Similarity has been cited, generally anecdotally, as a significant factor shaping the outcomes of language contact. A detailed investigation of long-term contact among more than a dozen related Lowland Mayan languages has yielded specific examples of contact-induced language changes that, I argue, were facilitated by the systematic similarities shared by these languages because of genetic relatedness. Three factors that seem to have been particularly relevant in the Mayan case are 1) the high degree of overlap in linguistic structure, which would have allowed significant interlingual conflation, the collapsing of language boundaries at points of similarity between the languages, 2) the paradigmatic interchangeability of particular elements of related languages without the need for adaptation or accommodation, which facilitated the borrowing of various kinds of linguistic material, particularly bound morphemes, that in other contexts have been found to be highly resistant to borrowing, and 3) contact-induced drift, parallel secondary developments in more than one language that were triggered by contact-induced innovations but subsequently proceeded along similar paths of change after contact because of the preexisting structural similarities that the languages shared as a result of their common inheritance. I argue that these processes of change are much less likely, if not impossible, in situations of contact between unrelated languages, and suggest specific ways in which contact between genetically related languages can be qualitatively different from contact between unrelated languages.
for the purposes of the present study, it is immaterial whether the two systems are ‘languages,’ ‘dialects of the same language’, or ‘varieties of the same dialect’…The mechanisms of interference [i.e. processes of contact-induced change] would appear to be the same whether the contact is between Chinese and French or between two subvarieties of English used by neighboring families (Weinreich, 1953: 1).
My own research on language contact phenomena involving related Mayan languages in the Lowlands of present day Guatemala, Southern Mexico and Belize, tells a different story, one in which the fact of genetic relatedness has had an enormous impact on both the degree of contact-induced language change in Mayan languages, and perhaps even the mechanisms through which those changes came about. In this article, I will describe three circumstances of contact-induced change involving Lowland Mayan languages and consider what we can infer about the processes of language change involved. I argue that, in each case, common linguistic inheritance has opened up pathways of linguistic change and interchange that would be much less likely to occur, or even impossible in situations of contact between unrelated languages.
Similarity in general has long been seen as a key factor in constraining or facilitating various outcomes of language contact. Scholars from Meillet, Jakobson and Weinreich, up to the present day have advanced the common-sense assumption that the transfer of linguistic elements between languages is likely to be easier if the languages have similar typological profiles or structural predispositions. In a review of this idea, Campbell (1993: 92) suggested that this ‘structural compatibility’ proposal “is aimed in the right direction, since in principle it ought to be easier to borrow constructions that are similar to existing ones (or at least do not conflict with the borrowing language’s basic structure) than structures that go against the typological grain of the borrowing language.” Campbell and others since (e.g. Matras, 2009: 221) have cautioned that this is only tenable as a general tendency not a universal rule. Still, even as a tendency, the notion is clearly relevant to situations of contact among related languages, in which the ‘facilitating’ potential of overlapping grammatical characteristics is at its extreme. The relevance of shared structure asserts itself repeatedly in the study of contact among Lowland Mayan languages and raises the question of whether the high degree of overlap between linguistic systems in contact has led to processes of contact-induced language change that are not found in typical situations of contact between unrelated languages.
Early scholars (e.g. Meillet, 1914) were of the opinion that certain types of grammatical transfer from one language to another were only possible between very similar languages, perhaps only dialects of a single language. In spite of this assumption, contact between related languages, and specifically how that relatedness might affect the processes and outcomes of language contact, has received relatively little attention in language contact research. Dialect contact studies (Trudgill, 1986, 2010; Kerswill, 2002) have generated rich empirical data and theoretical models of new dialect formation, but for the most part do not engage with theories of borrowability or mechanisms of linguistic transfer that dominate language contact studies. The lack of detailed research on contact between related languages may be, in part, a simple consequence of logistics: contact phenomena are much easier to recognize when they come from an unrelated language. They require less extensive data and labor-intensive analysis. Methodologically and empirically, studies of language contact between languages of the same family present unique challenges. They require detailed linguistic reconstructions, often spanning the length and breadth of the language family, in order to distinguish similarities due to contact from similarities due to common inheritance with any degree of confidence.
The present study considers contact phenomena that have affected at most a dozen related languages in the Mayan language family. Data from all 31 Mayan languages, including earlier documented stages of those languages, in Colonial manuscripts and ancient hieroglyphic texts, necessarily informed the work. I make use of data gathered during my own fieldwork on half a dozen Mayan languages, as well as archival and epigraphic sources, but the empirical core of the work is linguistic data gathered and published by over a dozen linguists in grammars over the last three decades. A cogent study of contact among Lowland Mayan languages would not have been possible without that rich corpus.
2. Overview of Mayan Languages and History
The Mayan language family includes around 31 languages spoken today in communities in Guatemala, Belize, and in the Mexican states of Chiapas, Tabasco, Campeche, Quintana Roo and Yucatan.1 Contemporary languages range in vitality from languages like Yukateko and K’iche’, with around a million speakers, to highly endangered languages like Itza’ and Mocho’ with elderly speakers numbering in the dozens. Two documented Mayan languages, Cholti and Chikomuseltek, have become extinct in the centuries since the Spanish Conquest. Fig. 1 shows the current geographical distribution of Mayan languages.
Lowland Mayan linguistic interaction took place in the historical context of the Classic Maya civilization, which thrived in the Lowlands of northern Guatemala between AD 250 and 900, after which warfare, drought, and social upheaval led to the abandonment of most Classic sites in the Southern Lowlands. During the Postclassic (AD 900-1520), unstable political relations predominated, and the political and cultural center of gravity, as well as apparently much of the population, shifted north to the area of Chichén Itzá, in the Yucatan. Hieroglyphic writing continued in use up until the time of the Spanish conquest, and some decades afterward, but the vast majority of surviving texts date to the Classic period and are written in a language that does not share many of the contact-induced innovations that have been identified in the modern Lowland languages. This suggests that the post-Classic period of social upheaval and more multilateral power relations is the likely impetus for much of the Lowland Mayan language contact phenomena, rather than Classic Maya language and culture.
Mayan languages are verb-initial, head-marking and mildly agglutinating. They follow a basic ergative pattern in person marking and show ergative traits in other grammatical domains as well, though most Mayan languages also have ergative splits in certain clause types, aspects or persons. Mayan languages have a productive arsenal of voice and valency-altering verbal suffixes. Voice alterations and other grammatical constructions are sensitive to animacy, definiteness and discourse organization. Proto-Mayan lexicon (Kaufman, 2003), phonemic inventory, and many aspects of syntax and morphology (Kaufman, 1990; Robertson, 1992) have been reconstructed. There is general agreement about the phylogenetic relationship of Mayan languages, though some points of disagreement remain, particularly regarding the placement of Wastek, Tojol-ab’al and the language of Maya hieroglyphs within the family. For an overview of recent research on Mayan historical linguistics, see Law (2013). The dominant model of the language family, proposed by Kaufman (1990), can be seen in Fig. 2.
Mayan languages have participated in a great deal of linguistic interaction. They are on the southern edge of the well-known Mesoamerican Linguistic Area (Campbell et al., 1986) and there is clear evidence of extensive linguistic exchange among different Mayan languages. Barrett (2002) and Campbell (1977) describe ongoing areal diffusion of several sound changes in Highland Guatemalan (K’iche’an, Mamean and Q’anjob’alan) languages. Of particular importance for this paper is the fact that Lowland Mayan languages have undergone particularly extensive grammatical transfer. Law (2009, 2011) presents evidence that contact among Lowland Mayan languages, including particularly Ch’olan and Yukatekan languages, has been exceptionally intensive, including borrowing of pronouns, aspectual affixes, and other morphological, syntactic and semantic elements. In the following pages, I will describe several of these linguistic outcomes in detail. I argue that the cross-linguistically unusual outcomes of language contact in the Lowlands are, at least in part, due to processes of contact-induced change that were available because of the inherited similarities that these languages share. These processes include structural conflations, paradigmatic interchangeability, and contact-induced drift.
3. Interlingual Conflation
A definitional difference between related languages and unrelated languages is that related languages share pervasive systematic similarities at various linguistic levels. A logical possibility that this opens up for bilingual speakers of related languages is that those languages can be conflated-the distinction between them can be collapsed-at those points of similarity. Bilinguals equate elements from one language with elements in another language in order to be able to establish commensurate meanings, what Weinreich referred to as interlingual identifications (1953: 7). When this process is carried out with related languages, the bilingual will find numerous points that are identical, allowing for the interlingual conflation of those identical elements.
Interlingual conflation is not limited to contact between related languages. Psycholinguistic and cognitive research (Paradis, 2004; Grosjean, 1998, 2001) suggests that bilinguals, regardless of the languages they speak, vary considerably both from person to person, and within a single individual over time with regard to the degree of cognitive separation that is maintained between linguistic systems. Ross’ (1996) process of metatypy might be considered a consequence of interlingual conflation with unrelated languages, and Matras (1998) argued for a similar process at a pragmatic level for various frequently borrowed ‘utterance modifiers’. Nevertheless, the potential for this process to impact the outcomes of language contact is arguably orders of magnitude greater when the languages involved are related.
3.1. Agent Focus
One example of how interlingual conflation has shaped the outcomes of contact in Lowland Mayan languages is the areally spread loss of what has been called the ‘agent focus antipassive’, or simply the ‘agent focus’. In several contemporary Mayan languages (and in proto-Mayan) focusing of arguments follows an ergative pattern. Transitive patients and intransitive subjects are extracted from the clause and moved to a focus position before the clause. Transitive agents are also fronted for focus, but they cannot be extracted from the main clause for focus without verbal derivation. The derivational morphology used varies somewhat from language to language but generally involves reflexes of either pM *-Vn or pM *-Vw, both of which reduce the valence of the main verb from two arguments to one. In example (1), from Q’anjob’al, the third person pronominal agent naq is in focus position preceding the main clause, and the verbs are derived with the agent focus suffix -(o)n.2
In this example, both arguments remain semantically associated with the verb, but syntactically, the derived verb is intransitive and only inflects with one person marker (In Q’anjob’al, the absolutive). In the languages of the Maya Lowlands, including members of the highly divergent Yukatekan and Ch’olan-Tseltalan subgroups, the agent focus construction has been lost, so that agents may be extracted to focus position without any verbal derivation.3 For example, in the following sentences, the agent of the transitive verb in the Yukatekan language Lakantun (in-na’ ‘my mother’, example 2), and the Ch’olan-Tseltalan language Tseltal (ch’o ‘mouse’, example 3), is in focus position, but the matrix verb has the same morphological form that it would have in a non-focussed structure.
Epigraphic linguistic data show that the Ch’olan language recorded in the vast majority of hieroglyphic texts maintained this structure for some time after Ch’olan and Tseltalan, not to mention Ch’olan-Tseltalan and Yukatekan languages had separated. In example (4), below, from the Classic Maya site of Copan (approx. AD 750), the matrix verb is derived with the suffix -(V)w and the third person pronominal agent haa’ is in focus position.
The use of the agent focus construction in the language of Maya hieroglyphs reinforces the hypothesis that the loss of the agent focus was an areal phenomenon. It was lost in perhaps a half dozen languages (which later differentiated into 10 languages). Within a single language, this innovation can be understood as the result of the regularization of a somewhat marked pattern. In proto-Mayan, all core arguments other than the agent of a transitive verb could be focused without verbal derivation, but focusing the agent required derivation. The loss of the agent focus made extracting arguments for focus consistent across argument types. The crucial point here is that transferring that regularized structure to another language presupposes an equation of the whole system of focusing. Bilinguals would have been able to conflate the languages’ identical constructions for focusing objects and subjects and may easily have assumed that the agent would be focused in the same way.
The main point for this paper is that the process through which the agent focus construction was lost, in contact with another Mayan language that lacked the constructions, will be quite different from the process through which such a structure would be lost in contact with, for example, Spanish. Spanish lacks the agent focus construction, but does not provide the same backdrop of interlingual conflations that Mayan languages provide one another because they were related. Barrett (2008) provides interesting indirect support for the relevance of that background of interlingual conflations in his sociolinguistic analysis of Spanish influence on dialects of Kaqchikel, a Mayan language that maintains obligatory agent focus construction. He found that in Kaqchikel dialects with heavy Spanish influence, verb-initial word order was infrequent and Agent-Verb-Object had become the primary word order. However, in spite of the fact that Spanish does not have anything like the Mayan agent focus, and this change affected the context in which the agent focus construction occurs, all dialects of Kaqchikel, regardless of the degree of contact with Spanish, maintain the agent focus structure. This is consistent with the proposal here: because Spanish and Kaqchikel do not share enough morphosyntactic structure, the agent focus, just one of a multitude of differences between the languages, was unlikely to be singled out and removed.
3.2. Numeral Classifiers
Another example of a contact-induced change that can be understood as deriving from an interlingual conflation of structures is the development of a complex system of numeral classifiers in several Lowland languages. Yukatekan, Cholan-Tseltalan, and, in a more restricted form, Greater Q’anjob’alan and Poqom (Santos Nicolás and Benito Pérez, 1998: 177) have a system of obligatory numeral classifiers that are used in conjunction with numbers and must agree with some feature (material, shape, configuration, texture, function, etc.) of the object being counted. These are particularly salient in Cholan-Tseltalan and Yukatekan languages, which have large and somewhat open inventories of numeral classifiers. For example, Keller (1955) identifies 78 numeral classifiers in Chontal, Bergqvist (2008: 77-79) numbers numeral classifiers in Northern Lakantun at around 90, and Tozzer (1921: 290-292) gives 80 for Colonial Yukatek. In his famous monograph on Tseltal numeral classifiers, Berlin (1968) identifies more than five hundred numeral classifiers. Arcos Lopez (2009), using a more refined categorization that distinguishes numeral classifiers from measure words, i.e. words that correspond to a unit of measure, identified 42 measures words and 140 numeral classifiers in Chol. Thus, even removing measure words from the count, the set of numeral classifiers in Chol is rather large. The only other Mayan languages with obligatory numeral classifiers, the Q’anjob’alan languages, have much smaller inventories. Some varieties of Q’anjob’alan reportedly have as many as 15, while others have as few as three numeral classifiers, one for humans, one for non-human animates, and a generic number suffix for all others.
Clearly both the use of a system of numeral classifiers and the development of such a large set of classifiers is an areal phenomenon particular to the Maya Lowlands. However, other Mayan languages have two similar features: a highly developed set of mensuratives, which follow numerals and have a similar function, and the use of a single generic number suffix of the form -Vb’ (-iib’, -ab’, or -eb’ ), used on numbers greater than one when a mensurative is not used. This generic number suffix occupies the same morphosyntactic slot as numeral classifiers in other languages. It is likely that these common features are the source from which the Lowland pattern of classification arose (see Kaufman, 1990: 97). In Q’anjob’alan languages, the suffix, -eb’, a reflex of the proto-Mayan generic number suffix, seems to be the base for the numeral classifier system, used for inanimates. In Tojol-ab’al, the -eb’ suffix is always used, even with measure words and a set of some 20 ‘numeral classifiers’, possibly mensuratives (Furbee-Losee, 1976: 118). In Tseltal and Tsotsil, -eb’ or -ib’ is still used productively for words that do not otherwise have numeral classifiers (Haviland 1981: 165; Polian, 2004: 140).
Once the system of numeral classifiers was established in a Mayan language, the process of transferring it into another language, as with the loss of the agent focus construction, would have been shaped by the preexisting shared structures. The development of numeral classifiers differs from the loss of the agent focus in that it is not a case of regularization or hypercorrection, but the transfer of a pattern of obligatorily categorization that derived from identical morphosyntactic strategies for dealing with numerals. The transfer of this pattern presupposes the conflation of those morphosyntactic possibilities, since by common inheritance, numerals in Mayan languages are not free morphemes, and may co-occur with more than just the generic suffix. In addition, in many of the Mayan languages with large inventories of numeral classifiers, those classifiers are derived from positional roots, a distinctive word class common in Mayan languages. Not only did the overlapping structural preconditions facilitate the spread of this form, but the shared inventory of word classes provided a readily transferrable source from which to coin new classifiers in each language.
4. Paradigmatic Interchangeability
Numeral classifiers provide a good transition point to consider another process of contact-induced language change that is particularly available in contact among related languages. Interlingual conflations, which can expand to cover more and more of the languages as they converge through contact, lead to morphosyntactic frames for particular morphemes that are identical. The adoption or spread of numeral classifiers led to convergence in structures for expressing quantification. These conflated structures, in turn, appear to have facilitated the direct borrowing of several specific numeral classifiers, since these languages now shared a paradigmatic position, a set of morphemes that occurred in a particular morphosyntactic frame with a particular function (Table 1).4
These numeral classifiers are bound morphemes, a morpheme type generally held to be highly resistant to borrowing (e.g. Matras, 2007) but because of the parallel morphosyntactic frame in which numeral classifiers occur in each language, the fact that these were bound morphemes was not an obstacle. Forms from a donor language could readily be deployed in a recipient language without phonological or morphosyntactic accommodation. In terms of the amount of work needed to incorporate these new morphemes into the recipient language, numeral classifiers would have been no more troublesome than lexical items. Because the syntagmatic context of a particular morphological paradigm or word class is identical in both donor and recipient, individual members of those paradigms or classes, even when not native to a language, can be straightforwardly deployed and are interchangeable with members of the class that are native to the recipient language. I argue that this paradigmatic interchangeability across languages facilitates borrowing of linguistic features that are, in other contexts, highly resistant to transfer.
Evidence of the ease with which normally transfer-resistant bound morphemes were able to circulate across language boundaries can be seen in the extensive borrowing of bound ergative and absolutive person markers described in Law (2009). Analysis there, and additional detail described in Law (2011: 180-193), identifies several waves of areally shared innovations in the semantic organization of the system of person marking. This research also found that in two subgroups, Cholan and Yukatekan, there was evidence that the actual phonological forms of more than half of both paradigms of person markers were exchanged through contact. Of the nine forms that were not borrowed, all but two were already identical because of common inheritance. One of the two innovative forms that was not borrowed, the first person prevocalic form kaw-, found in Ch’orti’ and Ch’olti’ was borrowed by two Yukatekan languages, Itza’ and Mopan, at some later point.
In the case of Yukatekan and Cholan person markers, structural similarities, both inherited and shared through contact, created identical morphosyntactic frames in both subgroups that facilitated later rounds of areal diffusion. Languages in both subgroups maintain the proto-Mayan distinction between ergative and absolutive person markers. As in proto-Mayan, Yukatekan and Cholan ergative person markers are verbal prefixes that have distinct prevocalic and preconsonantal allomorphs. Absolutive person markers in both subgroups differ from proto-Mayan in that the marker immediately follows the predicate regardless of aspect or predicate type. This similarity between Yukatekan and Cholan languages in terms of absolutive marking patterns is the result of an earlier round of contact that essentially re-aligned the morphosyntactic frames of the two subgroups after they had already diverged significantly. The end result of this combination of genetic inheritance and prior convergence was that the languages had virtually identical morphosyntactic frames. This structural isomorphism allowed person markers from one language to be interchangeable with corresponding person markers in the other language without the need for structural accommodation. While socially, such a move would likely have been significant, structurally, it was simple.
The concept of paradigmatic interchangeability casts Wichmann and Hull’s (2009) findings concerning loanwords and word classes in the Mayan language Q’eqchi’ in an interesting light. Among other things, they found that virtually all of the loanwords in Q’eqchi’ from Spanish were nouns (18.5% vs .2% verbs), while there was essentially no difference in the numbers of verb loans and noun loans in Q’eqchi’ from other Mayan languages (4.3% nouns and 4.3% verbs).5 Mayan languages are not at all isolating, and verbs are highly integrated morphosyntactically. They have obligatory inflections for person, highly productive voice and mood verbal suffixes, aspectual preclitics and an array of postverbal clitics. In addition, verbs in Mayan languages are overwhelmingly of the phonemic shape CVC, and all of the languages have separate sets of inflectional and derivational morphology for those rare verbs that have a different shape. Because the arsenal of verbal dressings is the same in both donor and recipient, that structural integration does not impede verbs from being borrowed just as frequently as nouns.
Chance similarities between unrelated languages or extensive convergence through contact may lead to paradigmatic interchangeability in certain linguistic domains, but in related languages, those interchangeable elements will be much more common. In some cases, like Mayan languages, this may lead to outcomes of language contact that can differ markedly from typical cross-linguistic tendencies.
5. Contact-induced Drift
A final way in which inherited similarity has shaped the outcomes of language contact in Lowland Mayan languages has to do with the shared linguistic structures that preceded individual instances of linguistic diffusion. At different points in the history of Lowland Mayan languages, it appears that areally shared innovations triggered the same series of developments in more than one language. If we trace the probable evolution of these innovations in each language, they appear to be logically independent developments. Because the preexisting structural conditions that led to those parallel outcomes were the same in the relevant languages, there is no need to infer contact between languages to explain the resulting similarities. In some cases, we can not establish conclusively whether a parallel series of developments in more than one related language is the result of contact or not. In other cases, the history of interaction (or lack thereof) between languages that share the innovation makes contact an unlikely explanation. In these cases, the languages developed in similar ways because their shared genetic inheritance predisposed development in those directions, something similar to Sapir’s famous notion of language ‘drift’ (Sapir, 1921: 171-178). Unlike Sapir’s concept, however, the impetus for these developments was a contact-induced innovation.
5.1 Innovation of the Dual Category
A good example of contact-induced drift is the innovation of a dual form in the Yukatekan language Lakantun and in the Q’anjob’alan languages Q’anjob’al, Akateko and Chuj. These are the only Mayan languages with a dual distinction in the first person plural. The first person ergative markers for the Yukatekan subgroup (Table 2) and the Q’anjob’alan subgroup (Table 3) can be seen below.
Not only do Lakantun and the Q’anjob’alan languages share the abstract grammatical category of dual, but the form with which they express it are all reflexes of the proto-Mayan first person plural ergative *qa- and absolutive *-o’ŋ. Since these languages are from very different subgroups (Yukatekan has been argued to be one of the earliest branches to separate from proto-Mayan), the similarity of both form and function for this term, which cannot be reconstructed to proto-Mayan, would seem to be the result of language contact. However, the geographical distribution of these languages and the known history of their speakers offer no evidence that these groups were ever in direct contact with one another. Q’anjob’alan languages are spoken primarily in the western highlands of present-day Guatemala. Lakantun is spoken in communities in the Selva Lacandona region of northeastern Chiapas, Mexico (see Fig. 1). But if we consider how the dual developed in these languages, and why it did not develop in other languages of the region, some interesting patterns emerge.
In addition to a dual category in the first person plural, Lakantun, Q’anjob’al, Akateko and Chuj all also make grammatical distinctions between inclusive and exclusive (as do several other Lowland Mayan languages). Those distinctions are innovative. The only semantic distinction in the first person in Proto-Mayan was between singular and plural. Inclusive and exclusive are expressed with the reflex of the proto-Mayan first person markers plus postclitics. The inclusive is marked with reflexes of the proto-Mayan first person plural *qa- /*-o’ŋ and the postclitics -hek (Q’anjob’al, Chuj) -wex (Akateko), and -(e)ex (Lakantun), all of which derive from a second person plural (the imperative, in the case of -hek, and the absolutive second person plural in the case of -wex and -(e)ex). The exclusive is marked in Q’anjob’alan with a reflex of the proto-Mayan first person plural, and the postclitics -(h)on (Q’anjob’al, Akateko), -hoŋ (Chuj), which are derived from the (still productive) first person plural absolutive in each language. In contexts that use the absolutive, this results in a sort of reduplication, though in some cases the position of each morpheme is different. Lakantun uses a combination of the first person singular person marker in(w)-/ -en, plus the generic third person plural suffix -o’/-oob’, also used to mark the third person plural.
The semantic logic behind the recruitment of these particular clitics is fairly transparent. The use of second person plural clitics for inclusive corresponds to the core meaning of that category: ‘us [and] you(pl)”. In the exclusive, the Q’anjob’alan strategy was an emphatic ‘us’, through reduplication, while the Lakantun strategy was to transparently express the category’s core meaning ‘me (and) them’ with the first person singular + third person plural. In terms of the development of the dual category, the underlying semantic logic of these marked forms, and particularly the inclusive, becomes important. The exclusive form would have restricted the meaning of the archaic first person plural to an inclusive meaning. As the inclusive form literally meant ‘me and you (pl)’, the unmarked form would have been further restricted to ‘me and you (sg)’, that is to say, dual. Thus, the dual meaning can be seen as a logical consequence of the development of marked forms for both inclusive and exclusive, and the recruitment, in Q’anjob’al, Akateko, Chuj and Lakantun, of an inclusive form with a transparent literal ‘you (pl)’ meaning.
The development of the dual can be seen as a natural, independent development in each language, with no need to motivate the parallel innovation with a historically implausible contact scenario. However, we must still explain the trigger innovations, the use of marked forms for both inclusive and exclusive and the use of a second person plural form in the inclusive, as well as the innovation of an inclusive/exclusive distinction in the first place. In fact, all of these triggering innovations have clear areal (and not genetic) distributions that extend far beyond the four languages discussed here, and include a large group of languages with a great deal of documented linguistic interaction (see Fig. 3).
Fig. 3 shows the probable geographic distribution of languages with an inclusive/exclusive distinction around the time of the Spanish Conquest. The languages form a large continuous swathe along the northern and western edges of the Maya region. Lakantun and the Q’anjob’alan languages are near opposite ends of this swathe. Two languages with an inclusive/exclusive distinction, Mam and Tektiteko, are markedly different from the other languages, and likely developed the distinction somewhat independently. Of the other languages with inclusive/exclusive, the Yukatekan languages Yukatek, Mopan and Itza’ only have a marked form (using a postclitic) for the inclusive. The remaining nine languages use postclitics to mark both inclusive and exclusive. This is summarized in Table 4, below.
Given the developmental scenario proposed above, it makes sense that the languages that do not mark the exclusive (Yukatek, Mopan and Itza’) would not have developed a dual category: rather than being restricted in meaning to ‘me and you (sg)’, the unmarked first person plural would have been restricted to a non-inclusive meaning, which is what happened in these languages, where the old unmarked form indicates exclusive. Of the nine languages that have a marked form for both inclusive and exclusive, four have developed a dual category, as previously discussed. The question, then, is why the other five languages (Tojol-ab’al, Tseltal, Tzotzil, Chol and Chontal) did not develop a dual category with the archaic first person plural. The fate of that archaic first person plural in these languages provides an intriguing clue. As the result of another areally diffused innovation, in all of these languages the old first person plural became unmarked for number and replaced *in- /*w- as the marker of first person singular. Because of this areally shared loss of the plural meaning for that form, no unmarked form remained to be restricted to a dual meaning, and that line of development was effectively blocked in those languages.
The scenario proposed above for the development of the dual distinction in two different, geographically separated, but genetically related languages incorporates two mechanisms of change that deserve comment. First, the dual depended on the preceding innovation of an inclusive/exclusive distinction marked with postclitics. This innovation, which triggered the subsequent innovation of a dual category, entered these languages through contact with other languages in the region. Second, the parallel consequences of that contact-induced change—the innovation of a dual category in Q’anjob’al, Akateko, Chuj and Lakantun—were parallel because the structural features that led to that outcome were identical in both languages as a result of their common genetic inheritance. This combination of factors is what I have here referred to as contact-induced drift.
5.2 Aspectual suffixes
The innovation of a dual category described above is striking because contact is not a viable explanation for the parallel development. Not all cases of contact-induced drift are so tidily discrete. There are several innovations that affected Lowland Mayan languages with clear histories of linguistic interaction, but whose details of development seem to derive from preexisting structural features of the languages, shared due to common inheritance. These innovations result from areally shared innovations, but do not themselves need to be explained with reference to contact. One example of this kind of parallel development is the innovation of a set of aspectual suffixes in Ch’olan and Yukatekan languages. In Proto-Mayan, and most contemporary Mayan languages, aspectual distinctions are marked before the verb with various prefixes, preclitics, and pre-verbal particles or auxiliary verbs, as can be seen in examples 5(a) and 5(b) below, from K’iche’, where completive and incompletive aspects are marked with the prefixes x- and k- respectively.
In Ch’olan and Yukatekan languages, however, in addition to these preverbal aspectual markers, and in some cases in their stead, there is a set of verbal suffixes that correspond roughly to completive and incompletive aspects.10 In the Cholan language Chontal for example, intransitive verbs use the suffix -e(l) for intransitive verbs in the incompletive (6a) and -i for intransitive verbs in the completive (6b), -tä(l) (incompletive) and -wän (completive) for positionals (7), and, as shown in example 8, for transitive verbs, -e’ (incompletive) and -i (completive). In Chontal, these suffixes, along with the person marker used, if the verb is intransitive, can be the sole indicator of aspect. Other languages use the suffixes in conjunction with the various aspectual prefixes and preverbal particles typical of Mayan languages.
The impetus for this development was the adoption of a particular type of ergative split that was shared through contact among Lowland Mayan languages (Law, In press). Lowland languages have a split in the pattern of person marking along aspectual lines, in which completive clauses follow the conservative ergative pattern, while incompletive clauses (and, in some languages other non-completive clauses) follow a nominative pattern. A subset of Lowland languages also adopted, through contact, the use of a historically non-finite verb form in split contexts, with the derivational suffix -Vl. In Yukatekan languages, this suffix was vowel-harmonic, matching the vowel of the verb, while in Cholan languages it was an invariant -el, likely the proto-Mayan shape of the suffix as well. The distribution of Lowland split-ergativity can be seen in Fig. 4. For a discussion of the development of this feature, see Bricker (1981), Law et al. (2006), and Law (In press).
In languages that adopted it, this second feature, the use of the -Vl derived non-finite verb form, was, I argue, the trigger that led to the development of a whole system of aspectual suffixes. Because each language involved inherited the crucial conditioning linguistic features from proto-Mayan, the extensive secondary consequences of adopting that infinitival verb form in a novel context played out in parallel ways. The most immediate of these was that because the new incompletive form was, at least historically non-finite, what had previously been a marker for finite intransitive verbs, pM *-ik (Ch’olan and Yukatekan -i), was restricted in its distribution to appear only in the completive aspect. Cognate suffixes exemplified in (9) for the Mamean language Teko (-ik), and in (10) for the Q’anjob’alan language Chuj (-i), as well as example 5, above, for K’iche’ (-ik), illustrate the fact that cognates of this suffix across the language family are independent of aspect.
Because the intransitive *-ik suffix, and reflexes, only inflected finite intransitive verb forms, it was in de facto complementary distribution with the non-finite *-el derivational suffix. When *-el became associated with the incompletive aspect, the *-ik suffix was restricted to completive contexts, as seen in examples 11 and 12, below, effectively producing a set of contrastive verbal suffixes for intransitive verbs associated with aspect.
The key point for this paper is that this restriction of the *-ik suffix to the completive aspect was an automatic and immediate consequence of the contact-induced use of *-el in the incompletive. The fact that all of the languages that adopted the novel, areally shared incompletive form also restricted the *-ik suffix to the completive is due to the structural predisposition common to all of these languages because they are all related. It is telling that the only Ch’olan language that did not adopt the non-finite verb form in the incompletive, Ch’orti’, also lacks aspectual suffixes, using the same verb form in both completive and incompletive (example 13).
The development of aspectual suffixes was not limited to intransitive verbs, however. All of the languages that developed a contrast between completive and incompletive for intransitives extended the contrast to transitive verbs (examples 14 and 15) as well.
For transitive verbs, the morphological elements recruited to fill this position vary, but because of common genetic inheritance, language internal pressures again led to parallel results. Because we have an established history of contact between the languages involved, it is perhaps impossible to entirely exclude contact as a factor in this series of develpments. However, in each language, internal structural factors alone are more than sufficient to motivate the innovation and in any account should not be ignored. The parallel developments that resulted from the areal spread of split ergativity were parallel because of the structural similarities that these languages shared prior to contact because of their common genetic inheritance.
These examples of contact-induced language change all attest to the important role of inherited similarity in conditioning the outcomes of language contact. The processes that I have highlighted here are not intended to be an exhaustive account of the ways that genetic relatedness can come into play in contact situations. In fact, these processes each have somewhat different relationships to more general processes of contact-induced language change. Of the processes discussed here, only contact-induced drift would arguably be entirely unique to contact between related languages. This is in part definitional, since particular tendencies of development that are not limited to a group of related languages would qualify as examples of grammaticalization, rather than drift (though this too has been argued to at times be triggered by contact - Heine and Kuteva, 2003, 2005).
As mentioned briefly in the first section, interlingual conflation - the collapsing of distinction between languages - at certain levels, has been argued to be an important process in several well known situations of convergence involving unrelated languages (Gumperz and Wilson, 1971; Ross, 1996, 1999; Matras, 1998). However, in these cases, interlingual conflation is the outcome of contact, not the process through which other contact-induced changes were facilitated. However, it is still the case that bilinguals can, and do, identify and conflate linguistic features that are shared, even when those similarities are due to chance, or the result of convergence through contact, rather than common inheritance. However, much like the difference between chance correspondences between lexical items (Yukatek /hol/ and English /hol/, for example, both of which mean a cavity or opening in something) such are arguably not nearly as pervasive as conflation points between closely related languages. The ‘relatedness effect’, with respect to interlingual conflation then would be a matter of much greater frequency of occurrence, as the process itself is not unique to contact between related languages.
A case can be made that this ‘relatedness effect’ can also be seen in instances of paradigmatic interchangeability, which facilitates the direct incorporation of foreign linguistic material because shared morphosyntactic and phonemic structural features eliminate the need for any sort of accommodation for the new form. As with interlingual conflation, there is no reason to expect that this process would be unavailable in contact between unrelated languages. However, in contact among related languages, I would argue that paradigmatic interchangeability is not only more frequent, but more importantly is more likely to apply at all levels. This can lead to cross-linguistically atypical linguistic outcomes because structurally integrated features that are unlikely to be shared through contact in other situations are no more resistant to transfer than other, less structurally embedded features. Thus, while cross-linguistically, verbs are less likely to be shared than nouns (Haspelmath, 2008), Wichman and Hull’s (2009) data for Q’eqchi’ show that this is not the case in this language when the donor language is Mayan as well. And while there is a global tendency in language to resist the transfer of bound morphology, numerous instances of borrowed bound morphemes have been identified in the Maya Lowlands. Thus, we can hypothesize that the effect of paradigmatic interchangeability is not just higher levels of borrowing, but also a change in the relative proportions of types of linguistic elements that are shared through contact.
The preceding discussion has presented evidence of three processes of contact-induced change in which genetic relatedness arguably played a crucial role. These proposed processes can account for a great deal of the cross-linguistically unusual grammatical borrowings that have been identified in Lowland Mayan languages (Law, 2009, 2011, forthcoming). The thrust of this analysis is that relatedness can have a real impact on the outcomes of linguistic contact, and on the processes that lead to those outcomes. Researchers who investigate such contact situations may benefit from particular attention to how that inherited similarity has influenced the processes and outcomes of contact.
BarrettRusty. 2002. The Huehuetenango Sprachbund and Mayan Language Standardization in Guatemala. In AndronisMaryDebenportErinPychaAnne and YoshimuraKeiko (eds.) Proceedings of the 38thChicago Linguistic Society: The Panels309–318. Chicago: Chicago Linguistics Society.
HaspelmathMartin. 2008. Loanword typology: Steps toward a systematic cross-linguistic study of lexical borrowability. In StolzThomasBakkerDik and Salas PalomoRosa (eds.) Aspects of language contact: New theoretical methodological and empirical findings with special focus on Romancisation processes43–62. Berlin: Mouton de Gruyter.
HullKerry. 2005. An Abbreviated Dictionary of Ch’orti’ Maya. A final report for the Foundation for the Advancement of Mesoamerican Studies Inc. Downloadable at http://www.famsi.org (accessed May 27 2013).
Kaufman Terrence (with John Justeson). 2003. A Preliminary Mayan Etymological Dictionary. Foundation for the Advancement of Mesoamerican Studies. Downloadable at http://www.famsi.org/reports/01050/index.html (accessed May 27 2013).
LawDanny. 2011. Linguistic Inheritance, Social Difference, and the Last Two Thousand Years of Contact Among Lowland Mayan Languages. Ph.D dissertation. The University of Texas at Austin.
Mateo ToledoEladio. 2008. The Family of Complex Predicates in Q’anjob’al (Maya); Their Syntax and Meaning. Ph.D dissertation University of Texas at Austin.
PolianGilles. 2004. Éléments de grammaire du Tseltal. Thèse de doctorat Université Paris III - Sorbonne Nouvelle.
RossMalcolm D. 1999. Exploring metatypy: how does contact-induced typological change come about? Keynote talk given at the Australian Linguistic Society’s annual Meeting Perth 1999. Downloadable at http://chl.anu.edu.au/linguistics/projects/mdr/Metatypy.pdf (accessed May 27 2013.
* The research on which this article is based was made possible by a Jacob Javits Fellowship from the U.S. Department of Education, and was written with support from a Colas Postdoctoral Fellowship at Vanderbilt University. The author would particularly like to thank Patience Epps, John Huehnergard and Na’ama Pat-El who organized the symposium on contact and genetic relatedness that was the impetus for this paper, as well as the symposium participants, particularly Claire Bowern, Marianne Mithun, Alex Magidow, and Craig Melchert, who offered valuable feedback on an earlier version of this paper during that symposium.
1 One Mayan language, Wastek, is spoken in a region hundreds of miles from any other Mayan language, in the Mexican state of San Luís Potosí.
2 Grammatical glosses use the following abbreviations: 1 = 1st person, 2 = 2nd person, 3 = 3rd person, ABS = absolutive, AF = agent focus, AP = antipassive, C = ‘set C’ person marker, CL = classifier, COMPL = completive, CT = Completive Transitive, DET = determiner, DIST = distal, EP = epenthetic segment, ERG = ergative, EXCL = exclusive, FOC = focus, INC = incompletive, INCL = inclusive, INTR = intransitive, NEG = negation, PL = plural, POS = positional, PROX = proximate, SG = singular, TOP = topic.
3 The ‘agent focus’ verbal derivation and accompanying syntactic structure may be used in these languages, but it is not obligatory. See Aissen (1999) for a discussion of this in Tsotsil and Norcliffe (2009) for Yukateko.
4 Abbreviated language names in Table 1 are as follows: tse = Tseltal, tso = Tsotsil, chl = Chol, chn = Chontal, yuk = Yukateko, lak = Lakantun, mop = Mopan, itz = Itza’, cht = Cholti’, toj = Tojol-ab’al.
5 Part of this disparity may be related to how verb ‘borrowing’ is defined in the study. Speakers of Mayan languages frequently incorporate Spanish verbal meanings through the adoption of a Spanish infinitive verb form that is the complement of a ‘light verb’, such as ‘to do’ or ‘to make’. As Wichmann and Hull’s data suggest, the direct importation of Spanish verbs as verbs, rather than infinitives is vanishingly rare in Mayan languages.
6 See Maxwell (1982: 137-140) for more details.
7 Note that Craig (1977: 109) has y- for this, but it must be a mistake, since she has j-ibanh ‘on top of us’ on page 110 of the same source.
8 Though note the use of Popti’ -an, mentioned in Craig (1977: 278).
9 The unmarked form in dialects without a dual/exclusive/inclusive distinction (including Ixcoy and Soloma, see Raymundo González et al., 2000: 55).
10 The exact range of aspects that followed a nominative pattern varies somewhat across languages. Most split-ergative Mayan languages have the split in at least the incompletive and progressive aspects. For clarity, I will here focus on the incompletive aspect. For details, see Law (2011: 246-252).