Description and determination in Tanti Dargwa NPs

1 Introduction

The semantics of nominal expressions (henceforth, NPs)^[1] often consist of at least two components, which I call here the descriptive and determinative components. The descriptive component is mostly associated with various properties assigned by the nominal and normally includes the head and a range of elements commonly described as modifiers. The determinative component relates the description to its context, either providing the reference in the universe of discourse (a function arguably fulfilled by grammaticalized articles) or interpreting it in relation to its immediate syntactic context (as do at least such quantifiers as ‘most’ and ‘every’, which make it impossible to interpret an NP without its syntactic environment; cf. Partee 1995 inter alia). For example, in the NP the black cat the expression black cat constitutes the description, while the definite article the provides determination. The determinative component has scope over the descriptive component and usually cannot be interpreted independently of the latter.

This model is widespread in formal semantic works since Barwise and Cooper (1981) (developing work by Montague (1973)). Parallels can be found in other frameworks (cf. the typologically oriented work by Rijkhoff (2002, 2008, who distinguishes between the discourse-referential layer and “descriptive modifiers”), even though the details of ascribing elements to layers vary between different approaches.

Standard Average European shows some evidence of such representation in syntax.^[2] The scope of the determinative component is recognized by the rules of its expression: the position of determiners is defined relative to the subconstituent of the NP expressing the descriptive component. For example, articles appear in the periphery of a subconstituent (as in English) or adjoined to its element showing head properties (as in Swedish) or placed in second position (as in Bulgarian), or sometimes in several of these positions (cf. Plank 2003). While such rules of placement may just as well be described as resulting from the mechanisms of grammaticalization (cf. Himmelmann 1997 for details), this does not contradict the semantic perspective, as grammaticalization processes may develop a compositional semantic structure (see Lander 2015a for discussion).

However, this picture is complicated by at least two facts. First, it is not obvious that the description/determination contrast is universally reflected in the syntactic structure of an NP since many languages lack any obvious determiner position in the syntactic structure. This question is widely debated for articleless languages such as Russian (cf. Lyutikova 2018). Second, even for languages that have been described as having such a position, the distinction between description and determination has been shown to be gradient; see, for example, Plank (1992) for German and Van de Velde (2009) for English, among others. In this article, I look at several phenomena which reflect this distinction in Tanti Dargwa and argue that even when such contrast exists, it need not be associated with the phrase structure. The relevant data also support the idea that the description/determination distinction is gradual rather than strict and discrete.

Section 2 provides background information on Tanti Dargwa and its NPs. Sections 3–5 describe three phenomena that reflect the description/determination contrasts. The implications of the findings presented here are discussed in Section 6. The last section provides conclusions and hypotheses which deserve further investigation.

2 Background on Tanti Dargwa NPs

Tanti Dargwa belongs to the Dargwa branch of the East Caucasian (alias Nakh-Daghestanian) family. While Dargwa is often considered a single language in the literature (cf. Eberhard et al. (eds.) 2020 inter alia), its “dialects” are quite different from each other. This is confirmed by recent descriptions of Dargwa varieties like Sumbatova and Mutalov (2003), Daniel et al. (2019), and Forker (2020); see also an earlier description of Standard Dargwa by Abdullaev (1954), which attempted to compare the standard variety with the “dialects”. A brief description of Standard Dargwa in English is given by van den Berg (2001), but there are also several articles devoted to Dargwa in Russian; cf. Temirbulatova (2011) for a bibliography. The Tanti variety was described by Sumbatova and Lander (2014), but their account of NPs was preliminary only and did not include much of the data presented below. The data discussed here mostly come from elicitation, though I also used some examples from texts (marked as such).

The Dargwa varieties are similar to most other East Caucasian languages in that they are morphologically ergative and show both head marking (agreement) and dependent marking (case) at the clause level, and only dependent marking at the phrase level. Some syntactic relations are reflected in gender marking, which in Dargwa usually contrasts between masculine, feminine, and neuter/non-human in the singular and human and non-human in the plural and is largely determined semantically (Lander 2015b; Sumbatova 2018). As for word order, East Caucasian languages are left-branching: they have postpositions rather than prepositions, the predicate tends to appear clause-finally (even though the languages allow much freedom in word order), and the possessor normally precedes the possessee (but see Section 5).

Not surprisingly for a left-branching language, in Tanti Dargwa complex NPs the noun which designates the referent usually (yet not necessarily) occurs finally. This noun has at least two head properties. First, it serves as a morphosyntactic locus of case marking (1), even when it is followed by other material belonging to the same NP (see Section 5). Other elements of the NP (called adnominals henceforth) are in principle capable of taking case morphology, but only when the noun is absent, as shown in (2). Because of these head properties, below I refer to this noun as the head noun.

Second, no rules are found that order other immediate constituents of the phrase relative to each other. In particular, apart from the quasi-incorporating construction discussed in Section 3 below, the relative position of adnominals is free. In fact, we have already seen this in the headless NP presented in (2), which shows that the order of an adjective and a numeral can be reversed. Example (3) shows a similar effect in the presence of a head noun: a quantifier and a relative clause can be combined in either order, without any semantic difference (though (3a) has a homonymic structure where har ‘every’ combines with dars-li-ja ‘lesson’, thus giving the meaning ‘the pupil who will come to every lesson’).

While it may be that there is an unmarked order in Tanti Dargwa NPs and the order of elements can be determined by such factors as emphasis and heaviness, I do not have enough data for such claims, partly because my corpus does not contain many NPs with several modifiers. Hence I assume that the word order in Tanti Dargwa NPs does not necessarily reflect the distinction between the determinative and descriptive components. This is especially obvious in (3), where the relative clause belongs to the descriptive component, while the quantifier manifests the determinative component having scope over the description. This pair of examples suggests that in Tanti Dargwa NPs word order does not need to reflect scope relations.

3 Attributive suffixes and the types of adnominals

In Tanti Dargwa NPs some adnominals take one of the three suffixes called “attributive” in Dargwa studies (see, for example, Sumbatova 2013; Sumbatova and Mutalov 2003):

1. the neutral attributive suffix -se (4),

2. the contrastive suffix -il (5), which overtly marks that the adnominal element contrasts the relevant subset of some set with other elements that do not possess the relevant property,^[3]

3. the plural suffix -te (6), which often (yet not always) bears some contrastive effect (see Section 4 for details).

The (in)capability of an element of the NP to take an attributive suffix (either neutral, contrastive, or plural) is a convenient criterion for distinguishing between three types of adnominal elements. Type I expressions do not take attributive suffixes and include demonstratives, indefinite pronouns, numerals, and most expressions of quantity. Type II contains expressions whose heads normally take attributive suffixes and includes participial relative clauses, adjectival phrases (probably a subclass of relative clauses; cf. Sumbatova and Lander 2014), and non-genitive adnominal NPs. Type III is constituted by possessor phrases and other adnominals which can easily function both with and without attributive suffixes, although sometimes the presence of these markers is associated with additional semantic effects.

It is easy to see that Type I consists of those elements that in languages with a grammaticalized syntactic category of determiners often fulfill the determinative function. Type II, on the contrary, includes adnominals whose primary function is to assign properties, and which serve as modifiers proper. Curiously, however, there is no clear one-to-one correspondence between the presence/absence of an attributive suffix and the semantic class of the expression. For example, while -aqal ‘many, much’ does not take attributive suffixes, kam ‘few, a little’ takes them.

Of special interest is Type III, which is not that homogeneous. First, it includes the adjective ‘other’ and some quantifiers like ‘several’ and ‘every’, which do not appear with the neutral suffix -se, yet can (but need not) be marked with either the contrastive suffix (7) or the plural attributive suffix (8), also implying contrastive semantics in this case (see Section 4):

Second, attributive markers are optionally suffixed to possessors. Both referential (9) and non-referential (10) possessors take contrastive and plural attributive suffixes with the expected contrastive semantics. When referential possessors are used without nominal heads and take case marking of the whole NP in addition to the genitive, the contrastive or plural suffix appears by default (11).

Non-referential possessors take the neutral attributive suffix without any semantic effects (12), but the appearance of the neutral attributive suffix with referential possessors results in a non-trivial semantic shift: the possessive relation is now thought to be terminated (13) (see Lander 2011 for more examples).

The semantics of the terminated possessive relation is sometimes expressed by grammatical markers crosslinguistically (cf. Nordlinger and Sadler 2004), but as far as I am aware, its expression by dependent-marking means has not been attested outside of Tanti Dargwa. In Section 6 it is suggested that this may be an epiphenomenon of the fact that the possessor takes a non-standard function here.

It is worth mentioning that there is one construction where Type II expressions lack attributive suffixes. Lander (2014) showed that in Tanti Dargwa adjectives and relative clauses can be quasi-incorporated into the nominal head, in which case they are syntactically deficient (e.g., cannot modify coordinate constructions), lose their special attributive marking, and may undergo stress shift. Cp. the quasi-incorporating and non-incorporating patterns in (14):

As argued by Lander (2014), the attributive marking in Tanti Dargwa represents an unmarked situation, while its absence is the marked choice (for example, it is much less frequent and is only allowed with a limited set of possible heads).^[4] In what follows, I abstract away from this very specific quasi-incorporating construction and consider only constructions where adnominals syntactically, morphologically and morphophonologically constitute full-fledged autonomous expressions.

4 Plural marking

The number category in Tanti Dargwa can be expressed on various elements of the NP. In general, it contrasts singular with plural: the singular value is unmarked, while the plural value can be expressed by several affixes. For head nouns, the choice of the affix is partly lexically determined, but adnominals only use the suffix -te (also found with many nouns). Furthermore, the information on number appears in gender marking contrasting between masculine, feminine and non-human in the singular and human and non-human in the plural. The gender category operates mainly at the clause level (mostly on predicates) but sometimes appears at the NP level as well: e.g., some quantifiers such as li<g>il ‘all’ and g-ajaqala ‘half’ contain gender affixes (indicated here as g).

Count head nouns describing plural referents are normally marked for plural (15), though almost always lose plural marking in constructions with numerals and certain quantifiers (16) (but see footnote 6 for a counterexample). Mass nouns do not take plural marking (if they are not interpreted as count nouns, e.g., as a number of kinds), but the whole mass NP behaves as plural: it triggers plural agreement on the verb and plural marking of non-head elements that take number marking (17).^[5]

For Type II adnominals, the appearance of plural marking depends on whether they are contrastive or not. Non-contrastive Type II adnominals normally appear with the neutral attributive suffix (cf. (15) above) and are only marginally found with the plural suffix, as in (18). Contrastive Type II adnominals are typically marked for plural (17), and only rarely allow number-neutral contrastive marking (19). As a result, plural marking on Type II adnominals almost always implies contrastive semantics.

Among Type I expressions, demonstratives are marked for number, even in constructions with numerals shown in (20) and (21).^[6] Other members of Type I do not usually take number marking proper, but they often imply the semantic number without overt marking (as do most quantifiers) or by taking gender prefixes (22).

The expressions of Type III, i.e., the possessors, ‘other’ and ‘every’, which only optionally take attributive suffixes (see Section 3), behave like Type II expressions in that number marking on them usually implies the contrastive semantics.

Thus, the appearance of plural marking on non-heads depends on the hierarchy of adnominals demonstratives > contrastive adnominals > neutral adnominals: the higher an adnominal is in this hierarchy, the more likely it is marked for plural. Moreover, given the fact that other Type I are normally also defined with respect to number (see above), I propose a more general hierarchy (23), which also appears elsewhere in the Tanti Dargwa grammar, as we will see below in Section 4 for the right-periphery construction. The top of this hierarchy is now occupied not only by demonstratives but also by Type I indefinite pronouns and quantity expressions whose interpretation normally result in the assignment of singular or plural to the referent. Hence this hierarchy can be read as follows: the higher an adnominal is in (23), the more likely it is either marked for number or defined for number in any other way (e.g., in accordance with its semantics).

It is worth noting that the plural marking on adnominals in Tanti Dargwa is not likely to be considered a kind of agreement. Unlike the canonical agreement as defined in Corbett (2006), the plural marking in Tanti Dargwa NPs is not redundant: it does not require the presence of the expected controller of agreement, i.e., the head noun, the latter need not be marked for plural (as is seen in mass NP examples like (17) and in constructions with numerals like (20) and (21)), so the value of the number category is based on the properties of the referent. Given this, I consider the Tanti Dargwa number to be a category of the whole NP which can be expressed on various elements of the phrase.

5 Right-periphery construction

While the order of adnominals preceding the head is free, there is yet another position in Tanti Dargwa NPs: some adnominals appear after the head noun in the right-periphery construction. Since the construction is not widespread, its functions remain hypothetical. Presumably, it is occupied either by heavy adnominals (especially heavy relative clauses) or by expressions refining the information provided before in the NP.^[7] As shown by the contrast between (24a) and (24b), the postmodifier is likely to form a constituent with the rest of the nominal expression, since it cannot be separated from it. This construction should be differentiated from patterns like (24c), where there are two coreferential NPs, which can be separated and must bear two case markers.

The posthead position can be filled by Type II expressions and never by Type I expressions, as illustrated for the proximate demonstrative in (25) and for the quantifier ‘all’ in (26). Contrastive Type II expressions, however, are often (cf. (27a) with the plural suffix implying contrast, which was considered infelicitous), yet not always (27b), prohibited to appear in this position.

Among Type III adnominals, genitives can appear to the right of the head noun, even without attributive marking:

Hence in the right-periphery construction we find a picture that is very similar to that of plural marking. The possibility of this pattern corresponds to the position of an adnominal in the hierarchy (23): the lower an adnominal is, the easier it appears in the post-head position.

6 Implications for the description/determination dichotomy

The results we have seen so far are summarized in Table 1. Formal tests suggest that adnominals do not constitute a homogeneous category. Rather there are two poles among adnominals, viz. Type I adnominals and non-contrastive Type II adnominals, and there are adnominals with intermediate characteristics, sharing some of the properties with one of these poles and some with the other (note that the mutual ordering of the intermediate columns in the table, i.e., contrastive Type II adnominals and Type III adnominals, should not be interpreted as if one of them is closer to the left pole and the other one to the right pole). We will now look at how this relates to the description/determination dichotomy.

The contrast between Type I expressions and non-contrastive Type II expressions in Tanti Dargwa clearly reflects this dichotomy. Type I includes those elements that directly affect the reference and relate the denotatum of the NP to the context: demonstratives, indefinite pronouns, and quantifiers. Type II comprises the expressions of properties that are prototypically used to restrict the denotatum of the NP but not to “project” it into a broader context.

The intermediate behavior of certain adnominals presented in Table 1 fits well into this picture. Rijkhoff (2008: 798) recognizes the expressions of ‘other’ as a member of the class of “discourse-referential satellites”, which “provide the addressee with information about the referent as a discourse entity”, i.e., fulfill a determinative function. Yet as shown by Cinque (2015), such expressions can also just specify the type of the referent, in which case they have a more adjective-like function. This duality of the semantics of ‘other’ may be manifested in its expression; see also Lander and Maisak (forthcoming) for East Caucasian languages in general. The intermediate status of cara ‘other’ in Tanti Dargwa probably reflects this duality of its semantics.

Adnominal possessors prototypically establish the reference of the NP (cf. Haspelmath 1999; Langacker 1995), i.e., serve a determinative function. Still, possessor expressions can be used just for providing characteristics of the referent (Koptjevskaja-Tamm 2004). The unusual effect of the neutral attributive suffix, whose appearance with genitives results in the semantics of a terminated relation, correlates with the fact that a terminated relation is less likely to unambiguously establish the reference of the possessum, so the possessor marked with this suffix does not function as a determiner anymore. It is not surprising, however, that the appearance of this suffix does not have any semantic effects with non-referential genitives, which already describe properties and lack the determinative function.

The intermediate status of contrastive modifiers is not surprising either. While ascribing properties, these modifiers usually rigidly determine the set of referents of the NP contrasting them with other individuals, even though they still can be used in contexts that do not require specific reference (as in existential constructions; cf. (6) and (17) above). This predilection to determination is reflected in that contrastive modifiers behave similarly to determiners in some respects. Not surprisingly, Rijkhoff (2002: 173) claimed that contrastive adjectives (such as red in No, I want the RED apple) may function at the same “localizing” level of NP as articles, demonstratives, etc. Still, Bakker (2009: 104) in her detailed description of NPs in Ancient Greek noticed that it is more accurate to think of such modifiers as combining descriptive and discourse (i.e., determiner-like) functions.^[8]

We thus find that postulating the description/determination distinction makes sense for the description of Tanti Dargwa NPs. At the same time, the data provided above have interesting implications for our understanding of this contrast.

The first implication is that the contrast between determiners and descriptive modifiers is not categorical, in accordance with earlier suggestions made by Plank and Van de Velde as mentioned in the introduction. Indeed, I have observed that ‘other’, possessors, and contrastive property expressions can show properties of both determiners and modifiers. The only way to account for this discretely and to retain the strong description/determination dichotomy is to propose that some elements function either as determiners or as modifiers depending on the context. This does not work well, however, since the three tests which are reflected in Table 1 do not always correlate. For example, even when possessors appear without attributive suffixes (and hence are expected to behave as determiners), they are nonetheless able to appear in the post-head position (and hence behave as Type II modifiers). Of course, one can hypothesize that the mixed properties of the adnominals under discussion reflect just an intermediate stage of their development from modifiers to determiners (cf. Van de Velde [2009] for such an approach to some elements of English NPs), but crucially, these mixed properties are inherently related to their semantic potential.

The second implication is that the functional differences between description and determination may almost exclusively (leaving the marginal right-periphery construction aside) affect rules that belong to the domain of morphology like plural marking and the presence of attributive suffixes. Hence, relatively free word order within the NP (probably associated with a flat syntactic structure) does not imply the absence of such difference.

7 Conclusions and open issues

In this article, I discussed formal differences observed among adnominals in Tanti Dargwa NPs and related them to the differences associated with the descriptive component and the determinative component in the semantics. There is indeed evidence that in this language semantically determiner-like elements behave differently from adnominals that mostly fulfill the descriptive function. Still, this difference (probably better reflected as a hierarchy (23)) does not tie any class to any syntactic position, even though there are some positions that are restricted to one of these types.

Indeed, the Tanti Dargwa NP cannot be described as a completely internally flexible NP (i.e., a well-established phrase without a rigid internal structure) in terms of Louagie and Reinöhl (this issue), since it has a dedicated head position, admits phrasal quasi-incorporation into the head (hence restricting the position of modifiers in some constructions) and may contain an optional “post-head” position. Even then, the core of the Tanti Dargwa NP is internally flexible as concerns its word order. While modifier-like adnominals have a wider distribution than determiner-like adnominals (because they have access to additional marginal positions, either being quasi-incorporated or appearing at the right periphery), at the core of the NP they are not contrasted with determiner-like elements syntactically (though they are contrasted morphologically). Hence, I conclude that the distinction between description and determination, which was also shown above to be gradual rather than discrete, need not be associated with word order. Below I will briefly survey some open issues related to this study.

First, the descriptive observations on the behavior of determiner- and modifier-like elements presented above do not add anything to our understanding of the mechanisms that govern the behavior of adnominals in Tanti Dargwa NPs. For example, we see that Tanti Dargwa has a neutral attributive suffix that only appears on modifier-like adnominals. Judging from the effects of “terminated relation” which this suffix shows on definite/specific possessors, it may demote an adnominal to a non-determinative function. It is not clear, however, that this suffix should always be analyzed to derive modifiers from something else.

Further, how can we interpret the appearance of plural marking on various constituents if it is not necessarily agreement but a category of the whole NP? In fact, for many languages being a locus of marking of a phrasal category can be considered a head property – and this is probably in line with the general inclination of determiner-like elements to have head properties.^[9]

The fact that the right-periphery construction does not extend to determiner-like elements could receive a functional explanation: determiners and determiner-like elements like demonstratives and numerals are more likely to appear earlier crosslinguistically independently of the branching direction (Hawkins 1983), and this may be related to the aspiration to establish the reference or the category of the phrase as early as possible (cf. Hawkins 2014 on similar principles; Haspelmath 1999 on preposed possessors). Uta Reinöhl (p.c.) suggested that if we assume that the right-periphery position is filled by a kind of disambiguating afterthought containing grammatically optional information, we should find there mostly modifier-like elements rather than determiner-like elements. In either case, it is expected that the posthead position in Tanti Dargwa is reserved for functions that are not related to determination.

Curiously, for some other Daghestanian languages Testelec (1998a: 274) claims that the postposition of a modifier may denote “that the postposed modifier is focused, contrasted, or restrictive” (see also some discussion in Forker [2021]; Testelec [1998b]), but in Tanti Dargwa, on the contrary, at least the contrastive modifiers are not likely to be placed to the right of the head noun. I tentatively attribute this variation to two different ways of resolving the conflict between the inclination of reference-establishing means to be placed earlier and a motivation to use of a non-standard position for focusing an adnominal. Yet, of course, developing an accurate description of the processes involved here – given the fact that in Tanti Dargwa the determiner-like adnominals need not appear at the very beginning of the NP – needs more research.

In general, the interpretation of the phenomena discussed in this article requires a much more fine-grained typological perspective than we have. Even if we give up the idea that the determinative and descriptive semantics should be strongly associated with syntactic structures proper, we are still left with the question of what kind of phenomena can reflect this (possibly gradual) distinction. Hence I consider the data presented here a stimulus for further investigations.

Abbreviations

ad

localization ‘at’

advloc

adverbial locative

atr

attributive marker

card

cardinal numeral

caus

causative

com

comitative

contr

contrastive

conv

converb

cop

copula

dat

dative

erg

ergative

el

elative

ess

essive

exst

existential

gen

genitive

hpl

human plural

in

localization ‘in’

indef

indefinite pronoun

inf

infinitive

ipf

imperfective

iq

indirect question

lat

latve

lv

light verb

m

masculine (singular)

n

non-human (singular)

neg

negation

npl

non-human plural

obl

oblique stem

pfv

perfective

pl

plural

pret

preterit

prev

preverb

prs

present

sg

singular

super

localization ‘on’

th

thematic