Rethinking Split Ergativity In Chol Author(s): Jessica Coon Source: International Journal of American Linguistics , Vol. 76, No. 2 (April 2010), pp. 207253 Published by: The University of Chicago Press Stable URL: https://www.jstor.org/stable/10.1086/652266 JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at https://about.jstor.org/terms The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to International Journal of American Linguistics This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms RETHINKING SPLIT ERGATIVITY IN CHOL1 Jessica Coon Massachusetts Institute of Technology All Mayan languages exhibit ergative–absolutive systems of agreement. However, Chol and some other members of the family display aspectually based splits: while perfective clauses show the expected ergative–absolutive alignment, imperfective and progressive (hereafter “nonperfective”) clauses show a nominative–accusative alignment (Larsen and Norman 1979, Bricker 1981, and Quizar and Knowles-Berry 1988). In addition to this difference in agreement, there are also important differences in stem morphology between perfective and nonperfective stems. I argue in this paper that nonperfective forms in Chol are formally possessed nominals. The main syntactic predicate is in fact the aspect marker, which shows absolutive agreement with its sole argument, the nominalized clause. Under this analysis, all predicates in Chol show an ergative–absolutive pattern of agreement. The illusion of split ergativity results from the nominalization of the notional predicate, along with the fact that ergative and genitive morphemes are identical in Mayan languages. This proposal builds on the suggestion in Larsen and Norman (1979) that the majority of splits in the Mayan family may be reduced to subordination. [Keywords: Chol, Mayan, split ergativity, morphology, nominalization] 1. Introduction. While all Mayan languages exhibit a basic ergative– absolutive pattern of agreement, about a third of the languages—including Chol—have aspectually based splits (Larsen and Norman 1979). As in other ergative languages with aspectual splits (see Dixon 1994), Chol shows an ergative–absolutive agreement pattern in clauses marked with perfective 1 I am especially grateful to my Chol consultants Virginia Martínez Vázquez, Matilde Vázquez Vázquez, Doriselma Gutiérrez Gutiérrez, and Nicolás Álvaro Arcos, without whom this work would not be possible. Special thanks to David Pesetsky, Roberto Zavala, Norvin Richards, John Haviland, Sabine Iatridou, Masha Polinsky, and two anonymous reviewers for reading and commenting on earlier versions of this paper. Thanks also to Judith Aissen, Marcel den Dikken, Patrick Grosz, Pedro Gutiérrez Sánchez, Hilda Koopman, Gilles Polian, Omer Preminger, Andrés Salanova, Kirill Shklovsky, Peter Svenonious, Juan Vázquez Álvarez, and audiences at M.I.T.’s Ergativity Research Seminar, the Mayan Symposium at SSILA 2008, WCCFL 27, and CIESAS-Sureste for helpful discussions on this and related material. It goes without saying that all remaining shortcomings are my own responsibility. The Chol data presented here were collected in Chiapas, Mexico, with support from M.I.T.’s Ken Hale Fund for Field Research, an NSF Dissertation Improvement Grant (#0816923), as well as research funds granted to Noam Chomsky by Ik-Hoon Lee. This material is based on work supported under a National Science Foundation Graduate Research Fellowship. The proposal presented here is extended in my M.I.T. dissertation, currently in progress. [IJAL, vol. 76, no. 2, April 2010, pp. 207–53] ç 2010 by The University of Chicago. All rights reserved. 0020–7071/2010/7602–0002$10.00 207 This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms 208 international journal of american linguistics aspect, but a nominative–accusative pattern in nonperfective (imperfective and progressive) clauses. This contrast is shown in (1)–(3). In the perfective constructions in (1) we observe that the intransitive subject shows the same agreement marking as the transitive object. The subject of the transitive clause patterns differently, resulting in an ergative–absolutive pattern. Here and throughout, I follow traditional Mayan literature in using the theory-neutral labels “set A” and “set B” to refer to the person morphemes, discussed in detail in 3 below.2 Perfectives (1a) Tyi prfv a-mek’-e-yoñ. a2-hug-tv-b1 ‘You hugged me’. (1b) Tyi prfv wäy-i-yoñ. sleep-itv-b1 ‘I slept’. Turning to the nonperfectives in (2) and (3), we find that the single argument of imperfective and progressive intransitives patterns with the subject of the transitive clauses: both take the set A marker. This results in what appears to be nominative–accusative pattern. Imperfectives (2a) Mi impf a-mek’-oñ. a2-hug-b1 ‘You hug me’. (2b) Mi impf a-wäy-el. a2-sleep-nml ‘You sleep’. 2 Unless otherwise noted, the data presented here are from my fieldnotes. Chol is written in a Spanish-based practical orthography. ~ = [?]; ä = [I] b = [∫]; ch = [tS]; j = [h]; ñ = [ˆ]; ty = [t3]; x = [S]; y = [j]; C’ = ejective consonant. Abbreviations in glosses are as follows: 1, 2, 3 = first, second, and third person; a = set A (ergative/genitive); abs = absolutive; aff = affirmative; apasv = antipassive; appl = applicative; b = set B (absolutive); caus = causative; dep = dependent suffix; erg = ergative; gen = genitive; itv = intransitive verb; loc = locative; nc = numeral classifier; nml = nominal suffix; pasv = passive; pl = plural; pos = positional suffix; prfv = perfective; prep = preposition; prog = progressive; rep = reportative; suf = suffix; tv = transitive verb. This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms split ergativity in chol 209 Progressives (3a) Choñkol prog a-mek’-oñ. a2-hug-b1 ‘You are hugging me’. (3b) Choñkol prog a-wäy-el. a2-sleep-nml ‘You are sleeping’. It is significant that Chol does not have dedicated nominative or accusative markers. Rather, we find the set A marker—reserved for transitive subjects in a strictly ergative system—used on intransitives as well. This type of pattern is called “extended ergativity” in the terminology of Dixon (1979) and does not exactly correspond to more familiar nominative–accusative patterns in which the nominative is unmarked. This will be relevant to the argument put forth below. A central claim of this paper is that Chol does not in fact show a split in the behavior of verbs. Rather, all main syntactic predicates exhibit an ergative–absolutive alignment pattern. It is argued that the stems in the nonperfective forms in (2) and (3)—amek’oñ and awäyel—are formally possessed nominals: set A marks both possessors and external subjects. Mi (imperfective) and choñkol (progressive) are argued to be the main syntactic predicates of their sentences. The appearance of split ergativity is the result of the nominalization of the notional predicate, along with the fact that ergative and genitive are identical in the Mayan language family. Throughout the paper I use the unqualified term “predicate” to refer to the syntactic predicate of a clause—that is, the element which takes nominal arguments. The claim is that mi and choñkol are the syntactic predicates of the forms in (2) and (3), and that the encyclopedic meaning (meaning from “open class” elements) of the clause is located in the nominalized stem forms. By “nominal” I mean the element which serves as the argument of the syntactic predicate, whether this be a bare noun like waj ‘tortilla’ or a nominalized clause like amek’oñ ‘your hugging me’ in (3a). Though referring to both bare nouns and nominalizations alike as “nominals” simplifies the differences in behavior between various types of nominalizations—some of which show more clause-like behavior than others (see Abney 1987 and Koptjevskaja-Tamm 1993 and the discussion in 4 below)—this terminology will be useful below. I begin in 2 with an overview of previous work on ergativity and split ergativity in the Mayan language family and then continue with a review of Chol basics in 3. In 4, I present evidence that the nonperfective stem forms This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms 210 international journal of american linguistics in examples like those in (2) and (3) are formally nominal, and then argue in 5 that the nonperfective aspect markers are predicates. In 6, I offer concluding remarks and discuss the possibility of extending this type of analysis to non-Mayan languages. 2. Ergativity and split ergativity in Mayan languages. 2.1. Mayan ergativity. The Mayan language family consists of about 30 languages spoken by over six million people located throughout Mesoamerica and northern Central America. Mayan languages may be divided into six subgroups: Huastecan, Yucatecan, Greater Tzeltalan, Greater Q’anjob’alan, Greater Mamean, and Greater K’ichean (Kaufman 1976). Despite significant grammatical diversity within the family, all Mayan languages exhibit ergative patterns of verbal inflection (Larsen and Norman 1979). Examples from a few different languages are given in (4)–(6).3 Kichee’ (K’ichean) (4a) x-at-u-ch’ay-oh comp-b2-a3-hit-fin ‘He hit you’. (4b) x-at-war-ik comp-b2-sleep-fin ‘You slept’. (4c) a-keej a2-horse ‘your horse’ (Larsen and Norman 1979:347) Akatek (Q’anjob’alan) (5a) chi-in-q’oj-ach ey-toj inc-a1-throw-b2 dir:down-dir:thither b’ey-W at-b3 ti’ prox an cl.1s ‘I’ll throw you down from here’. (5b) tol chi-ach-kam eyman so inc-b2-die quickly ‘So you die quickly’. 3 Glosses from languages other than Chol are in some cases slightly modified from those used by the original authors. Here asp = aspect; comp = completive; dir = directional; ds = directional suffix; fin = phrase-final suffix; inc = incompletive; prox = proximal demonstrative; s = singular. The numeral 7 represents a glottal stop. This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms split ergativity in chol (5c) in-chee a1-horse 211 an cl.1s ‘my horse’ (Zavala 1997:443–44) Mam (Mamean) (6a) ma asp chin b1 ok dir t-tzeeq’a-n-a a2-hit-ds-2s/1s ‘You hit me’. (6b) ma asp chin b’eet-a b1 walk-1s ‘I walked’. (England 1983a:2) As the above examples illustrate, grammatical relations in Mayan languages are head-marked on the predicate with two sets of morphemes, traditionally labeled “set A” and “set B.” The objects of transitive verbs in the (a) forms are marked the same as the single arguments of the intransitive verbs in the (b) forms—all are marked with a set B morpheme. Subjects of the transitives, in contrast, are marked with set A morphemes. This same set A morpheme is used to mark the possessors of nouns, as shown in (4c) and (5c). The set A and set B terminology is common in Mayan linguistics, where set A corresponds to ergative and genitive, and set B corresponds to absolutive. 2.2. Split ergativity. While all Mayan languages show a basic ergative– absolutive pattern of person marking, some exhibit what is known as split ergativity. That is, in certain contexts grammatical relations are marked following a nonergative pattern. The factors which condition splits in the world’s languages appear to be relatively consistent and include tense or aspect, nominal features (e.g., person, animate), semantics of the predicate, and the distinction between main and subordinate clauses (Comrie 1978 and Dixon 1994). Just one of the Mayan languages, Mocho ~ (also known as Motocintlec), exhibits a split conditioned by nominal features (Larsen and Norman 1979:353). Larsen and Norman note that in the rest of the Mayan family splits are triggered by three kinds of factors: (7a) Occurrence in subordinate clauses (7b) The presence of a focused constituent immediately preceding the verb (7c) Particular tenses or aspects Larsen and Norman (1979:353) In fact, they suggest, all three may be reduced to subordination. What Larsen and Norman call “pre-verbal focussed” constituents have been shown This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms 212 international journal of american linguistics in recent work to be complex predicate constructions, and thus also constitute cases of subordination (Mateo-Toledo 2003). That the aspects which trigger the apparent nominative–accusative system are also complex clause constructions will be the subject of the remainder of this paper. Larsen and Norman add that in all of these instances, the split involves the extension of the ergative (set A) morpheme to certain intransitive subjects. As discussed in 1 above, this pattern is what Dixon (1979) refers to as “extended ergativity.” That is, it is not the case that there are distinct nominative and accusative morphemes. Rather, the use of the ergative (set A) marker, is extended to mark subjects of certain intransitives. The argument put forth for Chol below lends further support to Larsen and Norman’s suggestion that all of the cases in (7) are instances of subordination (see also Bricker 1981 on Yucatec and Mateo-Toledo 2003 and Mateo Pedro 2009 for similar proposals for nonergative patterns in Q’anjob’al). Specifically, I argue below that the nonperfective aspectual morphemes in Chol serve as matrix predicates and take nominalized clauses as complements First, let us briefly examine each of the types of split in (7). 2.3. Splits and subordination. 2.3.1. Subordinated clauses. In languages of the Q’anjob’alan subgroup, such as Jakaltek, main clauses show the expected ergative–absolutive pattern of agreement, while subordinate clauses like those in (8) show a nominative–accusative pattern. In the subordinate clauses in (8), both subjects of transitives and subjects of intransitives are marked with the set A morpheme. Jakaltek (Q’anjob’alan) (8a) x-W-w-ilwe asp-b3-a1-try [hach b2 hin-kol-ni ] a1-help-suf ‘I tried to help you’. (8b) sab’ early ichi start [ha-munlayi ] a2-work ‘You started to work early’. (Craig 1977:617) This pattern is also discussed in Mateo-Toledo (2008) for Q’anjob’al. 2.3.2. Preverbal adverbs. In some languages of the Mamean and Q’anjob’alan subgroups, the appearance of certain adverbial elements before the verb triggers a nominative–accusative pattern. This can be seen in Ixil. In (9a), the intransitive verb wat ‘sleep’ appears with the set B morpheme expected of intransitives and the adverb jojli ‘face-down’ appears phrase-finally. In (9b), in contrast, the adverb appears preverbally and ‘sleep’ appears with a set A marker, normally reserved for transitive subjects. This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms split ergativity in chol 213 Ixil (Mamean) (9a) wat o7 sleep b1.pl jojli face.down ‘We slept face-down’. (9b) jojli face.down ku-wat-e7 a1.pl-sleep-suf ‘We sleep/slept face-down’. (Ayres 1983:39) In addition to the difference in person marking, the verb form in (9b) takes the suffix -e7 found in intransitive dependent clauses. While the person marking on transitives remains the same, the transitive verb root also appears with a special dependent suffix when the adverbial element appears clause initially (see Ayres 1983:40). As proposed by Larsen and Norman (1979), the appearance of the dependent marker on these verb forms suggests that the preverbal adverb in fact belongs to a higher clause; the lower verb is subordinated and marked as such. This type of split is then reducible to another instance of subordination: (9a) is a simple clause with an adverbial, while (9b) is a complex clause. Jojli in (9b) is not “focused” but is instead serving as the matrix predicate, embedding the predicate ‘sleep’. Mateo-Toledo (2003) and Mateo Pedro (2009) provide a similar analysis of analogous Q’anjobal constructions discussed further below; see also Pascual (2007). 2.3.3. Aspect. Finally, we turn to aspect-based splits, the focus of this paper. Aspect-based splits are found in languages of the Yucatecan group, in the Cholan branch of the Greater Tzeltalan group, as well as in Ixil (Mamean) and Poqomam (K’ichean) (Larsen and Norman 1979). In all of these languages, an ergative–absolutive pattern is found in perfective or completive aspects, while nominative–accusative patterns are found in nonperfective or noncompletive aspects. This is shown for the Yucatecan language Mopan in (10) and (11). In (10) we find an ergative–absolutive pattern in the perfective forms—the intransitive subject takes the set B marker, also used to mark transitive objects. Progressive forms like those in (11), in contrast, show a nominative–accusative pattern. Here the subject of the intransitive patterns with the subject of the transitive in taking set A morphology. Mopan (Yucatecan) (10a) in-lox-aj-ech a1-hit-suf-b2 ‘I hit you’. This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms 214 international journal of american linguistics (10b) lub’-eech fall-b2 ‘You fell’. (11a) tan prog in-lox-ik-ech a1-hit-suf-b2 ‘I am hitting you’. (11b) tan prog a-lub’-ul a2-fall-suf ‘You are falling’. (Larsen and Norman 1979:353–54) Again, the nominative–accusative patterning forms in (11) may be connected to subordination. Note that just as with the splits involving adverbial elements discussed for Ixil, the Mopan verb roots in (11) appear with different suffixes (-ik and -ul ) from the ergative-patterning forms in (10). Furthermore, while no overt tense or aspect marker appears on the perfective forms in (10), the progressive aspect is marked with tan. Larsen and Norman (1979:355) note that the tenses or aspects which condition a nominative–accusative pattern are always overtly marked. They note further that some of these morphemes may be historically traced to verb roots. They conclude that nominative– accusative constructions “are to be analysed diachronically as higher verbs with sentential subjects, that is, as instances of subordination.” This pattern is also discussed in Bricker (1981:85) for Yucatec. Drawing on historical evidence, Bricker notes that: “Evidence that the aspects that govern the nominative–accusative pattern of pronominal inflection are really the main verbs of complement constructions can be found in Classical Yucatec, where several of the aspectual ‘particles’ frequently appear as fully inflected auxiliary verbs.” Below I argue that this generalization is true for Chol, not just diachronically, but synchronically as well. That is, the imperfective and progressive aspect markers which trigger nominative–accusative patterning function as the main syntactic predicate of the clause, while the notional predicate is a subordinated nominal form. Larsen and Norman (1979) conclude (setting aside the person-based split in Mocho ~): “From the perspective of Mayan comparative grammar, to explain the nature of split case-marking it would be sufficient to account for why ergative (set A) prefixes are used to cross-reference intransitive subjects in subordinate clauses.” Noting that set A marks not just ergative but also genitive in Mayan languages, they speculate that intransitive verbs with set A subjects, such as the Mopan form in (11b), are in fact possessed nominals. This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms split ergativity in chol 215 While they do not discuss this in detail, the reason that Larsen and Norman propose that only intransitive forms are nominalized may be connected to the fact that only intransitives show overt nominal morphology in languages like Chol, or the fact that in order to account for the split it is only necessary to propose nominalizations for intransitives, since all transitives take both set A and set B marking and it is thus in the intransitives that the splits become apparent. Bricker (1981:87) notes that nominalization is “a plausible explanation for the ergative split in Yucatec Maya if only intransitive complements are considered, but it is not immediately obvious in the case of transitive complements.” Though in the end she does suggest a nominalization analysis for transitives in Yucatec, when considering cross-linguistic data she concludes (based on incomplete morphological evidence) that the Cholan languages do not show nominalization of transitives (Bricker 1981:101). Below I provide evidence that the nominalization analysis is correct not only for Chol intransitives but also for transitives. 2.4. Proposal. Under the proposal presented in this paper, the Chol forms in (1) and (2) above are represented as in (12) and (13), where I temporarily switch to the labels erg, abs, and gen to clarify the proposal. In the perfective forms in (12) we find the expected ergative–absolutive agreement pattern. Here the verb roots also appear with vocalic suffixes, discussed below. Chol perfectives (12a) Tyi prfv a-mek’-e-yoñ erg2-hug-tv-abs1 ‘You hugged me’. (12b) Tyi prfv wäy-i-yoñ. sleep-itv-abs1 ‘I slept’. In contrast, in the imperfective forms in (13) (along with progressives from 3 above), the stems amek’oñ and awäyel (in brackets) are possessed nominals; the set A marker in these forms is not ergative but instead is a genitive marker coindexing a possessor. The stems lack the vocalic suffixes found in (12). The true syntactic predicate in these forms is the aspect marker mi. Just as with other one-place predicates in the language (for example, 12b), mi shows set B (absolutive) agreement with its single argument, the possessed nominal phrase. The availability of absolutive within the nominal phrase in (13a), here the first-person -oñ, is discussed below. As in other Mayan languages, third-person set B is null. This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms 216 international journal of american linguistics Chol imperfectives (13a) Mi-Wi impf-abs3 [a-mek’-oñ]i. gen2-hug-abs1 ‘You hug me’ (lit., ~ ‘Your hugging me occurs’). (13b) Mi-Wi impf-abs3 [a-wäy-el]i. gen2-sleep-nml ‘You sleep’ (lit., ~ ‘Your sleeping occurs’). The appearance of a nominative–accusative (or “extended ergative”) pattern is then the result of two facts: (1) what has been previously considered the main verb in nonperfective constructions (Warkentin and Scott 1980, Vázquez Álvarez 2002, and Gutiérrez Sánchez 2004) is a subordinated nominal form, and (2) ergative and genitive are identical in Mayan languages. Chol’s split is thus not a split in the behavior of verbs but rather a split in whether the encyclopedic “verbal” information is encoded syntactically in predicates (as is the case with perfectives) or in subordinated nominal forms (as in nonperfectives). Under this analysis, all predicates in Chol show an ergative–absolutive pattern of agreement; no special rules of case assignment or agreement are needed. While it remains to be seen whether this analysis may be extended to other Mayan languages, the discussion presented in Larsen and Norman (1979), Bricker (1981), and Mateo-Toledo (2003) makes such an extension seem promising. I briefly discuss the possibility of extending this type of analysis to non-Mayan languages in 6 below. 3. Chol background. Chol is spoken in the northern part of the state of Chiapas, Mexico, by approximately 150,000 people. Chol, together with Chontal and Chortí, belongs to the Cholan subgroup of the Greater Tzeltalan family, which also includes Tseltal and Tzotzil (Tzeltalan subgroup) (Kaufman and Norman 1984). Many of the similarities found between Yucatecan and Cholan languages—such as the ergative split discussed in this paper— have been proposed to be the result of historical contact (Kaufman and Norman 1984 and Law, Robertson, and Houston 2006). Chol may be divided into two major dialect groups: Tila and Tumbalá. Data in this paper come from the Tila dialect of Chol, though initial work suggests that the arguments presented here may be extended to the Tumbalá dialect as well. In this section I examine in detail person marking and stem forms in Chol. 3.1. Person marking. Chol head-marks grammatical relations on the predicate with a set of ergative/genitive prefixes, labeled “set A,” and absolutive suffixes, labeled “set B,” shown in table 1. Glides are inserted to break This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms split ergativity in chol 217 TABLE 1 Chol Person Morphology First person Second person Third person Set A k-/ja(w)i(y)- Set B -(y)oñ -(y)ety W up vowel clusters; first-person k- becomes j- (IPA [h]) preceding velar consonants. Plural agreement is also possible, though it is omitted here for simplicity. The third-person set B marker is null and is generally left unglossed in the examples below. As in other Mayan languages, nominal arguments (subjects, objects, and possessors) may be dropped, context permitting, as shown in (14). The full pronouns joñoñ (first person) and jatyety (second person) are used only for emphasis. (14a) Tyi i-ch’äx-ä. prfv a2-boil-tv ‘You boiled it’. (14b) Tyi i-k’el-e-yoñ. prfv a3-see-tv-b1 ‘He saw me’. Nominals do not show case marking, as shown by the boldfaced arguments in (15). Depending on the context, bare nouns may be interpreted as definite or indefinite, singular or plural. Basic constituent order is VOS/VS (Vázquez Álvarez 2002 and Coon 2010). (15a) Tyi prfv i-mäñ-ä koya ~ a3-buy-tv tomato wiñik. man ‘The man bought tomatoes’. (15b) Tyi prfv majl-i go-itv wiñik. man ‘The man left’. In Chol, as in many Mayan languages, we find set A and set B morphology on both nominal and verbal forms. Set A prefixes in Chol mark transitive subjects as in (16a), unergative subjects as in (16b), and possessors as in (16c). Unergative constructions in Chol require the transitive light verb cha ~l (see Gutiérrez Sánchez 2004). This verb takes a nominal object, such This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms 218 international journal of american linguistics as k’ay ‘song’ in (16b). Chol unergatives are thus best considered a subclass of transitives. Set A morphemes (16a) Tyi k-mek’-e-yety. prfv a1-hug-tv-b2 ‘I hugged you’. (16b) Tyi prfv k-cha ~l-e a1-do-suf k’ay. song ‘I sang’. (16c) k-chich a1-sister ‘my sister’ In contrast, set B morphemes mark transitive objects as in (17a), unaccusative subjects as in (17b), and the theme argument in stative or so-called nonverbal predicate constructions, as shown in (17c) and (17d ).4 Set B morphemes (17a) Tyi i-jats’-ä-yoñ. prfv a3-hit-tv-b1 ‘He hit me’. (17b) Tyi majl-i-yoñ. prfv go-itv-b1 ‘I went’. (17c) X- ~ixik-oñ. cl-woman-b1 ‘I am a woman’. (17d) Chañ-oñ. tall-b1 ‘I am tall’. For Chol, we may generalize as follows across nominal and verbal stems: set A morphemes mark all external arguments; set B morphemes mark all internal arguments. The apparent exception are subjects of nonperfective unaccusatives, which are marked with set A—the source of the apparent split. I argue below, however, that the set A morphemes found on nonperfectives are in fact possessors and thus conform to the generalization. 4 The examples in (17c) and (17d ) show that stative predicates also follow the ergative pattern found in perfectives, though I do not discuss these here. This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms split ergativity in chol 219 3.2. Aspect. Chol grammatically encodes at least three basic aspects: perfective, imperfective, and progressive.5 Like some of the other languages of the Mayan family, for example, Jakaltek (Craig 1977) and Mam (England 1983b), Chol does not have grammaticalized tense morphology. Instead, temporal notions like past and future are marked via adverbs like wajali ‘back then’, abi ‘yesterday’, and ijk’äl ‘tomorrow’. Compare, for example, the sentences in (18). The imperfective sentence in (18a) receives a habitual or ongoing interpretation. In order to talk about an ongoing event in the past, a temporal adverb is used, as in (18b). There is no change in inflection on the stem. (18a) Mi impf i-jap a3-drink kabäl a.lot kajpej. coffee ‘He drinks a lot of coffee’. (18b) Wajali back.then mi impf i-jap kabäl a3-drink a.lot kajpej. coffee. ‘Back then, he drank a lot of coffee’. The imperfective embedded under a past tense adverb in (18b) may be contrasted with the perfective in (19), used to describe a bounded act of drinking. (19) Tyi i-jap-ä prfv a3-drink kabäl a.lot kajpej. coffee ‘He drank a lot of coffee’ (e.g., last night). The main claim of this paper is that the imperfective and progressive markers mi and choñkol (and their allomorphs discussed below) are predicates. As noted above, I refer to Chol’s imperfective and progressive aspects jointly as “nonperfective” aspects; this term does not include stative predicates (which never appear with aspectual morphology and show the ergative pattern also found in perfectives). In contrast, the perfective aspect marker tyi (proposed by Law, Robertson, and Houston 2006:442 to be a borrowing from Yucatec) is simply an aspectual particle.6 I argue that this division is the source of the apparent ergative split. 5 Further distinctions may be made with the addition of certain particles, such as the prospective keje or kej, from the verb kejel ‘being’. This particle is used in conjunction with the imperfective aspect to indicate an event that is going to take place. I do not discuss these aspectual markers here, but see Vázquez Álvarez (2002) for a more detailed discussion of Chol aspect. 6 It is an open question whether the aspect marker tyi is related to the homophonous preposition, discussed below. This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms 220 international journal of american linguistics TABLE 2 Chol Stem Forms Transitive a-root-T-b a-root-(e ~)-b Perfective Nonperfective Intransitive root-i-b a-root-el 3.3. Roots and stem forms. As mentioned above, in addition to differences in person marking, perfective and nonperfective forms differ in stem-forming morphology. Transitive roots appear with a harmonic vowel suffix (represented “-T”) in the perfective aspect, and with no suffix or the suffix -e ~ in the nonperfective aspects.7 Intransitive roots appear with the suffix -i in the perfective and the suffix -el in nonperfective forms. This is schematized in table 2, with examples shown in (20) and (21).8 In table 2, “B” represents the absolutive argument, whether it be a first- or second-person clitic or a third-person nominal form. Transitive stems (20a) Tyi prfv i-kuch-u a3-carry-tv ixim. corn ‘She carried corn’. (20b) Mi impf i-kuch ixim. a3-carry corn ‘She carries corn’. Intransitive stems (21a) Tyi jul-i-ety. prfv arrive.here-itv-b2 ‘You arrived (here)’. (21b) Mi impf a-jul-el. a2-arrive.here-nml ‘You arrive (here)’. 7 The -T suffix is identical to the root vowel except in cases where the root vowel is [a] and the final consonant of the root is a nonfricative consonant, as in the root jap ‘drink’. In these cases, the vowel suffix is ä (IPA [I]). Compare, for example: tyaj-a ‘find-tv’ and jap-ä ‘drink-tv’. The light verb cha ~l discussed above belongs to a subclass of transitives which take vowel suffixes (not necessarily harmonic) in the perfective aspect and -Vñ suffixes in the nonperfective aspects, which I simply gloss ‘suf’. See Vázquez Álvarez (2002) and 4.4 below. 8 Chol has a third class of verb stem-forming roots, positionals, not discussed here for reasons of space. Coon and Preminger (2009) argue that the analysis presented here for intransitives may be extended to cover positional stem forms as well. This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms split ergativity in chol 221 The vocalic suffixes on the perfective stems in (20a) and (21a) appear on all nonstative predicates in Chol. Stative predicates (also known as “nonverbal predicates” in Mayan literature), such as those in (22), do not appear with the vowel suffix and do not take aspect morphology. I propose that the suffixes -T and -i in perfective forms above occupy an eventive v0 (verbal or verbalizing) head and gloss them ‘tv’ and ‘itv’ respectively. Stative predicates (22a) K-om a1-want sa ~. pozol ‘I want pozol’. (22b) Y-ujil a3-know la-k-ty’añ. pl-a1-word ‘He knows Chol’ (lit., ‘He knows our words’). The nonperfective stems, proposed below to be formally nominal, lack these v0 suffixes. Intransitives like (21b) above appear with the suffix -el. Suffixes of the form -Vl are found on nominals throughout Chol (Warkentin and Scott 1980) and other Mayan languages. Bricker (1981:86) notes, for example, that “-V1, the suffix that is used with [subordinated intransitives] in the nominative–accusative part of the verb system in Modern Yucatec, looks suspiciously like a nominal suffix.” I gloss the corresponding Chol suffix ‘nml’ for ‘nominal’. Nonperfective transitives may appear with no suffix, as in (20b), but also frequently appear with the suffix -e ~, as in (23).9 (23) Mi impf i-kuch-e ~ a3-carry-dep ixim. corn ‘She carries corn’. This suffix is descended from the Proto-Mayan dependent suffix (Kaufman and Norman 1984:100). This suggests that ikuch(e ~) ixim is not in fact the matrix predicate in (20b) or (23). Compare the imperfective construction in (23) with the embedded clause in (24a), where the dependent suffix is again optional. As expected, the suffix -e ~ never appears on perfective forms like (24b) (irrespective of whether the transitive suffix -T is present or not), which are argued not to involve embedding. 9 This suffix appears to be possible only with third-person absolutive arguments (whether they are null or overt), not with first- or second-person clitics. This could be a morphological effect, related to a ban on combining certain suffixes, though more work is needed here. This content downloaded from 132.174.25ff:ffff:ffff:ffff on Thu, 01 Jan 1976 12:34:56 UTC All use subject to https://about.jstor.org/terms 222 international journal of american linguistics TABLE 3 Nonperfective Stem Forms Transitive Intransitive (24a) Y-om a3-want set a-[root obj] set a-[root-el] [i-kuch-e ~ ixim]. a3-carry-dep corn ‘She wants to carry corn’. (24b) *Tyi i-kuch-(u)-e ~ prfv a3-carry-tv-dep ixim. corn 4. Nonperfective forms are nominal. In this section I argue, based on distributional evidence, that the nonperfective stem forms from table 2 above are formally possessed nominals; the set A morpheme marks agreement with the possessor. The stem forms in question are repeated in boldface in table 3. Here I distinguish between two types of nominalized transitive stems: transitives with NP objects (TV+NPs) on the one hand, as in (25), and transitives with full DP objects (TV+DPs), as in (26). The former category may include objects with adjectives like kolem ‘big’ in (25b), so long as no determiner or demonstrative is present. TV+NP nominalizations (25a) Choñkol prog [k-jap a1-drink kajpej]. coffee ‘I’m drinking coffee’. (25b) Mi impf [i-chuk a3-catch kolem big chäy]. Fish ‘He catches big fish’. The latter category includes objects with determiners and demonstratives, as in (26a); first- and second-person set B pronominal objects (recall that third-person set B is null), as in (26b); proper names with the clitic aj-, which I analyze in Coon (2010) as a determiner, as in (26c); and possessed objects, as in (26d ).10 10 Though possessed nominals in Chol need not necessarily be full DPs (see Coon 2009), they involve a functional nominal layer above n0. This accounts for their behavior discussed in 4.2. For simplicity, I refer to them in the TV+DP category. This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms split ergativity in chol 223 TV+DP nominalizations (26a) Choñkol prog [i-jap a3-drink jiñi det kajpej]. coffee ‘She is drinking the coffee’. (26b) Mi [k-mek’-ety]. impf a1-hug-b2. ‘I hug you’. (26c) Choñkol prog [i-k’el a3-watch aj-Maria]. det-Maria ‘She is watching Maria’. (26d) Mi impf [k-wuts’ a1-wash k-pislel]. a1-clothes ‘I wash my clothes’. As the examples above illustrate, all of these forms may appear in nonperfective stems, and I argue that all are nominalizations. The difference lies in the level at which the stem is nominalized. The TV+DP forms are argued to be themselves full DPs, while the TV+NP forms are bare nPs. Intransitives pattern with TV+NP nominals. TV+DP nominals have a more limited distribution. They always appear with a set A possessive marker and appear optionally with the suffix -e ~. TV+NPs and intransitives, in contrast, may appear with or without a set A possessor in certain contexts and never appear with the suffix -e ~. (I assume that forms like 23 above, which take the suffix -e ~ but appear with a bare nominal object, involve the structure of TV+DPs, discussed in n. 19.) The proposed structural difference will be shown to have consequences for the behavior of these nominal stems, predictable from independent properties of Chol grammar. 4.1. Intransitive and TV+NP nominalizations. In this section, I show that intransitive nonperfective stem forms (root-el ), as well as nonperfective transitive forms with bare NP objects (root + NP), pattern distributionally with nominals. Like other Chol nominals, these forms may serve as the argument of a predicate, appear with determiners and adjectives, appear as agent nominals, appear possessed, trigger agreement morphology, and serve as the complement of a preposition.11 4.1.1. Embedded clauses. The nonperfective stems from table 3 also appear in Chol embedded clauses, such as those in (27). 11 See also Polian (2008) for nominality tests in Tseltal Mayan. This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms 224 international journal of american linguistics (27a) K-om a1-want [wuts’ wash pisil] / [wäy-el]. clothes sleep-nml ‘I want to wash clothes/to sleep’. (27b) Tyi prfv i-jop’-o a3-try-tv [juch’ ixim] / [ts’äm-el]. grind corn bathe-nml ‘She tried to grind corn/to bathe’. Note that the stem forms in (27) appear here in contexts where we find “infinitive” forms in languages like English. Indeed, the intransitive root-el forms have been analyzed as infinitives in Chol (Vázquez Álvarez 2002) and other Mayan languages, for example, Tzotzil (Aissen 1987), Jakaltek (Craig 1977), and Mam (England 1983b). In Chol, however, these forms appear not only in embedded contexts but in what have been analyzed as main clause nonperfective forms as well, as in (25) and (26) above. If we simply call these forms “infinitives,” we are left with the question of what role they play in clauses like those in (25) and (26), as well as why they appear inflected with the set A morpheme in certain contexts. The proposal put forth in this paper is that these stem forms never serve as matrix predicates. The formal identity between the nonperfective stem forms in (25) and (26) and the embedded clauses in (27) reflects their function in both instances as nominalized clauses serving as arguments of a higher predicate: the aspectual predicates mi and choñkol in (25) and (26) and kom and ijop’o in (27). In the terminology of Givón (2001:26), this makes Chol a “nominalizing” embedding language. In nominalizing languages, also found in the TibetoBurman, Turkic, Carib, Quechua, and Uto-Aztecan families, nonfinite subordinate clauses are nominalized. In Ute (Uto-Aztecan), for instance, the nominality of embedded clauses is manifested via genitive marking on the subject, a nominal suffix on the verb, and object case marking of the embedded clause (Givón 2001:27). While Chol shows agreement rather than case marking, we find all three of these properties as well. Subjects trigger set A (genitive) agreement (otherwise unexpected on intransitives). The embedded forms like those in (27) lack the vocalic suffixes found on matrix verbs and instead appear with a nominal suffix -el (intransitives) and no suffix or the dependent suffix -e ~ (some transitives). Finally, the higher predicate forms in (27), kom and ijop’o, show the expected null third-person set B agreement with their internal arguments. 4.1.2. Causatives. Chol has a causative suffix -(i )sä which forms transitive stems from certain intransitive roots (see Vázquez Álvarez 2002 and Gutiérrez Sánchez 2004), as shown in (28b). This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms 225 split ergativity in chol (28a) Tyi prfv wäy-i sleep-itv ñeñe ~. baby ‘The baby slept’. (28b) Tyi i-wäy-isä prfv a3-sleep-caus ñeñe ~ baby jiñi x- ~ixik. det cl-woman ‘The woman made the baby sleep’. This causativizing suffix, however, is impossible on transitive roots. In order to form a causative from a transitive, a periphrastic construction involving the verb äk’ ‘give’ is used. The noncausative use of äk’ is shown by the constructions in (29). Set A agreement is with the subject; set B agreement is with the primary object (recipient). (29a) Tyi y-äk’-e-yoñ [k-waj] prfv a3-give-appl-b1 a1-tortilla jiñi det x- ~ixik. cl-woman ‘The woman gave me tortillas’. (29b) Tyi prfv k-äk’-e a1-give-appl [sa ~] pozol k-alobil. a1-child ‘I gave my children pozol’. To add a causer to a transitive construction, the nonperfective transitive stem forms from table 3 appear in the secondary object position of äk’, as seen in (30). The set A marker is optional on the bracketed clause; when present it agrees with the primary object, as in (30a). (30a) Tyi prfv y-äk’-e-yoñ [k-wuts’ a3-give-appl-b1 a1-wash pisil] clothes jiñi det x- ~ixik. cl-woman ‘The woman made me wash clothes’. (30b) Tyi prfv k-äk’-e a1-give-appl [kuch ñeñe ~] x-k’aläl . carry baby cl-girl ‘I made the girl carry the baby’. Some nominal argument is obligatory in this position. That is, the stem forms above are obligatorily ditransitive.12 For example, if we remove the bracketed form in (30a), the secondary object is interpreted as a null pronoun and the sentence is read as ‘The woman gave it to me’. This suggests that the bracketed forms in (30) are indeed filling this nominal argument slot. 12 The root äk’ can take the harmonic vowel suffix found on transitives and appear in constructions with a single object. The suffix -e found in the perfective double-object constructions is likely the applicative suffix -be with the consonant deleted and is glossed as such here. This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms 226 international journal of american linguistics In both the embedded clauses from the preceding section, and here in causative constructions, we find the nonperfective stem forms serving as arguments of a predicate. The perfective stems are impossible in these contexts, as shown by the examples in (31). As with the other ungrammatical perfective forms shown below, the presence or absence of person marking on the bracketed form has no effect on grammaticality. (31a) *K-om a1-want (31b) *Tyi prfv [wuts’-u wash-tv pisil ] / [wäy-i]. clothes sleep-itv k-äk’-e [kuch-u a1-give-appl carry-tv ñeñe ~ baby x-k’aläl]. cl-girl 4.1.3. Determiners and adjectives. The determiner jiñi indicates definiteness or salience and appears preceding a noun. While bare nominals may be in some cases interpreted as definite in Chol, nominals with jiñi always receive a definite interpretation. Examples are shown in (32). (32a) Mach neg weñ good [jiñi det waj]. tortilla ‘The tortilla isn’t good’. (32b) Ma’añ neg mi impf k-mul-añ a1-like-suf [jiñi det arus]. rice ‘I don’t like the rice’. The nonperfective stem forms from table 3 can also appear with the determiner, as illustrated in (33a) and (33b). The determiner is unable to appear with perfective stems, as in (33c). (33a) Mach neg weñ good [jiñi det jap lembal] / [jiñi drink liquor det wäy-el]. sleep-nml ‘The liquor-drinking/the sleeping isn’t good’. (33b) Ma ~añ neg mi impf k-mul-añ a1-like-suf [jiñi det juch’ grind ixim] / [jiñi corn det uk’-el]. cry-nml ‘I don’t like the corn-grinding/the crying’. (33c) *Mach weñ neg good [jiñi det jap-ä drink-tv lembal ] / [jiñi wäy-i]. liquor det sleep-itv This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms split ergativity in chol 227 As shown by the examples in (34), nonperfective stems like uk’el ‘cry-nml’ can be modified in the same way as canonical nouns like koya ~ ‘tomato’.13 (34a) Tyi prfv k-mäñ-ä a1-buy-tv [kabäl a.lot koya ~]. tomato ‘I bought a lot of tomatoes’. (34b) Mach weñ neg good [jiñi det kabäl uk’-el]. a.lot cry-nml ‘A lot of crying isn’t good’. 4.1.4. Possession. We saw above that the stem forms from table 3 above may appear as arguments of the verb. They may also appear possessed in this position. Recall that possessive marking is identical to ergative marking in Mayan languages—both are marked with set A. In a possessive phrase, the set A marker appears on the possessum and agrees with the possessor. Possessors appear after the possessum. Examples of Chol possessive phrases are shown in (35) (see Coon 2009 for more on Chol possessive constructions). (35a) Baki where añ loc [aj-chich a2-older.sister proj]? ‘Where’s your older sister?’ (35b) Tyi prfv chäm-i die-itv [ij-wakax a3-cow wiñikj]. man ‘The man’s cow died’. Though speakers vary as to how readily they accept such constructions, the nonperfective stem forms from table 3 may in some contexts also appear possessed in argument position, as illustrated in (36a) and (36b). The perfective stems are always impossible in these constructions, as shown by (36c). (36a) Mach neg weñ good [i-wuts’ a3-wash pisil]. clothes ‘Her clothes-washing isn’t good’ (i.e., she does a poor job washing). (36b) Mach neg weñ good [i-wäy-el ñeñe ~]. a3-sleep-nml baby ‘The baby’s sleeping isn’t good’ (i.e., the baby doesn’t sleep well). 13 While kabäl may also be used as an adverb (not unlike English a lot), the fact that it appears between the determiner and the noun in (34b) suggests that it is being used adjectivally in this case. This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms 228 international journal of american linguistics (36c) *Mach weñ neg good [i-wuts’-u a3-wash-tv pisil ]. clothes-tv In addition to appearing possessed, the nonperfective stem forms may also appear as possessors and, as such, trigger set A agreement. In the forms in (37), the nominal intransitive form ujtyel ‘finish’ appears with third-person set A agreement. The notional subjects, however, are second and first person, respectively. Here, the nominalized nonperfective predicates serve as the possessor of the form ujtyel, also nominal. It is this stem which triggers the set A agreement. As in the cases above, the possessor follows the possessum. The main predicate is the progressive aspect marker choñkol, discussed below. (37a) Choñkol prog [yi-ujty-el a3-finish-nml [k-ts’äm-el]i]. a1-bathe-nml ‘I’m finishing bathing’ (lit., ~ ‘My bathing’s finishing is occurring’). (37b) Choñkol prog [yi-ujty-el a3-finish-nml [a-k’ux a2-eat waj]i]. tortilla ‘You’re finishing tortilla-eating’ (lit., ~ Your tortilla-eating’s finishing is occurring’). Compare, for example, the bracketed forms in (37) with the complex possessive constructions given in (38). (38) y-uskuñ a3-older.brother a-mama a2-mother ‘your mother’s older brother’ As with the constructions discussed above, perfective stems are ungrammatical here. 4.1.5. Agent nominals. The proclitics aj- and x- (historically masculine and feminine noun class markers respectively) appear on many Chol nominals, as illustrated by the examples in (39).14 (39a) Tyi prfv chäm-i die-itv [aj-ts’o ~]. cl-turkey ‘The turkey died’. 14 These clitics also appear on proper names in the language, though their distribution makes it clear that the gender component has been lost. Speakers from the city of Tila, for example, generally use x- on all proper names (xMaria, xPedro), while speakers from nearby El Campanario tend to use aj- (ajMaria, ajPedro). These clitics show further differences in behavior when appearing on proper names (where I gloss them ‘det’), discussed in Coon (2010). This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms split ergativity in chol (39b) Tyi prfv i-k’ux-u a3-eat-tv waj tortilla jiñi det 229 [x- ~ixik]. cl-woman ‘The woman ate tortillas’. These clitics also appear on the nonperfective stem forms from table 3, resulting in nominals with the meaning ‘one who X-es’, as shown in (40a) and (40b). As expected, perfective stems are impossible here, as shown by the ungrammatical (40c). (40a) [Aj-chuk cl-catch chäy] fish jiñi det wiñik. man ‘The man is a fisherman’. (40b) Añ loc kabäl many [aj-ts’äm-el-ob] cl-bathe-nml-pl tyi prep ja ~. water ‘There are many bathers in the water’. (40c) *Añ loc kabäl many [aj-ts’äm-i-yob] cl-bathe-itv-pl tyi prep ja ~. water 4.1.6. Prepositions. Like other nominals in Chol, the nonperfective stem forms in table 3 appear as complements of the preposition. Chol has one all-purpose preposition, tyi. As shown in (41), tyi can receive a variety of interpretations, depending on context. (41a) Tyi prfv majl-i go-itv [tyi prep klase]. school ‘She went to school’. (41b) Añ-oñ loc-b1 [tyi prep otyoty]. house ‘I’m in the house’. (41c) Tsajñ-ety return-b2 [tyi prep Salto]. Salto ‘You returned from Salto’. The imperfective stem forms from table 3 above—but not the perfective stem forms—can all appear in the same position as the uncontroversial nouns klase ‘school’, otyoty ‘house’, and Salto (name of town). Nonperfectives are shown in (42a)–(42c); the ungrammatical perfective stems are given in (42d ). This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms 230 international journal of american linguistics (42a) Tyi majl-i prfv go-itv kuch carry [tyi prep si ~]/ wood [tyi prep wäy-el]. sleep-nml ‘She went to wood-carry/sleep’. (42b) Añ-oñ [tyi wuts’ loc-b1 prep wash pisil]/ [tyi clothes prep uch’-el]. eat-nml ‘I’m clothes-washing/eating’. (42c) Tsajñ-ety return-b2 [tyi prep juch’ grind ixim]/ corn [tyi ts’äm-el]. prep bathe-nml ‘You returned from corn-grinding/bathing’. (42d) *Tsajñ-ety return-b2 [tyi prep juch’-u grind-tv ixim]/ corn [tyi prep ts’äm-i]. bathe-itv Set A markers are possible on the nominalized complements of prepositions so long as they do not corefer with the subject of the higher predicate, as shown by the contrast in (43). The forms in (43a) are grammatical without the possessors and naturally receive the interpretation given in the gloss. Further work is needed to understand this constraint. (43a) *Tsajñ-oñ return-b1 [tyi k-ts’äk-äñ-tyel] / [tyi k-ch’äm ja ~]. prep a1-cure-suf-pasv.nml prep a1-get water ‘I returned from my curing/my baptism’. (43b) Tsajñ-oñ return-b1 [tyi a-ts’äk-äñ-tyel]/ [tyi a-ch’äm ja ~]. prep a2-cure-suf-pasv.nml prep a2-get water ‘I returned from your curing/your baptism’. 4.1.7. The structure of TV+NP and intransitive nominalizations. In the above sections, we found that nonperfective intransitive stems, such as wäy-el ‘sleep-nml’, along with nonperfective transitives with NP objects, like wuts’ pisil ‘wash clothes’, appear in the same contexts as nominals. While previous authors have suggested that only the intransitive forms are nominal (Larsen and Norman 1979 and Bricker 1981), in Chol at least we find evidence for the nominality of transitives as well, despite the lack of any overt nominalizing morphology. I propose that the TV+NP forms like kuch si ~, wuts’ pisil, and juch’ ixim in (42) above are compounds, not unlike English forms such as wood-carrying, clothes-washing, and corn-grinding.15 In Mithun’s (1984) classifica15 Vázquez Álvarez (2002) labels these forms in their clearly nominal function described in 4.1 “incorporation antipassives,” noting the nonspecific and indefinite nature of the objects. He does not extend the analysis to nonperfective forms as I do here. This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms 231 split ergativity in chol tion, these are instances of composition by juxtaposition, in which “the V and the N are simply juxtaposed to form an especially tight bond” (Mithun 1984:849). The NP in the TV+NP nominalizations may appear with adjectives, but determiners are impossible on the compound object, as shown by the forms in (44).16 The availability of adjectives in these forms is reminiscent of “pseudo noun incorporation” discussed, for example, for Niuean in Massam (2001). (44a) Aj-[chuk cl-catch (*jiñi) kolem det big chäy] jiñi wiñik. fish det man ‘The man is a big-fish-catcher’. (44b) Tyi prfb majl-i go-itv tyi prep [mel make (*jiñi) det säsäk white waj]. tortilla ‘She went to white-tortilla-make’. 16 In many languages, the resulting compound behaves like an intransitive: “The V and N remain separate words phonologically; but as in all compounding, the N loses its syntactic status as an argument of the sentence, and the VN unit functions as an intransitive predicate” (Mithun 1984:849). In Chol, however, the resulting form behaves not as an intransitive predicate but as a nominal (also noted for Chontal in Knowles-Berry 1984:153). Looking elsewhere in Chol grammar, we find that this is in fact not surprising. It seems that in general in this language, only predicates with syntactic internal arguments may inflect as verbs (see Hale and Keyser 2002). For example, Chol has a suffix -oñ which appears on certain transitive roots and is cognate with the antipassive suffix in other Mayan languages (Vázquez Álvarez 2002). Like other antipassives, it suppresses the direct object. The resulting form, however, may not inflect as a regular intransitive, as shown in (ia). Instead, it must appear with the nominal suffix -el in light verb constructions like (ib). These Chol -oñel forms are not productive and appear with only a subset of transitive roots. (ia) *Tyi prfv wuts’-oñ-i-yoñ. wash-apasv-itv-b1 ‘I washed’. (ib) Tyi prfv k-cha ~l-e a1-do-tv wuts’-oñ-el. wash-apasv-nml ‘I did washing’. The same is true for unergatives like k’ay ‘song’, as in (16b) above. These are often referred to as “verbal nouns” and may not directly inflect but instead appear in light verb constructions. In some cases, these verbal nouns may take a verbalizing suffix to form a transitive stem (e.g., ‘sing a hymn’); the resulting form contains a syntactic internal argument and may then directly inflect as a predicate. This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms 232 international journal of american linguistics nP n0 ÷ | | -el wäy sleep Fig. 1.—Intransitive nominalization. nP ÷ n0 | -W ÷ ÷ | | kuch carry si ~ wood Fig. 2.—TV+NP nominalization. The proposed structure of these forms is given in figures 1 and 2. In the intransitive in figure 1, the root wäy ‘sleep’ combines directly with the nominal or nominalizing n0 head, -el. I take no stand here on the question of whether wäy should be represented as a verb root or as underspecified with respect to lexical category (as proposed in Coon 2004). In the transitive in figure 2, the roots kuch ‘carry’ and si ~ ‘wood’ are merged. Again, I leave as an open question what the correct representation of the two roots and the resulting projection is. One possibility is that kuch si ~ is then merged with a null n0 head, as shown in (46). Under this analysis, both types of nominalization discussed in this section are bare nPs and are thus free to appear with determiners, possessors, and adjectives, as described above. 4.2. TV+DP nominalizations. In this section I turn to TV+DP nominalizations, which in certain respects behave differently from the intransitive and TV+NP nominalizations discussed in 4.1. I argue that despite these differences, both TV+NP and TV+DPs (in brackets below) are possessed nominal forms, and when they appear in nonperfective forms, they serve as arguments to the aspectual predicate: This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms 233 split ergativity in chol (45a) V+NP nominal as complement to choñkol Choñkol k-[ch’il ja ~as]. prog a1-fry banana ‘I’m frying bananas’. (45b) V+DP nominal as complement to choñkol Choñkol i-[tsep jiñi koya ~]. prog a3-cut det tomato ‘He’s cutting the tomatoes’. Recall that TV+DP nominalizations include transitive roots with a determiner-bearing object, as in (45b), as well as transitive roots with pronominal clitic objects, proper name objects, or possessed objects (see 26 above). Like the nominalizations in the preceding section, TV+DP nominalizations are able to appear as arguments of predicates (46a and 46b), possessed (46c), and may trigger agreement as possessors (46d ). In all of the forms in (46), the set A marker on the TV+DP nominal is obligatory, a fact which I return to below. (46a) Y-om a3-want [i-mek’ a3-hug aj-Maria]. det-Maria ‘She wants to hug Maria’. (46b) Tyi y-äk’-e-yoñ prev a3-give-appl-b1 [k-mek’-ety]. a1-hug-b2 ‘She made me hug you’. (46c) Mach uts’aty [a-jats’-oñ]. neg good a2-hit-b1 ‘Your hitting me isn’t good’. (46d) Choñkol prog [k-wuts’ [yi-ujty-el a3-finish-nml a1-wash jiñi det pisil]i]. clothes ‘I’m finishing washing the clothes’. TV+DP nominalizations, however, are impossible with determiners and adjectives (47a), as complements of the preposition tyi (47b), or as agent nominalizations (47c). The presence or absence of the set A markers has no effect on the ungrammaticality of these forms. (47a) *Mach neg weñ good [jiñi det kabäl a.lot a-jats’-oñ]. a2-hit-b1 Intended: ‘A lot of hitting me isn’t good’. This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms 234 international journal of american linguistics (47b) *Tyi prfv majl-i go-itv i-k’el jiñi wakax]. a3-watch det cow [tyi prep Intended: ‘He went to look at the cows’. (47c) *[Aj-i-chuk cl-a3-catch ili det chäy] fish jiñi det wiñik. man Intended: ‘The man is a catcher of these fish’. The ungrammaticality of the forms in (47), I argue, is a natural consequence of the fact that these nominalizations are themselves full DPs, rather than bare nPs. If this is correct, we straightforwardly explain their inability to appear with adjectives and determiners, as in (47a)—adjectives appear below D0 and two D0 layers are impossible (discussed below). The fact that (47b) is impossible is connected to an independent fact about Chol: the preposition tyi is unable to appear with full DP complements, as shown in (48) (similar facts are noted for Tzotzil [John Haviland, personal communication]). If the VP+DP nominalizations are full DPs, we also explain their inability to appear as complements of the preposition. (48) Tyi prfv majl-i go-itv [tyi prep (*jiñi) det otyoty]. house ‘She went to the house’. Finally, the ungrammaticality of (47c) is also explained by the inability of the clitics to appear on full DPs, as shown in (49). (49) Tyi prfv chäm-i die-itv [aj-(*jiñi) ts’o ~]. cl-det turkey ‘The turkey died’. In the section that follows, I compare the Chol forms discussed above with nominalizations in English and argue for an analysis in which these forms begin as full verb phrases and are nominalized at a higher level. This analysis explains the different distribution of the TV+DP forms in Chol and ties the Chol facts to similar data in other languages. 4.3. Nominalized clauses. Nominalizations in the world’s languages come in a variety of forms and behave differently with respect to case marking of arguments as well as distributional properties (see Abney 1987, KoptjevskajaTamm 1993, and Borsley and Kornfilt 2000). Compare, for example, the English forms in (50), discussed in Borsley and Kornfilt (2000:104). (50a) [John’s repeated criticism of the book] was annoying. (50b) [John’s criticizing the book repeatedly] was annoying. This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms 235 split ergativity in chol DP DPi Du | John’s D0 VP DPi Vu | e V0 DP | criticizing the book Fig. 3.—Mixed extended projection. In (50a), the nominal criticism requires a PP object and the modifier repeated appears in its adjectival form. In the “poss-ing” construction in (50b), in contrast, the object the book appears with no preposition, as with regular finite verbs, and the modifier repeatedly appears in its adverbial form. Nonetheless, both of these constructions serve as sentential subjects. Borsley and Kornfilt (2000), following Abney (1987), propose the structure in figure 3 for the poss-ing nominal. Here, we begin with a verb phrase, but the verb phrase combines not with verbal functional projections (I0 or T0) but with the nominal functional projection D0. They term nominalizations like the one in (50b) mixed extended projections. Spec,VP contains an empty category—either a trace as proposed by Borsley and Kornfilt or a PRO as defended by Yoon (1996)—bound by the possessor in Spec,DP. This structure accounts for the properties of the nominal form in (50b) above. The object requires no special marking, as it is a regular verbal object. Assuming that the adverb modifies the VP, we predict the appearance of an adverbial modifier. Finally, since the form is ultimately a DP and contains no verbal functional projections, we predict its ability to appear as a sentential subject. While the nominal criticism of the book is free to appear with determiners, criticizing the book may not, as shown in (51). This parallels the Chol facts discussed above: while TV+NP forms appear with determiners, TV+DP forms may not. We can capture this fact, Borsley and Kornfilt (2000) propose, by stipulating that overt determiners may not combine with VPs. (51a) We discussed [this/that/the criticism of the book]. (51b) *We discussed [this/that/the criticizing the book]. I propose that the Chol TV+DPs are, like the English poss-ing construction in (50b), full verb phrases which are nominalized higher in the clause, This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms 236 international journal of american linguistics DP D0 nP nu DPi | pro 1pron n0 vP vu DPi | PRO v VP | -e ~ -dep V0 | ki-mel a3-make DP jiñi det waj tortilla Fig. 4.—Chol TV+DP construction. as shown in figure 4. The verb phrase contains a null subject, here represented as PRO. The presence of this null subject in the specifier of VP accounts for why set A (genitive) agreement is obligatory on the TV+DP forms but optional on the TV+NP forms discussed in 4.1: the possessor is needed to control the null argument in Spec,VP. Following the analysis proposed in Coon (2010), I assume that Chol possessors are generated not in DP but in the specifier of a lower projection, here represented nP. The possessum–possessor order is achieved by raising of the possessed XP to a functional position between nP and DP, not represented here for simplicity.17 This analysis maintains that both TV+NPs and TV+DPs are nominals, but with different structures.18 The TV+NPs discussed in 4.1 above are bare roots which combine directly with nominal morphology. The resulting form is a bare nP and is thus permitted to combine with adjectives, determiners, and Chol’s preposition tyi. Full DP objects are not licensed in these com- 17 The fronting of the possessum over the possessor analogous to the obligatory predicatefronting found in the verbal domain (recall that basic order in Chol is VOS/VS). This gives us the welcome result that while nonperfective clauses are proposed to be nominal, they have the same basic VOS order as perfective clauses. See Coon (2010) for details and for further parallels between the verbal and nominal domains. 18 It is an open question whether intransitive nominalizations may in some cases also involve more structure, for instance, a PRO subject. This content downloaded from ffff:ffff:ffff:ffff:ffff:ffff:ffff on Thu, 01 Jan 1976 12:34:56 UTC All use subject to https://about.jstor.org/terms 237 split ergativity in chol pound-like structures and instead require the structure found in TV+DP forms. What permits the presence of a full DP object in a TV+DP form? I assume that v0 is responsible for licensing internal “absolutive” arguments (i.e., assigning them abstract case) and propose that the suffix -e ~ is a realization of a dependent v0 (optionally realized as -W). This suffix differs from regular transitive v0 (the suffix -T), discussed above, in that it requires a PRO specifier. Evidence that the -e ~ suffix is a realization of v0 (and so is not present in TV+NP forms) comes from the fact that it may only appear on stems which appear with set A agreement, as shown by the contrast in (52).19 (52a) Y-om a3-want [TV+DP i-jap-e ~ a3-drink-dep jiñi det kajpej]. coffee ‘He wants to drink the coffee’. (52b) Y-om a3-want [TV+NP jap-(*e ~) drink-dep kajpej]. coffee ‘He wants to drink coffee’. As a final test, consider the transitive light verb cha ~l. While cha ~l takes bare NP complements, some speakers disprefer full DP complements, as shown by the contrast in (53). (The nominal soñ is freely able to appear with a determiner when serving as the argument of other predicates.) (53a) Tyi prfv k-chal ~l-e a1-do-suf [NP soñ]. dance ‘I danced’. (53b) %Tyi prfv k-chal ~l-e a1-do-suf [DP jiñi det soñ]. dance In a similar vein, TV+NP forms—which are argued to be bare nPs— are possible as complements to the light verb, while TV+DPs are ungrammatical. (54a) Tyi k-chal ~l-e prfv a1-do-suf [NP k’el watch wakax]. cow ‘I watched cows’. 19 The object in (52) may also be a bare nominal: kajpej. One possibility is that in this structure it involves a null determiner. Since bare NPs in Chol may receive definite interpretations, this is not implausible. Alternatively, we could propose that bare NPs may appear in either TV+NP or TV+DP-type nominalizations, but DPs are always banned from TV+NPs. This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms 238 international journal of american linguistics (54b) *Tyi prfv k-chal ~l-e a1-do-suf [DP k’el-ety]. watch-b2 ‘I watched you’. In this section, we have seen that despite the distributional differences between the intransitive and TV+NP transitives, on the one hand, and the TV+DP transitives, on the other, all exhibit nominal properties and all appear in nonperfective constructions like those in (55). Their different distributions are predicted by the structures proposed above. In 5, I argue that the true predicates in the forms in (55) are the aspect markers mi and choñkol. (55a) Mi y-uk’-el impf a3-cry-nml ñeñe ~. baby ‘The baby cries’. (55b) Choñkol k-pets’ prog a1-squeeze alaxax. orange ‘I’m squeezing oranges’. (55c) Mi a-mek’-oñ. impf a2-hug-b1 ‘You hug me’. 5. Mi and choñkol as predicates. Above, we found evidence for the nominal nature of nonperfective stem forms. The behavior of the nonperfective stems parallels that of other uncontroversial nouns in combining with determiners, adjectives, possessors, the preposition tyi, serving as the argument of a predicate, triggering agreement, and forming agent nominals. The fact that TV+DP forms are necessarily full DPs explains their inability to appear in some of these constructions. Perfective stems, on the other hand, do not appear in any of these contexts and exhibit no nominal behavior. The true syntactic predicates in nonperfective constructions, I propose, are the aspectual morphemes: mi (imperfective) and choñkol (progressive), as shown in (56).20 Like other one-place predicates in the language, mi and choñkol show set B agreement with their single arguments. The set A agreement marker is in fact the genitive and agrees with the possessor—a null pro in (56a) and ñeñe ~ in (56b). 20 In the Tumbalá dialect of Chol, the progressive marker is woli. Initial data suggests that woli behaves syntactically like the Tila Chol choñkol. This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms split ergativity in chol 239 Chol nonperfectives (56a) Mi-Wi [a-k’el-oñ]i. impf-abs3 gen2-watch-abs1 ‘You watch me’ (lit., ~ ‘Your watching me occurs’). (56b) Choñkol-Wi prog-abs3 [i-wäy-el ñeñe ~]i. gen3-sleep-nml baby ‘The baby is sleeping’ (lit., ~ ‘The baby’s sleeping occurs’). Recall that under the account offered here, there is no “split” among predicates: all syntactic predicates in Chol show an ergative–absolutive agreement pattern. In both the perfective and nonperfective intransitives in (57), for example, the predicate shows set B agreement with its argument. What differs is the nature of the argument. In nonperfective constructions like (57a), the argument is what has typically been treated as the verb phrase. In perfectives, the arguments of the predicates are the theta role recipients of the root, as in more familiar languages like English. (57a) Mi-Wj impf-abs3 [i-wäy-el ñeñe ~]j. gen3-sleep-nml baby ‘The baby sleeps’. (57b) Tyi wäy-i-Wj [ñeñe ~]j. prfv sleep-itv-abs3 baby ‘The baby slept’. Ergativity, then, is found in all Chol predicates. The appearance of accusativity in the nonperfective stems reflects the fact that all nonperfective stems are possessed nominals. The notional subject is expressed as a possessor—marked with the set A morpheme—in both transitives and intransitives. The fact that possessors are marked with the same set of morphemes as external subjects gives the appearance of accusativity (see Coon 2010 for structural similarities between possessors and external subjects). 5.1. A closer look at the aspect markers. If this analysis is correct, the nonperfective aspect markers (mi and choñkol ) are radically different from the perfective morpheme (tyi ). Indeed, other evidence supports this conclusion. First, I review phonological facts about the aspect markers. Both mi and tyi have fuller CVC allomorphs, muk’ and tsa ~, shown in table 4. The minimal word requirement in Chol is CVC and so the larger forms are necessary when the aspect markers are used to host clitics. As the progressive marker already meets the CVC requirement, it does not have a distinct form. This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms 240 international journal of american linguistics TABLE 4 Aspect Markers Perfective Imperfective Progressive Short Form tyi mi choñkol Larger CVC Form tsa ~ muk’ choñkol The use of these forms with second-position clitics such as -äch (affirmative) and -bi (reportative) is shown in (58). While the larger CVC forms are required in certain phonological contexts, they are always possible and have no known effect on meaning. That is, the forms in (58) are grammatical with or without the clitics.21 (58a) Muk’-äch k-ts’äm-el. impf-aff a1-bathe-nml (*mi-äch) ‘I indeed bathe’. (58b) Tsa ~-bi prfv-rep majl-i go-itv tyi prep Tila. Tila (*tyi-bi) ‘It’s said she went to Tila’. There is another phonological fact about the imperfective marker mi relevant to this discussion, illustrated in (59). In noncareful speech, mi and the following set A clitic combine to form a phonological word: mi + k § mik; mi + a § ma ~; mi + i § i § mi ~. The allomorph muk’ is required when no set A marker is present, as will be shown below. (59a) Mi k-ts’äm-el. fl Mik ts’ämel. impf a1-bathe-nml ‘I bathe’. (59b) Mi a-ts’äm-el. fl Ma ~ ts’ämel. impf a2-bathe-nml ‘You bathe’. (59c) Mi i-ts’äm-el. fl Mi ~ ts’ämel. impf a3-bathe-nml ‘He bathes’. 21 This contrasts with data reported in Coon (2004), cited in Law, Robertson, and Houston (2006), in which it was reported that muk’ may only be used with clitics. Though most speakers tend to use mi when no clitics are present, muk’ is also possible in its bare form. Further work is needed on the phonology of these forms and the factors governing their use. This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms split ergativity in chol 241 Returning now to the difference between the nonperfective and perfective aspect markers, we observe that both the imperfective form muk’ and the progressive choñkol can appear directly with situation-denoting nominal complements such as ja ~al ‘rain’ and k’iñijel ‘party’. This is impossible with the perfective forms, as shown by the contrast in (60c) (60a) Muk’ impf ja ~al tyi rain prep k-lumal. a1-land ‘It rains in my country’. (60b) Choñkol prog k’iñijel tyi party prep aw-otyoty. a2-house ‘There’s a party going on at your house’. (60c) *Tsa ~ k’iñijel prfv party tyi prep aw-otyoty. a2-house Intended: ‘There was a party at your house’. The intransitive and TV+NP nominal forms discussed in 4.1, like ts’ämel ‘bathe’ and juch’ waj ‘grind masa’, can also appear in this position without their set A marking and receive an impersonal or generic interpretation, as shown in (61). TV+DP nominals, as in other contexts, require the set A marker and thus do not appear in impersonal constructions. (61a) Muk’ impf ts’äm-el tyi bathe-nml prep ja ~. water ‘Bathing occurs in the water’. (61b) Choñkol prog juch’ grind waj masa tyi k-otyoty. prep a1-house ‘There is masa-grinding going on at my house’. While either mi or muk’ is possible when the following stem carries a set A marker, as in (62), muk’ is required in forms like (61a) and (60a). I connect this to the phonological facts discussed above. (62) {Mi/Muk’} k-ts’äm-el tyi impf/impf a1-bathe-nml prep ja ~. water ‘I bathe in the water’. Note that the only difference between the construction in (61a) and the form in (62) is the presence of the set A marker. In both, I argue, the imperfective morpheme mi/muk’ serves as a predicate and takes a nominal complement. The This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms 242 international journal of american linguistics difference lies in whether or not the nominal complement appears with a possessor (the notional subject) or not. In the forms in (61), no possessor is present and we are left with an impersonal interpretation. 5.2. Raising predicates. The imperfective and progressive aspectual predicates are able to appear in what have been called “raising” constructions (Robertson 1980; 1992), shown in (63a)–(63c).22 In these constructions, the notional subject serves as the argument of the one-place aspectual predicate and triggers the set B marker expected of intransitives. The nominalized verb stem appears obligatorily subordinated by the preposition tyi.23 As predicted, this construction is impossible with the perfective aspect marker, as illustrated by the ungrammaticality of (63d ). (63a) Choñkol-oñ prog-b1 [tyi prep uk’-el ]. cry-nml ‘I am crying’. (63b) Muk’-ety [tyi impf-b2 prep päk’ plant bu’ul ]. bean soñ] dance jiñi det ‘You plant beans’. (63c) Choñkol prog [tyi prep wiñik. man ‘The man is dancing’. (63d) *Tsa ~-ety prfv-b2 [tyi prep wäy-el ]. sleep-nml Intended: ‘You slept’. It appears that only external subjects may be raised: this construction is possible with transitive forms like (63b), unergative “verbal nouns” like soñ in (63c), and with “ambivalent” intransitives like uk’el in (63a). Ambivalents form a subclass of intransitives which may behave either like unaccusatives, as in (64a), or as unergatives, as in (64b) (Vázquez Álvarez 2002). That is, they appear in their nominal forms as the complement of the light verb cha ~l 22 I follow Robertson in adopting the term “raising,” though he notes that he uses the term for convenience and that it is not meant as a description of an actual grammatical mechanism (Robertson 1992:77). Similarly, I do not have evidence at this point bearing on whether true raising is taking place here, or whether we are dealing with some other phenomenon, e.g., control. I nonetheless continue to use this label to describe the marking of the lower-clause subject on the matrix (aspectual) predicate. 23 TV+DP nominalizations may not appear in this construction, as predicted by the fact that the preposition tyi does not take full DP complements (see 4.2). This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms split ergativity in chol 243 (see 3.1).24 They also appear in nonperfective raising constructions, as in (64c). True (nonambivalent) unaccusatives like majlel ‘go’, in contrast, are unable to appear in the light verb construction, and similarly do not permit raising by the nonperfective predicates, as in (65b) and (65c). I conclude that while unaccusatives assign only an internal thematic role, ambivalents may assign either an internal or an external thematic role. (64a) Mi k-wäy-el. impf a1-sleep-nml ‘I sleep’. (64b) Mi k-cha ~l-eñ impf a1-do-suf wäy-el. sleep-nml ‘I sleep’. (64c) Muk’-oñ tyi wäy-el. impf-b1 prep sleep-nml ‘I sleep’. (65a) Mi impf k-majl-el. a1-go-nml ‘I go’. (65b) *Mi k-cha ~l-eñ impf a1-do-suf (65c) *Muk’-oñ impf-b1 majl-el. go-nml tyi majl-el. prep go-nml Note that in the raising constructions, the predicates muk’ and choñkol behave like other one-place predicates in the language in showing set B agreement with their single argument (an ergative–absolutive pattern). Any other nominals must be introduced with the preposition tyi. Compare, for example, the raising construction in (66a) with the stative predicate in (66b). 24 The forms in (64a) and (64b) have different semantic properties as well. While (64a) may be used to describe the act of accidentally dozing off, (64b) requires that the sleeping be a premeditated event. Further work is needed to determine the semantics of the form in (64c). While antipassive -oñel forms and verbal nouns (discussed briefly in n. 16) may appear in place of wäyel in constructions like (64b) and (64c), they are impossible in nonraising forms like (64a). This may be connected to the fact that they are not productive nominalizations, or we might propose that nonraising constructions require internal arguments. This might suggest that nonraising constructions always involve TV+DP nominalizations, in some cases with null determiners, though further work is needed on this issue. This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms 244 international journal of american linguistics (66a) Muk’-oñ tyi impf-b1 prep ts’äm-el. bathe-nml ‘I bathe’. (66b) Buch-ul-oñ seated-pos-b1 tyi prep siya. chair ‘I’m seated in the chair’. The inability of the perfective tsa ~ to serve as a predicate in the raising constructions discussed in this section provides additional support for a formal division between the perfective and nonperfective aspect markers. Under the proposed analysis, the nonperfective markers always behave as predicates. In a regular (nonraising) imperfective construction like the one in (67a), the argument of the imperfective predicate is the possessive phrase its’ämel jiñi alob ‘the child’s bathing’. In the raising construction discussed in this section and exemplified in (67b), the stem ts’ämel appears unpossessed and subordinated by the preposition tyi. The argument of the imperfective predicate here is the notional subject jiñi alob ‘the child’, raised from the subordinated clause. In both cases, the predicate shows an ergative– absolutive agreement pattern, agreeing with its sole argument via the set B morpheme. (67a) Mi-Wj [i-ts’äm-el impf-b3 a3-bathe-nml jiñi alob]j. det child ‘The child bathes’ (lit., ~ ‘The child’s bathing occurs’). (67b) Muk’-Wj impf-b3 tyi ts’äm-el [jiñi alob]j. prep bathe-nml det child ‘The child bathes’ (lit., ~ ‘The child is at bathing’). To review, I provide structures for perfective and imperfective intransitive constructions below. In a perfective construction like (68), the aspect marker is simply an aspectual particle. The root wäy combines with a v0 head (realized as -i ) and takes an internal (set B-marked) argument, as shown in figure 5. In an imperfective construction, in contrast, the aspect marker serves as the main predicate. The root wäy combines with n0 (realized as -el ), and a possessor appears in Spec,nP, triggering set A marking on the nominal form wäyel.25 This is illustrated in figure 6. For simplicity, I omit the set A and set B person markers on these structures and remain neutral as to 25 This structure is slightly simplified, as the possessum should appear before the possessor (evident when the possessor is overt). Here I follow the analysis in Coon (2010), in which the possessum NP fronts to a DP-internal inflectional projection above the possessor. This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms split ergativity in chol 245 AspP Asp0 vP | tyi prfv v0 | -i -itv VP V0 | DP | wäy sleep pro 1pron Fig. 5.—Perfective intransitive. PredP Pred0 nP | mi impf DPi | pro 1pron nu n0 | ÷ | -el -nml wäy sleep Fig. 6.—Imperfective intransitive. whether these are agreement morphemes or pronominal clitics. Transitives pattern similarly, modulo independent differences: an external argument added to the perfective construction in figure 5, and a compound form or a full verb phrase under the n0 as the nominal in figure 6. (68) Perfective intransitive Tyi wäy-i-yoñ. prfv sleep-itv-b1 ‘I slept’. (69) Imperfective intransitive Mi k-wäy-el. impf a1-sleep-nml ‘I sleep’. This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms 246 international journal of american linguistics 5.3. Further properties of nonperfective predicates. In the preceding sections, I argued that in nonperfective constructions—like (70) below—the syntactic predicate is the aspect marker—mi or choñkol. In (70), mi behaves like other one-place predicates in Chol in taking a single argument and showing set B agreement (null third person in this case) with its argument. The argument of mi is a possessed nominal phrase, shown here in brackets. We found evidence above that stem forms like wejlel ‘fly’ behave distributionally as nominals. The notional subject is a grammatical possessor, here aj-loro ‘parakeet’. Like other possessors in the language, it follows the possessum and triggers set A agreement. (70) Mi [i-wejl-el impf a3-fly-nml aj-loro]. cl-parakeet ‘The parakeet flies’. Here I note some important differences between the predicate mi and other intransitive predicates in Chol. Recall from 4.1.3 that nominalized intransitive stem forms like wejlel can appear with determiners like jiñi. The complement of the nonperfective predicates mi/muk’ and choñkol, however, may not appear with a determiner. Compare the forms in (71). Under the analysis outlined above, both buchul ‘seated’ in (71a) and mi in (71b) are one-place predicates. They both show set B agreement with their sole arguments; in both cases here the argument is third person and so agreement is null. However, while the argument of buchul is free to appear with a determiner, the argument of mi/muk’ may not. (71a) Buchul seated-pos [jiñi det x- ~ixik]. cl-woman ‘The woman is seated’. (71b) *Muk’ impf [jiñi det i-wejl-el a3-fly-nml aj-loro]. cl-parakeet ‘The parakeet flies’. A further difference is found in topicalization. Though subjects canonically appear postverbally, they may also be fronted to preverbal topic or focus positions for emphasis (see Aissen 1992). Again, although the argument of the one-place predicate buchul may appear fronted, as in (72a), this is impossible with the argument of mi/muk’, as shown in (72b). (72a) [Jiñi det x- ~ixik] cl-woman buch-ul. seated-pos ‘The woman is seated’. This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms split ergativity in chol (72b) *[I-wejl-el aj-loro] a3-fly-nml cl-parakeet 247 muk’. impf ‘The parakeet flies’. Though these facts point to certain restrictions on the raising predicates mi/muk’ and choñkol, I propose that they do not change the status of these forms as predicates. As discussed in 4.3, the impossibility of DP objects is also found with the light verb cha ~l (for some speakers). We saw above in 4.2 that the preposition tyi is also not permitted to take full DP complements. The impossibility of fronting in (71b) may be connected to phonological factors, such as the tendency for the set A marker to attach to the aspect marker, or a ban on stranding the aspect marker (the preposition tyi may also not be stranded), though further work is needed here. We may conclude that the nonperfective predicates, like the preposition tyi and the light verb cha ~l, have more restrictive selectional requirements than predicates like buchul ‘seated’, but they are predicates nonetheless. The aspectual predicates mi/muk’ and choñkol may not combine with the perfective aspect marker tyi, as shown in (73a). This is predicted by the fact that stative predicates in Chol are generally unable to appear with aspectual morphology, as shown in (73b). Temporal adverbs must be used instead (see 3.2). (73a) *Tyi [PredP prfv choñkol prog k-mel waj]. a1-make tortilla ‘I was making tortillas’. maystraj-oñ]. teacher-b1 (73b) *Tyi [PredP prfv ‘I was a teacher’. Finally, while the complements of the nonperfective predicates cannot be topicalized, it is fine to extract arguments out of the nonperfective complements, as shown in the interrogative constructions in (74b) and (74c). (74a) Mi [i-choñ impf a3-sell waj tortillas x- ~ixik]. cl-woman ‘The woman sells tortillas’. (74b) Maxkii Who Mi impf [i-choñ a3-sell waj tortillas ti]? ‘Who sells tortillas?’ This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms 248 international journal of american linguistics (74c) Chukii what mi impf [i-choñ a3-sell ti x- ~ixik]. cl-woman ‘What does the woman sell?’ The extractability of maxki out of what has been proposed here to be a nominal phrase is in fact predicted, as possessors of internal arguments can always undergo extraction in Chol (see Coon 2009 on Chol and Aissen 1996 on Tzotzil), as shown by (75). (75) Maxkii who tyi prfv chäm-i die-itv [i-wakax ti]? a3-cow ‘Whose cow died?’ The extraction of the internal argument of choñ ‘sell’ is also not surprising if we assume that wh-words like chuki ‘what’ are full DPs. If this is correct, then the nominal in (74c) necessarily involves a full verb phrase (see 4.3). In this case, chuki originates as the complement of V0—not as the complement of a noun. In the terms of Chomsky (1981), the trace of the wh-word is thus properly governed and extraction is permitted. 6. Conclusion. I have argued for a new way of thinking about the apparent split ergativity found in Chol. Returning to the forms from (12) and (13) above, repeated here in (76) and (77), we find that in the perfective aspect, the subject of the intransitive bears the same marking as the object of the transitive, resulting in the ergative–absolutive pattern found throughout the Mayan family. In the nonperfective aspects, the subject of the intransitive patterns with the subject of the transitive. As noted above, however, this is not a typical nominative–accusative pattern—there are no distinct nominative or accusative morphemes, and the “nominative” is not unmarked, as in many nominative–accusative patterns. Rather, in the terminology of Dixon (1979), this is a case of extended ergativity—the set A marker has been extended to mark the subjects of certain intransitives. Chol perfectives (76a) Tyi a-mek’-e-yoñ. prfv erg2-hug-tv-abs1 ‘You hugged me’. (76b) Tyi wäy-i-yoñ. prfv sleep-itv-abs1 ‘I slept’. This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms split ergativity in chol 249 Chol imperfectives (77a) Mi-Wi [a-mek’-oñ]i. impf-abs3 gen2-hug-abs1 ‘You hug me’. (77b) Mi-Wi [a-wäy-el ]i. impf-abs3 gen2-sleep-nml ‘You sleep’. I argued above that the set A marker in the nonperfective forms in (77) is not in fact an ergative marker but rather the genitive. Specifically, I argue that the apparent split in Chol arises from two factors: (1) nominalization of the verb forms in the nonperfective aspects and (2) the fact that ergative and genitive are marked identically in Mayan languages. This analysis builds on and extends proposals made in Larsen and Norman (1979) and Bricker (1981) for splits in other Mayan languages. First, while these authors suggest that it is only the intransitive forms which are nominalized, I argue that, at least in Chol, we find evidence for nominality in both intransitives and transitives. In the transitives, we distinguished between two types of possible nominalizations, which allowed us to predict their different behavior in other contexts. Second, while Larsen and Norman (1979) suggest that nominalization provides a historical explanation for the split, I provide evidence that the Chol nonperfective aspect markers behave today as predicates in permitting situation-denoting arguments and participating in raising constructions. The discussion of split ergativity in the Mayan family provided in Larsen and Norman (1979) suggests that it may be possible to extend this analysis to other Mayan languages. Looking now beyond the Mayan family, Dixon (1994:99) notes that in the tense- or aspect-based splits found in languages around the world, “the ergative making is always found either in past tense or in perfective aspect” (though languages vary in exactly where the split is made). While further work is needed to determine whether aspect-based ergative splits represent a unified phenomenon throughout the world’s languages, the fact that these splits are always made along the same lines makes such a unification desirable. The analysis presented in Laka (2006) for Basque progressives in many ways foreshadows the analysis for Chol presented here, suggesting that a biclausal analysis of nonperfectives extends beyond the Mayan family. Most Basque transitive constructions appear as in (78a), with ergative marking on the subject. The object appears in the absolutive (zero-marked) This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms 250 international journal of american linguistics case. In the progressive, however, the ergative marker is absent, as shown in (78b). Furthermore, we find differences in agreement between progressive and nonprogressive forms. While the auxiliary in (78a) shows agreement with both the subject and the object, the progressive in (78b) form shows agreement only with the subject. Basque (78a) emakume-a-k woman-det-erg ogi-ak ja-ten d-it-u bread-det.pl eat-impf 3abs-pl-have.3erg ‘The woman eats (the) bread’. (78b) emakume-a woman-det [ogi-ak bread-det.pl ja-ten] ari da eat-impf prog 3abs.is ‘The woman is eating (the) bread’. (Laka 2006:173) Though I do not review details of the analysis here, Laka (2006) proposes that the absence of ergative patterning in (78b) is due to the fact that progressive constructions are in fact biclausal: the main verb is ari, which takes a locative PP complement. Under this analysis, emakumea ‘the woman’ does not take ergative marking because it is the single argument in its clause. Furthermore, this analysis explains the differences in agreement. The progressive auxiliary does not agree with the object ogiak ‘bread’ because it is not in the same clause. While Laka does not explicitly extend this analysis to the aspect-based split in Hindi, she notes that the differences in case marking and agreement found between Basque nonprogressives and progressives are parallel to those found between Hindi perfectives and nonperfectives. Laka’s analysis of the Basque progressive in (78b) parallels the structure of Chol raising constructions (available only in the nonperfective aspects) discussed in 5.2 above. In the Chol construction in (79), the progressive marker serves as the main predicate, taking the subject as its only argument. The verb root ts’äm ‘bathe’ appears in a nominal stem form, subordinated by the preposition tyi. (79) Choñkol-Wi prog-b3 [PP tyi prep ts’äm-el ] bathe-nml jiñi det alobi. child ‘The child is bathing’. However, as we have seen above, Chol has recourse to another type of nonperfective construction. In (80), the aspectual predicate choñkol takes the nominalized verb form directly as an argument. The notional subject appears as the possessor of this argument. In both (79) and (80), we have a This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms split ergativity in chol 251 biclausal construction in which the aspectual morpheme serves as the syntactic predicate. (80) Choñkol-Wi prog-b3 [NP i-ts’äm-el a3-bathe-nml jiñi det alob]i. child ‘The child is bathing’. In both Chol and Basque, we may account for the appearance of a nonergative pattern in certain aspects by analyzing these constructions as involving subordination. In both languages, the main predicate in “split” forms is the aspect marker. It shows the expected ergative pattern, and the appearance of accusativity is connected to the fact that what has been analyzed as the verb is in fact a subordinated nominal. The presence of subordination in nonperfective constructions has been noted elsewhere. As discussed in Laka (2006), in unrelated languages around the world, the progressive frequently involves a locative construction (Bybee, Perkins, and Pagliuca 1994). Mateu Fontanals and Amadas Simon (1999) and Demirdache and Uribe-Etxebarria (2000) argue for universal connections between locative spatial relations and nonperfective constructions, and Mateu Fontanals and Amadas Simon (1999:160) propose that progressive constructions “must be regarded as implying an unaccusative structure over that structure assigned to the verb in the lexicon.” Further work is needed to understand variation between splits (progressive vs. nonprogressive in Basque, perfective vs. nonperfective in Chol and Hindi), as well as to determine whether other languages may be captured under this type of analysis. Nonetheless, biclausal analyses of split ergativity— like that for Chol advanced in this paper—may provide an explanation as to why we always find the appearance of a nonergative pattern in the nonperfective forms: ergative case is not assigned because we are dealing with a structure in which the matrix verb is intransitive. In Chol, the presence of set A possessive marking on both transitive and intransitive subordinate forms further obscures the underlying ergativity. REFERENCES Abney, Steven. 1987. The English noun phrase in its sentential aspect. Ph.D. dissertation, Massachusetts Institute of Technology. Aissen, Judith. 1987. Tzotzil Clause Structure. Reidel: Dordrecht. . 1992. Topic and focus in Mayan. Language 68:43–80. . 1996. Pied-piping, abstract agreement, and functional projections in Tzotzil. Natural Language and Linguistic Theory 14:447–91. Ayres, Glenn. 1983. The antipassive “voice” in Ixil. IJAL 49:20– 45. This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms 252 international journal of american linguistics Borsley, Robert D., and Jaklin Kornfilt. 2000. Mixed extended projections. Syntax and Semantics, vol. 32, The Nature and Function of Syntactic Categories, ed. Robert D. Borsley, pp. 101–31. New York: Academic Press. Bricker, Victoria R. 1981. The source of the ergative split in Yucatec Maya. Journal of Mayan Linguistics 2:83–127. Bybee, Joan; Revere Perkins; and William Pagliuca. 1994. The Evolution of Grammar: Tense, Aspect, and Modality in the Languages of the World. Chicago and London: University of Chicago Press. Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris. Comrie, Bernard. 1978. Ergativity. Syntactic Typology: Studies in the Phenomenology of Language, ed. Winfred P. Lehmann, pp. 329–94. Austin: University of Texas Press. Coon, Jessica. 2004. Roots and words in Chol (Mayan): A distributed morphology approach. B.A. thesis, Reed College. . 2009. Interrogative possessors and the problem with pied-piping in Chol. Linguistic Inquiry 40:165–75. . 2010. VOS as predicate fronting in Chol. Lingua 120:354–58. Coon, Jessica, and Omer Preminger. 2009. Positional roots and case absorption. New Perspectives in Mayan Linguistics, M.I.T. Working Papers in Linguistics, ed. Heriberto Avelino, Jessica Coon, and Elisabeth Norcliffe, pp. 35–58. Cambridge, Mass.: The M.I.T. Press. Craig, Colette Grinevald. 1977. The Structure of Jacaltec. Austin: University of Texas Press. Demirdache, Hamida, and Myriam Uribe-Etxebarria. 2000. The primitives of temporal relations. Step by Step, ed. Roger Martin, David Michaels, and Juan Uriagereka, pp. 157– 86. Cambridge, Mass.: The M.I.T. Press. Dixon, R. M. W. 1979. Ergativity. Language 55:59–138. . 1994. Ergativity. Cambridge: Cambridge University Press. England, Nora. 1983a. Ergativity in Mamean (Mayan) languages. IJAL 49:1–19. . 1983b. A Grammar of Mam, a Mayan Language. Austin: University of Texas Press. Givón, Talmy. 2001. Syntax: An Introduction. Amsterdam: John Benjamins. Gutiérrez Sánchez, Pedro. 2004. Las clases de verbos intransitivos y el alineamiento agentivo en el chol de Tila, Chiapas. M.A. thesis, CIESAS, México. Hale, Ken, and Samuel Jay Keyser. 2002. Prolegomenon to a Theory of Argument Structure. Cambridge, Mass.: The M.I.T. Press. Kaufman, Terrence. 1976. Archaeological and linguistic correlations in Mayaland and associated areas of Meso-America. World Archaeology 8:101–18. Kaufman, Terrence S., and William M. Norman. 1984. An outline of proto-Cholan phonology, morphology, and vocabulary. Phoneticism in Mayan Hieroglyphic Writing, ed. John S. Justeson and Lyle Campbell, pp. 77–166. Albany: Institute for Mesoamerican Studies, State University of New York. Knowles-Berry, Susan M. 1984. A descriptive grammar of Chontal Maya (San Carlos dialect). Ph.D. dissertation, Tulane University. Koptjevskaja-Tamm, Maria. 1993. Nominalizations. London: Routledge. Laka, Itziar. 2006. Deriving split ergativity in the progressive: The case of Basque. Ergativity: Emerging Issues, ed. Alana Johns, Diane Massam, and Juvenal Ndayiragije, pp. 173–96. Dordrecht: Springer. Larsen, Tomas W., and William M. Norman. 1979. Correlates of ergativity in Mayan grammar. Ergativity: Towards a Theory of Grammatical Relations, ed. Frans Plank, pp. 347–70. London and New York: Academic Press. Law, Danny; John Robertson; and Stephen Houston. 2006. Split ergativity in the history of the Ch’olan branch of the Mayan language family. IJAL 72:415–50. This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms split ergativity in chol 253 Massam, Diane. 2001. Pseudo noun incorporation in Niuean. Natural Language and Linguistic Theory 19:153–97. Mateo Pedro, Pedro. 2009. Revisiting split ergativity in Q’anjob’al. Ms., University of Kansas. Mateo-Toledo, B’alam. 2003. Ergatividad mixta en Q’anjobal (Maya): Un reanálisis. Proceedings of the Conference of Indigenous Languages of Latin America 1. Mateu Fontanals, Juame, and Laia Amadas Simon. 1999. Extended argument structure: Progressive as unaccusative. Catalan Working Papers in Linguistics, pp. 159–74. Barcelona. Mithun, Marianne. 1984. The evolution of noun incorporation. Language 60:847–94. Pascual, Adán F. 2007. Transitividad y dependencia sintáctica y discursiva en Q’anjob’al. M.A. thesis, CIESAS, México. Polian, Gilles. 2008. Reducción de la finitud y formas no-finitas en tseltal. Ms., CIESASSureste. Quizar, Robin, and Susan M. Knowles-Berry. 1988. Ergativity in the Cholan languages. IJAL 54:73–95. Robertson, John. 1980. The Structure of Pronoun Incorporation in the Mayan Verbal Complex. New York: Garland. . 1992. The History of Tense/Aspect/Mood/Voice in the Mayan Verbal Complex. Austin: University of Texas Press. Vázquez Álvarez, Juan J. 2002. Morfología del verbo de la lengua chol de Tila Chiapas. M.A. thesis, CIESAS, México. Warkentin, Viola, and Ruby Scott. 1980. Gramática Ch’ol. México: Instituto Lingüístico de Verano. Yoon, James H. S. 1996. Nominal gerund phrases in English as phrasal zero derivations. Linguistics 34:329–56. Zavala, Roberto. 1997. Functional analysis of Akatek voice constructions. IJAL 63:439–74. This content downloaded from 132.174.254.72 on Fri, 05 Mar 2021 22:04:13 UTC All use subject to https://about.jstor.org/terms