Propping up predicates : Adjectival predication in

In TłĮchǫ Yatıı ̀ (Dene, aka Athapaskan), copulas appear obligatorily with adjectives predicated of animate subjects, but are barred from appearing with adjectives predicated of inanimates. I propose that this asymmetry arises from a requirement to realize grammatical agreement for person, and that animate nouns alone bear a person feature. Unlike verbs, adjectives in this language cannot inflect; hence copulas are inserted in adjectival predicates as a rescue strategy to avoid ungrammaticality.


Introduction
This article examines an asymmetry in the behaviour of predicative adjectives in Tłı̨chǫ Yatıı ̀ (also known as Dogrib), a Dene (Athapaskan) language of the Northwest Territories, Canada.Although, as in other Dene languages, the majority of property predicates are expressed with stative verbs, there is a small class of adjectives that have distinct properties.Only a subset of these adjectives can occur as predicates, and when they do they sometimes occur with a copula.Although the occurrence of copulas with predicative adjectives may appear to be optional, I argue that this is not the case.Copulas are obligatory with adjectives predicated of animate subjects and barred with those predicated of inanimate subjects: I argue that animate subjects bear a formal person feature that inanimates lack, and that the copula that appears with adjectival predicates realizes these features inflectionally, a requirement of the grammatical system that I formalize as a constraint.This realization is otherwise impossible to fulfill due to adjectives' lack of inflection.23 4

Assumptions
The article's arguments are couched in the framework of Minimalism (Chomsky, 1995;  1998), assuming that morphological agreement is an operation motivated by a requirement to check formal features: functional heads bear uninterpretable features that are checked against, and valued by, the interpretable features of their agreement targets.I supplement this view of agreement with a constraint requiring all Φ-features, whether interpretable or uninterpretable, to be checked.That is, an unmatched feature, even if interpretable, will cause the derivation to crash.

TłĮchǫ Yatıì and its morphosyntax
Tłıchǫ Yatıì (formerly known as Dogrib) is a Dene language spoken in the communities of the Tłıchǫ Government and in nearby Yellowknife, Ndilǫ and T'èʔehdaà (Dettah), Northwest Territories.
In common with other Dene languages, Tłıchǫ Yatıì has SOV constituent order and a highly synthetic verbal morphology showing both subject and object agreement.The verb consists of a monosyllabic stem to which are prefixed numerous inflectional and derivational morphemes, in general also monosyllabic. 5The Dene verb is typologically unusual in a number of ways.First, the lexical material is discontinuous: the minimal verb, referred to in the literature as the theme, consists of the stem at the right edge, plus, in most cases, one or more lexical and derivational prefixes, referred to as thematic.Second, these thematic elements occur outside (leftward of) the inflectional prefixes for agreement and aspect, which immediately precede the stem.Thirdly, object agreement occurs outside of subject agreement. 6n example of a representative Tłıchǫ verb appears in (2).
(2) Verb morphology in Tłıchǫ Yatıì Naxısınıyats'eehtı.naxı-sınì-ya-ts'e-e-Ø-h-tı 2pl.obj-thm-thm-1pl.sbj-cj-ipfv-clf-speak.ipfv 'We judge you.'  (MLBW 2012)   This verb is inflected for subject and object agreement and aspect and has two lexical portions: the thematic prefixes sınì-and ya-, which together etymologically mean something like 'prepared words', and the stem tı (etymologically 'speak'), which together with the classifier morpheme (h-in this case) comprise the verb theme sınìya-h-tı 'judge'. 7s can be seen by comparing the morpheme-by-morpheme gloss with the word, Tłıchǫ verbal morphology is highly fusional. 8In particular, the conjugation marker, aspect/mode, subject agreement and classifier are typically pronounced as one syllable.9Henceforth, therefore, I will gloss examples as in (3), abstracting away from the conjugation marker and classifier: (3) Simplified glossing for verbal morphology Naxısınìyats'eehtı.naxı-sınìya-ts'eeh-tı 2pl.obj-thm-ipfv1pl.sbj-judge.ipfv'We are judging you.' (MLBW 2012)   Rich systems of verbal agreement morphology are pervasive within the Dene language family.However, in contrast, there exists a class of predicates in Tłıchǫ Yatıì that is uninflecting, hosting no aspect or agreement morphology.The class of predicative adjectives in Tłıchǫ Yatıì reveals facts about agreement and predication that provide a window into the realization of Φ-features.

Adjectives in Tłıchǫ Yatıì
Adjectives in Tłıchǫ Yatıì are a small lexical class.10They lack inflectional morphology, unlike nouns, which inflect for possession, or verbs, which inflect for aspect/mode and subject and object agreement.

Identifying adjectives
There are reliable diagnostics for distinguishing adjectives from stative verbs (which also denote properties) and from adverbs (which also are uninflected).

Adjectives versus stative verbs
Most concepts expressed in English by adjectives are expressed in Tłıchǫ Yatıì by stative verbs.Nevertheless, adjectives do exist as a class, and can act either as predicates or as modifiers of nouns.They can be distinguished readily from stative verbs by their lack of inflectional morphology.
In (4) we can see that the adjectives eya 'sick/painful' and edı 'warm/feverish' have no inflectional morphology: rather, this morphology occurs instead on the copula, marked for first-person subject agreement (4a) and perfective aspect (4b).Further, adjectives often occur as bare predicates, without even a copula, such as edza 'cold ' and ehkw'ı 'correct' (4c-d).By contrast, the stative verbs etedeht'ı ̨ 'be poor/pitiful' and ełèak' à 'be wrinkled' bear first-person subject agreement (4e) and perfective aspect (4f) on the verb word itself, and no copula appears:11 (4) Adjectives contrasted with stative verbs It is lack of inflection both for agreement and for aspect that sets adjectives apart from verbs.

Adjectives versus adverbs and nouns
Adjectives are distinguished from the class of adverbs by occurring as complements only of copulas and the psych verb nı ̨wǫ 'think/believe/consider', or as modifiers, and by taking complements of their own; they are distinguished from attributive nouns by appearing after the nouns they modify rather than before. 14The adjective ahxe 'rich' appears with the copula elı ̨ (5a), while in (5b), edza 'cold' is the complement of the psych verb ts'ı ̨hwhǫ 'we think'.Adjectives are barred as complements of wegaat'ı ̨ 's/he appears, is seen as' in (5c), which takes clausal complements (5d).In (5e) we see the adjective ezhı ̨ 'crazy' appearing with the copula, and in (5f), ezhı ̨ taking computer ghǫ 'about computers' as a complement.In (5g, h) the adjectives deèdlı ̨ı ̨ 'original/authentic' and edza 'cold' appear attributively, following the nouns they modify, in contrast to the noun nı ̨htł'e 'paper/ book' (5i), which (in compound nouns) precedes the noun it modifies.
( The morphological and selectional criteria outlined above allow us to identify the following (almost certainly not exhaustive) list of adjectives.1 5 Note that morphosyntactic criteria alone can identify these adjectives.Semantic relatedness is not characteristic of them as a class: they do not correspond to any of the semantic classes that have often been considered canonical, and "typically associated with both large and small adjective classes": dimension, age, value, colour (Baker,  2003; Dixon, 2004).In addition, many of the adjectives in Table 1  In the next section, we will look at the occurrence of copulas with P-adjectives, demonstrating that a copula appears with such an adjective if and only if the subject is animate.

Asymmetries in copula behaviour
While adjectives are a small class, and P-adjectives smaller still, they are disproportionately illuminating for our picture of agreement and predication in Tłıchǫ Yatıì.
There is an asymmetry in the behaviour of adjectival predicates; this asymmetry occurs when P-adjectives occur either as predicates or as modifiers, and gives us a window into the agreement mechanisms of the language, and subject-predicate relations in general.

Asymmetries in predicates
Recall from section 3 that while adjectives bear no inflectional morphology, such morphology does occur on copulas that appear with adjectival predicates: (8) As mentioned at the beginning of this paper, copulas appear with P-adjectival predicates if and only if the subject of the clause is animate. 17We saw examples of this pattern in (1); in (10) we see a further example.Ìį ̨zha 'shy/ashamed' may be predicated of chekoa 'children' only if the predicate includes the copula elı ̨ (a); without the copula (b), the sentence is ungrammatical.The same is true of eya 'sick' and its subject tłı ̨ '(the) dog' in (10c, d). ( 10 In ( 11) and ( 12) the presence or absence of the copula determines the interpretation.The inclusion of the copula forces an animate interpretation, as in (11a) and (12a), whereas the bare adjectives in (11b) and (12b) are interpreted as predicated of inanimate subjects (a body part or the weather, respectively).
( Conversely, if the subject is inanimate, the inclusion of a copula is ungrammatical, as in (13).
Bare adjectival predicates (a) are the only option when the subject is inanimate, and P-adjectival modifiers of inanimate nouns must likewise be bare (c).Adding a copula makes the sentence ungrammatical (b, d).
[ Copulas, then, are necessary for P-adjectival predication of animate subjects, and barred for inanimate subjects.The next section demonstrates that this fact is also true when P-adjectives are relativized and used to modify nouns.

Asymmetries in modifiers
All verbs and predicative adjectives in Northern Dene languages can be turned into modifiers by the addition of a derivational suffix, which in Tłıchǫ Yatıì takes the form of an extra mora on the final vowel: Since modification by P-adjectives or relative clauses is derived from predication, it is unsurprising that the asymmetry in modificational P-adjectives should parallel that in predicational ones.
The role of the copula with P-adjectives is the same whether the adjective is used as a predicate or a modifier: the copula rescues the clause from ungrammaticality by providing inflectional realization of features for an uninflecting adjective.The next section proposes that this asymmetry has its origin in two conditions of the language: a requirement to realize the Person features of subjects in agreement morphology, and the presence of such features only on animate nouns.

Animacy and agreement
The rich morphological system of verbs in Tłıchǫ Yatıì, like those of other Dene languages, marks agreement with the person and number of subjects and objects.I propose, however, that agreement is not manifested with inanimate nouns, and that this fact is explained by a lack of Person on such nouns.

Verbal subject agreement
Agreement with a verb's subject is marked in the affixal region immediately to the left of the verb stem, as in (17).However, as mentioned in section 3.1.1,this marker is a zero morpheme for third-person singular subjects, as in (17b The immediate impression that one receives from this is that agreement with all thirdperson subjects is null, regardless of animacy.This is incorrect, however: there are discernable differences between the facts of agreement with inanimates and with animates.
We have a further parallel, then, between verbal and adjectival modification, just as between verbal and adjectival predication, the distribution of the copula mirroring the distribution of plural number agreement:

Person agreement and apostrophizing inanimates
We have seen that inanimates do not trigger plural subject agreement.In fact, it is arguable that the majority of inanimates cannot trigger person agreement at all.It goes without saying that in ordinary discourse, inanimate nouns cannot be first or second person.However, when an artificial context is created in which inanimate nouns are apostrophized as the personifications of natural forces, treating them as animate and agentive entities, second-person agreement may become acceptable, but with only a small subset of inanimates. 20The sentences in ( 22 20 These "honorary animates" show other unusual syntactic behaviour as well; see Rice & Saxon (2005:  714-716) and Willie (2000: 365) for discussion. 21Other inanimates behave differently when apostrophized in this way.Most inanimate nouns trigger no subject agreement even in these contexts: a sentence like "Meat, why aren't you cooked yet?" was judged ungrammatical by my consultants when the verb was inflected for second-person agreement.Both consultants whom I asked rephrased the sentence without agreement.Similarly, adjectival predicates of most inanimates did not trigger copula insertion, even in apostrophizing contexts.

Subject animacy
Animate plural agreement-V A + plural agreement-copula Inanimate V A

Relativized clause Adjective
Subject animacy Animate plural agreement-V A + plural agreement-copula-μ The only direct object agreement markers that may refer to inanimate objects are the class of anaphors, of which there are several in Tłıchǫ Yatıì: the reflexive, reciprocal, and disjoint.The first needs no explanation here.The second indicates a plural object coindexed with a plural subject where the verb has an interpretation of mutuality ( 27).The third indicates an object that is not coindexed with the subject (28). 24None of the anaphors is specified for number, except that semantically, of course, the reciprocal is only compatible with subjects referring to two or more entities.
( Notice that the anaphors in these examples are compatible with both animate (27a, 28a) and inanimate (27b, 28b, c) objects.Therefore we can say that whereas the other object agreement markers impose an animacy restriction on the objects to which they refer, the only restrictions affecting anaphors are those of coindexing.The properties of the object agreement markers are summarized in Table 4.

Object marker Restrictions
we-  Inanimate nouns may not be referred to with the direct object markers we-or gı-and may not be specified for plural. 25When the subject of a transitive verb is non-third-person, an animate object may be referred to with we-, or gı-if it is plural; however, an inanimate object may only be marked by by zero, or by the disjoint anaphor if the verb's subject is third person.In other words, inanimate nouns occupy the least privileged spot with respect to direct object marking.

Postpositional agreement
In Tłıchǫ Yatıì, as in other Dene languages, postpositions may show inflection for the person and number of their arguments.This fact is demonstrated in (29), where a firstperson singular pronominal argument of gha 'for' (a) contrasts with a third-person plural argument (b).( 29 We see, therefore, that not only are inanimate nouns uniquely underprivileged in the subject and object agreement systems, but as objects of postpositions as well.

Analysis
The data in the previous section have shown that inanimate nouns pattern in a way radically different from animates.Animates, but not inanimates, may trigger plural agreement for subject and object on the verb and for objects of postpositions.A small subclass of inanimates referring to natural forces may trigger second person agreement on verbs when apostrophized in artificially coerced contexts, but the great majority of inanimates do not.

Personless inanimates
Since Benveniste (1974), it has been recognized that nouns may behave as though they lack Person.In his original formulation, nouns in general had this characteristic: in other words, third person was in fact the manifestation of Personlessness.This analysis is appealing because it allows us to reduce the complexity of person inflection: Anderson (1992:  92), for example, treats it as the interaction of features [me] and [you], where third person is {[-me], [-you]}. 26n recent years it has been recognized that this picture requires finer-grained distinctions.Béjar (2003), in analyzing the agreement systems of Georgian and Nishnaabemwin, finds a necessity to distinguish between different kinds of third person.I suggest that the agreement system of Tłıchǫ Yatıì displays such a distinction as well, and that it is animacy that distinguishes the two kinds, inanimate nouns being truly personless while animates bear a [person] feature but no further specification.
Animate nouns can be speakers, listeners or referents, while inanimates are ordinarily restricted to being referents.This fact hints at a basic semantic difference between the two.In Tłıchǫ Yatıì, this semantic difference is reflected in morphological and syntactic structure: animates can trigger various morphological inflections for person and number that inanimates cannot.I posit the system in Figure 1, where [person] primarily distinguishes animates from inanimates: those nouns that trigger inflectional agreement from those that do not.Nouns that bear [person] are licensed to bear [participant], which separates discourse participants from all other referents; [speaker] then separates first from second person.This geometry expands upon that of Harley & Ritter (2002: 486), on which it is partly based, by adding the feature [person] to separate animates from inanimates.
The facts presented in this paper might lead one to suppose that inanimate nouns also lack Number.While this is a tempting analysis, I believe that it is incorrect.The Dene system of classificatory verbs, by which sets of contrasting suppletive verb roots are distinguished according to the properties of their subjects or objects, is sensitive to the number of subjects and objects, even when these are inanimate nouns.I therefore conclude that the absence of inflectional Number realization with inanimate subjects is a consequence of their lack of Person, since Number and Person inflection are fused in this language.

Effects in the morphosyntax
The system in Figure 1 plays out in the morphosyntax of Tłı̨chǫ Yatıì in the following manner.
Inflectional number marking is directly dependent upon the [person] feature.Without [person], inflectional number is unlicensed, resulting in the observed fact that inanimate nouns fail to trigger number agreement in three different domains: verbal subject agreement, object agreement and postpositional agreement. 27Without [person], person inflection is impossible on verbs with subects drawn from the majority of inanimate nouns (roughly, those to which agency cannot be attributed).Finally, it is the need to agree with a [person] feature that leads to the insertion of copulas with adjectival predicates of animate subjects. 28

Copula insertion
The presence of copulas in adjectival predicates of animate subjects and their absence from such predicates of inanimate subjects can be explained by a lack of Person features on inanimates, as discussed in the previous section.The current section explores the reasons for the insertion of copulas (rather than another verb) and the probable locus of insertion.
The Personlessness of inanimate nouns would have no consequences for predicates unless there were a requirement for Person, if present, to be realized in agreement.I formulate this requirement as in ( 31): (31) Morphological Realization: Φ-features must be realized in agreement morphology at spellout.
If ( 31) is a correct description of the principles at work, we expect that the semantic contribution of any element inserted to satisfy Morphological Realization must be minimal or zero; otherwise, the compositional semantics of predicates of animate and inanimate subjects would differ, which does not appear to be the case.That being so, the semantically minimal element that is capable of hosting Φ-agreement must be a copula, as any other verb has more complex semantics.
A question that arises is where in the clause structure this insertion occurs.Word-order data can resolve this question.It is clear that the copula, when it is present, is inserted inside TP, because it occurs before future and modal marking in linear order: 29 The next question is whether the copula is merged with the adjective at A (amounting to incorporation), or in some intervening projection.The first cannot be correct.
Copulas in this language are fully verbal, with inflection for aspect and Φ-agreement.This suggests a merger at v or higher, if v is a category-forming head that forms verbs (Halle & Marantz 1993).Furthermore, adjectives' very lack of inflection would 28 As indicated in footnote 17, section 4.1, the set of nouns that trigger the animacy effects detailed in this paper varies from language to language within the Dene family, and even intralinguistically from speaker to speaker.An anonymous reviewer wonders whether the three main phenomena described and analyzed here (verbal subject agreement, object agreement, and copula insertion with adjectival predicates) might be sensitive to different categories for some languages or speakers: for example, copula insertion might be triggered by nouns referring only to humans, and subject agreement by those referring to humans and animals.This may be so: in Dëne Sułıné, a close relative of Tłıchǫ Yatıì, it is apparently possible to say łı ̨ ʔeya 'the dog is sick', without a copula (Cook, 2004: 295).Whether the phenomena investigated herein are unified across Dene languages as a Person distinction, and what set of nouns bears Person in a given language, is a topic for further research. 29I argue in a separate paper (Welch 2015) that future is realized at T (or INFL) in Tłıchǫ Yatıì, and that past marking is not (and in fact is optional and adjunctive).
seem to preclude an inflected item of category A, which is what would result from the copula merging into A.
On the other hand, merging them at v or Asp yields the result that their aspect and agreement inflection occurs outside of A, in accordance with the facts.
The locus of morphological agreement seems to be Asp rather than T in this language.As we have remarked already (sections 2 and 6.1), aspectual marking is typically fused with subject agreement marking; tense marking is not.Since copulas with AP predicates realize agreement morphology, and since word-order evidence points to their insertion inside TP and outside AP, I conclude that they are merged in Asp, as in Figure 2.

Conclusion
Adjectives pose a unique problem to the computational system of Tłıchǫ Yatıì in being unable to agree with the Person feature that is proper to animate subjects.Copulas are therefore inserted at Asp in order to realize agreement morphologically.The failure of inanimate subjects to trigger inflectional person and number agreement on verbal predicates is paralleled by their failure to trigger the insertion of a copula (which, like other verbs, is marked for person and number agreement) with adjectival predicates.
Much recent work has assumed without comment that inanimacy is realized syntactically by a lack of formal Person (e.g., Adger & Harbour 2007; Alexiadou & Anagnostopoulou  2006; Bartošová & Kučerová 2015; Demonte, Fernández-Alcalde, & Pérez-Jiménez 2011;  Ghomeshi & Massam 2015; Piriyawiboon 2007; Piriyawiboon 2013; Richards 2008;  Rooryck 2000).The patterns of adjectival and verbal agreement in Tłıchǫ Yatıì are evidence that these assumptions are in fact correct for some languages.The Tłıchǫ data are part of a broader pattern of instantiation in the Dene language family as a whole (Lochbihler, Oxford & Welch 2015), so personlessness as the syntactic realization of inanimacy may well be a family-wide phenomenon.
Inflection on copulas with adjectival predicates a. Eya ehłı̨ t'à, edı ehłı.The clauses in (9) clearly demonstrate that uninflected P-adjectives are capable of being predicates without the support of a copula.What forces the appearance of a copula in (9a, b)?
Just as in predication, a P-adjective modifying an animate noun requires the appearance of a copula, in which case the extra mora appears on the copula instead of on the adjective itself.Thus (15a, c), where copulas appear with adjectives modifying the animate nouns chekoa 'child' and dǫ 'person', are grammatical, while (15b, d), which lack copulas, are not:19 In (16), we see that the converse is also true: copulas are ungrammatical with P-adjectives modifying the inanimate nouns bò 'meat' and dzęę̀ 'day(s)'.The clauses in (16b, d), which are versions of (16a, c) with an added copula, are ungrammatical. ).

Table 3 :
Verbal and adjectival predication and modification.

Table 4 :
Restrictions on object markers.