2.1. The syntax of the verbal template or complex

Yucatec is a null subject and object language in which a verbal template or complex may instantiate a fully-fledged clause that lacks overt lexical arguments.¹ Such clauses consist of an aspect/mood (am) marker followed by the verb. The verb itself is composed of three main parts: a root and transitivity marker (verb stem), a suffix known as “status” in the Mayanist literature (^{Kaufman 1990}), which will be explained in more detail below, and a set of morphemes that cross-reference verbal arguments, labeled Set a and Set b, respectively. The examples in (3-5) provide illustrations of these basic components of Yucatec clause structure using a transitive and intransitive verb in the three main inflectional patterns.

(3)

(4)

(5)

As can be observed above, each sentence contains an aspect/mood (am) marker at the left edge. In (3), the incompletive/indicative pattern, the argument-marking morpheme known as “Set a” in the Mayanist literature cross-references both transitive (3a) and intransitive (3b) subjects and immediately follows the aspect marker. Next comes the verb stem. (3a) contains the root líi’ (‘raise’) and the transitivizing causative morpheme -s while (3b) contains the intransitive root wen (‘sleep’). Attached to the right of the verb stems is a suffix called “status” whose form is sensitive to aspect/mood, transitivity, and within different types of intransitive verbs, root class. It is -ik for transitive verbs in incompletive/indicative paradigm and -Vl (where V = harmonic vowel that is identical to the vowel of the root) for the intransitive root class to which wen belongs. Finally, the morpheme -ech, glossed as “Set b”, cross-references the object of the transitive verb and follows the status suffix.

The other two inflectional patterns, completive (4) and subjunctive (5), differ in important ways from that of (3). Not only do the am markers and status suffixes have different meanings but the alignment of the Set a and Set b morphemes is ergative-absolutive rather than nominative-accusative. As can be observed in (4) and (5), the Set b morpheme -ech that cross-references the object of the transitive verb in the a-examples also marks the subject of the intransitive verb in the b-examples.²

In order to establish a working hypothesis about clausal syntax that will form the basis of the subsequent analysis of non-verbal predication in Yucatec, I will closely follow work by ^{Aissen (1992}; 1996; 2011), ^{Pye (2011}), ^{Coon (2013}; 2017), Coon et al. (2014) and ^{Gutiérrez Bravo (2015}b), introducing some novel adaptations and simplifying assumptions to these proposals. Based on these works, the clausal architecture of Mayan languages can be organized as follows: a verb stem that may take an argument as its complement, a v layer that is instantiated by status suffixes and may introduce an argument in its specifier, a subject agreement projection, AgrP (Pye 2011; Gutiérrez Bravo 2015b), and an Infl layer in which am markers appear. This is schematized in (6). Following Aissen (1992; 1996; 2011), I assume that lexical/thematic specifiers project right while functional specifiers project left.

(6)

There are a number of morpho-syntactic operations that derive correct morpheme order and account for the formal licensing of arguments. These are spelled out in (7), and illustrated below with some concrete examples.

The basic derivational operations for the transitive clause in (4a), repeated in (8a), are shown in (8b) and (8c) respectively.

(8)

The derivation for a clause with an intransitive verb in (5b) is shown below in (9).

(9)

Evidence for locating the Set a morpheme in Agr (in line with ^{Gutiérrez Bravo 2015}b, which differs from ^{Coon 2017}’s claim for Ch’ol) comes from the fact that it is separable from the rest of the verbal complex. Certain monosyllabic quantifiers and adverbs may intervene between the Set a morpheme and the verb as shown in (10) (see ^{Blair & Vermont-Salas 1967}; ^{Bohnemeyer 2002}; ^{Verhoeven 2007}).

Though no concrete claim about the syntactic position of the adverbs highlighted in bold in (10) has been made in the literature as far as I am aware, a plausible way of accounting for their position is to claim that they are generated in a projection between Agr and vP. This would explain why the Set a morpheme appears separate from the verb stem.

Turning now to the licensing of the Set b morpheme, the primary evidence that it is licensed via different means in transitive and intransitive clauses comes from comparing embedded sentences that exhibit subject control with those that do not. For instance, in (11) below, the matrix predicate k’áat (‘want’) may take a clausal complement that exhibits subject control (11a) or one in which the matrix subject is distinct from the embedded one (11b) (see ^{Bricker 1981} and subsequent work).

In the former, no am marker is possible, the transitive verb is marked for subjunctive status and both Set a and Set b morphemes appear. In the latter, the subordinating particle káa appears. This particle has been analyzed as an am marker since it is in complementary distribution with other am markers (see ^{Chan Dzul 2010}; ^{Gutiérrez Bravo 2015}b) and the embedded transitive verb appears in subjunctive status with both Set a and Set b morphemes. The fact that all arguments of a transitive verb can be licensed in the absence of an am marker (=Infl) supports the idea that they are both licensed by a source other than Infl, mainly transitivity.

Intransitive verbs, on the other hand, show a different pattern. In control environments like (12a), the verb appears in a bare form without a Set a or Set b morpheme. In non-control environments like (12b), the presence of the subordinating particle káa correlates with a fully inflected intransitive verb -it appears in subjunctive status and contains a Set b morpheme.

If káa is an instantiation of Infl, the pairs of sentences in (12) support linking the formal licensing of the Set b morpheme in intransitive clauses to the presence of Infl. In fact, it can be demonstrated that all intransitive verbs that have Set b morphemes have Infl. For example, for all root classes inflected for subjunctive status, either an am marker or a subordinating particle káa (‘that’) or chéen/kéen (‘when’) is present when Set b appears. In completive status, the correlation of the presence of Infl with Set b is not as transparent. This is because the aspect marker in intransitive completive sentences can be either aspirated or null. Consider the completive sentences in (13) that have verbs from different intransitive root classes.

The verb bin (‘go’) belongs to a root class in which an aspirated completive am marker appears and the status suffix is null while the verb meyaj (‘work’) belongs to a root class in which there is arguably no overt am marker and the status suffix is -naj.³ This would appear to question the correlation between the presence of Infl and Set b. One way of maintaining the generalization is to hypothesize that the completive am marker of intransitive verbs can be aspirated or null but that Infl is always present. This is supported by examining contexts that involve embedding under the perfective connective káa/ka’aj/ka (roughly ‘when’/‘and then’). ^{Bohnemeyer (1998}: 486) notes that this connective “occurs exclusively with the perfective ( our completive, cp) am marker (t- with transitive verbs and h with intransitives); combination with any other aspect-mood marker is straightforwardly ungrammatical.” In (14a), an example taken from Bohenmeyer (1998) shows that any verb following káa must have a completive am marker. This author represents the intransitive completive am as aspiration, which is italicized below. In other cases, an intransitive verb may appear embedded under káa/ka’aj/ka without an am marker as in (14b).

Bohnemeyer analyzes the particle as a connective that requires the event it appears with to be aspectually completive. It plays a crucial role in integrating bounded events into the structure of the larger discourse. This plausible analysis of káa would require the presence of an aspect marker even if it is not realized phonetically. The fact that an intransitive of any root class may appear after káa, regardless of whether there is an aspirated am marker or not, is evidence in favor of the claim that Infl is always present when Set b is present. Returning to the data in (12), we can now offer a generalization about how Set b morphemes are licensed differently in transitive and intransitive clauses. In transitive clauses it is transitivity itself that licenses both the subject and object while in intransitive clauses it is finite Infl that licenses the intransitive subject. The same type of pattern shown in (12) has been observed in other ergative languages by ^{Legate (2008}) and ^{Coon (2013}) and analyzed as “absolutive as default” (abs=def) since absolutive case is not licensed by a single source, but rather varies according to whether the verb is transitive or intransitive. I will adopt this generalization for the licensing of Set b in Yucatec given the discussion above.

Finally, let us consider the evidence for treating Set a and Set b morphemes as agreement and pronominal clitic, respectively. The main evidence for making this distinction is outlined in ^{Grinevald & Peake (2012}), ^{Coon et al. (2014}) and Coon (2017). The first piece of evidence comes from comparing the position of Set a and Set b morphemes in different Mayan languages. Set a morphemes appear uniformly in the preverbal position or prenominal position (possessors) in all Mayan languages. On the other hand, Set b morphemes appear in different positions across the family. For instance, in Yucatec they appear to the right of the verb stem or any other predicate, but in other Mayan languages such as Q’anjobal, they appear to the right of the pre-verbal am marker as in (15).

The fixed position of Set a morphemes has been taken as one piece of evidence that they are more like agreement affixes while the variable position of Set b morphemes supports treating them as clitics.

A second piece of evidence that Set b morphemes should be treated as pronominal clitics is that in many Mayan languages, they appear to be reduced forms of strong pronouns. The table below illustrates the similarities between strong pronouns and Set b morphemes in Yucatec, and contrasts these with Set a morphemes (see ^{Coon 2017} for similar information on other Mayan languages).

(16)

As can be observed in the table above, Set b morphemes are reduced forms of strong pronouns while Set a morphemes are not. This situation reflects a tendency across languages that have both agreement morphology and pronominal clitics. While agreement morphemes are not necessarily formally related to anything in the functional domain of nouns, argumental clitics are usually formally related to determiners or pronouns. This situation is exhibited in Romance languages, where verbal agreement morphology is not related in any systematic way with determiners or subject pronouns whereas the clitic system is formally related to determiners and object pronouns (see ^{Ormazabal & Romero 2013} for detailed discussion of Spanish across dialects).

While a detailed account of agreement and cliticization is beyond the scope of this paper, an explicit proposal about how Set a and Set b morphemes actually appear in the positions they do is important for the sections to come on non-verbal predication. I will follow ^{Pye (2011}) and ^{Gutiérrez Bravo (2015}b) in assuming that the Set a morpheme appears as a reflex of agreement relation between the Agr head and a pronoun or dp in its c-command domain. Specifically, Agr acts as a probe that needs its phi-features valued and once it finds a goal, those features are spelled out as a Set a morpheme that matches the phi-features of the goal (^{Chomsky 2000}; 2001).

Turning now to Set b, following some recent work on pronominal clitics, I assume that the Set b morpheme is a generated as a determiner in argument position. It may be bare (17a) or adjoined to a strong pronoun or full dp as in (17b) and is licensed by an agreement relation with either transitive v or Infl (see discussion above).

(17)

However, the agreement relation itself is not necessarily a trigger for movement of this determiner to the head that has licensed it. Instead it cliticizes to the right edge of the lexical head of the predicate that thematically licenses it, which in Yucatec is the complex v-v head formed after the verb raises to v to pick up the status suffix.

It should be noted that this disconnect between formal/Case licensing and clitic position is similar to how object clitics have been analyzed in Spanish if we combine the classic work of ^{Uriagereka (1995}) with more recent observations by ^{Ormazabal & Romero (2013}). For example, in (18), the object clitic lo is licensed by agreement with a transitive v, but it does not cliticize to the verb. Instead it moves to a higher position, labeled f by Uriagereka (1995), which is above t, and cliticizes to the auxiliary.

The point here is that pronominal clitics can be formally/Case licensed through a relation with one functional head but cliticize to a different head for morpho-phonological reasons. While certain details have been glossed over here for reasons of space (see ^{Harizanov 2014}; ^{Kramer 2014}; ^{Preminger 2014} for recent work), I believe that this working hypothesis provides a plausible account of argument licensing and morpheme ordering in Yucatec clauses headed by a verb.

2.2. Pre-verbal positions: focus and topic

A detailed discussion of the word order properties of overt pronominal and lexical arguments of verbs is far beyond the scope of this work (see, for instance, ^{Durbin & Ojeda 1978}; ^{Briceño Chel 2002}; ^{Skopeteas & Verhoeven 2005}; 2009; ^{Gutiérrez Bravo & Monforte 2010}, among others). The point that is relevant for non-verbal predicate sentences is that the neutral word order for intransitive subjects is post-verbal as in (19a). Pre-verbal subjects of intransitive verbs may appear in focus and topic positions. For instance, in (19b) and (19c), a focalized or wh subject appears immediately to the left of the am marker while in (19d), the subject is in a topic position that is marked with the enclitic -e’.

I assume that focalized and wh-subjects move from their post-verbal base position to the specifier of InflP and that topicalized subjects are base-generated in the specifier of cp as shown in (20) (following ^{Aissen 1992}; ^{Gutiérrez-Bravo & Monforte 2011}; ^{Skopeteas & Verhoeven 2012}; Gutiérrez-Bravo 2017).

2.3. The internal structure of other types of phrases: nps, aps, pps

The internal structure of the noun phrase in Yucatec is described in ^{Briceño Chel (1996}), ^{Bricker et. al. (1998}), ^{Lois & Vapnarsky (2003}), ^{Verhoeven (2007}), ^{Butler (2012}) and especially ^{Lehmann (2002} [1998]). The examples in (21) serve as illustrations of some basic components of the noun phrase.

(21a) shows that the head noun may be preceded by an adjective and a numeral. The entire noun phrase is flanked by a definite determiner at its left edge and a deictic clitic at its right edge. (21b) shows that a head noun may be marked for plurality by the suffix -o’ob (see ^{Lucy 1992}; ^{Lehmann 2002} [1998]; ^{Butler 2012} on the general optionality of the plural morpheme).

Much more attention has been paid to the properties of possessors and other nominal arguments and the morphological reflexes they trigger (see especially ^{Bricker et. al. 1998}; ^{Lehmann 2002}[1998]; ^{Lois & Vapnarsky 2003}). Consider the examples in (22) below. In these cases, a possessor is marked by a Set a morpheme that appears in the pre-nominal position just as it does in the verb phrase. The presence of the possessor may also trigger a suffix on certain noun classes but not others. For instance, in (22a), there is no suffix on the possessed noun otoch (‘house’) but in (22b), the suffix -el appears on the inalienably possessed noun k’i’ik’ (‘blood’).

Vl (= vowel + l) suffixes like -el in (22b) mark different types of relations among the head noun and another np within the noun phrase. For instance, in (23), the head noun kaaj (‘town’) may be marked with -al when the other noun is interpreted as an inhabitant of the town (23a) while -il appears if the other noun names the town as in (23b).

As shown below in (24), the Set a marker can be separated from the head noun by adjectives and possessive classifiers, just like verbal heads can be separated from it by adverbs.

Taking this discussion into consideration, I propose that the internal structure of the noun phrase is as in (25).

(25)

As in the verbal domain, I propose that nouns raise to the projection n. The n head introduces nominal arguments such as possessors and is the locus of relational suffixes. Its specifier is right-projecting just like v. The Agr projection that triggers Set a agreement morphemes in the verbal domain is also present in the extended np (^{Pye 2011}). Finally, like certain monosyllabic adverbs and quantifiers in the verbal domain, I propose that adjectives are merged between Agr and nP. A sample derivation for (23a) is given below.

The morphological and syntactic parallelism between verb and noun phrases is very clear based on these properties. A further parallelism is that there is a dp-internal position for interrogative possessors (see ^{Aissen 1996} and ^{Lehmann 2002} [1998]). Interrogative possessors precede the head noun and the entire dp constituent appears in the pre-verbal focus position as shown in (27).

I follow ^{Aissen (1996}) in assuming that the interrogative possessors in such cases move within the extended projection of the noun phrase to spec, dp and then the entire dp is pied-piped to spec, InflP. The internal structure of a dp with an interrogative possessor is shown in (28).

(28)

The syntactic properties of adjectival and prepositional phrases are not nearly as well studied as verbs and nouns in any Mayan languages (see ^{Coon 2016}: 539 for a brief discussion on Mayan languages generally and ^{Bricker et. al. 1998}: 371-382; ^{Lois & Vapnarsky 2003}: 72-76; ^{Verhoeven 2007} for morphological and syntactic descriptions of adjectives in Yucatec). For the purposes of this paper, it will suffice to point out two properties of adjectives. First, gradable adjectives may be modified by degree terms, which appear before the head a as in (29a). Second, some adjectives may take arguments or adjuncts that are typically introduced by the preposition ti’ as in (29b).

Rounding out the discussion of internal structure of phrases headed by categories other than verbs are pps. Mayan languages typically have a very limited inventory of prepositions and express spatial relations with a set of nouns called relational nouns (see ^{Coon 2016} for an overview). Yucatec has two prepositions, ti’ and ich(il), which are followed by a nominal complement. Example (29b) contains a pp with the preposition ti’, which functions as the stimulus argument of the adjective sajak (‘scared’). As is the case with interrogative possessors of nouns, interrogative complements of prepositions are fronted within the pp and then the entire pp is pied-piped to a pre-verbal focus position as shown in (30) (see ^{Gutiérrez-Bravo 2015}a and references therein for details).

3.1. Clauses with non-verbal predicates: descriptive properties

Clauses that contain non-verbal predicates consist of a predicate to which a Set b morpheme attaches as an enclitic, as in (31).

Their defining characteristics are detailed in a number of previous works (see ^{Armstrong 2010}; Mateo Toledo 2011; ^{Pye 2011}; ^{Vapnarsky 2013}; ^{Coon 2014}). I will focus on two properties here: the position of Set b clitics and the lack of am marking and status suffixes.

As we saw above, the Set b clitic always appears in the same position with respect to verbal predicates in Yucatec -it attaches to the right of the status suffix on the verb. In non-verbal predicates, it attaches phrase finally in predicate nouns and adjectives with pre-nuclear modifiers as shown in (32).

On the other hand, in cases where a predicate noun or adjective has a post-nuclear argument or adjunct, the Set b morpheme nearly always cliticizes to the head noun or adjective, thus “splitting” the head of the non-verbal predicate from its argument or adjunct (see ^{Lehmann 2002} [1998]: 43 for details). This is shown in (33).

I have found very few cases in which the Set b morpheme may cliticize phrase-finally in the presence of a post-nuclear argument or adjunct. One such example is shown in (34), where the Set b morpheme cliticizes to the possessor sajlu’umkeep rather than the head noun aal.

Other cases of such phrase-final attachment involve the connective bey (‘like’, ‘=as’). It may take nominal complements in similative constructions of the form you are like a dog. According to the informants consulted for the context in (35), the similative interpretation is possible regardless of the position of the Set b clitic. That is, it does not have to appear on the head adjective bey but may also appear phrase-finally, after soots’ (‘bat’).

^{Vapnarsky (2013}: 71) reports that for such interpretations in her data, the Set b clitic appears on the head bey and if it appears on the noun that follows, the structure is interpreted as a basic np predicate with a causal conjunction of the type since/because you are a bat. One potential way of accounting for this observed difference is to claim that for some speakers like my consultants, bey may form a complex head with the nominal complement similar to the English expressions cat-like or bear-like, and this complex head acts as the locus of cliticization. Moreover, this could be extended to the example in (34), which could be taken as a complex head similar to the English expression son-of-a-gun (and variants).⁵ In sum, the data in (32-35) simply corroborate the claim in §2.1 that a Set b clitic must cliticize to the right of the lexical head that thematically licenses it. For verbs, this is the complex v-v head formed by the verb stem and status suffix and for non-verbal predicates it is the lexical head of the predicate (e.g. 32) or the complex head formed by a noun and a relational suffix (e.g. 33a) or by multiple heads within the predicate (e.g. 34-35).

Prepositional phrases headed by ti’ and ich(il) cannot be directly predicated of an argument and instead require the adjectival existential predicate yaan (see ^{Bricker 1981} for description of its origin and ^{Lehmann 1993}, 2002 [1998]; ^{Verhoeven 2007}; and ^{Vapnarsky 2013} for descriptions of its adjectival properties).⁶Yaan appears in at least three types of non-verbal predicate sentences: existentials, locatives and possessives (see ^{Freeze 1992}; ^{Myler 2016} for a detailed discussion of these cross-linguistically and Lehmann 2002 [1998]; Verhoeven 2007; and Vapnarsky 2013 for detailed discussions of Yucatec). As can be observed in (36), the position of the Set b clitic is on the adjectival head in each case while the location/possessor can be clause final (36a-b) or between yaan and the subject (36c).

A second property of non-verbal sentences is that they cannot appear with am markers and are uninflected for status as shown in (37b).

The lack of am marking and status suffixes in non-verbal predicates establishes a basic distinction between dynamic predicates that describe changes in time and non-dynamic or stative ones like non-verbal predicates.

One exception to this generalization is discussed in ^{Vapnarsky (2013}), who expands on earlier observations reported in ^{Lehmann (1993}) and ^{Verhoeven (2007}), showing that non-verbal predicates may take a type of status suffix called extra focal or manner focus status in the literature on Yucatec (^{Bricker 1981}; Lehmann 1993; ^{Bohnemeyer 2002}). Extra focal status suffixes appear on verbs when there is a focalized constituent with a manner interpretation to the left of the verbal complex, arguably in spec, InflP (see §2.2 above). In completive aspect, transitive verbs are marked with -il and intransitive verbs with -ik as shown in (38).

Predicate adjectives may appear in this construction and are marked with -il, which is homophonous with the extra focal status marker of completive transitive verbs, as in (39).

More generally though, non-verbal predicates can themselves be used as a type of manner focus, in which case yaan appears with an extra focal status suffix. This may either be the -ik of intransitive verbs or the -il which is used with non-verbal predicates as can be observed in (40).

Finally, the extrafocal status suffix -il appears on yaan in constructions where certain focused non-verbal predicates are used as “accidental conditions on the participants” (^{Vapnarsky 2013}: 67, citing Andrade 1955 and ^{Lehmann 1993}). Some notable examples from Vapnarsky’s work are given in (41).⁷

To sum up, am marking is universally absent in non-verbal predicate sentences while status-marking is limited only to extra focal contexts, where a manner adverb or the non-verbal predicate itself occupies a focus position to the left of the core elements of the clause.

3.2. Theoretical questions associated with small clauses

Non-verbal predicate sentences in Yucatec are examples of matrix small clauses since they consist of a subject, a predicate and nothing else. Small clauses have generated many important questions for syntactic theory over the years (see ^{Citko 2011} for an overview). Three of them are relevant for the objective of the current work: (i) establishing what the syntactic structure for relating the subject and predicate is, (ii) determining how the sole argument of the predicate is formally licensed and (iii) diagnosing the presence of higher functional material in the clause headed by the non-verbal predicate. I will consider each of these questions in turn, discussing previous literature that engages with them. There are essentially three ways of relating a non-verbal predicate to its subject. The first is to say that the predicate projects the label of its lexical category when it merges with the subject (Stowell 1983). On this view, the subject-predicate relation is established between a specifier position, where the subject is generated, and the rest of the material within the phrase. This view is criticized by ^{Bowers (1993}), who provides a wealth of data in favor of relating subjects and non-verbal predicates via a dedicated predicational head, Pr. Many theories of non-verbal predication share with Bowers the idea that the subject and predicate are related as a specifier and complement of a particular head, but differ as to its label and whether its sole purpose is to establish predicational relations (^{Adger & Ramchand 2003}; ^{Baker 2003}; Den Dikken 2006; Citko 2008). Finally, a third possibility is explored in ^{Moro (1997}; 2000), who revives ^{Williams’ (1975}) original small clause proposal for complements of verbs like consider. He claims that small clauses are exocentric, symmetric projections that contain a subject and a predicate. The order of the subject and predicate is determined by a movement operation. Either the subject or predicate raises to a higher functional projection from which it c-commands and precedes the other member of the small clause. These three approaches are represented abstractly in (42) below. The label f is a placeholder for different types of heads that have been proposed in the literature that mediate the subject-predicate relation.

(42)

The literature on non-verbal predication in Mayan languages has approached this phenomenon from different perspectives. ^{Armstrong (2010}) proposes that non-verbal predication in Yucatec should be analyzed with structure like (42b). Following work by ^{Bowers (1993}), ^{Baker (2003}) and Den Dikken (2006), he reasons that if predication relations are mediated by a head, which is null in most cases, there should be at least some instances of this head having a phonological realization.

It is important to note that this head is not to be confused with a copular verb, which exists in some languages as an anchor for inflectional morphology generated above the core predication relation (see ^{Baker 2003}; ^{Benmamoun 2008} and references therein for discussion) Instead, it should be realized by a head in instances that lack higher inflectional material. Such relators or copular particles have been identified in certain types of embedded small clauses in European languages like English and Spanish, where a preposition arguably mediates the relation between the subject-predicate relation selected by the matrix verb, as in (43).

^{Armstrong (2010}) claims that Yucatec has one such element, the term bey (‘like’, ‘as’, ‘thus’), which was introduced in the previous section (see ex. 35). The idea outlined there is that the examples in (44) are both derived from an underlying structure like (42b).

In (44a), the head f is null and the Set b clitic attaches to the predicate. In (44b), the head f is realized by bey and the Set b clitic attaches either to the f head or to the predicate. Armstrong is aware that these are not semantic equivalents and claims that there must be two semantically distinct f heads in order to capture the difference in interpretation. While this idea is a possible analysis of bey, another analysis is to simply treat bey as an adjective that takes a nominal complement, similar to yaan, which takes pp complements as shown above. Indeed, like yaan, bey takes the suffixes -tal and -chaj, which derive verbs from adjectives. On this view, the two examples in (44) would instantiate different types of non-verbal predications, one with an np predicate and another with an ap predicate. Thus, the conclusion is that bey does not provide definitive evidence for distinguishing between the abstract representations in (42). ^{Vapnarsky (2013}) has further shown that there is no other potential candidate for a relator or copular particle in Yucatec that might occupy the f position in (42b).

Alternatives to this approach have been pursued by ^{Pye (2011}) and ^{Coon (2014}). Pye (2011) claims that non-verbal predicates are instantiations of a stateP projection that is embedded under another projection that is the locus of finiteness in the clause.

(45)

In this approach, an entire phrasal projection, such as dp or ap, may be treated itself as the instantiation of the head state, which is a predicate that takes an argument as its syntactic sister. It seems to me that this approach is attempting to reconcile the fact that there is little evidence for a relator or copular particle with the idea that predication itself is licensed by something external to the non-verbal predicate. The main problem here is that there is no precedent for inserting an entire phrase into some head that then projects its label like state is purported to do.

^{Coon (2014}) proposes something similar that avoids these problems, building on ideas in ^{Baker (2003}) and ^{Sabbagh (2011}). Baker (2003) claims that the head f in (42b) is Pred and that it is universally required in non-verbal predications because non-verbal categories are unable to license an argument without it. Basically, the defining property of verbs with respect to other categories is that verbs project specifiers and take arguments without the help of a mediating head. Nouns and adjectives, on the other hand, require Pred in order to be converted into argument-taking expressions. Sabbagh (2011) revises Baker’s strong claim, proposing that in some languages there is no distinction between lexical categories based on argument versus non-argument-taking capacity. Instead, all categories can take an argument directly and thus all types of lexical categories can directly instantiate Pred. Sabbagh’s (2011) claim is based on data from Austronesian languages and Coon (2014) applies it to Ch’ol Mayan. The proposed representations for the Ch’ol examples in (41) are shown in (42). Following ^{Vapnarsky’s (2013}) proposal for Yucatec, the main idea is that all classes of roots are predicates and this can be captured by claiming that they project a Pred label. Different types of inflectional material above Pred ultimately derive differences in dynamic versus stative interpretations and the possibility of combining with aspect markers.

(47)

This proposal does not encounter the problems mentioned for ^{Pye (2011}) since it is a root rather than an entire phrase that is inserted under the Pred head. However, it raises some theoretical questions. Consider the pair of examples in (48). The a-example has a non-verbal predicate and the b-example a verbal one. In (48), there is little obvious difference between the morpho-syntactic properties of the possessor in (48a) and the transitive subject in (48b): both appear after the predicate and trigger Set a agreement.

Even though ^{Coon (2014}) does not discuss the details of how to capture such parallels in the paper, if we take §2 as our base, the conclusion is that the possessor is licensed dp-internally and triggers Set a through a relation with a dp-internal Agr projection. On the other hand, transitive subjects are licensed in v, externally to the verbal predicate and trigger Set a through an agreement with Agr in a higher clausal position. If the predication relation established between a noun and its argument and a verb and its internal argument were identical as in (46), it is not clear how to capture the different ways in which nominal possessors and transitive subjects are licensed with respect to themes. In (48a), the noun + possessor is predicated of the theme argument whereas in (48b), it is the other way around, the verb + theme is predicated of the agent. The only way to capture this difference would be go back to ^{Pye’s (2011}) proposal and claim that the dp u jmeenil le kaaja’ is inserted under Pred, which leads to the same set of problems. Otherwise, we would have to say that possessors are added to predicate nouns in a fundamentally different way than they are added to non-predicate nouns: they would merge with PredP in the former and with np in the latter. This seems like an undesirable way to integrate possessors into nominal predications.

The only proposal that has not been explored in the formal syntactic literature in non-verbal predicates in Mayan languages is the one that can perhaps most easily capture its properties. If we propose that the relation between subject and predicate is an exocentric, small clause constituent as in (42c), none of the problems faced by the previous proposals described above arise. I propose that all non-verbal predicates in Yucatec license their argument as a syntactic sister of a small clause, following ^{Moro (1997}; 2000). The basic representation for (48a) is shown in (49). On this view, there is no linking element between subject and predicate, which accounts for the lack of copular particles in the language. Moreover, predicates can be phrasal, which solves the problems highlighted with the proposals in ^{Pye (2011}) and ^{Coon (2014}).

(49)

^{Moro (1997}; 2000) claims that one member of the small clause must raise to a higher functional projection, mainly spec, tp. The motivations for raising are linked to the extended projection principle (epp), which states that the specifier of tp must have some phonological material in it and linearization, which in Moro’s (2000) proposal is dictated by the linear correspondence axiom (lca, see ^{Kayne 1994}), which states that in order for x to linearly precede y, x must asymmetrically c-command y. By raising one member of the small clause to spec, tp, both of these grammatical principles can be satisfied. While this works well for the languages Moro investigates, English and Italian, there are some complications applying it to Yucatec. First, since there is no am marking in non-verbal predicates, it is not obvious whether Infl is actually present. Second, in the most common types of non-verbal predicate sentences, the predicate precedes the subject, which is enclitic on the head of the predicate. I give supporting evidence below for the presence of higher functional projections above sc and then propose how to derive the word order in (48a) from a structure like (49).

Both ^{Pye (2011}) and ^{Coon (2014}) propose that the predicational core is embedded under some functional projection (see Mateo Toledo 2011 for similar conclusions based on an analysis of Q’anjobal). For Pye (2011) it is the locus of finiteness and for Coon (2014) it is stativity. Following the discussion in §2, these two projections correspond roughly to what I have labeled Infl and v. Since am marking and status suffixes are generally absent in non-verbal predicates, it is important to adduce evidence in favor of representing one or both of these projections in non-verbal predictions.

As for status marking, we have already seen that it is possible in the manner focus construction, but absent in all other cases. To simplify representations, I will assume that status marking is present in manner focus contexts but absent in all others. It is highly probable that its presence is triggered by a dynamic reading attributed to the non-verbal predicate that is associated with an episodic interpretation in manner focus contexts. I will not discuss the issue further here (see ^{Vapnarsky 2013}; ^{Coon 2014} for a more detailed discussion).⁸

The next question concerns the presence of Infl. As discussed above, all non-verbal predicate sentences lack am marking. This means that the presence of Infl above the proposed sc would always be null. Recall from §2.1 that certain completive intransitive verbs may lack an overt am marker but still have the distribution of other completive sentences that have such overt aspect marking. Thus, it is not without precedent to propose that there is a null Infl in non-verbal sentences (see Mateo Toledo 2011 for similar arguments from Q’anjobal). In order to establish an argument that it exists despite lacking a phonological exponent, I will compare and contrast an intransitive verb with the completive am marker j with a non-verbal predicate. Side-by-side, these two sentences look strikingly similar.

While (50a) and (50b) are clearly different in that one is dynamic and the other is stative, they share an important set of semantic and morpho-syntactic properties that can be linked to the presence of a projection responsible for finiteness/aspectuality. First, if both have this projection, we expect it to be the target of a certain class of temporal modifiers. Indeed, temporal modifiers of completive verbs include those that refer to periods prior to speech time such as jo’oljeak (yesterday) and appear in either the pre-predicate topic position or in a post-predicate position as shown in (51).

Non-verbal predicates also admit temporal adverbs such as ka’ach(ij) (‘formerly’). This adverb is primarily used with stative predicates in order to describe a state that held at a point prior to speech time but no longer does (see ^{Bohnemeyer 2002}). It appears in the same positions as other temporal adverbs.

^{Bohnemeyer (2002}) analyzes these adverbs as topic time shifters. They establish a relation between the topic time, or temporal reference interval of the clause, and coding time, the time interval of the act of speech. I suggest that without some kind of higher projection, such as Infl, to establish what the temporal reference interval of the clause is, there would be nothing for these adverbs to modify.⁹

A second reason to believe that there is a null stative counterpart to the dynamic am marker j is that both types of clauses behave identically with respect to negation. ^{Gutiérrez Bravo (2015}b: 138, ex. 29) shows that negation is generated above the projection that houses am markers, which in our case would be above Infl, as in (53b). In completive aspect, this high clausal negation position is filled with ma’ and triggers the presence of the enclitic -i’ to the right of the verbal predicate.

Non-verbal predicates behave in the same way as completive intransitive verbs with respect to negation as shown in (54).

One might ask how the parallelism between (53a) and (54) could be taken as evidence of Infl. I suggest that by positing the presence of null Infl in (54), clausal negation can be given a uniform syntactic and semantic analysis. Essentially, it selects InflP as its complement and negates the time interval for which the event/state described by the predicate is specified to occur/hold.

Taking these observations into consideration, I propose that non-verbal predicate sentences instantiate the following basic structure.

(55)

Proposing that Infl exists in non-verbal predicate sentences not only allows for a unified account of the topic time shifter ka’ach(ij) and sentential negation in verbal and non-verbal sentences, it also means that there is no need to introduce a new formal licensing mechanism in order to explain why the subject appears as a Set b clitic in the latter. Recall from §2.1 that all intransitive subjects are licensed through an agreement relation with Infl, and it is no different for non-verbal sentences.

Finally, I claim that Infl in (55) has an epp feature that requires phonological material in its specifier, thus triggering movement. While the original formulation of the epp (extended projection principle, see ^{Chomsky 1981}) was to ensure the head purportedly responsible for establishing subject-predicate relations, namely ip, had a subject or some other material to occupy its specifier, the nature of the epp has changed significantly in the years since (^{Alexiadou & Anagnostopoulou 1998}; ^{Adger & Ramchand 2003} and ^{Bowers 2010} for a discussion of different strategies employed cross-linguistically for satisfying the epp). I will assume that the epp on Infl in Yucatec is a purely formal feature that can be satisfied by moving an xp to its specifier and, in keeping with the idea presented in §2.2, the xp in the specifier of Infl is interpreted by the semantic component as the focus of the sentence.¹⁰ Summarizing this information, the general properties of Infl in Yucatec are described in (56).

3.3. Summary

In this section, I have presented the main descriptive properties of non-verbal sentences in Yucatec, contrasting them with clauses headed by verbs as well as the internal structure of phrases headed by non-verbal categories. I then discussed a number of theoretical issues raised by small clauses with a specific eye toward Mayan languages. I showed that all of the existing formal syntactic accounts of non-verbal predicate sentences are lacking in some respect and proposed that by treating non-verbal predicates as part of a symmetric small clause that takes the subject as a syntactic sister, many of these problematic issues disappear. Finally, I presented evidence for including an Infl node above the small clause, which is the locus of finiteness and stative aspect. I claimed that this Infl node is endowed with the interpretable feature [-dyn] and that it has two uninterpretable features, [uϕ] and [epp], which drive the syntactic derivation of non-verbal sentences.