Competing constructions in Kaqchikel focus contexts

1 Introduction

The present study uses experimental techniques to investigate the distribution of the agent focus (AF) construction in Kaqchikel (ISO 639-3: cak, Glottocode: kaqc1270), a Mayan language belonging to the K’ichean subgroup. The usual generalization is that in Eastern Mayan languages there are restrictions on the direct relativization, questioning, and/or focusing/clefting of ergative arguments (see e.g., Aissen 2017a; Campbell 2000; Dayley 1981; Larsen and Norman 1979; Stiebels 2006; Trechsel 1993) (for comments on Western Mayan see Section 5). To circumvent these restrictions, Mayan languages commonly use either antipassives or AF, a construction unique to Mayan (described below). In addition to using AF or antipassive constructions to relativize, question and focus/cleft ergative arguments, in Kaqchikel these constructions are also used in indefinite argument contexts (‘a girl hugged the boy’), existential indefinite contexts (‘someone hugged the boy’), and negative indefinite contexts (‘no one hugged the boy’), only when the relevant argument is the subject of a transitive verb (all six environments are henceforth discussed together as “focus contexts”).^[1] Examples and a discussion of each of these focus contexts are provided in Section 3.

The AF construction in Kaqchikel, as in other Mayan languages, has been characterized as morphologically intransitive but syntactically transitive (see e.g., Aissen 1999: 452). AF in Kaqchikel has the following key characteristics:

1. The verb only cross-references one of its two core arguments, which can be either the agent or the patient depending on the respective person and number of the participants;

2. One of the arguments must be third person (AF is not available with two local person arguments);

3. The construction is indicated by a verbal suffix which is -o for root transitives and -n for derived transitives;^[2]

4. It is used in focus contexts involving A (ergative) NPs, and cannot be used when focusing O or S (absolutive) NPs;

5. The patient NP is not oblique, and it can be overt or pro-dropped (syntactically present but unpronounced).

An example of AF in Kaqchikel is given in (1a), where the agent is focused, which contrasts with the transitive construction in (1b), where the patient is focused.

In addition to the AF construction, Kaqchikel has an antipassive construction which only appears in the same limited syntactic environments as AF constructions. This antipassive shares a marker with AF – -o for root transitive verbs, -n for derived transitive verbs – but unlike AF, the patient appears in an oblique phrase and the verb invariably cross-references the agent. It also can be used with two local persons.^[5]

AF and this antipassive construction both function the same way with respect to questioning, relativizing, etc., and speakers can use both constructions interchangeably in the tasks reported on here (although AF is approximately four times as frequent as the antipassive in these contexts in Kaqchikel; see this dataset and Stiebels [2006: 513] on other Mayan languages). As such, both structures will be discussed together here under the umbrella of “AF/AP”.

There is an extensive body of formal syntactic literature on AF which informs expectations about how AF operates in Mayan languages. To frame the discussion in generative terms, some Mayan languages have an extraction asymmetry wherein ergative arguments are unable to undergo A-bar extraction, which involves movement of the transitive subject from the postverbal subject argument position to the preverbal focus position (Mayan languages are verb-initial). Several analyses have derived this pattern from the way nominal arguments are case-licensed (see e.g., Coon et al. 2014). Such case-licensing explanations consider AF a necessary “last resort” (see e.g., Campana 1992; Coon et al. 2014: 219; Ordóñez 1995; see also Aissen [2017a] for a summary and comments), which precludes the possibility that it could be optional in those contexts where it appears.^[6] Although she proposes a different analysis, Aissen (2011: 10) states with respect to K’iche’ that “It is a general property of [K’iche’] that the transitive form is possible (when A is extracted) only when the AF form is impossible”, which likewise indicates that AF and transitive verbs should be in complementary distribution. This same type of generalization is also found in the descriptive literature for Kaqchikel (García Matzar and Rodríguez Guaján 1997: 379).

Despite the above statements, Stiebels (2006: 510–511) mentions that AF/AP constructions are not always obligatory in at least three Mayan languages (Mam, Poqomam, Poqomchi’), such that the direct relativization, questioning, etc. of the ergative argument of a transitive verb is sometimes possible in some of these focus contexts. Additional evidence along these same lines comes from Heaton et al. (2016) who found that while questioning the subject of transitive verb in Kaqchikel almost always involves the use of an AF/AP construction, they rarely appear in picture elicitation contexts when relativizing transitive subjects.

The present study builds on the findings of Heaton et al. (2016) to provide a more complete picture of the use of AF/AP constructions in Kaqchikel by documenting the incidence of AF/AP versus transitive constructions in four additional focus contexts. This paper describes the results of four experimental tasks designed to elicit AF/AP clauses in contexts which permit them. The data presented here address the following questions:

1. With what frequency do Kaqchikel speakers use AF/AP constructions in focus contexts? Are there any contexts in addition to relativization in which a transitive construction is more frequent than AF/AP constructions?

2. Are there any other factors which influence the use/non-use of AF/AP constructions?

3. What generalizations can be made about contexts where AF/AP constructions appear less frequently?

4. How do the facts for Kaqchikel compare to descriptions of the distribution of AF/AP constructions in other Eastern Mayan languages?

An explanation of the design of all four tasks is presented in Section 2, and each task and their results are detailed in Section 3. Section 4 provides a summary of results across tasks and presents possible explanations for the observed patterns. Evidence of similar phenomena in other Mayan languages is given in Section 5, and Section 6 offers concluding remarks.

2 Task design

There are six primary focus contexts in which AF/AP constructions appear in Kaqchikel: wh questioning A arguments, relativizing A arguments, focusing/clefting A arguments, indefinite A, existential indefinite A (‘someone/something’), and negative indefinite A (‘no one/nothing’) contexts. The expectation based on the literature to date has been that AF/AP constructions are required in all of these contexts in Kaqchikel. To elicit subject and object wh questions and relative clauses, Heaton et al. (2016) used picture description tasks which carefully manipulated the context to make specific comparisons. The present study uses the same basic methodology to investigate which construction(s) are used in the remaining four contexts. The benefit of picture elicitation is that it allows the researcher to control contextual variables while avoiding translation, and allows the speaker to provide a more naturalistic response.

The stimuli for all four tasks were designed to invite single-sentence responses to each image. The question was, when presented with the opportunity to use an AF/AP construction, would speakers use one, or would they (and could they) use a fully transitive construction. The possible responses to a sample existential indefinite A subject item are given in (2).

As suggested by the glosses for (2b), if it were possible to use a transitive verb to say ‘someone is waking the girl’, k’o nuk’asoj ri xtän could be interpreted as having either a focused agent or a focused patient. Some Mayanists have argued that the function of AF/AP is to disambiguate between an agent and a patient interpretation in just these sorts of situations (see Section 4.2 for more discussion). This cannot be the whole story, since AF/AP constructions are used with local persons in Kaqchikel. However, since the goal was to encourage the production of AF/AP clauses, all of the scenarios depicted in the picture elicitation tasks involved singular third persons acting on singular third persons. If the primary function of AF/AP constructions is to disambiguate between an agent and a patient interpretation, we would expect them to appear in 3 > 3 contexts like these.

All four tasks involved the same basic design which manipulated both the animacy of the patient (human vs. inanimate) and the focused argument (the agent vs. the patient). Each test consisted of 20 target items, five from each of the four conditions generated by the combination of these two variables, illustrated in Table 1.

The two conditions which targeted focused patients are essentially control items, where transitive verbs are expected and AF/AP verbs are ungrammatical. However, passives were also acceptable in focused patient conditions, and the rate of use for passives is discussed where relevant to the results. Animacy was included as a parameter because it is relevant to discussions of ambiguity (see Section 4.2), even though the animacy of the patient was found not to be a significant factor in Heaton et al. (2016).^[7]

In all tests, one block of items with a focused agent was presented first in order to avoid priming effects from the expected use of transitive verbs in patient conditions. However, there were no discernable priming effects of this kind; responses in the initial block of focused agent items were not notably different from responses to those focused agent items which appeared after focused patient items. The four conditions in Table 1 were interspersed as blocks with respect to matched versus mismatched animacy. The five items in each of the four conditions were randomly interspersed within their condition. Each item was introduced by a prompt that where possible included neither a transitive verb nor an AF/AP construction, likewise to avoid any potential priming effects. The prompts for each task are described in the following subsections.

All tasks were conducted by the researcher in Kaqchikel. Interviews took place either in speakers’ homes or in their places of work, and speakers were invited to participate via existing social networks. Thirty native Kaqchikel speakers participated in all four tasks. Since Heaton et al. (2016) found age to be a significant factor, speakers were equally distributed across three different age categories: ten speakers between 20 and 30, ten speakers between 31 and 50, and ten speakers over the age of 51.^[8] These categories broadly capture changes in the educational landscape of rural Guatemala, where older people were less exposed to compulsory Spanish-medium education.

In preparation for each task, speakers were asked to describe pictures in a single sentence in response to a prompt. Participants were trained on the format for each task using test items that involved intransitive scenarios. All four tasks were presented in the same session, but with small breaks between each task. Each individual task took speakers between five and 10 minutes to complete. The four tasks were presented in the following order: indefinite arguments (‘a(n)’), focus words (‘only’), negative indefinite items (‘no one/nothing’), then existential indefinite items (‘someone/something’). This order reflects a progression from simple to more complex tasks (see Section 3), while also separating the indefinite and existential indefinite tasks which used similar prompts and images. After all four tasks were completed, participants were asked several short follow-up questions: if they had predominantly used AF/AP constructions in focused agent conditions, could they also use a transitive verb (and vice-versa)? If so, is there any difference in meaning/use? Do they perceive one as being more common/acceptable than the other?

Responses were coded as either AF/AP (1) or transitive (0). Responses that were infelicitous, ungrammatical, or did not contain all the necessary elements (or were questionable on these points) were omitted from the count. See the accompanying documentation for more information.

3 Individual tasks and results

3.1 Focus word task (xa xe ‘only’)

Focus in Kaqchikel can be indicated in several ways, including focus particles (e.g., ja ‘foc’) and other focus words (e.g., xa xe ‘only’). This task used xa xe ‘only’, which avoided serious overlap with cleft-type focus constructions/relative clauses (‘this is the man that …’), which were already tested by Heaton et al. (2016). The focus word task involved sentence correction, where the prompt included an incorrect statement, and speakers were instructed to correct that statement by saying ‘No, only the [agent/patient] …’. A sample item is given in Figure 1.^[9]

Figure 1:

Focus word task: Human→human, focused agent.

Prompt:

(3)

k ’o	jun	ala’,	jun	xtän,	chuqa	jun	achin.	Chupam	ri	achib’äl,
Exist	one	boy	one	girl	also	one	man	in	det	picture,
n-Ø-ki-jïk’	ri	xtän	che	ka’i’.
ncompl-3sg.abs-3pl.erg-pull	det	girl	prep	two
‘There’s a boy, a girl and a man. In the picture, the two of them are pulling the girl.’

Target response:

(4)

manäq,	xa xe	ri	ala’	n-Ø-jik- o	/ (*)n-Ø-u-jïk
no	only	det	boy	ncompl-3sg.abs-pull- af	ncompl-3sg.abs-3sg.erg-pull
ri	xtän.
det	girl
‘No, only the boy is pulling the girl.’

The expectation was that speakers would use an AF/AP construction in the focused agent conditions, and that a transitive verb would be ungrammatical, indicated by (*) in (4) and subsequent examples. Responses involving transitive verbs in focused agent conditions would be unexpected.

3.1.1 Results

The number of responses that involved the use of AF/AP constructions versus fully transitive constructions in agent conditions with xa xe ‘only’ by age group are given in Figure 2. Most notably, transitive verbs are used much more frequently than AF/AP constructions by members of all age groups.^[10]

Figure 2:

Frequency of AF/AP versus transitive verbs in agent conditions with xa xe ‘only’ by age.

Example (5a) provides a typical response to focused agent items which involved the use of a transitive verb (as in the majority of responses), while (5b) provides an example of the corresponding response involving the use of an AF construction (as in a minority of responses). This particular item depicted a woman pushing a boy on a swing set while a man stood nearby with his hands in his pockets.

(5)

Typical transitive response:

a.

xe	ri	te’ej	n-Ø-u-chokomi-j	ri	ala’.
only	det	mother	ncompl-3SG.ABS-3sg.erg-push- tr	det	boy
‘Only the mother is pushing the boy.’ [associated audio-5a-Heaton.wav]
Typical AF response:

b.

xa xe	ri	ixöq	n-Ø-chokomi- n	ri	ala’.
only	det	woman	ncompl-3sg.abs-push- af	det	boy
‘Only the woman is pushing the boy.’ [associated audio-5b-Heaton.wav]

Recall that the focused patient conditions were essentially control items where AF/AP constructions are not possible and transitive verbs are expected. However, there were 38 instances of passives in patient conditions for the xa xe ‘only’ task, one of which is given in (6).

(6)

xa xe	ri	xtän	n-Ø-q’et- ëx	r-uma	ri	xta	ala’.
only	det	girl	ncompl-3sg.abs-hug- pass	3sg-obl	det	dim	boy
‘Only the girl is being hugged by the boy.’ [associated audio-6-Heaton.wav]

This use of passives in patient conditions suggests that transitive verbs are indeed commonly used to indicate that the preverbal NP is the agent in this context, since some speakers feel the need to disambiguate via passivization when the preverbal NP is the patient. This contrasts with few to no attestations of passives in other contexts where AF/AP constructions consistently appear; see for example the results of the existential indefinite task in Section 3.2.1.

3.2 Existential indefinite task (k’o ‘exist’)

Although the English translation of this construction involves an indefinite pronoun ‘someone/something’, the corresponding target construction in Kaqchikel does not involve an overt pronoun. To express the meaning ‘someone/something’, Kaqchikel uses k’o ‘to exist’ followed by the main verb (as in [8]). To elicit an indefinite, non-specific referent, participants were shown pictures of transitive actions where one person or item was almost entirely obscured by a ‘wall’, as in Figure 3. Participants were trained to begin their statements with k’o as part of the format of the task.^[11] The prompt for each item was simply ‘what is happening with ___?’, a statement which in Kaqchikel uses a passive verb.

Figure 3:

Existential indefinite task: Human→human, focused agent.

Prompt:

(7)

a chike	n-Ø-b’an-täj	r-ik’in	ri	ala’?
wh	ncompl-3sg.abs-do-pass	3sg-with	det	boy
‘What is happening with the boy?’

Target response:

(8)

k’o	n-Ø-chokomi- n	/ (*)n-Ø-u-chokomi-j	ri	ala’.
exist	ncompl-3sg.abs-push- af	ncompl-3sg.abs-3sg.erg-push-tr	det	boy
‘Someone is pushing the boy.’

3.2.1 Results

Results show that AF/AP constructions are by far the dominant strategy used with existential indefinite subjects for all age groups (Figure 4). Although transitive constructions were occasionally used in agent conditions by speakers in every age category, when asked about those responses afterward, most speakers rejected the grammaticality of such statements. As such, it seems AF/AP constructions are very common and perhaps mandatory for most speakers in this context. Additionally, passive constructions were unattested in patient conditions.

Figure 4:

Frequency of AF/AP versus transitive verbs in agent conditions for the existential indefinite context by age.

An actual response typical of the use of AF in this context is given in (9a) (the most common strategy), while an actual response to the same item involving a transitive verb is given in (9b) (as in 20 responses). This particular item depicted an obscured figure picking a flower.

(9)

Typical AF response:

a.

k’o	n-Ø-k’uq- un	la	kotz’i’j.
exist	incompl-3sg.abs-pick- af	dem	flower
‘Someone is picking the flower.’ [associated audio-9a-Heaton.wav]
Minority transitive response:

b.

(*)k’o	n-Ø-u-k’üq	ri	kotz’i’j.
exist	incompl-3sg.abs-3sg.erg-pick	det	flower
‘Someone is picking the flower.’ [associated audio-9b-Heaton.wav]

3.3 Negative indefinite task (majun [achike ta] ‘no one/nothing’)

The Kaqchikel structure for ‘no one/nothing’ involves the negative ma(n) plus jun ‘one’, optionally followed by the wh word achike ‘who/what’ and the irrealis particle ta, literally something like ‘not one (who) X’. To elicit ‘no one/nothing’ statements, the negative indefinite task involved two contrasting pictures, one of which depicted a transitive action, the other of which depicted the same action but lacked either the agent (for focused agent items) or patient (for focused patient items). A sample item targeting the agent is provided in Figure 5, along with the prompt and target response.

Figure 5:

Negative indefinite task: Human→human, focused agent.

Prompt:

(10)

w awe’	ri	ala’	n-Ø-ch’eqeb’-äx	r-oma	ri	xtän.	Po	re
Here	det	boy	ncompl-3sg.abs-soak-pass	3sg-obl	det	girl	but	this
ala’	re’,	ma-jun	(achike	ta)…
boy	this	neg-one	wh	irr
‘Here a boy is being soaked by a girl. But this boy, no one ….’

Target response:

(11)

ma-jun	(achike	ta)	n-Ø-ch’eqeb’a- n
neg-one	wh	irr	ncompl-3sg.abs-wet- af
/ (*)n-Ø-u-ch’eqeb’a-j	(ri	ala’).
ncompl-3sg.abs-3sg.erg-wet-tr	det	boy
‘No one is soaking the boy.’

The prompt includes a passive verb in the first sentence to avoid priming a transitive or AF/AP construction. Participants were trained to expect their utterance to begin with majun (achike ta) ‘no one/nothing’, which was also provided at the end of the prompt to help dissuade speakers from providing a passive response. The figures in the background of the second image represent possible agents for the action, likewise to help retain a two-participant interpretation of the depicted scenario.

3.3.1 Results

All speakers who participated in the study used AF/AP forms exclusively in negative indefinite subject contexts across every age group.^[12] The follow-up questions also corroborated that AF/AP constructions are mandatory in this context, and there were only six instances of passive verbs in patient conditions. The total numbers of responses by age and construction type are given in Figure 6.

Figure 6:

Frequency of AF/AP versus transitive verbs in agent conditions for the negative indefinite context by age.

An actual response typical of the use of AF in this context is given in (12). In this particular item, the first picture showed a girl burying a boy in the sand at the beach. In the second picture, a boy was lying in the sand, unburied, with two other children in the background.

(12)

ma-jun	achike	ta	n-Ø-muq- un .
neg-one	wh	irr	ncompl-3sg.abs-bury- af
‘No one buried him [in the sand].’ [associated audio-12-Heaton.wav]

3.4 Indefinite argument task (jun ‘one, a(n)’)

As noted by Broadwell (2000: 3–4), indefinite subjects of transitive verbs appear preverbally and trigger the use of AF (13a). Additionally, indefinite transitive subjects cannot appear post-verbally (13b).

(13)

a.

jun	tz’i’	x-Ø-b’a’- o	ri	a	Juan.
one	dog	compl-3sg.abs-bite- af	det	clf	Juan
‘A dog bit Juan.’

b.

*n-Ø-u-tïj	ri	saq’ul	jun	ak’wal .
ncompl-3sg.abs-3sg.erg-eat	det	banana	one	child
Intended: ‘A child is eating the banana.’

These facts were corroborated by the author with several Kaqchikel speakers in elicitation, and similar textual examples like (14) are not uncommon. In this context, the weaver is new to the discourse, and the listener does not know the woman.

(14)

jun	ixöq	x-Ø-kem- o	pe	la	qa-po’t.
one	woman	compl-3sg.abs-weave- af	dir	dem	1pl.poss-blouse
‘A woman wove our blouses.’ [associated audio-14-Heaton.wav]

As such, indefinite A contexts merit investigation as focus contexts in Kaqchikel.

The task for indefinite As, like the existential indefinite A task, involved the description of a single picture in which a figure in an action was obscured (cf. Figure 7). However, the ‘wall’ that covered the figure was smaller in this task, which encouraged participants to use an NP like ‘a girl’ or ‘a man’, since they could identify some but not all of the features of the figure (although any indefinite NP was accepted). Participants were trained to consider obscured figures indefinite.

Figure 7:

Indefinite task: Human→inanimate, focused agent.

Prompt:

(15)

a chike	n-Ø-b’an-täj	chupam	ri	achib’äl?
wh	ncompl-3sg.abs-do-pass	in	det	picture
‘What is happening in this picture?’

Target response:

(16)

jun	ala’	n-Ø-tij- o	/ (*)n-Ø-u-tïj	ri
one	boy	ncompl-3sg.abs-eat- af	ncompl-3sg.abs-3sg.erg-eat	det
saq’ul.
banana
‘A boy is eating the banana.’

It is important to note that the indefinite A context differs in key ways from the other contexts discussed here. Unlike indefinite A NPs which must appear preverbally and which involve the use of AF/AP constructions (as in 13), indefinite Ps do not have to be preverbal; in fact, indefinite Ps are almost always postverbal, and preverbal indefinite Ps are sometimes judged ungrammatical. This is a point of departure from the other focus contexts discussed here, where questioned, relativized, clefted, etc. objects must be preverbal. Compare the grammatical postverbal indefinite patient in (17a) with the ungrammatical postverbal wh patient in (17b):

(17)

a.

n-Ø-u-tïj	jun	saq’ul	ri	ak’wal.
ncompl-3sg.abs-3sg.erg-eat	one	banana	det	child
‘The child is eating a banana.’

b.

*n-Ø-u-tïj	achike	ri	ak’wal?
ncompl-3sg.abs-3sg.erg-eat	wh	det	child
Intended: ‘What is the child eating?’

Additionally, it is possible in Kaqchikel for both arguments of a transitive verb to be indefinite, in which case the expectation is that the indefinite subject will appear preverbally, the verb will show AF/AP morphology, and the indefinite patient will be postverbal, as in (17c). This follows from the fact that an indefinite P NP need not appear in the preverbal focus position.

c.

jun	ak’wal	n-Ø-tij- o	jun	saq’ul.
one	child	incompl-3sg.abs-eat- af	one	banana
‘A child ate a banana.’

From the point of view of the other tasks, utterances like (17c) are akin to simultaneous agent and patient condition items, which was an impossibility in the other three tasks. Since definiteness of the patient was immaterial to the definiteness and syntactic position of the agent, responses in patient conditions were tallied according to the relevant variable, the definiteness of the agent.

3.4.1 Results

Despite speakers’ stated grammaticality judgments which favored AF/AP constructions, all speakers used transitive verbs in contexts with indefinite A arguments (Figure 8). In fact, the use of AF/AP clauses was rare for speakers of all ages, and non-existent for the youngest speakers (although all speakers agreed AF/AP constructions were also possible during post-test questioning).

Figure 8:

Frequency of AF/AP versus transitive verbs in agent conditions for the indefinite argument context by age.

A sample response from this task involving a transitive verb (the majority strategy) is given in (18a), while one of the nine actual responses involving AF is given in (18b). This particular item depicted an obscured person holding scissors, with a cut-up sheet of paper to their right.

(18)

Typical transitive response:

a.

jun	ala’	x-Ø-u-qupi-j	la	wuj.
one	boy	compl-3sg.abs-3sg.erc-cut- tr	dem	paper
‘A boy cut the paper.’ [associated audio-18a-Heaton.wav]
Minority AF response:

b.

jun	ti	winäq	n-Ø-qupi- n	ri	wuj.
one	dim	person	incompl-3sg.abs-cut- af	det	paper
‘A person (dim.) is cutting the paper.’ [associated audio-18b-Heaton.wav]

These data show that AF/AP constructions are clearly not obligatory when a transitive subject is indefinite; to the contrary, they are rarely used.

There were also eight instances of passives in the indefinite P conditions, an example of which is given in (19).

(19)

jun	xta	ala’	n-i-muq	pa	arena	roma	jun	ti	xtän.
one	dim	boy	incompl-3sg.abs-bury. pass	prep	sand	3sg-obl	one	dim	girl
‘A little boy is getting buried in sand by a little girl.’ [associated audio-19-Heaton.wav]

The proportion of passive responses in object conditions is smaller with indefinite patients than with focused patients in xa xe conditions (6 vs. 16 %, see Section 3.1.1). This is likely related to the fact that indefinite objects are rarely if ever preverbal.

4 Analysis

4.1 Summary of findings

AF/AP and transitive constructions in these additional four focused A contexts are clearly used to different extents. The results of the four tasks discussed here as well as the two tasks from Heaton et al. (2016), which now provide a more complete picture of the use of AF/AP constructions in these contexts in Kaqchikel, are summarized in Figure 9.

Figure 9:

Summary of use of AF/AP versus transitive constructions with a focused preverbal agent, as percentages.¹³

In three of the six contexts examined here – A wh questions, negative indefinite A contexts, and existential indefinite A contexts – a majority of speakers consistently or exclusively used AF/AP constructions. Speakers (even those who sometimes produced responses with transitive verbs in agent conditions) also consistently judged transitive constructions to be ungrammatical in these contexts.

The results of the wh question, existential indefinite, and negative indefinite tasks contrast with the results for the other three contexts – relative clauses, focus word, and indefinite contexts – in which speakers of all ages frequently used transitive constructions in focused agent conditions, to the point where transitive constructions were the most common way of expressing those types of propositions. There is also evidence of differences between these three focus contexts, where AF/AP constructions are more common with focus words and relative clauses than with indefinite As.^[13]

To identify meaningful differences between the test results for these different focus contexts, the data were fit to a Bayesian logistic regression model (estimated using six chains of 10,000 iterations and a warmup of 1,000) with weakly informative priors. Results show a meaningful difference between the pattern exhibited for A wh questions (intercept) and all other tests (0 % of the posterior distribution within the region of practical equivalence, with a range of [−0.18, 0.18] and a 95 % confidence interval) with the exception of existential indefinite As, which as shown in Figure 9 have very similar rates of AF/AP usage. There is also a strong negative effect for the patterns of use for AF/AP constructions in focus word, relative clause, and indefinite A contexts (in order of increasing strength) compared to the other three contexts. The magnitude of these differences is illustrated in Figure 10, where Type = 1 represents AF/AP and Type = 0 represents transitive constructions.

Figure 10:

Conditional effects of the focus context on AF/AP use.

However, it is important to remember that although less frequent than transitive constructions, AF/AP constructions continue to be possible in A relative clause, A focus word, and indefinite A contexts. But the fact that none of these three contexts consistently involves AF/AP is surprising, given that the expectation from the literature was that AF/AP is always required in the contexts in which it appears (cf. Dayley 1981; García Matzar and Rodríguez Guaján 1997; Trechsel 1993).

5 Examples beyond Kaqchikel

As noted in Heaton et al. (2016: 45), comparative evidence suggests that this type of variability in the use of AF/AP constructions in different contexts appears elsewhere in Mayan. However, it is useful to say more about the specific nature of that variation and how it relates to the proposed analysis. For Mam, England (1983: 214–218) describes a pattern very similar to what we find here for Kaqchikel in that antipassivization is required for wh questions, but optional for argument focus and relative clauses. Relatedly, Ayres (1983: 31–33) notes for Ixil that while AF (referred to as TSI ‘transitive subject indexing’) is mandatory in wh questions and in contrastive focus situations, it is optional in relative clauses. As for other K’ichean languages, Mondloch (1981) includes examples such as the utterances in (30) which show that AF/AP constructions are possible but not mandatory in A wh questions.

Mondloch (1981: 245) states that although AF and AP constructions are the more common strategies for forming A wh questions, at times transitive constructions are used and context must disambiguate. The same variation in the use of AF/AP constructions and transitive verbs in A wh questions in K’iche’ is recorded by Larsen (1987: 45). Mondloch also notes with respect to relativization that “sentences […] with relativized ergative subjects, even though they do occur, are often judged unacceptable by reflective [K’iche’] speakers” (1981: 228). He provides examples of transitive clauses with relativized ergative subjects like (31a) which contrast with the use of AF in relative clauses, as in (31b).

This suggests that K’iche’, like Kaqchikel, Ixil, and Mam, allows ergative arguments to be relativized, even if the AF/AP form is often considered the preferred (or only) form in elicitation.

Along these same lines, Du Bois (1981: 243, 252–253) describes AF/AP constructions as optional in A wh questions and relative clauses in Sakapultek, and Barrett (1999: 266–267) notes that AF/AP constructions are not obligatory when relativizing A arguments in Sipakapense. For Tz’utujil, Dayley (1985: 348) states that although AF/AP constructions are almost always used when relativizing A arguments, there are textual examples where transitive verbs also appear in that context. Can Pixabaj (2007) provides examples of relative clauses in Uspantek where an A argument can be relativized using a transitive verb, which she claims is because there is no ambiguity between which argument is the agent and which is the patient. She also describes both transitive and AF/AP constructions as being possible in focus word contexts, even when there is the possibility of ambiguity. She additionally notes that there is a tendency to interpret the first element in the sentence as the agent, which tracks with the data and explanation presented here for Kaqchikel.

The facts for the remaining Eastern Mayan languages are less well-documented. Smith-Stark (1978: 171) describes AF/AP constructions in Poqomam as obligatory in wh questions with third person patients, optional in the relative clause and focus word contexts in which they appear, and optional if the agent is a negative indefinite pronoun and the patient is a third person. Although I have not located any statements about the possibility of transitive verbs where we would otherwise expect AF/AP constructions in Poqomchi’, there are some transitive examples where Brown (1978: 147) claims the preverbal agent is focused (although they could be topics). I have found no mention to date that suggests the use of AF/AP constructions in Q’eqchi’, Achi, Tektitek, or Awakatek are not obligatory.

Given the above statements and examples, it seems that it is in fact common in K’ichean, and possibly Eastern Mayan in general, for there to be variability in the distribution and use of AF (and other specialized constructions) and transitive constructions, both between different syntactic contexts and within a particular context. Specifically, for the majority of the languages discussed above, both AF/AP and transitive verbs can be used to relativize A arguments. However, there are also significant differences between Kaqchikel and other closely related languages. For example, while AF is optional in Kaqchikel with focus words, it is purportedly mandatory in the same context in closely related Tz’utujil. Also, while the use of AF is obligatory in A wh questions in Kaqchikel and Tz’utujil, AF appears to be optional when questioning A arguments in K’iche’ and Sakapultek. Because patterns in the use of AF are not uniform across Eastern Mayan, the explanation proposed here for Kaqchikel with respect to the divide between constructions that are introduced by a lexical NP and those that are not cannot apply in the same way even to other closely related languages. This complicates the identification of the underlying motivations for AF, but it is worth noting that this is perhaps not unusual; information and argument structure differences between even closely related Mayan languages has been identified in other work (see England and Martin 2003), and competition approaches allow the same pressures to lead to different grammaticalized outcomes in different languages (cf. e.g., Bickel 2007: 241; Dryer 1998; Haspelmath 1999).

It is also relevant to provide a brief summary of the facts for Western Mayan languages as they pertain to the observations made here. Although many Western Mayan languages do not have syntactic restrictions on ergative arguments, some still have AF constructions. The most prominent example is Tsotsil, for which Aissen (1999) has suggested that AF serves an ‘inverse’ function. In Tsotsil, AF is used when the agent is focused and the patient is more animate, definite, or individuated than the agent. When that is not the case, transitive verbs tend to be used. Although the characteristics of AF in Tsotsil differ in significant ways from AF in Kaqchikel, it is notable that both languages are more likely to allow transitive verbs when the subject is a clear agent.

There is also a significant literature on the syntax of focus in Yucatec (Bricker 1978; Gutiérrez-Bravo 2017; Gutiérrez-Bravo and Monforte 2011; Norcliffe 2009; Norcliffe and Jaeger 2016; Tonhauser 2007; Verhoeven and Skopeteas 2015; inter alia). Yucatec has an AF-type construction that appears when questioning, relativizing, or focusing the agent of a transitive verb, but transitive constructions are also common when relativizing A arguments (cf. Bricker 1978). Gutiérrez-Bravo and Monforte (2011: 261) state that “there is no structural unity between the contexts and conditions where AF is observed” and that “it is problematic to instead assume that AF [in Yucatec] is triggered by a specific syntactic configuration shared by agent focusing, interrogative fronting, and relativization”. Rather, they argue that AF is used in relative clauses along with passivization to disambiguate subject and object interpretations. While I argue against any direct avoidance of ambiguity in Section 4.2.2 as a motivating factor for the Kaqchikel data, the relevant point is that there is evidence from elsewhere in the family that information structure and other contextual factors have bearing on the use of AF, beyond the structural constraints of A-bar movement.

In addition to these data from Mayan languages, Tsunoda (1988: 43–45, 48) provides some evidence that it is perhaps not uncommon for there to be exceptions to syntactic restrictions on ergative arguments. His data come primarily from Pama-Nyungan languages, namely textual studies of Dyirbal and Warrungu. In Warrungu, coreferential NP deletion involving an ergative argument requires an antipassive in the vast majority of cases. However, coreferential NP deletions involving an ergative argument involve transitive verbs rather than antipassives about 5–10 % of the time (1988: 43). There are also a handful of examples in Dyirbal coordination and what Tsunoda calls “sentence-sequence” which exhibit the same type of variability (1988: 44). Although examples in these languages of transitive verbs in contexts where they are not expected are not nearly as extensive as those documented here for Kaqchikel, they do constitute evidence outside of Mayan for the availability of multiple constructions in contexts commonly involving restrictions on ergative arguments.

6 Conclusions

The data presented here provide novel evidence regarding the distribution of AF/AP constructions in Kaqchikel in the primary syntactic contexts in which they appear. Findings show that AF/AP constructions are consistently used in wh questions, existential indefinite contexts, and negative indefinite contexts. While AF/AP constructions appear in relative clauses, focus word constructions, and indefinite A contexts, these contexts also involve the use of transitive verbs more than half the time.^[19] There is also evidence of age-grading where younger speakers are more likely than older speakers to use transitive verbs.

An important observation about this distribution in Kaqchikel is that focus contexts which are introduced by a specific lexical NP, particularly one that has already been established in the discourse, are likely to permit the use of transitive verbs. I suggest that this is a grammatical manifestation of general, cross-linguistic preferences related to both (a) processing economy, which favors rapid identification of the actor/agent role as well as the use of more frequent constructions (here transitive AVO), as well as (b) discourse pressures related to reference tracking and information structure where given arguments are more likely to be encoded as ergative NPs. Preverbal lexical NPs provide information as to whether that argument is a possible agent, regardless of the form of the verb, a tendency which has also been observed in several other Mayan languages. This set of general functional principles could motivate the distribution of constructions we see here, both between and within particular syntactic contexts. A competition account is particularly suited to these data, since the interaction of these factors create probabilistic tendencies rather than rules (although the patterns they create may grammaticalize into rules), and we see variation not only in which constructions are available but which are preferred across the six focus contexts investigated. The claim is that these discourse-level and semantic cues have shaped the syntactic possibilities in this domain in Kaqchikel. This type of processing approach is compatible both with analyses of Mayan syntax that would view this as an innovation, where ergative extraction restrictions were historically more uniform, as well as with views where ergative extraction restrictions originated in a more restricted set of contexts and then spread to other, similar contexts in different ways in different languages.

Finally, comparative evidence was provided which suggests that this type of variation in the use of AF/AP constructions is not uncommon in Eastern Mayan, and perhaps elsewhere as well. While it may be possible to make a generalized account for the Kaqchikel data, it is important to our broader understanding of AF to note that these constructions have grammaticalized in different ways in different languages, and serve different information-structural purposes (cf. Zavala [1997] on Akatek). Investigating broader similarities and differences requires collecting more comparable descriptive facts across Mayan. However, a level of caution should be exercised in investigating this phenomenon, since there is a documented tendency in multiple languages for the judgments provided in standard elicitation to contrast with use in other modes of communication, and to pay attention to the information-structural factors as well as the syntactic ones. I also encourage researchers working in this area to use multiple genres and methods (e.g., texts, picture elicitation, experimental designs, historical documentation) and investigate the potential for dialectal, age, diachronic, etc. variation with a variety of speakers. Only once we have a larger base of comparable data can we work towards a more complete explanation for the distribution of AF/AP constructions in Eastern Mayan and beyond.