—that is, in which yut-a:l is the first element after the preposition in a prepositional phrase in the content phrase. The order of terms in the phrases specifying contents are therefore largely consistent overall and may be fully consistent within a prepositional phrase.

The order among adjectives is likely to relate to the semantics or relevance of these modifiers; in many languages, semantically defined classes of adjectives occur in a fixed sequence except when departures have a specific contextual motivation. The distributional properties of tzi:h(i:l) and yut(a:l) indicate that these words precede both ingredient and plant-type modifiers of kakaw. If ingredient modifiers precede plant-type modifiers, as suggested above, then tzi:h(-i:l) and yut(-a:l) are not plant-type modifiers. They could both be ingredient modifiers, or they could both be modifiers of a third category that itself precedes both ingredient and plant-type modifiers. Another likely example of such a modifier in the PSS is <7a-ch'a> for proto-Greater Tzeltalan *7ach' ‘new’ or pre-Ch'olan *7a:ch' ‘wet’ in K8713 <ti 7a-ch'a ka₂-CACAO-wa> (Zender Reference Zender2002).

As with the other terms in these phrases, when the word tzi:h (-i:l) appears with kakaw, it is most often the only modifier of kakaw. When another modifier appears with tzi:h, it is almost always te7-e:l; yut(-a:l) does not occur in these cases. When tzi:h(-i:l) appears without kakaw, usually none of the other terms occurs with it, but te7(-e:l) or yut(-a:l) appears in about a third of these cases. In about a third of the phrases with te7-e:l but not 7ixi:m, tzi:h occurs before it; tzi:h rarely or never occurs when 7ixi:m appears without te7-e:l (and rarely accompanies the phrase 7ixi:m te7-e:l). One possible interpretation is that cacao was more often uncooked than was maize in these drinks. Another is that the descriptive terms associated with the word kakaw might affect a reader's default interpretation of which particular variety of cacao was being alluded to.

In contrast, yut(-a:l) appears at about the same rate before 7ixi:m kakaw, before te7-e:l kakaw, and before 7ixi:m te7-e:l kakaw—in each case, about once in four instances. This word yut(-a:l) has one somewhat puzzling but potentially revealing association. Sixteen (70%) of the 23 cacao vessels in Mora-Marín's database whose owner is labelled <7i-tz'a-ti> are among the 77 with yut(-a:l) as the first modifier, while only seven are among 124 that lack it. This association is statistically significant (p=0.001). In colonial Yukateko, the adjective <idzat> is attested, with the meanings ‘clever, crafty, wise’. There may be a connection between possession of a lucky charm and cleverness.

(8) Another possible modifier in the phrase specifying contents, each time preceding <CACAO-wa>, is <LORD TREE-le>~<LORD-wa TREE-le>~<LORD TREE>. The first word is most likely 7aja:w ‘lord’. It could form a compound with te7-e:l—thus, 7ajaw=te:7 or 7ajaw=te7-e:l. A Mayan etymon *7a:ja:w=tye:7 ‘white sapote (Casimiroa edulis)’ is known from all Central Mayan branches, but it names a highland fruit tree. The corresponding word in Yukateko has a different referent. We do not know whether white sapote is a plausible ingredient of any cacao drink.

Along with the Central Mayan *7a:ja:w=tye:7 ‘zapote blanco, matasano // white sapote (Casimiroa edulis)’, as attested in Tzeltal 7aja[w]=te7 and Tzotzil 7aj=te7, we have the Ch'ol form käk=te7=pa7 (“cacao tree [of] gully”) ‘zapote de agua (fruta parecida a la del zapote)’. The terminological connection between certain types of sapote with both 7aja:w and kakaw suggests that <LORD TREE> as an ingredient of a cacao drink might reflect the Central Mayan term *7a:ja:w=tye:7, even though it is a highland tree.

In any case, it is feasible to analyze te7-e:l as the same one that specifies a variety of cacao, and to treat <7AJAW-(wa)> as a general modifier meaning ‘lordly’. In all cases, <LORD> is the first element of the sequence, consistent with being a member of the same class as tzi:h(-i:l) and yut(-a:l).

(9) Another possible rare modifier is <k'in>, which appears at most once in the corpus, on K3472. The sign is not entirely clear, and Stuart (personal communication 2005) points out that parts of this vase have been repainted. The sign at issue occurs in a spelling <ta k'i-?k'in>. Repainting seems to have affected parts of the preceding and following glyph blocks, since the ubiquituous fine lines due to root damage do not cross thickly applied black paint outlining parts of some signs, but we see no evidence of repainting of the <ta k'i-?k'in> sign group. The identification of the uncertain sign as <k'in> is due to Mora-Marín (Reference Mora-Marín2003), who marks the transcription as uncertain. However, the surviving details appear to be consistent with <k'in> and not with any other sign that is known to take a <k'i> superfix. If the sign's identification is valid, it seems likely to spell the Lowland Mayan word *k'ihn ‘hot’. This would make it a general modifier appropriate to the drinks made with cacao. It could, however, simply represent the word *k'i:n ‘sun’, which would have to be a once conventional but no longer current usage as the name of a type of cacao plant or drink.

(10) Stuart (Reference Stuart and McNeil2006:194) identifies place names as another class of modifiers for ‘cacao’, suggesting that they indicate cacao from particular locations (so far, all in the eastern Peten). There are two clear examples and one likely example. One of these is the only modifier of kakaw in its phrase, and two of them immediately follow yuta:l and precede kakaw.

In summary, we distinguish three classes of terms that precede kakaw in the phrases describing the contents of the drinking vessels: a class whose members appear immediately before the word kakaw and specify a biological variety of cacao (III); a class of nouns whose members appear before biological variety labels and specify ingredients of the drink made with cacao (II); and a general class consisting of all other terms, all of them seemingly general types of modifiers (I). They can be charted as follows:

Only one of these modifiers seems to require that the ‘cacao’ referred to in the PSS is the drink itself rather that the ingredients—<k'in> k'ihn (if for ‘hot’ and not ‘sun’)—and this may not be a correct reading of the original text. Certain of these modifiers seem to require that the cacao be an ingredient rather than the drink: <tzi(-hi)> tzi:h ‘raw, uncooked’; probably place names (presuming that they are places where a particular kind of cacao comes from); and <k'an> (if it is not for k'a:n ‘prized’). Other mentions of cacao provide no evidence that shows whether kakaw in those instances refers to an ingredient or to a prepared drink.

Table 7 presents the attested combinations of kakaw with its associated terms.

Table 7. Epigraphic phrases referring to cacao and cacao drinks

^aTo EpM te7-e:l kakaw may be compared: (1) Yokot'an (Chontal Mayan) te7el käkäw ‘guoguo, gogo’=Salacia belizensis ‘spindle tree’ [family]; Fusanus Hippocratacea [family] – an exact match to EpM /te7-e:l kakaw/; (2) Tecpatan Chiapas Soke kuy=kakawa7 ‘cacaté (tipo de fruta como nuez silvestre que es amarga)=Oecopetalum mexicanum; and (3) Soteapan Gulf Sokean kuy kaakwa ‘cacao’, “cacahua”. Salacia belizensis and Oecopetalum mexicanum are different genera, but both bear the name “tree cacao” in some languages.

On a few cacao vessels (/y-uk'.ib'... kakaw/) we find the expression <JAGUAR × CACAO>, a superimposition of jaguar markings on the <CACAO> logogram, which we interpret as b'ahlam kakaw ‘pataxte’ on the basis of forms with this meaning in Q'eqchi7 and in Ayapa Gulf Sokean that literally mean “jaguar cacao”; see “The Mije-Sokean Hypothesis”, above. All examples are of 7ixi:m te7-e:l b'ahlam kakaw. We do not know whether te7-e:l b'ahlam kakaw is the Epigraphic Mayan word for pataxte or for a “tree-type” variety of pataxte. In either case, it may relate to the origin of the later Ch'olan term, *b'ahlam=te:7, for pataxte. All of the examples refer to the drinking vessel being for a mixture of maize with pataxte, but since the few examples known to us seem to come from a single scribe or scribal school, the lack of variation in these rare references to pataxte is probably misleading. Other examples with dark markings on the face or body of the CACAO logogram, which may also be intended to represent a word for pataxte, occur with a broader range of ingredients.

A Proposed Mayan Source for Nawa chokola:tl

Dávila Garibi (Reference Dávila Garibi1939) proposed that the Nawa word chokola:tl for the drink chocolate originated, in part, in a Mayan language. This proposal was accepted and elaborated by Coe and Coe (Reference Coe and Coe1996:118–119) with data from Yukateko and colonial Kaqchikel. It has been uncritically accepted by some scholars—for example, by Tedlock (Reference Tedlock2002:170)—but it is demonstrably false. It depends on a misunderstanding of the Kaqchikel sources and on a lack of understanding of the history of a Yukateko word meaning ‘hot’.

Yukateko ‘chocolate’=“hot water”. Coe and Coe (Reference Coe and Coe1996:118) cite and endorse Dávila Garibi's (1939) suggestion that Spaniards combined a putative “Maya” (Yukateko) <chocol> (meaning ‘hot’) + “Aztec” <atl> (meaning ‘water’) to form a new word for chocolate. The linguistic elements of their argument are (1) the fact that the colonial Yukateko term for the drink chocolate is attested in the earliest sources as <chacau haa>, an expression with the literal content “hot water”; (2) the erroneous belief that there existed a Yukateko form <chocol>* meaning ‘hot;’ (3) the speculation that there might have been a Yukateko expression <chocol haa>* with the same literal content and thus the same application as the attested <chacau haa>; and (4) the speculation that Spaniards who knew enough Yukateko and Nawatl substituted <atl> for <haa> in this hypothetical form to produce a Nawa neologism chokol=a:-tl. Finally (5), Coe and Coe (Reference Coe and Coe1996:118) cite a supposed K'ichee7 chokola'j ‘to drink chocolate together’ as somehow relating to this proposal, while admitting that how the terms could relate is unclear (in fact, this is really a Kaqchikel word, Coe and Coe's phonological representation of it is faulty, and it means ‘to do something in common’).

Coe and Coe (Reference Coe and Coe1996:pp) cite Miguel León-Portilla (personal communication to Coe and Coe) as considering this a “reasonable explanation”. Tedlock (Reference Tedlock2002:170) endorses Coe and Coe's opinion that the word chocolate incorporates a Mayan word chokol meaning ‘hot’. Nonetheless, the only factual element of this proposed etymology is the first point—namely, that there was a Yukateko form <chacau haa> ‘chocolate’ that is literally “hot water”.

There are several difficulties with this proposal, but the most important involves the supposed Mayan word chokol*. (Except as noted, data cited in this section are from Bricker et al. Reference Bricker, Yah and de Po'ot1998, for Yukateko; Ulrich and Ulrich Reference Ulrich and de Ulrich1976, for Mopan, Hofling with Tesucun Reference Hofling and Tesucún1997 for Itzaj; and Canger Reference Canger1969 for Lakantun.) Modern Yukateko has a word chokoh ~ chokow ‘hot’; underlyingly it is /chokow/, as demonstrated by the derived form /chokwil/ ‘heat’. This is in fact the same word as the Motul's <chacau>. Proto-Mayan *tiqaw ‘hot’ is reflected in every branch of the family but Wastekan. Sometime in the ancestry of proto-Yukatekan, the *t shifted to ch; this is a regular change in Yukatekan that occurred whenever *t (or *ty) was followed by *i or *e. It also underwent the Yukatekan shift of *q to k (which is incidentally found in all Mayan languages that migrated into the lowlands out of the Mayan homeland in highland Guatemala). It cannot be determined which change happened first. Sometime after the change of *t to ch, pre-Yukatekan *chiqaw or *chikaw underwent vowel assimilation, leading to proto-Yukatekan *chakaw. Short a before final w sporadically changes to o in some Mayan languages. Yukateko chokow and Itzaj chokoh arose in this way from earlier *chakaw, while the proto-Yukatekan vowels descend normally to Lakantun chäkäw and Mopan chäkäj. (Justeson Reference Justeson1985 notes that the shift of *w to j is also attested in Mopan käkäj from *kakaw. This may be a regular change, as no other disyllabic words surviving in Mopan have word-final w. Final w in this word also shifts to h in Yukateko and Itzaj, although it resurfaces as w in some prevocalic contexts.)

The timing of the development of chokow in Yukateko can be roughly determined through the occurrences of this word in the sources cited in Table 8. These citations suggest that forms like chokow were well established but still in variation with chakaw in the mid- to late eighteenth century, while there is no evidence for any form but chakaw in the seventeenth century. To this extent, then, a suggestion that Yukateko chokow might have had a role in the origin of the Nawa word chokola:tl would be anachronistic. As for chokol*, no such form exists in the meaning ‘hot’ in any Mayan language.

Table 8. Colonial citations of Yukateko forms descended from chakaw ‘hot’

The Yukatekan historical development behind chokow and chokoh is sufficient to eliminate this proposal, but there are other difficulties with it. One is its consequence that Nawa chokola:tl is a Nawatization of the Spanish word chocolate; we show in “Evidence concerning the History …”, above, that this is implausible. Another difficulty for Dávila Garibi's hypothesis is created by Dakin and Wichmann's (2000:62) plausible argument that the Nawa term was originally chikola:tl, thereby rendering inconsistent the comparison between the Mayan vowels and the Nawa vowels and rendering irrelevant the compared words themselves.

One feature of this proposal—the notion that the word for this Mesoamerican drink came from Spaniards and was then widely adopted by Nawas—is of at most minor concern. Such a process would provide a lexical distinction that did not fully or customarily exist in indigenous languages between cacao as a plant/kernel and the drinks that are made from it. There are many parallels, such as the adoption by English of Old French words for animals as food (beef, pork) in contrast to native words for animals on the hoof (cow, swine). Nonetheless, Coe and Coe (Reference Coe and Coe1996) attempt a rationale for Spaniards' creating a word for this drink, a word that by chance was then widely adopted by Mesoamericans (specifically by Nawas and only occasionally by other Mesoamericans). They note the existence of a Nawa word that could have such a meaning, citing Molina (Reference Molina1571) for Nawa <cacahuaatl> /kakawa=a:tl/ ‘cacao water’ (which they misspell as cacahuatl). They go on to suggest that this word was eliminated because the first two syllables remind one of caca, the Spanish word for ‘poop’ (Coe and Coe Reference Coe and Coe1996:119). This argument is invalid because Spanish borrowed from Nawa the words cacao, cacahuatal ‘cacao plantation’, and cacahuate ‘peanut’ (cited in the same connection in Dakin and Wichmann Reference Dakin and Wichmann2000:62a), all involving the same morpheme /kakawa/, as in /kakawa=a:tl/; and Spanish also borrowed unrelated words containing this sequence, such as cacalote ‘crow’.