LIGHT VERBS AND SPLIT ERGATIVITY IN THE WESTERN CHOLAN LANGUGES

This paper is a discussion of a complex verbal construction in Western Cholan languages and how this construction interacts with the split ergative systems found in these languages. The Cholan languages all display split-ergative systems based on aspect. In addition to this split system, Vázquez claims that Chol has properties of split intransitivity as an agentive/non-agentive language. This perspective would mean that Chol has accusativity that is aspect-based (split-ergativity) and lexically/semantically based (split-intransitivity). This characterization is rendered problematic by the fact that these person markers attach to a light verb cha’len which, by itself, is a transitive verb. Moreover, complex constructions in Mayan languages have often been analyzed (historically as well as diachronically) as involving nominalization. In order to evaluate the status of cha’len it is useful to compare the Chol examples with similar cases in its closest relative, Chontal of Tabasco, as well as other comparative and historical data. This comparative and historical approach reveals both languages moving closer to accusative-systems, a process that is being accelerated through contact with Spanish.


INTRODUCTION.
The Western Cholan languages consist of two languages: Chontal from the state of Tabasco and Chol from the neighboring state of Chiapas.These two languages, t ogether with the Eastern branch found in Guatemala, form the Cholan subgroup, in itself part of the larger grouping of Western Mayan languages.The Cholan languages all display split-ergative systems based on aspect.In addition to this split system, Vázquez claims that Chol also has properties of split intransitivity as an agentive/non-agentive language.He bases this characterization on a group of intransitive verbs that he characterizes as semantically agentive; these verbs always take the Set A person markers.These markers, along with tense/aspect/mood markers, attach to the verb cha'len and are followed an uninflected complement that carries the semantic weight of the construction.
Vázquez characterizes cha'len in these constructions as a verbo ligero or light verb.Matthews defines light verbs as "a verb such as make in make a turn or take a look whose contribution to the meaning of the whole is less specific than in e.g.make a table or take a sandwich." (1997:208) A light verb in many respects seems to be like a modal or an auxiliary, but should be distinguished from them.As its name indicates, a modal changes the mode of a sentence.Payne describes mode as "the speaker's attitude toward a situation, including the speaker's belief in its reality, or likelihood." (1997:244).The constructions in Chol are clearly not modals.Auxiliaries, on the other hand, serve to mark tense or aspect but do not carry semantic information.Payne describes auxiliaries as "verbs in that they satisfy the morphosyntactic definition of verbs (whatever they may be for the language) e.g., they occur in the position of a verb and they carry at least some of the inflectional information (subject/object "agreement" and tense/aspect/mode marking) normally associated with verbs.However, they are auxiliary in that they do not embody the major conceptual activity, state, or activity expressed by the clause.They are semantically 'empty '." (1997:84) He lists various verbs that typologically tend to be grammaticalized as auxiliaries, including stative verbs (e.g.be, stand, sit), motion verbs and complement-taking verbs.In this last category he lists as typical verbs used as auxiliaries say, finish, start, permit, make, force and want.
Light verbs in Chol seem to be auxiliary-like in that they take the aspect and person markers; at the same time their function is different in that they are used only with certain verb-like complements, what Vázquez call the 'unergatives'.These construc tions are similar to auxiliary constructions in that the unergative complements need the help of a verb that is without semantic content in order to bear the necessary inflections.Such an auxiliary is rather different form the sort employed in European languages where the auxiliary is used to express tense/aspect/mood distinctions that the any given verb cannot express by itself.Such auxiliaries work with all lexical verbs.In Chol, however, we are not dealing with a tense/aspect/mood that cannot be inflected on the verb as the majority of Chol verbs directly take these inflections.What we have is a small class of verb-like complements that do not take any inflection and therefore require the assistance of an auxiliary.Thus the verbal phrase has two elements that seem to be evenly dividing duties: the helping verb that carries the grammatical information and the agentive 'verb' that carries the semantic information.This kind of auxiliary is sufficiently different from European style auxiliaries because its appearance is determined by the semantics of the lexical item rather than a need to express a finer grade of tense/aspect/mood.
Having established this difference, I will use Vázquez's term 'light verb' rather than 'auxiliary' to refer to cha'len / chen in these complex constructions.It should be kept in mind that Vázquez divides intransitive verbs into non-agentive and agentive, the latter being the group of verbs that always use cha'len to take their inflections.This characterization is rendered problematic by the fact that cha'len is in other contexts a normal transitive verb; one could therefore argue that the construction is a transitive verb taking a nominalized verb as a complement.The elements that carry the semantic meaning I will refer to as agentive complementizers, or ACs.The purpose of this paper will be to examine the status of these ACs and the effect they have on Western Cholan ergative systems.If they are indeed verbs and if there is an agentive class of intransitives in Chol -i.e. a set of intransitives that always takes accusative marking-then Chol could be seen as moving towards an accusative system of grammar.The language already has split ergativity defined by tense/aspect, and if there is also a system of split intransitivity then accusativity would seem to be creeping into Chol lexically as well as morphosyntactically.In order to evaluate this claim it is necessary to examine three sources of data: 1) data that determines transitivity in Chol itself; 2) data from a similar phenomenon that occurs in Chontal and, 3) Comparative data from other Mayan languages.

LIGHT VERBS IN CHOL 2.1 INTRODUCTION.
Mayan languages prototypically mark intransitive verbs with the Set B marker, the same marker used to indicate objects of transitive sentences.This grammatical relation of intransitive subject (S) with transitive object (O) is an ergative relation as opposed to an accusative relation.Set A markers are used to refer to the subjects of transitive sentences as well as to possess nouns.Chol follows this basic pattern and uses the following person markers: In Cholan languages any given verb will have a status marker suffix that will indicate if the verb is completive or incompletive.In addition to this suffix, a proclitic placed before the inflected verb provides additional aspectual information.The status suffix is obligatory, whereas the pre-verbal aspectual marker/adverb is not.Like all the Cholan languages, Chol presents a system of split ergativity based on aspect.Intransitives in the incompletive aspect take Set A marking and therefore display accusative agreement: (1) Intransitive Pronominal inflection: Completive Vázquez further tests these sentences by using a light verb with the non-agentive and the direct inflection with an agentive.In both cases the result is ungrammatical.These 'verbs' therefore always take the Set A marking.
In addition to these two classes of intransitive, describes a third class that he calls 'ambivalent '.These verbs can either take direct inflection -and Set B in the appropriate aspect-or they can take a light verb, in which case they never take Set B. As an example he offers wäy to sleep: ( One can see from this list that the unaccusative list consists entirely of verbs of motion (interestingly, all of them would take the auxiliary être in French) whereas the unergatives consist of some verbs that are semantically 'active' (i.e. the single partic ipant could be seen as initiating and controlling the activity) along with verbs indicating bodily functions.Although we could question the active nature of the latter from a purely semantic standpoint, from a typological standpoint such verbs seem to rather arbitrarily as a group fall into either the stative or active category in languages that make such a distinction.
Viola Warentkin and Ruby Scott of the Summer Institute of Linguistics have published a sketch of Chol grammar that provides an interesting perspective on Vázquez's description.They list two functions of cha'len, one of which is to express 'intransitive verbal concepts' along with a complement that they call a 'verbal noun'.The second function they list is surprising: they report that cha'len can apparently combine with any verb as another means of forming the progressive: "There is a second way of expressing the progressive aspect.It uses the transitive verb cha'len to do followed by a verbal noun."(1980:74-75) This information is intriguing in that the progressive tense inherently focuses on the action more than the result: because the action is in progress, the success of the activity and its accompanying outcome is unclear.
Moreover, the operation they describe treats verbs differently according to their transitivity.We will discuss this problem below when we compare Chol with Chontal.Vázquez, however, seems to imply that there are verbs that never accept cha'len.For example, he has tested the verbs in the agentive category and finds that they are ungrammatical when they receive direct inflection; in like fashion, he finds the use of the light verb ungrammatical with the non-agentives.The tests he gives us, however, are in the completive: It is clear that we also need to do these tests in the incompletive and especially the progressive.
Leaving aside this issue for the time being, we can portray Vázquez's description of Chol as a kind of 'creeping accusativity' in that the language is split-ergative as well as split-intransitive: Note that this description is based on Vázquez's classification of Chol as an agentive language.In order to examine his characterization we need to examine 1) the transitivity of the verb cha'len, 2) the morphosyntactic properties of ACs as nouns or verbs and 3) transitivity tests for cha'len verbal phrases.

TRANSITIVITY OF CHA'LEN.
Having summarized Vázquez's characterization of intransitivity in Chol, the question that immediately comes to mind is whether he is accurately characterizing [cha'len + AC] as an intransitive verb construction.Given that cha'len also functions as a commonly used transitive verb (with lexical content similar to the Spanish verb hacer), one could easily ask if the construction were not an [auxiliary + intransitive] construction, but rather a transitive verb with a nominalized verb as its argument.A similar argument has been employed for Yucatecan to show that accusative person marking in the incomple tive is explained by the verb being nominalized and the Set A marker acting as a possessor.Moreover cha'len is a transitive verb and Mayan languages in general usually mark transitivity-adjusting operations.
Vázquez in fact does seem to view these intransitives according to this characterization.For example, he analyzes a sentence such as "he sings" as "he does it, the singing": A transitive construction is being used to communicate an idea that, in Spanish or English, would require an intransitive verb.Thus cha'len seems to retain its transitivity.
Another interesting feature of cha'len is that it is used for the infrequent antipassive operations that occur in the language.A suffix -oñ is added to a transitive verb; this is a well-attested antipassive marker found not only in other Cholan languages but attested in Proto-Maya as well.It is interesting to note that the appearance of this marker always requires an accompanying -el nominalizer.The absolute antipassive, therefore, always appears in a nominalized form.Vázquez (2002:82-84)

ACS AS VERBS.
There is strong evidence for ACs as normal verbs that have gained some noun-like properties while retaining some of their status as verbs.For example, the existence of an ambivalent class of verbs seems to indicate that the there are some verbs in the processes of becoming ACs.Vázquez does not indicate any semantic distinction between verbs like way that take direct inflection or use cha'len.The difference between the ambivalent light verb and the agentive light verb is the wäy takes a nominalizing suffix.We could interpret this behavior to mean that wäy has only started taking the light verb relatively recently and is still transparent to speakers as a verb in need of nominalization.We can therefore speculate that the light verb construction will eventually supersede the direct inflection form.When this happens way will be stranded as an AC and speakers will no longer need the -el suffix to distinguish it as a noun.As we shall see, data from Chontal will corroborate this interpretation.An important distinction between Vázquez's unergatives and unaccusatives is that the latter all take the -el nominalizing suffix whereas the former do haphazardly.From this observation we can speculate that those with the suffix are in the process of becoming age ntives, and that those without any suffix are at the final stage of this process and have been stranded; i.e. they are only used as verbal nouns in light verb constructions.The steps might look like the following:

TRANSITIVITY TESTS FOR CHA'LEN VERBAL PHRASES .
Although ACs like k'ay no longer take direct inflection, their status as nominalized verbs is evident when the light verb constructions undergo valency-adjusting operations.A valency-increasing operation thus seems to treat ACs as more verb-like than their more transparently nominalized counterparts.

VALENCY-DECREASING.
If the AC constructions are intransitive, they should not be able to undergo passive or antipassive operations.At this point we should also note a gap in Vázquez's description of Chol.If we examine his list of unergatives, we can immediately see that while there are a few verbs that are prototypically intransitive -to run, for example-many of the verbs seem that they could be transitive in certain contexts.
In fact, verbs like to sing seem just as likely to be transitive as intransitive.It seems unlikely that Chol has no way to say, for example, 'he sings an old song"' or "he shouted the answer" with a clear direct object.Our study of these unergatives will be incomplete without such information.
Given this shortcoming, we can test the unergatives to see if the can passivize.Because cha'len takes the inflectional markers, we should expect it to take any passive derivation.As a root verb (Warkentin and Scott describe root verbs as CVC or CVCVN) it should take the passive suffix -tyel in the completive and -tyi in the completive.Vázquez's study has no example of a passivized cha'len either as a lexical verb or as a light verb.
As stated above, cha'len is used as a helping verb in antipassive constructions, although we do not find evidence of it itself being antipassivized it its hacer role.Lacking further data, we could speculate that the light verb constructions we have been discussing are in fact a kind of antipassive involving noun incorporation; the verbal noun is an indefinite, non-specific noun that is attached to the semantically empty light verb that carries all of the aspectual and person markers.Indeed, Vázquez's own depiction of valency-reducing operations seems open to this interpretation: Contrary to what Quizar and Knowles (1990) and Dayley (1990) affirm, Chol has two antipassive constructions: the absolutive antipassive and the antipassive of incorporation.The first represents institutionalized actions where the patient does not have thematic importance; in the second antipassive the patient is integrated into the verb, forming a compound.These two antipassive forms are also common in other Mayan languages. (2002:134) 2.5 CONCLUSION.From this discussion of Chol we have reviewed evidence both for and against the agentives as verbs and as nouns.It seems that in Chol these elements were originally verbs that have become verbal nouns; i.e. they have the semantics of verbs with some morphosyntactic properties of nouns, the most prominent of which is their role as a complement of a transitive verb cha'len.At this point we can turn to Chontal to provide a further comparative and diachronic perspective.

LIGHT VERB CONSTRUCTIONS IN CHONTAL. 3.1 INTRODUCTION.
Chontal also has a split ergative system in which intransitive verbs in the imperfective take Set A (accusative) marking.There is a further complication in that negation governs Set B marking for incompletive as well as completive.We can summarize the split as follows: In Chol all incompletive intransitives take Set A marking; moreover, the agentives take Set A whatever the aspect.From this point of view Chol could be said to be more "accusative" than Chontal.Chontal uses the following person markers: Like Chol, Chontal also has a verb chen meaning to do, to make that is used with a verblike complement (which I will still refer to as an AC) to create verbal phrases.This verb chen takes the person and the aspect markers while the AC is left uninflected save for a plural clitic that can come at the end of the entire verb phrase.The two main sources of grammatical information on Chontal are 1) Katherine Keller's 1900 SIL dictionary and grammatical sketch of Chontal and, 2) Susan Knowles 1984 dissertation.I will supplement their data with my own field data from the summer of 2003 with the Project for the Documentation of the Languages of Mesoamerica.3

Intransitive
3.2 KELLER'S DISCUSSION OF CHEN.Keller discusses chen, but without using the term light verb.Rather, her discussion focuses on the complement itself which she characterizes as a 'sustantivo verbal' or verbal noun.In her characterization these elements have the morphological and syntactic properties of nouns; the only feature that is verb-like is the semantics.There are three areas in which chen appears: 1. Verbs that are listed as intransitives and function syntactically as such despite having transitive inflection with chen.This category would correspond to Vázquez's "agentives." 2. To detransitivize intransitive verbs in order to focus on the general activity 3.All Spanish Loan words.
We are primarily interested in the first two categories.The first group consists of verbs only formed with chen, while the second can take chen or direct inflection.Keller describes what we have called agentives as verbal nouns that refer to intransitive actions; she points out that they prototypically refer to movements or repeated activities.It is important to note, therefore, that these inherently intransitive actions are only used in chen constructio ns.She lists the following constructions in her dictionary: give fruit Keller alludes to the difficulty in ascribing transitivity values to such constructions: 'Some constructions with a verbal noun are found in the dictionary with chen and are classified as intransitive in spite of chen having transitive verb inflection because the construction functions in the syntax like an intransitive verb.' (1984:474) What is clear, however, is that the ACs listed above do not take direct inflection and are intransitive in the sense that chen does not take a more specific, definite complement.
The second class of intransitive constructions is rather surprising.These verbs have transitive equivalents that in ma ny cases are root CVC verbs.Keller states that the purpose of the light verb construction is to focus attention on the activity itself: "There are some transitive verbs that use the verbal noun en the chen construction to call attention to the action in general.These constructions suppress the mention of a specific complement." (1984:474) According to Keller, the transitive verb is transformed into a verbal noun that no longer takes the verbal aspect marking.Below is a list of these verbal nouns alongside their transitive counterparts:

AC used with chen
Transitive These twenty four transitive verbs undergo a valency-reducing operation in which focus is taken off the object/result and put on the activity.In her discussion Keller does not use the terms antipassive or noun-incorporation, although what she describes seems to match these terms (we should bear in mind that her SIL grammar is attached to the end of a dictionary and is intended for more pedagogical purposes).The deverbalized noun (i.e.verbal noun) acts as a non-definite, non-specific complement that bears the semantic weight of the verb phrase; the verb chen, on the other hand, has no semantic content but carries the aspect markers as well as the person markers.Furthermore, she gives an example of wha t she calls re-transitivization by adding the incompletive status suffix -in: (24)7a laj-0=waj-in-0 (Knowles 1984:154) A2 pat-N=tortillas-B3 'You make totillas.' One is tempted to ask the question, however, if the above VP is really an intransitive; that is, it is an example of an object incorporation antipassive.Of course, one can ask the same question of Keller's reported chen lajwaj-to make tortillas.In Chontal there is an example of the intransitive construction is an example of the complex construction preserving the original verb.chen ak'ot 'to dance' has the direct inflection counterpart ak'otnan that is also intransitive.This intransitive form appears to carry the -n antipassive suffix as well as the usual -an suffix that accompanies derived verbs.However, Keller does not mention any transitive form of the verb-the expected direct inflection of ak'ot.From this unusual situation we can infer several possibilities: 1) the original transitive has been lost, 2) the transitive has not yet been elicited or, 3) the chen construction is not as intransitive as we thought.

PDLMA FIELD DATA CONCERNING
This last item in particular seems to corroborate the previously mentioned path for the growth of a split-intransitive class in Chol: Directly inflected verb?Ambivalent verb: cha'len and direct inflection?Agentive with suffix?Stranded: Agentive without suffix

CHA'LEN/CHEN IN OTHER MAYAN LANGUAGES .
I have not been able to find any example of chen or cha'len in the Mayan Etymological Dictionary; in fact there is nothing that even remotely resembles it.So far it seems that chen / cha'len is an areal feature of the lowlands rather than a feature common to the Greater Tzeltalan group or even the Cholan group; i.e. no evidence of it has been found in the Eastern Cholan languages in Guatemala.Itza' Mayan, however, has a remote past marker uchi which looks like the grammaticalized verbal phrase he did it.In Itza', the 3 rd person set A marker is u before a consonant.Itza' is in the Yucatecan group, a different branch of the Mayan language family but part of the Lowlands area.The Yucatecan languages share many areal features with the Cholan languages, including split-ergative systems determined by tense aspect.In Pacheco Cruz's Yucatecan dictionary úuchi is glossed as 'antiguamente' in former times.

SPLIT ERGATIVITY AND LANGUAGE SHIFT.
Probably the most important factor for determining the present and future status of split-intransitivity in Chol and Chontal has to do with language shift and high rates of bilingualism in Spanish.So far we have not discussed the role of cha'len/chen as an importer of Spanish verbs.These verbs are shorn of their final -r and follow the light verb with no other change or inflection: 7uchi senti wichu7jo7 we7e 'The dogs smelled the meat.' Now what is interesting for our purposes is that both transitives and intransitives are Chontalized this way and that large numbers of Spanish verbs are continuously coming into the language through this very productive process.In my own field data I seem to have an unending supply of them.To give the reader an idea of this phenomenon, we can count up all the intransitive chen constructions in Keller's dictionary and see ho w many of these are loan words: 6. CONCLUSION.Vázquez's characterization of Chol as spilt-intransitive language appears to have much merit to it, although there are many more contexts in which we would like to test its interaction with aspectual systems.His characterization seems applicable to Chontal as there is a cla ss of intransitive verbal nouns roughly equal in size to that in Chol that only takes chen.When we compare the two languages, however, we see that in most cases the individual items in these classes do not match up, indicating the dynamic and ongoing nature of these changes.Moreover, Vázquez's description of Chol has a third class of 'ambivalent s' that apparently display either pattern with no motivation.Chontal, on the other hand, has a class of verbs that take either construction in order to serve certain discourse needs.More research needs to be done on both languages to disambiguate historical change (what appears to be going on in Chol) from discourse function (what appears to be going on in Chontal).Our comparison shows a possible path whereby the light verb takes the nominalized form of the verb as an alternate to direct inflection; over time the verbalized noun becomes stranded in the light verb construction.The result of this process is a steady increase in intransitives that take Set A marking in all environments.This split-intransitivity, combined with splitergativity in the aspectual systems, presents a picture of Western Cholan moving slowly, over time, to an accusative system.As alluded to in the final section, however, this natural process seems to have been accelerated by the impact of Spanish.In the case of Chontal in particular it seems likely that a new generation of bilinguals/semi-speakers could produce a system that is almost entirely accusative.

( 9 )
mi a-cha'len-eñ-ø k'ay (Vázquez 2002:303) INC A2-hacer-SUF-B3 sing 'You sing.' Vázquez claims that there is a group of intransitive verbs called agentives that always take the set A markers and that Chol therefore fits Mithun's definition of an agentive/active language.Significantly, the verbs in the intransitive agentive class all use what he calls a light verb.This verb means to do or make and, when used by itself, is a typical transitive verb.According to Vázquez's analysis, the agentive intransitives are a combination of this light verb and the complement : They provide a list of eleven such verbal nouns:

3 MORPHOSYNTACTIC PROPERTIES OF ACS 2.3.1 ACS AS NOUNS.
Having characterized cha'len as a transitive verb, we need to examine next the status of the ACs.if they are nominalized verbs, they have no affix that wo uld mark a change in transitivity.In Chol nouns are typically nominalized with a suffix -el.So what is k'ay?Is it a verb or a noun?There are arguments for both.It would be interesting to find if this possession can occur in a light verb construction.For example, can we say "I sing his/her song"?The data on the grammaticality of this is lacking.Of course, such a construction would be transitive.Although ACs are possessed like nouns, they cannot function as the predicates of non-verbal sentences like normal nouns:As discussed above, transitive verbs can be detransitivized and nominalized through suffixes and then operate as a nominal complement in verbal phrases.ACs like k'ay seem unable to undergo this transformation.It should be noted that these elements never take Set B marking.To conclude, ACs appear to have some qualities of nouns while lacking others.
'I am a buyer.'Havinglost its verbal status, the antipassived nominal requires a light verb to form verbal sentences.It is interesting to note how Vázquez glosses such a construction: (14) Antipassive nominalization as complement of light verb (Vázquez 2002:267) mi k-cha'len-eñ-ø I have not found examples in Vázquez of attempts to passivize the light verb constructions.With the information we have we can conclude that the cha'len remains a transitive verb in all environments.2.
The above example could be glossed as 'I give you the falling'.In Chol therefore such a construction is a transitive construction with the verb serving as the nominalized complement.Agentives, however, act more like verbs in two ways: the light verb is no longer used and a tyi marker, employed to subordinate verbs, is used: (20)Agentive with Causative: Set A (Vázquez 2002:43) mi k-äk'-ety tyi k'ay INC A1-give-B2 SUBD sing 'I make you sing.' This noun-incorporation antipassive is similar to what we suggested at the end of the discussion of Chol concerning what Warkentin and Scott called progressive constructions.The main difference between the two is that they seem to suggest that all Chol verbs can take the light verbs, whereas Keller gives us a list of only twenty-four verbs that undergo this operation.I have not included in the above list a group of chen constructions which are clearly compounds.Keller also classifies these constructions as intransitive:Of these six compounds, the first two are straightforward combinations of a transitive root verb with a root noun.The third item, aläs t'an , is a combination of two verbal nouns that we have already encountered.The fourth item is a combination of a normal noun with the verbal noun t'an.The remaining two compounds have components that are less transparent.We shall see that Knowles has a different approach to analyzing these chen compounds.What is particularly intriguing is her discussion of the compounds that we had discussed above.Keller had listed chen lajwaj an intransitive verb meaning "to throw tortillas".Knowles, however, states that "the incorporation of the object into the verbal complex produces a nominal, not an intransitive verb" (1984:153) She gives the following examples: CHEN.My own field data from 2003 with the Project for the Documentation of the Languages of Mesoamerica provides more perspective on this problem.It should be kept in mind that the dialect represented in this data is much closer to that described by Keller than to that described by Knowles.Below is a list of all examples of chen in my data, excluding the Spanish loanwords: ACs in Chol that only take cha'len but take both direct inflection as well as chen in Chontal:We can see clearly the nominalizing suffix -el in the Chol example, but it looks like in Chontal the form has eroded phonologically after it was stranded because speakers will no longer need the -el suffix to distinguish it as a noun.This form is probably cognate with Kaqchikel oq'-to cry.4.The Chol verb chu7 -'to suck' is an interesting case.It needs a light verb in Chol, but in Chontal it is not only found without the light verb, but is one of the few verbs that detransitivize using the rare antipassive suffix.5.The body function verbs are restricted to cha'len in Chol but take both in Chontal and are possessed: