3.1. Predictions for determiner status

The proposal we laid out makes predictions for the behavior of a quantifier depending on its categorial status in the syntax. The following predictions are made if the quantifier is a determiner:

Building on Royer, Buenrostro & Jenks (Reference Royer, Buenrostro, Jenks and Mathieu2026), we show in the next three subsections that these predictions are fulfilled for jantak, but not for tzijtum. While not shown here, the remaining quantifiers in Table (17) pattern according to their respective category.

3.1.1. Ability to appear in the complement of a preposition

Chuj’s sole preposition t’a can only combine with nominal expressions (DPs or NPs), and never with verbal or nonverbal predicates. This fact thus predicts that a Det-Q should be readily found within a prepositional phrase (PP) as a complement to the head. As shown below, this prediction is borne out for jantak. Note that this is so regardless of the position of the PP: whether the PP is postverbal or preverbal, jantak can occur within the complement of t’a. Footnote ⁷

(19)

The predicted distribution is different for Pred-Qs, which, as nonverbal predicates, are expected to never occur in the complement of t’a. As shown below, the prediction is borne out for tzijtum, regardless of the position of the PP:

(20)

In Section 3.2.2, we delve into the necessary syntax for tzijtum to quantify over oblique arguments, which further supports its status as a nonverbal predicate.

3.1.3. Ability to appear in topic position

A third expectation regarding Det-Qs is that they should be able to appear within topicalized DPs. In Chuj, only DPs may be topicalized (see Maxwell Reference Maxwell1982, Reference Maxwell1987, Bielig Reference Bielig2015, and Royer Reference Royer2022c on topicalization in Chuj); when topicalized, DPs appear with the marker ha and resumption is obligatory. Fulfilling this expectation, the quantifier jantak can appear inside a topicalized DP:

(24)

On the other hand, the quantifier tzijtum, as a nonverbal predicate, should not be able to appear within the topic position. This is indeed borne out:

(25)

3.2. Predictions for predicative status

We make the following six predictions for quantifiers in Chuj that are syntactically instantiated as nonverbal predicates:

The first prediction above was already shown to hold for tzijtum: quantifiers like tzijtum, in contrast to those like jantak, must generally appear in a sentence-initial position, or at least before verbal predicates (nonverbal predicates can appear, for example, after modifiers like adverbs). Since Chuj is a predicate-initial language, the initial position of quantifiers like tzijtum is derived directly from their analysis as nonverbal predicates. Relevant examples are repeated below for convenience:

An anonymous reviewer wonders about our generalization, asking whether tzijtum may appear on the surface non-initially if the expression that it quantifies over is topicalized. This is indeed true, as shown in (28). Under the relevant topicalization – notice the marker ha and the resumption in (28) – the quantifier tzijtum appears linearly after a verbal predicate such as ixwila’. However, we need not amend our generalization, since the topicalized element is in a peripheral position:

(28)

Expanding on Royer, Buenrostro & Jenks (Reference Royer, Buenrostro, Jenks and Mathieu2026), we show in the following subsections that the rest of the predictions in (26) are fulfilled for tzijtum, but not for jantak. Again, while not shown here, the quantifiers in Table (17) pattern according to their indicated category.

3.2.2. Visible signs of relativization

A second diagnostic concerns visible signs of relativization. Recall that building sentences with Pred-Qs often requires a relativization component – for example, the equivalent of ‘Malin bought many books’ translates more literally to ‘The books that Malin bought are many’. Recall as well from Section 2 that argument relativization does not exhibit obligatory marking of any kind (9), but that the relativization of oblique arguments does (10): namely, the use of the relative pronoun b’aj.

Therefore, if tzijtum is indeed a nonverbal predicate, we predict that whenever it quantifies over an oblique argument, we will observe the relative pronoun b’aj . As shown below, this prediction is borne out:

(31)

In contrast, whenever jantak quantifies over an oblique argument to deliver a meaning such as ‘to many towns’, it must appear within a prepositional phrase (see example (19) for relevant data).

The data in (31) in conjunction with those in Section 3.1.1 provide further support for our claim that quantifiers like tzijtum can be nonverbal predicates. Crucially, however, (31) supports the even stronger view that they must be so. Otherwise, we could not explain the obligatory presence of the relative pronoun in sentences where tzijtum quantifies over an oblique argument.

3.2.3. Ability to participate in possessive ‘have’ sentences

In Mayan languages, possessive ‘have’ constructions are formed by combining an existential predicate with a possessed DP (Freeze Reference Freeze1992, Coon Reference Coon2016). An example with the basic existential predicate ay ‘ext’ is provided below:

(32)

Now, since Pred-Qs also involve existential quantification, we might expect them to pattern with the basic existential predicate to form sentences that would translate with possessive ‘have’ in English. As shown below, this is borne out. Quantifiers like tzijtum can combine with a possessed DP to serve as the sole predicate of a possessive ‘have’ construction – in other words, tzijtum is in the same natural class as existential ay .

(33)

We predict a different behavior for the quantifier jantak, when used as a Det-Q. Indeed, combining jantak with a possessed DP does not deliver the intended translation in (34a).Footnote ⁹ Instead, to utter the equivalent of (33), jantak must co-occur with ay ‘ext’ (34b), as expected if jantak is a Det-Q.

(34)

For completeness, also note that ay and tzijtum may not co-occur (no matter the order of ay and tzijtum) to deliver a possessive ‘have’ sentence:

(35)

3.2.5. Ability to combine with morphology reserved for nonverbal predicates

Our last diagnostic concerns morphology that appears exclusively on predicates, either nonverbal or verbal. This includes morphemes like -(h)am(a’), -(h)amlaj, -xom, and -xomlaj, which give rise to various subtle epistemic modal interpretations whose semantics is yet to be completely understood.

We exemplify this with the morpheme -(h)amlaj below. When used with nonverbal predicates, -(h)amlaj conveys something similar to speaker uncertainty (i.e. it is a possibility modal). As shown in (39–42), this morpheme can attach to all sorts of nonverbal predicates, including nominal, adjectival, existential, and stative predicates. It can never, on the other hand, appear on any morpheme of the extended nominal domain of arguments.

(39)

(40)

(41)

(42)

Since this morphology can only ever appear on predicates, and never on nouns in argument position or on any other element of the extended nominal domain (classifiers, determiners, quantifiers, and demonstratives), we make a prediction: quantifiers like tzijtum should be able to host morphemes like -amlaj, but quantifiers like jantak (when used as ‘many’) should not.Footnote ¹¹

As shown below, this prediction is borne out on both counts, irrespective of the positioning of jantak in relation to the verb.

(43)

(44)

3.3. Summary and analysis

To summarize the empirical takeaways of this section, we have expanded on diagnostics originally proposed in Royer, Buenrostro & Jenks Reference Royer, Buenrostro, Jenks and Mathieu2026 to show that Chuj distinguishes between two types of quantifiers. While quantifiers like jantak fulfill all diagnostics for determiner status (they are Det-Qs), quantifiers like tzijtum fulfill all diagnostics for nonverbal predicate status (they are Pred-Qs). Crucially, this is different from quantifiers like ‘many’ in languages such as English, which, as shown at the outset of this paper (1a–2a), exhibit both determiner and predicative uses:

1. (45)

   1. a. Lou bought many books.

   2. b. We are many.

Chuj, instead, employs separate lexical items for each quantifier type: e.g. tzijtum is only ever instantiated as a nonverbal predicate, whereas jantak – when used to convey ‘many’ – is instantiated as a determiner. For a summary of diagnostics, see Table (16) above.

Following Royer, Buenrostro & Jenks (Reference Royer, Buenrostro, Jenks and Mathieu2026), we thus conclude that there are two different types of quantifiers in Chuj, requiring distinct syntactic analyses. For Det-Qs, we assume that the quantifier occupies the head of the DP. This builds on Royer (Reference Royer2022b), who locates indefinite determiners in that position. Classifiers are then located in a Classifier Phrase (ClfP), which merges with the NP.

(46)	Syntax of Det-Qs
	a.	Ix-s-man	jantak	ch’anh	libro	ix	Malin.
		pfv-a3-buy	many det	clf	book	clf	Malin
		‘Malin bought many books.’					(repeated from (11a))
	b.

For Pred-Qs, we assume that they occupy the head of a PredP, taking a DP subject as their sole (internal) argument. There is no need for a copula: a quantifier such as tzijtum encodes predication itself.

(47)	Syntax of Pred-Qs
	a.	Tzijtum	ch’anh	libro	ix-s-man	ix	Malin.
		many pred	clf	book	pfv-a3-buy	clf	Malin
		‘Malin bought many books.’					(repeated from (11b))
	b.

We further assume that the PredP head in (47) can also be filled by nominal, adjectival, stative, and existential nonverbal predicates such as those displayed in (7a–7d).

Finally, note that while our treatment of the subjects of predicational clauses as internal arguments follows Coon Reference Coon, Huang, Poole and Rysling2014 and Royer, Buenrostro & Jenks Reference Royer, Buenrostro, Jenks and Mathieu2026 (instead of external arguments merged in the specifier of PredP; see Bowers Reference Bowers1993, Baker Reference Baker2003, den Dikken Reference den Dikken2006, Lyskawa & Ranero Reference Lyskawa and Ranero2022), other approaches to predicational clauses in Mayan like Armstrong Reference Armstrong2017 or Coon & Martinović Reference Coon, Martinović, Lam and Ozaki2023 would also be compatible with the empirical facts discussed here. All that is needed is for nonverbal predicates in general (including Pred-Qs) to consistently surface in sentence-initial position, be it because they are base-generated there (as in the current approach) or as a result of movement (Coon & Martinović Reference Coon, Martinović, Lam and Ozaki2023; for predicate initiality beyond Mayan, see papers in Lee-Schoenfeld & Ott Reference Lee-Schoenfeld and Ott2021).

Having now shown evidence that Chuj distinguishes syntactically among two types of quantifiers, we turn to the second goal of this paper: establishing the syntactic status of wh-quantifiers, which we will argue instantiate Pred-Qs.

4.1. Background on wh-items in Mayan

Previous research into Mayan wh-items like ‘what’, ‘which’, and ‘who’, including in Chuj (Kotek & Erlewine Reference Kotek and Erlewine2019, Royer Reference Royer, Caponigro, Torrence and Zavala2021, Reference Royer2022b), has almost unanimously assumed that they are part of the extended nominal domain (see, e.g. Smith-Stark Reference Smith-Stark1988, Aissen Reference Aissen1996, Gavruseva Reference Gavruseva2000, Broadwell Reference Broadwell2005, Coon Reference Coon2009, Reference Coon2010, Heck Reference Heck2009, Douglas, Ranero & Sheehan Reference Douglas, Ranero and Sheehan2017, Kotek & Erlewine Reference Kotek and Erlewine2019, Mendes & Ranero Reference Mendes and Ranero2021, Ranero Reference Ranero2021, Coon, Mateo Pedro & Preminger Reference Coon, Mateo Pedro and Preminger2014, Clemens & Polinsky Reference Clemens, Polinsky, Everaert and van Riemsdijk2017, Little Reference Little2020, Coon, Baier & Levin Reference Coon, Baier and Levin2021, Ewing Reference Ewing2022, Royer Reference Royer, Caponigro, Torrence and Zavala2021, Aissen & Polian Reference Aissen and Polian2024, Alonso-Ovalle & Royer Reference Alonso-Ovalle and Royer2024, Ranero & Royer Reference Ranero, Royer, Phadnis, Spellerberg and Wilkinson2024, Coon & Vázquez Álvarez Reference Coon and Álvarez2025). Under these analyses, wh-items with an overt nominal associate exhibit the syntax proposed for Det-Qs in this paper, insofar as they head the DP. At the clausal level, wh-movement to a clause-peripheral position would account for the word order in (48).

(48)	a.	Mach	ch’anh	libro	ix-s-man	ix	Malin?
		which	clf	book	pfv-a3-buy	clf	Malin
		‘Which book did Malin buy?’
	b.	Possible syntax for (48a)			(to be refuted for Chuj)

Under this kind of approach, wh-words exhibit the syntax standardly associated with such items in more widely studied languages, such as English. However, the above-mentioned literature has also established three somewhat surprising observations regarding wh-questions in Mayan, suggesting that a different syntax should be explored instead.

The first observation is that wh-items in Mayan must always be in sentence-initial position in order for interrogative semantics to arise. The following data exemplify this. The word for ‘who/which’ in Chuj, mach, cannot appear in a postverbal position:

Previous works interpreted the pattern in (49) as an indication that wh-in situ, even in echo questions, is forbidden, and thus that focus fronting in wh-questions is obligatory (see Royer Reference Royer, Caponigro, Torrence and Zavala2021 on Chuj, and Aissen Reference Aissen1996 on Tsotsil; Polian Reference Polian2013 and Polian & Aissen Reference Polian, Aissen, Caponigro, Torrence and Maldonado2021 on Tseltal; Vázquez Álvarez & Coon Reference Vázquez Álvarez, Coon, Caponigro, Torrence and Maldonado2020 on Ch’ol; Curiel Reference Curiel, Aissen, England and Maldonado2017 on Tojol-ab’al; AnderBois & Chan Dzul Reference Ander Bois, Dzul, Caponigro, Torrence and Maldonado2021 on Yucatec Maya; Duncan, Torrence & Mateo Pedro Reference Duncan, Torrence and Mateo Pedro2024 on Kaqchikel; Duncan Reference Duncan2010, Velleman Reference Velleman2014 on K’iche’; and Elkins & Brown Reference Elkins and Brownto appear on Mam).Footnote ¹² The assumption is thus that wh-items are generated in an argument position within a DP, but obligatorily move to the preverbal position. As Coon, Baier & Levin (Reference Coon, Baier and Levin2021: 289) put it: ‘There is broad evidence that the requirement for A’-elements to move is strong in Mayan. Mayan languages prohibit in situ wh-words in interrogative contexts […].’

The second observation regarding wh-items, in Chuj and more broadly in the family, concerns the impossibility of multiple wh-questions. In a nutshell, no matter the order of wh-items, multiple wh-questions are ill-formed (see same references as the one just cited above). This is shown below.

Given a DP-based analysis of wh-phrases, one could derive the ban on multiple wh-questions if there is only one left-peripheral position to which wh-phrases may move (as has often been assumed in the literature on Mayan; see, e.g. Aissen Reference Aissen1992, Coon, Mateo Pedro & Preminger Reference Coon, Mateo Pedro and Preminger2014, Coon, Baier & Levin Reference Coon, Baier and Levin2021). That is, if focus fronting is obligatory for interrogative meaning to arise (as suggested by data like (49)), and there is only one focus position in the syntax, then multiple wh-questions are expected to be impossible.

A final observation regarding Mayan wh-questions is what has come to be known as pied-piping with inversion (Smith-Stark Reference Smith-Stark1988, Aissen Reference Aissen1996, Coon Reference Coon2009, Ewing Reference Ewing2022). This phenomenon can be illustrated in Chuj with the relational noun et’ ‘with’, which can be characterized as an agreeing preposition, where the complement of the P controls genitive (Set A) agreement (i.e. y-et’ is decomposable into a3-with). First consider the data in (51), which shows that postverbal DPs linearly follow the relational noun. As shown in (52), this is in contrast with preverbal wh-words, which must linearly precede the relational noun (hence the term ‘inversion’):

(51)

(52)

In work that assumes that wh-words are DPs, what must be posited to account for these data is a two-step derivation: (i) an initial movement internal to an oblique/possessive phrase, followed by (ii) movement of the entire phrase (see, e.g. Aissen Reference Aissen1996, Coon Reference Coon2009, Little Reference Little2020, Ewing Reference Ewing2022).

In sum, the literature on Mayan languages has long noted certain properties of wh-question formation that might seem surprising at first glance. Explaining the ill-formed data in (49) and (50), and the apparent existence of pied-piping with inversion, all while maintaining a DP analysis of wh-expressions, rests on a number of stipulations: (i) the fact that wh-fronting is obligatory for interrogative meaning to arise, (ii) the fact that there is only one focus position in the syntax, and (iii) the fact that there is movement internal to oblique/possessive phrases whenever there is A’-extraction of such a phrase. In the next section, we show that a Pred-Q analysis of wh-items, in contrast, embodies a way to account for all three restrictions in one fell swoop.

4.2. An alternative: wh-items as predicative quantifiers in Chuj

The current paper has established the existence of two types of quantifiers in Chuj: determiner quantifiers (Det-Qs) and predicative quantifiers (Pred-Qs). Since wh-expressions are generally regarded as a subtype of existential quantifier (see, e.g. Karttunen Reference Karttunen1977, Kratzer & Shimoyama Reference Kratzer, Shimoyama and Otsu2002, among many others), a plausible approach to Chuj syntax would be to treat wh-items consistently as Pred-Qs (as explicitly suggested by Zavala (Reference Zavala1992) for Akatek and by Tonhauser (Reference Tonhauser, Anderssen, Menéndez-Benito and Werle2003, Reference Tonhauser2007) for Yucatec Maya; see Verhoeven & Skopeteas Reference Verhoeven and Skopeteas2015 and Gutiérrez-Bravo Reference Gutiérrez-Bravo2017 for criticisms; see also Ewing (Reference Ewing2022), who suggests this analysis for a subset of wh-items in Kaqchikel), in contrast with the works cited in Section 4.1 that almost unanimously treat them as Det-Qs. We will argue that doing so sheds new light on why wh-items exhibit the otherwise three surprising properties that were laid out.

Let us first consider what a wh-question would look like in Chuj if the syntax of the wh-item were that of a Pred-Q. An example for the same sentence as the one in (48a) is provided in (53a), now bracketed to reflect the alternative analysis we will defend. Again, we are assuming a head-external analysis of relative clauses in the language (see Partee Reference Partee and Austerlitz1975, Chomsky Reference Chomsky, Culicover, Wasow and Akmajian1977; Jackendoff Reference Jackendoff1977; see also De Vries Reference De Vries2018). Note, however, that other analyses of relative clauses are equally compatible with our proposal – e.g. head-raising approaches (Schachter Reference Schachter1973, Vergnaud Reference Vergnaud1974, Bhatt Reference Bhatt2002) or those that assume movement inside an internally headed relative clause (Bianchi Reference Bianchi2000, Bodomo & Hiraiwa Reference Bodomo and Hiraiwa2010, Hiraiwa Reference Hiraiwa, Everaert and van Riemsdijk2017).

(53)	a.	Mach	[	ch’anh	libro	[	ix-s-man	ix	Malin	]]?
		which.is		clf	book		pfv-a3-buy	clf	Malin
		Literally: ‘Which is the book that Malin bought?’
	b.

Under our approach, sentences with wh-items belong to a larger class of Pred-Qs, and thereby exhibit the same structure as the one independently needed for quantifiers like tzijtum. In particular, the wh-item takes a DP subject as its complement, which can be relativized. When relativized, the verb within the relative provides key information about the interrogated event.

Importantly, note that the translation of the example in (53a) showcases a wh-item taking an overt nominal complement in English. Translations of questions without nominal complements would thus necessarily map to headless relative clauses in Chuj, which are independently known to exist in the Mayan language family (see references in Caponigro, Torrence & Zavala Maldonado Reference Caponigro, Torrence and Maldonado2021, and Kotek & Erlewine Reference Kotek and Erlewine2019 and Royer Reference Royer, Caponigro, Torrence and Zavala2021 specifically for Chuj). An example, with a schema for the proposed structure, is illustrated below.

Regardless of whether the head of the relative is overt or null, this alternative analysis entails that wh-questions in Chuj always involve biclausality. Specifically, questions are formed by combining a wh-predicate with a single argument, resulting in a pseudocleft strategy as the means to ask a question. The only difference between (53a) and (54a) is that a null head or pronoun heads the relative clause in the latter case.

It is important to note at this juncture that what corresponds to the null pronoun in (54a) can be overtly instantiated by a pronoun in Chuj, resulting in a light-headed relative clause. Like other Q’anjob’alan languages, a set of noun classifiers is employed as third person pronouns (see, e.g. Buenrostro, Díaz & Zavala Reference Buenrostro, Díaz and Zavala1989, Hopkins Reference Hopkins2012, Royer Reference Royer2022a). A minimal pair for (54a) is provided below:

In Section 4.3.3, we will provide additional evidence for the analysis that questions like (54a) indeed involve headless relative clauses.

Crucially, if the syntax proposed previously is the only strategy available to form questions in Chuj, we have a unified explanation for the three properties exhibited by wh-questions in the language and Mayan more broadly (recall Section 4.1). Consider first the ban on wh-in situ. Under our analysis, the ban is actually an illusion, a misnomer. Since Chuj is a predicate-initial language, wh-items must appear in sentence-initial position – i.e. they are actually always in situ. What is ruled out is a postverbal placement for the wh-items, and this follows naturally because nonverbal predicates simply cannot appear in that position.

Turning now to the ban on multiple wh-questions, it can also follow from our analysis. Many languages ban cleft/pseudocleft stacking (Kiss Reference Kiss1998, Reeve Reference Reeve2012). This is the case in languages like EnglishFootnote ¹³:

1. (56) *Who is what is the thing that you saw?

If the ban on stacking (pseudo)clefts extends to Chuj, then we can understand why multiple wh-questions are not permitted, regardless of the placement of each wh-item.

Finally, under our approach, pied-piping with inversion becomes a misnomer as well. That is, it becomes possible – and analytically necessary – to account for the relevant data without appealing to any inversion at all. As nonverbal predicates, wh-items stay in situ (in the preverbal position), and no inversion is needed: an element such as yet’ below is instead fronted as a byproduct of the movement of the relative operator, and yet’ is thus expected to surface after the wh-item.Footnote ¹⁴

As was the case with the headless relative clause in (54a), it is important to mention that the covert relative head in (57a) can be instantiated by an overt classifier pronoun. For completeness, a relevant example with context, as well as the proposed syntactic structure, are provided in (58).

In sum, this section has offered an analysis of the syntax of wh-questions in Chuj that departs from the traditional view that considers them to be DP-internal elements. We extended the independently necessary analysis of Pred-Qs from Section 3.3, thus reanalyzing wh-questions as involving a pseudocleft structure in which the wh-item serves as a matrix nonverbal predicate. Under our approach, wh-questions never involve movement of a wh-item; instead, they tend to require combination with a relativized DP subject (see Sections 4.3.2 and 4.3.5 for predicative uses of wh-items without relativization).

Moreover, we argued that our analysis provides an appealing way to unify the basic syntax of wh-questions with three (only apparent) idiosyncrasies that they exhibit. If interrogatives exhibit the clausal syntax involved in Pred-Qs, then the three properties follow in one fell swoop. In fact, non-interrogative Pred-Qs are subject to the same three syntactic restrictions, which is exactly what we would expect: as shown via the distribution of our Pred-Q quantifier tzijtum, (i) Pred-Qs cannot follow verbal predicates (59a), (ii) multiple Pred-Qs in the same sentence are ill-formed (59b), and (iii) Pred-Qs give rise to cases of (perhaps wrongly named) pied-piping with inversion (59c).

In contrast, under the prevalent DP analysis of wh-items, it is not possible to uniformly account for the three properties. As discussed in Section 4.1, that approach requires stipulating independent reasons for each.

Now, in order to definitively adjudicate between the DP analysis of wh-items and our own predicative analysis, we employ the sets of diagnostics that we used to establish the distinction between Det-Qs and Pred-Qs in Section 3. As will become clear, wh-items in Chuj exhibit all the signals for predicative status (§4.3) and exhibit none of the testable signals for determiner status (§4.4). In a nutshell, then, we will conclude that wh-items in the language are nonverbal predicates and, as a consequence, that wh-questions involve the pseudocleft syntax proposed in this subsection, instead of the monoclausal syntax in Section 4.1 that is generally assumed in previous work on Mayan wh-questions.

4.3. Wh-items check predicative diagnostics

If wh-items are exclusively Pred-Qs, then they should fulfill the predicative diagnostics discussed in Section 3.2. Here, we show that this is correct. While we only provide examples from a handful of wh-items below, we note that wh-items behave alike. A full list of wh-items in San Mateo Ixtatán Chuj is provided below (see Kotek & Erlewine Reference Kotek and Erlewine2019, Royer Reference Royer, Caponigro, Torrence and Zavala2021, and Alonso-Ovalle & Royer Reference Alonso-Ovalle and Royer2024 for further descriptions).

Before delving into the diagnostics, we highlight that there is homophony between the wh-items jantak and jay-num.clf and the non-interrogative determiner quantifiers meaning ‘many’ (jantak) and ‘few’ (jay-num.clf). However, the interrogative versus non-interrogative counterparts for each form are syntactically in complementary distribution. When used as determiners in postverbal positions or within focus constructions (with the marker ha), jantak and jay-num.clf are necessarily non-interrogative, meaning ‘many’ and ‘few’ respectively. When used as nonverbal predicates, they necessarily lead to interrogative semantics (meaning ‘how many’).

The following examples illustrate this. When jantak is clearly employed alone in a predicational sentence – where it is necessarily instantiated as a nonverbal predicate – it can only deliver an interrogative meaning ‘how many’ (61a). In contrast, when jantak is employed postverbally (61b), or in a focus construction (61c), it can only deliver the meaning of ‘many’.

It may be possible to derive Pred-Qs from Det-Qs (or vice versa), with the goal of explaining this homophony and the observed semantic differences – such an exercise is left for the future. We simply note here that our main point is not affected: as we will show below, wh-items systematically behave like nonverbal predicates in Chuj.

4.3.1. Inability to appear in postverbal position

As amply discussed already, wh-expressions in Chuj cannot appear postverbally in questions. Examples are repeated below for convenience. While this has traditionally been taken as evidence that wh-in situ is not possible, a Pred-Q analysis of wh-items in questions predicts the ill-formedness of postverbal wh-items – in brief, whatever drives the predicate-initial nature of Chuj renders such a configuration impossible.

(62)	a.	Mach	ix-h-ila’?
		who	pfv-a2s-see
		‘Who did you see?’
	b.	*	Ix-h-il	mach?
			pfv-a2s-see	who
			Intended: ‘Who did you see?/‘You saw who?’

As was the case for the distribution of predicate quantifiers like tzijtum (see (28) above), when the subject of a wh-predicate is topicalized, the wh-item appears linearly after a verbal predicate (we thank an anonymous reviewer for pointing this out to us). However, the topicalized element is in a peripheral position, and thus our conclusion need not be amended:

(63)

4.3.3. Visible signs of relativization

If wh-items are nonverbal predicates, then they should show the hallmarks of relativization when a question is complex in involving a verbal event. This follows because the structural correlate of the denoted event will have to be embedded within a relative clause. This prediction is indeed borne out when we consider wh-items for oblique arguments. The relevant wh-questions require the presence of the (previously discussed) relative pronoun b’aj , which we also found with Pred-Qs.

A baseline sentence with an indirect object is provided in (66a), with the corresponding wh-question provided in (66b). While indirect objects appear with the preposition t’a when postverbal, questioning the indirect object requires a different structure without t’a and with a relative pronoun instead (see Section 4.4.1 for data on combining wh-items with t’a).

(66)

As already hinted at in Section 4.2, additional signs of relativization are present in questions involving headless relative clauses, which lack a clear nominal head (see Caponigro, Torrence & Zavala Maldonado Reference Caponigro, Torrence and Maldonado2021). The reason is that such clauses are often light-headed (Citko Reference Citko2004) in Chuj: while they lack an overt nominal head, they nonetheless exhibit a determiner, pronoun, or demonstrative. Examples with pronouns as light heads were provided in (55a) and (57a) above. We offer other examples below, one where the light-head is a demonstrative (67a) and another where it is the numeral/indefinite determiner jun (67b). For more on these kinds of relative clauses in Chuj, see Royer Reference Royer, Caponigro, Torrence and Zavala2021: §4.2.

(67)

While the light-heads are not obligatory in the examples in (67), they provide additional support for the idea that wh-questions in Chuj involve a structural element that is relativized.

4.3.4. Ability to appear in possessive-‘have’ constructions

If wh-items are a subtype of Pred-Q, then we would expect that some of them might appear in possessive ‘have’ constructions. While this diagnostic is not informative with every wh-item in the language (i.e. there might be independent reasons why an incompatibility could arise between wh-items and this construction), (68) shows that wh-items such as jay+num.clf ‘how many’ can indeed appear as the sole predicate here. Notice as well that an interrogative meaning arises:

(68)

4.3.6. Ability to be modified by predicate modifiers

The final predicative diagnostic discussed in Section 3.2 was the ability for Pred-Qs to co-occur with predicative modifiers, such as -(h)amlaj. Again, wh-items in Chuj readily co-occur with -(h)amlaj. When they do, the result is a meaning paraphrasable as ‘Who knows wh X’:

(71)

4.4. Wh-items do not check (testable) diagnostics for determiner status

Since wh-items are independently known to be bad candidates for topicality (wh-items map to focus instead; see, e.g. Aissen Reference Aissen2023), we will only consider as informative the other two diagnostics for determiner status.Footnote ¹⁶

4.4.1. Inability to co-occur with prepositions

The first diagnostic we consider is co-occurrence with a preposition. As shown in Section 3, only Det-Qs are expected to be found in the complement of the proposition t’a in Chuj. If wh-items are Pred-Qs, as we maintain, they should never co-occur with a preposition. This prediction is borne out: wh-items cannot co-occur with a preposition, irrespective of the order of the elements. Instead, the relativization strategy discussed in Section 4.3.3 is used.

(72)

4.4.2. Inability to appear in the possessor of a DP

The second testable diagnostic for determiner status is the ability to appear in the possessor of a possessed DP. Constructions in which a wh-item co-occurs immediately beside a possessum, no matter the word order, are indeed unacceptableFootnote ¹⁷:

(73)

The only way to convey the intended meaning in (73c) is with the example in (74). In this case, the wh-item appears in a sentence-initial position, as expected given the Pred-Q analysis, and the possessed DP remains in its base position. In this case, we assume that a null pronoun appears as the possessor of the possessive phrase.

(74)

A potential counter-analysis could be invoked to derive the example above. While our predicative approach predicts directly that the structure in (74) would be employed, one could alternatively call upon possessor sub-extraction from a DP. In light of similar data, such an analysis is defended by Little (Reference Little2020) for Ch’ol and Hedding & Yuan (Reference Hedding, Yuan, Duygu Demiray and Segal2025) for other Mesoamerican languages exhibiting pied-piping with inversion. We note, however, that sub-extraction analyses have been contested on independent grounds (Aissen & Polian Reference Aissen and Polian2024), and that there is no independent evidence for subextraction (of, e.g. numerals or adjectives) in Chuj – unlike in Ch’ol, where such independent evidence is actually available.

We therefore maintain that data like (73c) and (74) indicate that wh-items cannot appear in the possessor position of possessive phrases in Chuj, as expected if wh-items are exclusively instantiated as nonverbal predicates.

4.5. Summary

In this section, we considered whether wh-items, a subtype of quantifier, should be analyzed as Det-Qs or Pred-Qs in Chuj. We applied the diagnostics from Section 3 and concluded that wh-items are exclusively instantiated as nonverbal predicates in the language. The following table compiles the results of all the diagnostics:

We also argued that analyzing wh-items exclusively as Pred-Qs provides a unified account of three (seemingly) idiosyncratic aspects of wh-question formation in Chuj: (i) the apparent ban on wh-in situ, (ii) the ban on multiple wh-questions, and (iii) the phenomenon known as pied-piping with inversion. This conclusion is appealing on a theoretical level and far from trivial, since it runs against the status quo in the Mayanist literature, including previous work on Chuj (Kotek & Erlewine Reference Kotek and Erlewine2019, Royer Reference Royer, Caponigro, Torrence and Zavala2021), which had unanimously assumed a DP analysis for wh-items. As far as we can tell, such an analysis cannot account for these restrictions in a unified manner and must invoke independent stipulations.

Moreover, since the idiosyncrasies on wh-questions apply across the Mayan family (as far as we know), we suggest that the predicative strategy for question formation may well be widespread – an investigation that we hope will bear many fruits in the future. In the conclusion, we turn to suggestions for such elaborations.

5.1. Focus constructions and predication

Since this paper focused on the syntactic distribution of wh-words within questions, we intentionally set aside a construction whose syntax is generally treated in parallel with that of Mayan wh-questions: the construction involving the focusing of nominal expressions. Since Aissen Reference Aissen1992, a standard approach to focus constructions in this language family has been to assume that they involve A’-movement of a DP to the left periphery (see also Aissen Reference Aissen, Ostrove, Kramer and Sabbagh2017a and Royer & Coon Reference Royer and Coon2025 for overviews). Under such an approach, focus constructions are treated as monoclausal, and whatever focus marker is employed (in this case ha) is generally viewed as part of the focused DP (see, e.g. Royer, Ranero, Felipe Gómez & López García Reference Royer, Ranero, Gómez and García2025).

(76)	a.	[	Ha	jun	te’	onh	]	ix-s-man	ix	Malin.
			pv	indf	clf	avocado		pfv-a3-buy	clf	Malin
		≈ ‘It’s an avocado that Malin bought.’								(focused object)
	b.	Possible syntax for (76)								(to be refuted for Chuj)

Since wh-clauses involve pseudoclefts under our proposal, it is interesting to consider whether Chuj focus constructions could also be analyzed along those lines. This, again, would be in accordance with Tonhauser’s (Reference Tonhauser, Anderssen, Menéndez-Benito and Werle2003, Reference Tonhauser2007) approach to focus constructions in Yucatec Maya (see also Bricker Reference Bricker, Bricker and Norman1979, Bohnemeyer Reference Bohnemeyer1998, and for Ixil, Blunk Reference Blunk2008; see Gutiérrez-Bravo Reference Gutiérrez-Bravo2017, Verhoeven & Skopeteas Reference Verhoeven and Skopeteas2015 for criticisms). Under this type of approach, ha in focus constructions could be reanalyzed as a predicate: a copular morpheme occurring in special kinds of copular clauses (77b).

(77)	a.	[	Ha	jun	te’	onh	]	ix-s-man	ix	Malin.
			pv	indf	clf	avocado		pfv-a3-buy	clf	Malin
		≈ ‘It’s an avocado that Malin bought.’								(focused object)
	b.

As it turns out, there are various empirical reasons to favor a biclausal analysis of focus constructions in Chuj. The first is that focus constructions exhibit biclausal behaviors. For example, a focused nominal can bear negation in tandem with a verb that also bears negation. We indicate phonological deletion of ha in (78), since this segment is expected to be deleted given independent phonological rules governing vowel hiatus in the language; see Royer, Ranero, Felipe Gómez & López García Reference Royer, Ranero, Gómez and García2025. This would follow directly from extending our approach to focus configurations, since an example like (78a) below would be biclausal:

Since only predicates may be negated in Chuj, a monoclausal, focus-fronting approach could not provide a straightforward account for the data above.

Several other applications of the diagnostics for predicatehood established in this paper support this. Consider, for instance, the following two examples.

(79)

(80)

In (79), we see that focused oblique arguments behave like predicative quantifiers and wh-items in co-occuring with a relative pronoun, suggesting relativization is at play. In (80), we see that a modal particle that only ever appears on predicates appears on ha, suggesting that this element is itself a predicate. In fact, ha can never co-occur with predicative quantifiers or wh-items in questions, suggesting they indeed belong to the same natural class. These preliminary facts suggest that our analysis of wh-questions in Chuj should be extended to focus constructions in general.

Another well known property of constructions involving wh-questions, focus, and relativization in Chuj and other Mayan languages is that they exhibit what has been described as syntactic ergativity (Aissen Reference Aissen, Coon, Massam and Travis2017b; for Chuj, Royer & Coon Reference Royer and Coon2025). In particular, when an ergative subject is questioned, focused, or relativized, the verb cannot bear Set A agreement and it must appear with an agent focus suffix. Whereas prominent analyses have attributed this phenomenon to a consequence of DP movement in the relevant configurations (see, e.g. Coon, Mateo Pedro & Preminger Reference Coon, Mateo Pedro and Preminger2014; Coon, Baier & Levin Reference Coon, Baier and Levin2021), our analysis forces a reconsideration: what unifies the three constructions is the existence of relativization of an ergative subject. We thus open the door to reanalyzing the agent focus morpheme as a marker of agent relativization. Much more needs to be said – including what leads to the loss of Set A morphology – but we leave this for future exploration, both in Chuj and beyond.Footnote ¹⁸

5.2. Mayan languages beyond Chuj

Applying diagnostics akin to ours in other Mayan languages will put our proposal to the test, thus shedding light (or casting doubt) on whether our unified derivation of the properties of wh-items and wh-questions replicates beyond Chuj. If such replications are indeed observed, we will have further evidence that the grammar may provide predication as the sole strategy for wh-question formation, with cascade effects observed across the syntactic systems of individual languages.

For now, however, let us point out that at least one Mayan language appears to challenge a cross-Mayan unification. Ch’ol has a relativizing morpheme, the clitic =bä, that does not appear in wh-questions (Vázquez Álvarez Reference Vázquez Álvarez2011).Footnote ¹⁹ Consider for instance the following examples:

All else being equal, the above Ch’ol data are surprising if wh-questions should always involve pseudoclefts in that language as well; that being the case, we might expect the relativizing clitic =bä in (81c) to also appear in wh-questions.

While these facts pose a potential challenge to the pan-Mayan unification, they do not rule out a predicative analysis definitively. For example, it could be that Ch’ol employs more widely a secondary predication strategy analogous to the one discussed in Section 4.3.5 for wh-question formation, or another structure involving a predicative wh-item (see, e.g. Can Pixabaj & Aissen Reference Can Pixabaj and Aissen2021 on ‘how’ questions in K’iche’). Another possible avenue might be to reconsider the function of the clitic =ki, bolded in (81) and found on all wh-items in interrogatives. Indeed, to extend the pseudocleft analysis, we suggest preliminarily that this clitic might be reanalyzable as a relative marker, one sensitive to features of a matrix wh-predicate. Finally, note that one wh-word in Ch’ol, ba=ki=bä, has been claimed to involve the relativizing morpheme =b’ä (Vázquez Álvarez & Coon Reference Vázquez Álvarez, Coon, Caponigro, Torrence and Maldonado2020: 371), suggesting that relativization might be involved at some level.

While we can only provide speculative thoughts for the moment, we stress that there is a clear empirical motivation for a pan-Mayan unification: most Mayan languages we are aware of, Ch’ol included, exhibit the three properties identified at the beginning of this paper (3) – the ban on wh-in situ, the ban on multiple wh-questions, and the existence of pied-piping with inversion.

5.3. Broader cross-linguistic parallels

Naturally, our search for parallels or extensions should take us beyond the Mayan language family.

For starters, as influentially established by Smith-Stark (Reference Smith-Stark1988), pied-piping with inversion is an areal feature of Mesoamerica (see also Aissen Reference Aissen1996, Broadwell Reference Broadwell, Butt and King2006 and Hedding Reference Hedding2022: chapter 4). Therefore, if we are correct that pied-piping with inversion is a misnomer for Chuj, it may also be a misnomer for at least some other Mesoamerican languages. In light of this, we note that the two other properties that we identified at the outset of this paper in (3) – (i) the ban on wh-in situ and (ii) the ban on multiple wh-questions – are well documented for at least several Mixtec (Otomanguean) languages (Caponigro, Torrence & Cisneros Reference Caponigro, Torrence and Cisneros2013; Hedding Reference Hedding2022, Reference Hedding2025), as well as for Sierra Popoluca (Mixe-Zoquean; López Márquez Reference López Márquez, Caponigro, Torrence and Maldonado2021: 480–482). Identifying language-specific diagnostics for nonverbal predication in these languages, and applying them to wh-words and/or quantificational items in general, should prove fruitful in shaping future analyses of questions, foci, and consequently relativization in the Mesoamerican area. We also note that, in the context of defending the omnipredicativity hypothesis for Classical Nahuatl (Launey Reference Launey1994, Reference Launey2004), Launey (Reference Launey2004: 8) proposes a predicative analysis of wh-items and questions analogous to the one proposed for Chuj in this paper. Again, this may suggest that the types of structures proposed here are more widespread across Mesoamerica.

Second, many of the languages in the Austronesian family have also been argued to employ the pseudocleft strategy for wh-questions (see Potsdam Reference Potsdam2009, Chen, Polinsky & Potsdam Reference Chen, Polinsky, Potsdam and Palmerto appear, and references therein). A deeper exploration into our cluster of properties in (3) must be left for the future, but we note a couple of facts. First, in contrast to Chuj, wh-in situ does seem to be available in these languages (see fn. 13 in Potsdam Reference Potsdam2009), though subtle distinctions seem to be at play depending on the nature of the wh-expression (i.e. subject vs. non-subject in Malagasy; see Paul & Potsdam Reference Paul, Potsdam, Merchant and Simpson2012: 168), and at least some sentence-final wh-items may in fact be analyzed as predicates (see Kraft Reference Kraftto appear, §3). When it comes to pied-piping with inversion, some languages in the family such as Sasak do exhibit the phenomenon (Austin Reference Austin, Schulze and Warnk2006). Length considerations preclude us from further discussion, but we hope that our proposal spurs it.

Third, we find it thought-provoking that languages from a diverse range of language families have also been analyzed as involving pseudocleft structures for question formation, including Salish (Kroeber Reference Kroeber1991, Davis, Gardiner & Matthewson Reference Davis, Gardiner and Matthewson1993, Jelinek & Demers Reference Jelinek and Demers1994, Jelinek Reference Jelinek, Czaykowska-Higgins and Kinkade1998, Davis Reference Davis and Takeuchi2008, Reference Davis2009), Algonquian (Bloomfield Reference Bloomfield and Hoijer1946, Blain Reference Blain1997), and Tsimshianic (Beck Reference Beck2002, Davis & Brown Reference Davis and Brown2011, Brown Reference Brown2024; though see Brown Reference Brown2026 on Sm’algyax), and that all such languages seem to exhibit some or all of the same restrictions on wh-questions as the ones discussed in (3).

Finally, another potentially relevant language family to consider is Celtic, which has also been claimed to exhibit identical properties as the ones listed in (3) (see, e.g. McCloskey Reference McCloskey1979, Oda Reference Oda2002, Adger & Ramchand Reference Adger and Ramchand2005), including what Thoms (Reference Thoms2025: Section 4.2) recently describes as pied-piping with inversion (see Adger Reference Adger, Arche, Zwart, Demirdache and Borer2025 for an analysis of this phenomenon in Scottish Gaelic). This is in addition to the well known fact that wh-questions in Celtic languages appear with the complementizer otherwise employed for nominal relativization (see morpheme glossed as C.rel in (82b)), a fact that led previous authors such as McCloskey (Reference McCloskey1979) and Oda (Reference Oda2002) to propose that wh-questions in languages like Irish consistently involve cleft-like syntax. To illustrate, consider the following example from Scottish Gaelic, zooming in on the order of the preposition and DP in the declarative sentence in (82a) versus the order of the preposition and the wh-item in the interrogative in (82b).

(82)

It would be encouraging to uncover more evidence that the phenomena discussed across our paper indeed cluster together. Again, one potentially productive way of going about this would be to first establish language-internal diagnostics that single out predication specifically.

To conclude, note that many of the languages in the families mentioned above are described as predicate initial. It is worth continuing to explore how tight a link exists between predicate initiality and the range of the inventory of predicative wh-words in specific grammars (see Potsdam Reference Potsdam2009 for an exploration of this nature in Austronesian). It seems natural to us that acquirers would integrate wh-expressions as predicates if their target grammar reserves clause initiality precisely to predicates in baseline declaratives, with cascade effects arising as a direct consequence.