Introduction

The form and function of antipassive and antipassive-like constructions in Mayan languages have been of formal and comparative interest to Mayanists for decades, notably since Thomas Smith-Stark's seminal paper in 1978 on “facts and fictions” about Mayan antipassives. This chapter builds on this tradition on two fronts: first, it revisits the descriptive data on the forms and functions of antipassive-related voice suffixes in Mayan languages to provide a more complete historical picture of how these morphemes evolved within the family. Second, it presents new synchronic data on Kaqchikel which track a merger in progress through a number of Kaqchikel dialects. In the dialects where this merger is taking place, a formerly robust morphological distinction is being neutralized, a process which is reflected in historical mergers elsewhere in the family. In a broader view, this type of investigation furthers our understanding of language change, and highlights some of the contributions that ongoing language documentation can make to historical linguistics.

Section 1 provides background information on the two markers that are the focus of this chapter and the various constructions in which they appear. Section 2 provides details on these markers and constructions in each branch of the Mayan language family, and discusses proposed reconstructions for the two markers in Proto-Mayan. Section 3 presents new data which demonstrate that a merger is taking place in some Kaqchikel dialects, and Section 4 concludes. Data on Kaqchikel are from the author's fieldwork unless otherwise cited, while data on other Mayan languages are assembled from the literature.

Forms

There are two verbal suffixes that consistently appear as markers of antipassive and antipassive-like constructions in Mayan languages. These have been reconstructed in Smith-Stark (1978: 179) as *-(V)w and *-(V)n, in Dayley (1983: 86–7) as *-w and *-(V)n, and in Kaufman (1986) as *-(o)w and *-(o-)an, which for the sake of convenience will be referred to here as *-(V)w and *-(V)n (see Section 2.8 for comments on these reconstructions). The forms of these two morphemes do not vary greatly among the Mayan languages that have them; as the parentheses imply, sometimes a vowel is present, and that vowel may be short or long (for example, -oon in Tz’utujil vs. -Vn in Sakapultek).

Introduction

In Harris (2013), I showed how one kind of metathesis in Batsbi originates in derived transitive verbs. Metathesis is a phonological process by which two segments change order, for example XYZ becomes ZYX (where Y may be null). The goal of the present chapter is to show that in the speech of younger speakers the process I previously described has spread to intransitive verbs derived with the intransitivizing suffix -al.

In this chapter, I describe relevant aspects of the structure of verbs in Batsbi and briefly review the origins of metathesis in derived transitive verbs (Sections 2 and 3). I present in Section 4 a variety of morphological types of verbs, showing that metathesis does not occur in these. In Section 5, I present the facts of metathesis in derived intransitives, and in Section 6, I argue that metathesis spread from transitives to intransitives. Section 7 contains a brief conclusion.

Basic verb morphology in Batsbi

Batsbi is a native name for the language that the Georgians call Tsova Tush. It is spoken in a single village and is severely endangered, with the youngest speakers now around sixty, with a handful of exceptions.

For the purposes of this chapter we need morphological information primarily about the verb. Verbs may take a lexically governed gender–number prefix, illustrated in the left-hand column of (1). In the examples, d- is the default gender–number marker and stands in for the others. Not all verbs with gender–number prefixes occur in doublets of the kind in (1), but such examples are effective in showing that these prefixes are lexically governed. (In these examples, -ar forms the masdar, a declinable deverbal noun.)

Gender–number affixes are glossed as “CM,” as they are traditionally known in the Caucasus as class markers. These markers indicate the gender and number of the subject of an intransitive (in the absolutive or ergative case, see below) or the direct object of a transitive.

Introduction

There are many examples of archaeological and linguistic evidence each corroborating hypotheses of the other discipline. Scholars have explored this relationship between archaeology and historical linguistics in the Americas, where much of what we know about diachronic language change prior to European contact has come from thorough documentation of daughter languages. This thorough documentation enables historical reconstructions in the absence of written documents for societies that had no known writing systems. Most of the Indigenous languages of North America fall into this category, including the languages of this chapter: Algonquian languages, Mobilian Jargon, and its lexifiers (for example, Natchez). Our knowledge of Algonquian language change stems from the documentation efforts of early scholars like Leonard Bloomfield (1946). Campbell and Kaufman (1976) determined that the archaeological civilization known as the Olmec included speakers of Mixe-Zoquean languages (see also Campbell 1997: 12). Aztec (Hill 2017), Mayan, and Andean archaeological findings have also been connected to known language families, and are discussed further in Section 2. For some reason that remains unexplained, the ancient archaeology of North America has received less public attention than have the ancient histories and archaeology of Mesoamerica and South America. However, some of the ancient civilizations of Mesoamerica and adjacent regions in southern North America—for example, Aztec, Maya, and Zapotec—were contemporaneous with the complex interaction spheres of North America, including the Adena, Hopewell, and the Cahokia (Townsend 2004: 12). While extensive valuable scholarship has been done on the historical linguistics of North America (Goddard 1996; Campbell 1997; Mithun 1999), and the archaeology of North America (Theler and Boszhardt 2003; Carr and Case 2005; Spence and Fryer 2005; McCoy et al. 2017), there has yet to be much discussion attempting to connect the two fields.

Denny (1989: 91) proposed that speakers of Algonquian languages were connected to the mound-builder archaeological sites in the Northeast of North America. He cites Goddard (Denny 1989: 92) as proposing Proto-Algonquian to have split approximately 1000–500 bce (Denny 1989: 91). This would place Proto- Algonquian speakers at the time of the Middle Woodland Hopewell. Drechsel (1997) has been one of the few to engage in this important discussion. He proposed that Mobilian Jargon, a trade language of North America, was originally spoken in pre-European contact situations (1997: 292).

Introduction

This chaper deals with two topics that have been amply researched by Lyle Campbell: the application of philology to address matters of interest in the historical linguistics of American Indian languages, including Mayan hieroglyphic writing (Campbell 1984, 1990); and the scrutiny of loanwords in order to assess relative chronologies of culture historical and linguistic developments, including evidence of past contacts between different ethnolinguistic groups (Campbell 1970, 1972, 1986, 2000; Campbell and Kaufman 1976; Justeson et al. 1985). More specifically, this chapter addresses the question of the *k(’) > *ch(’) shift in the Mayan languages, focusing on the problem of its chronology and attestations in Epigraphic Mayan, and the nature of scribal practices and their connection to linguistic ideologies.

This question has recently resurfaced: Law et al. (2014) have put forth a proposal for a late application of this shift among already differentiated Ch’olan and Tzeltalan languages, which they argue is attested in real time primarily from the beginning of Late Classic (600–900 ce) period. Their proposal would revise the previous model, based on both comparative linguistic and epigraphic evidence, that called for the application of the shift among speakers of a largely undifferentiated community of Greater Tzeltalan speakers, by the first half of the Late Preclassic (400 bce–250 ce) period, possibly by c.200–100 bce (Kaufman 1976; Kaufman and Norman 1984; Justeson et al. 1985; Justeson and Fox 1989; Kaufman and Justeson 2007), and more conservatively no later than c. 250-400 ce, roughly corresponding to the first half of the Early Classic period (c.200-600 ce), when spellings of ka-ka-wa for proto-Ch’olan (pCh’) *käkäw “cacao,” a loanword from Mixe-Zoquean, are first attested. That etymon presents the right conditions to have experienced the shift, and yet, as Campbell (2000: 5) has noted, it did not, demonstrating that the shift must have occurred prior to its earliest occurrences. Law et al.'s (2014) proposal would revise the timing of shift to between c.600 and 900 ce, and would require that we accept that scribes ignored three phonemic contrasts (*q vs. *k, *q’ vs. *k, and *nh vs. n).

This chapter assesses and critiques Law et al.'s (2014) proposal.

The syntax of Meskwaki (Algonquian; spoken in Iowa) is sensitive to grammatical relations such as subject, object, secondary object, and oblique, but word order is not used to distinguish subject from object. In other words, Meskwaki is an example of the type of language proposed by Mithun (1987) in which none of the six permutations of subject, verb, and object familiar from Greenberg (1966) can be identified as the basic word order. Instead, Meskwaki word order is sensitive to a template including positions specialized for discourse functions such as topic and focus. The word order template is largely flat, with no verb phrase (VP) constituent grouping together a verb and its direct object and excluding the subject. We here present a corpus-based study of a set of clauses in which word order might be predicted to play a role in distinguishing subjects from objects: clauses in which the subject and object are both marked as obviative, used for third persons peripheral to the discourse. We show below that even in this context word order is not determined by grammatical relations. Instead, word order is sensitive to the relative ranking of the two obviatives in the discourse, rather than functioning to indicate which argument is subject and which is object.

A secondary goal of the chapter is to emphasize the value of older written texts for language documentation. For Meskwaki there is an extraordinary corpus of nearly 27,000 pages of traditional narratives and ethnographic information, written during 1911–18 in the Meskwaki syllabary (Goddard 1996) by mostly monolingual speakers of Meskwaki for Smithsonian ethnologist Truman Michelson. This corpus is stored at the Smithsonian Institution's National Anthropological Archives (NAA). As discussed in Goddard (1990b), these texts constitute a remarkably accurate record of connected speech by fluent speakers who were skilled narrators: they preserve an oral culture on paper. For linguists, the NAA corpus is invaluable for discovering the discourse conditions influencing the order of constituents in Meskwaki, a language with extremely flexible word order, or for investigating the discourse-based opposition within third person known as obviation (Thomason 2003). Both obviation and word order are extensively discussed below.

Background

Bolivia, a country with more than 11 million inhabitants, has no less than thirty-six languages. A large number of these languages are only spoken by small ethnic groups and are becoming less and less used. Before the small Bolivian languages become extinct, it is important they are documented in reference books and recordings. For the Bolivians this is of cultural historical and cultural political importance, while for the international linguistics community such a description is important because it gives insight into issues such as language development and what happens when languages are in contact with each other.

Bolivia is not only one of the poorest countries of Latin America, it is also a country with a huge cultural and linguistic diversity. In addition to Spanish, the thirty-six indigenous languages are or were recently spoken, with speaker numbers ranging from several million to less than five (Crevels and Muysken 2009, 2012, 2014, 2015). There are not only many languages, but these languages also belong to many different, genetically completely unrelated families. So far, geographical obstacles, the lack of roads, and the isolation of large parts of Bolivia have led to the preservation of parts of the cultural and linguistic heritage. Nonetheless, the majority of the languages of Bolivia are critically endangered. It is expected that only 10‒20% will survive in the next century, and in fact all indigenous languages are under pressure or severe threat in the country (Crevels 2012).

Despite the great cultural and linguistic wealth, the knowledge of Bolivian languages and cultures was very limited until recently. There is no national tradition of linguistic research, there are hardly any trained Bolivian linguists, and native speakers of indigenous languages with training and interest in their own language belong mostly to the two largest groups, the Quechua and the Aymara. Most of the research was and is done by foreigners, and the results, in so far as they have been published, are hardly available in Bolivia itself, since they are mostly written in English. They contain, moreover, a lot of technical linguistic terminology. Unfortunately, while Western scholars often achieve excellent research results with data acquired in the Third World, the local population, and particularly the people who collaborated in the projects, seldom get to see any results.

This chapter builds the general theory of the UG–iconicity interface, which consists of two general principles, with the Functional Iconicity Complementation Hypothesis (FICH) defining when the interaction is activated, and the Uniform Structure Mapping Principle (USM), how this interaction is carried out. In general, if a functional void of UG prevents a semantic/conceptual relation from mapping to a structural relation properly and thus halts an otherwise well-formed process of clause-generation, iconicity is called in to help finish the task. Intricate interactions between UG and iconicity may happen under the regulation of the USM so that the solution remains UG-compliant (and therefore has a proper interpretation). The scarce literature on the UG–iconicity relation is reviewed to provide the context in which the interface theory in this book is positioned and evaluated. The specific theory of the UG–iconicity interface is also compared with familiar examples in biology for further validation.

A synthetic approach to long-lasting theoretical debates and the necessary condition for implementing it are discussed with concrete examples, including the very theory of the UG–iconicity interface. The popular role models of linguistics, biology and physics, are shown with certain general properties which are argued to deserve attention in linguistic research and theorization.

Categorially obscure words of relations, sometimes accompanied by semantic obscurity too, characteristically occur between the connected constituents regardless of the specific word-order setting of a language. This chapter investigates two kinds of such connectors, the conjunctions and the linkers. It is argued that the functional void of UG responsible for the grammatical behaviors of connectors is that UG cannot project any lexical item without a categorial specification. If a lexical item does not acquire a category for some independent reason, its consequential lack of syntactic representation leads to unique grammatical behavior dependent on the specific semantic relation it encodes. In the case of conjunctions, the symmetric nature of ‘and’ and ‘or’, triggering linear iconicity while restricted by the isomorphic implementation of the USM, results in representing conjuncts on parallel planes of which one is the “default” due to computational cost. It is this simultaneous symmetry–asymmetry enforced by the UG-iconicity interface that explains a large set of apparently self-contradictory traits of coordination. Linkers from Mandarin Chinese, Chamorro and Cantonese are analyzed in a similar manner.

The SVCs have a cluster of grammatical traits not derivable exclusively from UG, some of which have been misrepresented/misunderstood in the UG literature. The first functional void of UG is proposed which limits the lexicalization of a semantic relation R to only those cases where R connects first-order entities. All the characteristic properties of SVCs result from this functional void collaborating with the proposed Serial Verb Parameter (SVP) in the theory of the UG–iconicity interface. It also explains a wide range of related cross-linguistic facts, from phrasal SVCs to compounds, the parametric variations between Chinese and Gbe languages in (dis)allowing ditransitive V1, why the resultative SVC acts in a particular set of ways different from other types and the subtle disparities between head-initial and head-final SVCs, as well as the full range of variations among Kwa languages in object- and verb-fronting. The relation between the linearization pattern of UG and the iconicity-induced word order of SVCs is shown to pattern with the “high” and “low” neural pathways underlying fear, with implications on the nature of redundancy in biologically based systems and suggestive of an evolutionary connection between these two linearization mechanisms in language.