On the nature of adjectives: evidence from Dinka

This paper presents a description of the adjective class in Dinka, and an exploration of its typological relevance to the existence of adjectives cross-linguistically. In Dinka, a subset of intransitive verbs can be defined by two morphophonological characteristics. The overwhelming majority of these are property concepts, despite the fact that the defining characteristics are unrelated to semantic properties. This calls into question whether or not properties provide a useful semantic context in the investigation of adjectives, even in languages where the adjective class is clearly a subclass of another lexical category. While the analysis is based primarily on the Bor South dialect of Dinka, it has been corroborated using evidence from other dialects. Therefore, it is likely that this characterization holds across the language


Introduction
The status of adjectives as a universal word class is widely debated.For that matter, the existence of universal word classes is equally contentious; opinions differ on whether the ways that different languages express the same concepts can or should be compared.One perspective posits crosslinguistically universal categories-nouns, verbs, and adjectives (cf.Dixon 2004)-arguing that every language must have at least these three categories.Adjectives can be a separate word class, or they can pattern with nouns or verbs.In the latter case, defining distinctions can be quite subtle.This perspective is based on the assumption of a semantically coherent set, which forms the basis for morphological or syntactic properties that are then shared within the set.According to Dixon (1982), this coherence takes the form of property concepts, which are a set of semantic types that are predictably lexicalized as adjectives cross-linguistically. Another perspective (cf. Dryer 1997;Haspelmath 2010) argues that lexical categories cannot be compared crosslinguistically-that comparative concepts based in meaning can be universal, but grammatical behaviour cannot be.Whereas the former perspective argues that subtle differences-as opposed to dramatically different syntactic behaviour, for example-can form the basis for postulating a lexical category or sub-category, the latter perspective argues that subtle differences cannotthat an adjectival subclass of verbs in one language, for example, should not be equated to a separate lexical category of adjectives in another language.In some sense, however, both views of lexical categorization converge on the same questions: are there items that pattern together both syntactically and semantically?Do the same items pattern together across languages?This paper will demonstrate that the adjective class in Dinka (West Nilotic, South Sudan) is a subclass of intransitive verbs, defined by two morphophonological characteristics: a low-toned finite form and an overlong 'attributive construct state' (ACS), which is the form that appears when a postnominal modifier is followed by another modifier.Though they are clearly distinct morphologically, the adjectives are identical syntactically to non-adjectival intransitive verbs, and lack many characteristics posited as universal distinguishers of an adjectival class.For example, both adjectival and non-adjectival intransitive verbs can directly modify nouns, in contrast to what has been suggested as a key difference between verbs and verb-like adjectives (Dixon 2004: 19).Since the adjective class is distinguished without this feature, the Dinka language calls into question assertions that adjectives can be distinguished or defined by their ability to directly modify nouns.The Dinka adjective class also calls into question whether there is a syntax-semantics connection in lexical categorization; the status of a lexical item as a 'property' (in contrast to 'event' or 'action') is not a predictor of adjectivehood, but it should be noted that most adjectives in Dinka are indeed properties-that is, they are stative in meaning, though they should be classified as adjectives, not stative verbs.Dinka adjectives provide evidence that there may be a syntax-semantics connection with regards to adjectives, but also that semantics should not be used as a factor in lexical categorization.
In order to distinguish adjectival intransitive verbs, this paper will also describe a range of intransitive verb morphology in Dinka.Description of the language has focused considerably less on intransitive verb paradigms than on transitive verb paradigms.In addition, references to 'adjectival verbs' have appeared in several descriptions of other grammatical phenomena in the Agar dialect of Dinka (Andersen 2014;2019;2020), but it has not been categorically stated whether the adjective class in Dinka is a subclass of verbs, whether there is more than one lexical category that contains properties, or whether there are multiple adjective classes in the language.Therefore, to show that adjectives are a subset of intransitive verbs, I must first present some intransitive verb morphology in Dinka, so that the adjectives can be both readily identified as intransitive verbs and differentiated as a subclass.Considering that the intransitive verbs have received little attention, this description is also valuable in relation to the study of Dinka morphology as a whole.
While dialect-specific work is, of course, necessary and important in its own right, I suggest that hypotheses about the morphosyntax of Dinka as a whole should not be made unless crossdialect evidence exists.This paper focuses primarily on data from the Bor South dialect of Dinka.
However, I also present evidence that my classification criteria for adjectives in Dinka hold crossdialectally, which means that the typological contribution of this paper towards a definition of adjectives can be regarded as coming from the Dinka language as a whole, as opposed to a specific dialect of the language.Any references to data from the Agar dialect of Dinka are from Torben Andersen's extensive work on Dinka and are cited as such.All other data-from the Bor South, Hol, and Ngok dialects-are my own; data were collected with two speakers of Bor South, one speaker of Hol, and two speakers of Ngok.
The remainder of Section 1 provides background information on Dinka.Section 2 provides a description of aspects of the inflectional morphology and syntactic properties of adjectival and non-adjectival intransitive verbs.Section 3 provides the classification criteria for adjectival intransitive verbs, examines whether vowel grade (cf.Andersen 1993) can be used to classify the adjectival intransitive verbs, and rejects the possibility of grammatical adjectives in other Dinka lexical categories.Section 4 provides a cross-dialect comparison; Section 5 consists of a discussion, and Section 6 contains concluding remarks.

Language background
Dinka is spoken in South Sudan, though there are also sizeable Dinka communities in neighbouring countries such as Kenya and Uganda, as well as in the United States and Australia.
Dinka is generally separated into four dialect clusters: Rek, Agar, Padang, and Bor (Roettger & Roettger 1989).The data for this paper are taken primarily from the Bor South dialect of Dinka, which is spoken on the east side of the White Nile, near the Bor municipality.It is a dialect of the Bor cluster of Dinka, which includes (from north to south) Hol, Nyarweng, Twic East, and Bor South.I will use Dinka to refer to Bor South Dinka throughout most of this paper, but readers should be aware of the fact that dialects of Dinka can and do differ-sometimes considerably-in grammatical phenomena, which will become evident from cross-dialect comparison in Section 4.
Dinka phonology can be summarized as follows.The language has 20 consonants; plosives (both voiced and voiceless) and nasals are found at five places of articulation: bilabial /p, b, m/, dental /t ̪ , d̪ , n̪ /, alveolar /t, d, n/, palatal /c, ɟ, ɲ/, and velar /k, g, ŋ/.Also present are liquids /l, r/ and semivowels /ɰ, j, w/; the language has no fricatives.Dinka generally exhibits consonant place harmony (Mackenzie 2016), though this may not be entirely true for the Bor dialects.Dinka has seven phonemic vowels: /i, e, ɛ, a, ɔ, o, u/.There is, additionally, a voice quality contrast between breathy and modal voice.All vowels exhibit this except for /u/, which is invariably breathy.Dinka also has a three-level vowel length contrast (Andersen 1987;Remijsen & Gilley 2008;Remijsen 2014).All vowels exhibit this three-level contrast, except that /a/ and /ɛ/ are not contrastive in short vowels.In addition, Dinka is tonal; all dialects for which descriptions exist have either three or four tonemes.The Bor South dialect has four tones: Low (L, v); High (H, v); Fall (HL, v); and Rise (LH, v) (Remijsen 2013).Finally, Dinka is largely monosyllabic, with the following syllable structure: C(w/j)V(V)(V)C, where parentheses represent optional syllabic elements.All following transcriptions are phonemic, rather than phonetic.Vowel length is represented by one (short) two (long) or three (overlong) vowel symbols; voice quality and tone are indicated on the first vowel symbol, as is the convention in recent literature on Dinka.
Dinka sentence structure is relatively constrained; an abridged version of the declarative root clause template is seen below.An optional conjunction is utterance initial.The next constituent is the topic, which is the noun phrase preceding the declarative prefix.The topic precedes the declarative prefix /ǎ/, which is followed by a finite verb; this finite verb may either be the main verb or an auxiliary.The finite verb is followed by either the subject or the object-whichever is not the topic-which is followed by a non-finite verb if the sentence expresses negation or is not present-tense.
(Conjunction) Topic Declarative Prefix-V fin Subj.Obj.V nonfin (Andersen 2007: 91) It should be noted that topic and subject are not the same in Dinka (Andersen 1991).Subject refers to the semantic subject, whereas topic refers to the noun phrase preceding the verb.In a transitive clause, this may be the semantic subject, semantic object, or a third argument such as the instrument.In intransitive verbs, it is generally the subject.

Inflectional morphology and syntactic properties of intransitive verbs
This section outlines the syntactic properties and the inflectional morphology of adjectival and nonadjectival intransitive verbs, emphasizing where they pattern together and where they diverge.
The section is organized in terms of the base forms (Section 2.1), forms that inflect for person and number (Section 2.2), nominalized forms (Section 2.3), forms that modify nouns (Section 2.4), and plural inflections (Section 2.5).I also examine forms used in negation (Section 2.6), comparison, superlatives, and intensification (Section 2.7).

Base forms
Intransitive verbs present a finite form and a non-finite form, as seen in ( 1).These forms are found in predicational phrases, for which adjectival and non-adjectival intransitive verbs display identical morphosyntax.The finite form is found in the present imperfective and past imperfective tenses,1 and is morphologically unmarked; for this reason, it is not marked in the glosses.Andersen (2020: 293) refers to it as the 'inflectionally unmarked' form in the Agar dialect.The non-finite form is used in phrases marked for present perfect, past perfect, habitual, and future.This pattern mirrors that of the transitive verb system.In ( 1), (1a) and (1c) display the finite form of an adjectival and non-adjectival intransitive verb, respectively; (1b) and (1d) display the corresponding non-finite forms.One can see that for the adjectival intransitive verb pjàt ̪,2 'good,' the difference between the finite and non-finite forms is a matter of vowel length; the non-finite form, pjàat ̪, is longer than the finite form.For the non-adjectival intransitive verb la̤ aŋ, 'pray,' the difference between finite and non-finite form is instead a difference in vowel quality.
( (2) shows the tenses not displayed in (1): past imperfective (2a-b), present perfect (2c-d), past perfect (2e-f), and future (2g-h).Examples use the adjectival intransitive verb 'good' and the non-adjectival intransitive verb 'pray.'Note that (2a-b) combine an auxiliary with the finite form of the verb, despite the fact that auxiliaries usually appear with non-finite inflections.
( For some intransitive verbs, the finite and non-finite forms are identical, meaning that the past imperfect and present perfect will be identical as well.It is not clear whether the resulting homophony poses any issues.In any case, this ambiguity is likely unique to the Bor South dialect, as other dialects use ǎ-ci ̤ instead of ěe for the present perfect (2c-d).
The finite and non-finite forms of several adjectival intransitive verbs and several nonadjectival intransitive verbs are displayed in Table 1.One sees that the verbs display a variety of alternations between finite and non-finite forms: possibilities include alternations in vowel quality (10), vowel length (5), vowel quality and vowel length (2), tone (3), tone and vowel quality (1,4) or tone, vowel length, and voice quality (9).Alternatively, the non-finite form may be identical to the finite form, as in ( 6), (7), and (8).
Additionally, just as with transitive verbs, the declarative prefix agrees with the preverbal subject for number in adjectival and non-adjectival intransitive verbs, lengthening for a plural topic, as seen in ( 3).
( Nominal predication is distinguished from intransitive verb predication in forms marked for tense-aspect as well.In nominal predicational phrases marked for tense-aspect, the inflected form is the copular verb, as opposed to the predicate, as seen in ( 5).
( Nominal predication requires a verb in addition to the predicate, whereas the predicates under investigation in this paper (i.e.adjectives) do not; they are clearly verbal.While this does not differentiate between adjectival and non-adjectival intransitive verbs, it supports the argument that the words discussed in this paper are verbs.

Inflections for person and number
Both adjectival and non-adjectival intransitive verbs inflect for person and number.There are two sets of inflectional paradigms, depending on the context; I call them Set A and Set B. The common characteristic between the sets is that they do not appear in main clauses.The verb must precede the subject (if the subject is overt, as in Set A inflections), and there is no syntactic topic.
Set A inflections are found following te̤ , which is a grammaticalized form of the word 'place,' but which functions as a relativizer.It can be interpreted as 'when,' 'where,' or 'how,' depending on the context.Set A inflections have only two phonological forms.As seen below for the adjectival intransitive verb pjàt ̪, 'good,' and the non-adjectival intransitive verb la̤ aŋ, 'pray,' first person singular is differentiated, while all other person forms are identical.Set A inflections are obligatorily followed by a pronoun, which allows differentiation of the syncretic forms.( 6) contains examples for an adjectival intransitive verb, and ( 7 While the vowel quality alternations in Set A and Set B are similar, they do not match with regards to person-for example, the second person inflection in Set A has vowel quality /ɛ/ while in Set B it has /a/.These differences in alternations should be investigated in the future, particularly in the context of the grade system laid out by Andersen (1993).What is important for this paper, however, is not the inflections themselves, but the fact that the alternations for adjectival intransitive verbs and non-adjectival intransitive verbs are similar.

Nominalization
Both adjectival and non-adjectival intransitive verbs can be nominalized, as seen in ( 10).The nominalized form of the verb is often identical to the non-finite form, as seen in Table 2. Since the forms are often homophonous, their status as actual nominalizations, as opposed to a single multifunctional inflection, might be questioned.The evidence for their status as nominalization comes from the construct state of the nominalizations.The construct state is the form a noun takes when it is followed by a modifier, though not every Dinka noun has a construct state.The nominalizations of several intransitive verbs and the construct states of those nominalizations are shown in Table 2.
The construct state of a noun is expressed by one or more of the following exponents in Bor South Dinka: nasalization of a coda consonant, a tone change, often to high tone, a vowel quality change, or a vowel length change.In the nominalized forms of intransitive verbs, the construct state is

No.
Gloss the relationship between the nominalized form of an intransitive verb and its form when followed by a possessor can only be expressed as 'construct state.'Accordingly, these nominalizations and construct states appear in environments that are normally occupied by a noun, as in (10).

Postnominal modification
In Dinka, modifiers follow the noun.The postnominal position is not diagnostic of any word class,  Andersen (2020) refers to as the construct state; to differentiate it from construct states described in the previous section, I refer to it as the attributive construct state, glossed acs.The attributive construct state is the form taken when a postnominal adjectival intransitive verb is followed by another modifying element.Note that the adjectival intransitive verb does not need to be directly postnominal; for example, in a chain of three postnominal adjectival intransitive verbs, the first two will take the attributive construct state, as seen in ( 14). ( 14) mwɔɲ bɛ̤ ɛɛr pjɛɛɛt ̪ di ̤it man.cstall\acs good\acs big 'The big, tall, good man' The final modifying element can be an intransitive verb, adjectival or non-adjectival, or a demonstrative, -e̤ /e̤ (proximal) or -e̤ ̌ (distal).The attributive construct state is invariably overlong (VVV).Non-adjectival intransitive verbs do not have an attributive construct state; they take the finite form in this position.Table 3 compares the forms of adjectival and non-adjectival intransitive verbs when they follow the noun (i.e. are noun phrase-final) and when they are followed by another modifying element (i.e. are noun phrase-medial).
One can see from the table that adjectival intransitive verbs lengthen when they are noun phrase-medial and non-adjectival intransitive verbs do not.Bor South Dinka does not specify the order in the post-nominal context between the individual subclasses of adjectival and nonadjectial verbs.However, there is a preference for adjectival verbs to precede non-adjectial verbs in a phrase with multiple modifiers.Therefore, (13c), in which the non-adjectival verb precedes the adjectival verb, is not ungrammatical, but (13d) is preferred.A similar phenomenon is reported by Andersen (2020: 300) for the Agar dialect.

Plural inflections
Five adjectival intransitive verbs in the elicited dataset have a plural form in Bor South Dinka.
The plural form appears both when the verbs are used as predicates and when they are used as modifiers following the noun.This is illustrated in ( 15 The singular and plural inflections of the five verbs are displayed in Table 4. The plural inflections of these verbs involve a shortening of the stem vowel.Therefore, this pattern of exponence cannot distinguish a plural form in stems whose finite form is short.It is also important to note that there is no plural attributive construct state, so while the plural inflections described here can be used as postnominal modifiers, they must be noun phrase-final in order to appear.Adjectival intransitive verbs that follow plural nouns and are followed by another modifier take the same attributive construct state as those that follow singular nouns, as seen in ( 16).The length alternation between singular and plural finite forms is not a productive pattern in the language.In fact, it is likely an older singular/plural contrast in West Nilotic, as Andersen (2014) hypothesizes based on a contrast also found in Anywa (Reh 1996: 258-260).In Anywa, a member of the Luo sub-group of West Nilotic languages, 'monovalent static' verbs have a plural derivation that involves consonant strengthening, tone shift, and vowel shortening (Reh 1996: 259)

Negation
Negation in Dinka involves the use of the auxiliary ci ̤,3 which appears directly after the declarative prefix, in the slot that is occupied either by a verbal auxiliary or a finite verb.The negated verb is moved to the position occupied by a non-finite verb; in clauses with intransitive verbs, this Intransitive verbs exhibit a tonal alternation in negated clauses; the tone of the verb changes to high.This tone change is also found in the transitive verb stems of the Agar (Andersen 1992(Andersen -1994;;2016), Luanyjang (Remijsen & Gilley 2008), and Bor dialects (Blum 2020).
However, the tone change is only possible in verbs with short stem vowels (V).This condition is clearest in the five stems that agree for number, as seen in ( 18).In a negative sentence with a singular topic, the stem vowel of the verb is not short, and therefore is identical to the finite form.In a negative sentence with a plural topic, however, the vowel of the verb shortens (as discussed in Section 2.5), and the tone of the verb becomes high.It should be explicitly noted that this alternation is due to vowel length, not to agreement for number, as it is seen in other short-stem intransitive verbs (i.e.ones that do not agree for number).Examples of other verbs that exhibit this alternation are seen in ( 19).(19a,c) show the uninflected forms of the verbs, and (19b,d)

Comparative, superlative, and intensification
Dinka marks neither comparative nor superlative constructions inflectionally.There is no superlative, and the comparative is expressed by the addition of an external argument using a preposition, as in (20).
(20) tìik ǎ-bà̤ aar é̤ mòc woman decl.sg-tallprep man 'The woman is taller than the man.' Intensification is expressed by the use of the intensifier are̤ êet.This is true for both adjectival and non-adjectival intransitive verbs, as seen in ( 21).

What defines adjectival intransitive verbs?
The descriptive analysis in Section 2 shows that adjectival intransitive verbs pattern with other intransitive verbs, as they behave identically to non-adjectival intransitive verbs in almost all of the contexts discussed.However, while I have consistently separated adjectival and non-adjectival intransitive verbs terminologically, I have not shown why they should be separated.This section shows primarily the defining characteristics of adjectival intransitive verbs, and also secondarily why vowel grade (cf.Andersen 1993;2017;2020) cannot be a defining feature of adjectives in Dinka.My classification is based on two conditions that lie entirely in the suprasegmental system of the language; it is not based at all on semantics.Andersen (2020) notes that the finite form of all adjectival intransitive verbs has a low tone in the Agar dialect, and I find this to be true for the Bor South dialect as well.This low tone is the first of the two conditions, which are as follows: The presence of these two conditions defines the adjectival subclass of intransitive verbs.An intransitive verb in Dinka can only be classified as adjectival if these two conditions are satisfied; both of them are necessary, and neither of them are sufficient alone.The differences between the two classes can be seen in Table 5.In the table, Nos.1-7 are adjectival intransitive verbs, which satisfy Condition 1 and Condition 2. Nos.8-14 are non-adjectival intransitive verbs, which may satisfy one of the conditions but do not satisfy both.
Non-adjectival intransitive verbs can fulfill Condition 1; their finite forms can be low, as seen in Nos. 8 and 9 in Table 5.With regards to Condition 2, however, non-adjectival intransitive verbs do not have attributive construct states (ACS); their form in this position (postnominal but followed by another modifier) is identical to the finite form of the verb.Therefore, the length of a non-adjectival intransitive verb in this position can only be overlong if the finite form is over long as well; in this case, the unmarked form of a non-adjectival intransitive verb may superficially appear to be an ACS.In contrast, the ACS of an adjectival intransitive verb is only identical to the finite form in rare cases, and these cases can only occur when the finite form of the adjectival intransitive verb is overlong.
Though most adjectival intransitive verbs have meanings that are canonically adjectivalthat is, distinguished as adjectives cross-linguistically (cf.Dixon 2004)-there are no semantic conditions for the classification of adjectival intransitive verbs.I find that meanings associated with adjectival classes in languages that have a small class of adjectives-dimension (long, big, small); colour (white, red, black); and value (good, bad)-are adjectival in Dinka.Other semantic types associated with adjective classes in languages that have a larger adjective class-physical properties, for example-can also be adjectival in Dinka.Human propensity, also cited by Dixon (2004) as a semantic type associated with medium/large adjective classes, generally involves the use of an adjectival verb, such as 'good,' 'bad,' or 'sweet,' along with a noun.For example, 'kind' translates literally to 'sweet of the heart.'However, not all properties are adjectival intransitive verbs, and not all adjectival intransitive verbs are properties.For example, as seen in properties denoting emotions, and both have low-toned finite forms.They are differentiated in subclass only because ɰo̤ òk lengthens to ɰo̤ òok, showing that it is an adjectival intransitive verb, whereas njàam does not.In contrast, 'break'4 should be classified as an adjectival intransitive verb, as it fulfills both conditions, but it is not a property.The overlap between properties and grammatical adjectives will be discussed in Section 6.
Given the available evidence, it is not clear that items within the semantic types of 'properties' will be classifiable as adjectival, nor is it clear that an item whose meaning is not a property will be non-adjectival.Additionally, properties whose semantic types Dixon (2004) associates with a large adjective class are sometimes adjectival in Dinka, but there does not appear to be regularity with regards to which ones are adjectival and which are not.This seeming semantic instability with regards to the adjectival subclass will be discussed in Section 6.
With regards to vowel 'grade, ' Andersen (2020: 303) claims that a characteristic of adjectival intransitive verbs in Dinka is a Grade 1 vowel in the finite form and a Grade 2 vowel in the attributive construct state form.The vowel 'grade' alternations-systematic combinations of vowel quality and vowel length-are a key aspect of the morphology of transitive verb inflections and derivations (Andersen 1992(Andersen -1994;;1993;2017) as well as the nominal system (Storch 2005;Andersen 2014;Ladd & Blum, to appear).Therefore, it makes sense to examine the vowel grade alternations for intransitive verbs.In contrast to Andersen (2020), however, I argue that vowel grade is not a defining characteristic of adjectival intransitive verbs.The verbs discussed in Andersen (2020) clearly are the adjectival intransitive verbs; they have a low tone in their finite (his 'unmarked') form and are overlong when preceded by a noun and followed by another modifier.But while Andersen (2020)'s claims about grade may be true for the Agar dialect, they are not true for the Bor South dialect, and therefore not true for Dinka as a whole.Because the intransitive verbs do not exhibit the same regular length alternations as the transitive verbs, it is not possible to consistently analyze their vowel grades.alternating between long (VV) and overlong (VVV), in the base paradigm.5The intransitive verb system does not exhibit this regular lexical quantity alternation.It is clear, however, that the same vowel alternations that appear in the transitive verb system (Andersen 1993;2017) and nominal system (Andersen 2002;2014) appear in the intransitive verb system.6

No. Gloss Finite form ACS
In Bor South Dinka, the grade of the ACS is not consistently a Grade 2 vowel, as seen in Table 8.Because the ACS is always overlong, it is by default either Grade 2 or 3, which invariably differ in vowel quality.In  Andersen (1993), but applies to the Bor South dialect; the switch in dialect involves the addition of a long-grade /a̤ a/ series.It should also be noted that the Grade 3 form of the /ṳ/ series is /wo̤ o(o)/ for Agar, and /wɔ̤ ɔ(ɔ)/ for Luanyjang; it is not clear which set appears in Bor South, and thus both Grade 3 forms of the /ṳ/-series are in parentheses. 6 an attributive construct state form with /ɛ/ can only be Grade 3; similarly, a root vowel /a/ can only have /ɛ/ as Grade 2.
While the grade of the ACS can be determined in a straightforward manner, classifying the grade of a given finite form is difficult.It is important to note that analysis of these vowel grades, particularly long (VV) vowels, is not possible without more than one inflection of a paradigm.
The vowel grades exist in the context of a verbal paradigm; /aa/ could be Grade 1 long just as easily as it could be Grade 3 short.Indeed, some adjectival intransitive verbs have a short finite form and an overlong attributive construct state form, which violates regular lexical quantity.
Several examples are presented in Table 9.
It is worth noting that the distinction between long (VV) and overlong (VVV) is potentially lost in Dinka preceding a coda /r/ (Remijsen & Gilley 2008).Therefore, one might consider the attributive construct state forms of 'white' and 'strong' to be both lexically short and phonetically overlong, and thus not a violation of regular lexical quantity.Even so, 'black,' 'light,' and 'weak' still pose a problem with regards to regular lexical quantity.The finite and non-finite forms of these three verbs fit a lexically short (V/VV) paradigm, and the attributive construct state does not.All three forms-finite, nonfinite, and attributive-should be taken into account, just as multiple forms are taken into account in the transitive verb system.
Otherwise, the alternation is analyzed solely on vowel quality, not on the combination of vowel quality and vowel length.In addition, even if lexical quantity between the base forms is treated as regular, and the overlong quality of the attributive construct state is accounted for via a morphophonological lengthening process (cf.Andersen 1993), it is still not possible to have both consistent lexical quantity and an invariably Grade 1 vowel in the finite form.This is because in many cases the finite and non-finite forms of adjectival verbs have the same vowel length and differing vowel quality.
To illustrate the importance of this matter to the classification of Dinka adjectives, I take the example of 'hot.'The finite form is le̤ et ̪, the non-finite form is lɛ̤ ɛt ̪, and the attributive construct state is lɛ̤ ɛɛt ̪.Relative to the attributive construct state, which can only be analyzed as Grade 3 long, the finite form can be analyzed as Grade 1 long.However, the non-finite form can only be analyzed as Grade 3 short, and relative to the non-finite form, the finite form would be Grade 2 short.This is a consequence of the grade system.While it is true that there is a consistent overlap in the grade system-for example, Grade 2 short of the /e/ root vowel will always be homophonous with Grade 1 long of the same root vowel-the system is described with these homophonous grades because it is so regular in the transitive verbal system, where a given form cannot be analyzed as multiple grades.In adjectival intransitive verbs, however, this is clearly not the case.A finite form might be Grade 1 relative to the attributive construct state form in some cases, and Grade 1 relative to the non-finite form in others.Therefore, while it is the case that the finite form seems to consistently have a vowel quality that could potentially be analyzed as Grade 1, depending on the other forms, a Grade 1 finite form cannot be taken as a defining property of adjectival intransitive verbs in Dinka.
In short, the vowel quality and vowel length alternations present in adjectival intransitive verbs are similar to those found in the transitive verb system.However, the patterns are different enough that postulating a classification on vowel grade is not possible.Having multiple analyses within the same lexical item-for example, the finite form as Grade 1/Grade 3-is not satisfactory given the regularity of the transitive verb system and the fact that this ambiguity is not consistent across lexical items.

Absence of non-verbal adjectives
It should also be noted that there is no evidence of grammatical adjectives in other lexical categories.Items whose meanings are properties also exist in the nominal class-however, they are not grammatical adjectives.For example, 'blue ' and 'true,' displayed in (22), both behave fully like nouns.
(22) a. ké̤ -pjàt ̪ ée ji ̀ic thing.sg;cs-gooddecl.sg.cop true.sg 'The good thing is true.' b. kṳ̌ uur ée mà-làak stone.sgdecl.sg.cop m-blue.sg'The stone is blue.' As can be seen in ( 22), jìic and mà-làak behave like the nominal examples in (4)-that is, when used as predicates they are preceded by the copular verb ée.Unlike intransitive verbs, when inflected for tense-aspect, the predicates of these sentences (jìic and mà-làak) do not change; the copular verb is inflected, as with other nouns in (5).These nouns, like other nouns, are also able to act as the subject of a sentence and to inflect for a construct state, as seen in ( 23).These nouns can also be the object of an existential predicate marker, which the finite forms of adjectival intransitive verbs cannot.In order for an adjectival intransitive verb to be in the existential context, it must undergo nominalization, as seen in (24c).

Cross-dialect comparison
Dinka dialects tend to retain the binary voice quality and ternary vowel length contrasts, but tone systems vary both in number of tonemes and tonal behaviour.Vowel length of specific items may vary across dialects, even though the three-way contrast is preserved.In addition, morphosyntactic phenomena vary cross-dialectally.The classification conditions presented in this paper rely on both tonal specification (Condition 1) and specification for vowel length (Condition 2).Therefore, in order to claim that these conditions hold for the Dinka language as a whole, other dialects must be examined, because it would not be surprising if cross-dialect variation invalidated the conditions in other dialects.However, at least three other dialects of Dinka-Agar (Andersen 2020), Hol, and Ngok-follow the patterns found in Bor South; Condition 1 and Condition 2 are satisfied by a group of intransitive verbs whose meanings are generally properties-specifically, lexical items that one expects to be adjectival in a given language if anything is, such as big/small, black/white, and good/bad.This is true despite cross-dialect variation involving, for example, the vowel length of specific items; these differences do not affect Conditions 1 and 2. Crossdialect evidence of the validity of these classification criteria, therefore, shows that this subclass of intransitive verbs is common to the language, independent of the many differences between dialects, and may have been present in an earlier stage of the language.
The Hol dialect follows the patterns laid out in this paper for Bor South almost exactly.One key difference between the Hol and Bor South dialects appears with the adjectival intransitive verbs whose finite form inflects for plural number.Recall that these have been observed to be common to West Nilotic in general (cf.Reh 1996, Andersen 2014, Remijsen & Ayoker 2020).
Singular and plural forms of adjectival intransitive verbs in the Hol dialect can be seen in Table 10.Just as in the Bor South dialect, the two items with a coda /r/ (Nos. 1 and 2) shift from overlong in the singular to short in the plural.However, it is worth noting that 'tall/long' has an overlong /ɛ/ vowel quality in the singular, whereas Bor South has /a/.'Red' behaves in Hol exactly as it does in Bor South, shortening from long in the singular to short in the plural.'Bad,' however, does not shorten but shifts to breathy voice quality, which is entirely unexpected in this context.'Rough' does not shorten as we expect it to.There is also an additional item: 'black' shortens in Hol.I take the lack of a singular/plural differentiation for 'black' in Bor South to be morphophonological; 'black' will shorten in the plural if it is able to, but it cannot in Bor South because the finite form is already short.These differences call into question the validity of the assertion, both in Section 2.5 of this paper and in Andersen (2014), that the few items with both singular and plural forms are evidence of an older contrast common to West Nilotic.However, a more comprehensive study of this phenomenon is clearly necessary, and it does not interfere with the classification of adjectival intransitive verbs.
The behavior of the plural finite forms differs in the Agar dialect as well.In Agar Dinka, a plural noun can appear with both the finite singular and finite plural (shortened) form of these adjectival intransitive verbs (Andersen 2014).Andersen hypothesizes that 'the plural form indicates that each member of a plural set is an individual' (Andersen 2014: 260), relying on a partial semantic explanation.However, it appears that this is not true in the Bor South dialect; a plural subject is ungrammatical if it appears with the singular finite form.
The Ngok dialect of Dinka also has a sub-group of intransitive verbs that satisfy Conditions 1 and 2; the finite forms are low-toned, and the attributive construct states exist and are invariably overlong.However, the specific items that fit the criteria differ slightly from the Bor cluster dialects, 7 emphasizing the subtlety of the classification system presented in this paper.in the plural in other dialects.In Ngok, these items differ depending on the context.Verb stems can shorten in the predicational context, which invariably involves the /ki ̤/ suffix, but shortening does not seem to be predictable in the same way that it is in other dialects.However, in the postnominal position (plural attributive in Table 12) the shortening appears on the same verbs as other dialects, shortening in the same way.
In summary, cross-dialect comparison of intransitive verbs confirms that the classification criteria of adjectival intransitive verbs can be used across dialects of Dinka, despite the fact that dialects can differ considerably both morphosyntactically and tonally.Number marking can differ cross-dialectally, but this has no bearing on the satisfaction of Conditions 1 and 2.

Discussion
It is clear that there is a sub-group of intransitive verbs that can be defined by the presence of two morphophonological properties: a low-toned finite form and an overlong attributive construct state.But while these two characteristics indicate that adjectives are a distinct subclass of intransitive verbs, they are syntactically identical to other intransitive verbs.In all dialects of Dinka examined in this paper, and to the best of my knowledge, there is no syntactic position that can be occupied by an adjectival intransitive verb that cannot be occupied by a non-adjectival intransitive verb.Therefore, the adjective class in Dinka poses a problem for claims (cf. e.g.

Dixon 2004
) that, cross-linguistically, there is a syntactic difference between adjectives and non-adjectives, particularly in their capacity to directly modify nouns.Moreover, the difference between adjectives and other intransitive verbs in Dinka is very subtle, exemplified by the fact that there is at least one item which cannot be classified as adjectival or non-adjectival because it is ambiguous.The finite and non-finite forms of the item 'fly,' as well as the form when it is postnominal and followed by another modifier, are all pa̤ aar.Since the stem has a coda /r/, it is not possible to tell whether there is an underlying length alternation between the finite and attributive construct state forms.Additionally, pa̤ aar is not used postnominally in the Ngok dialect, which suggests that it is not adjectival, but it is used postnominally in other dialects, so classification remains unclear.Fully answering this question will require a comprehensive overview of the intransitive verbal system in Dinka.
Dinka also presents an issue for the claim that 'property' does not predict adjectivehood.
Clearly, being a property does not mean that an item will be an adjective in Dinka, nor does not being a property mean that an item will be non-adjectival.However, the items in the Dinka adjectival subclass are overwhelmingly properties, despite the fact that the means of classification are entirely morphophonological.Moreover, the Dinka adjectives are of the semantic types that Dixon (2004) predicts will be adjectives if anything in a given language is a grammatical adjective.
In addition, calling these 'stative' verbs (as Reh (1996) does for Anywa) is not appropriate; there are other intransitive verbs that are stative in meaning and are not adjectival.For example, the verb 'sit' has a high-toned finite form, rɛ̤ ɛɛr, and therefore cannot be adjectival.What is an adjective if morphosyntactic or morphophonological criteria do not reveal some sort of semantic pattern?
This question relates to the larger issue of identifying adjectives cross-linguistically.
Determining whether or not a language has a class of adjectives may be in large part a matter of initial perspective.That is, if one begins an investigation assuming that there is a class of adjectives in every language-even subtly defined-one is likely to find that class.On the other hand, if one undertakes the description of a language with the understanding that adjectives must conform to specific, pre-determined criteria, a distinct class of adjectives might not be identifiable.This is the case for Korean, for example: Kim (2002) asserts that Adjective is an abstractly defined entity, and that the stative verbs of Korean do not fulfill the requirements to be a member of this class, because they do not directly modify nouns.Sohn (2004: 241) on the other hand, uses the same verbs to assert that the adjective class in Korean does exist, noting that adjectives can modify a noun via relativization (cf. Dixon 2004: 19).To classify adjectives, Sohn describes various other morphological criteria that allows them to be distinguished from other verbs.Additionally, both authors describe a class that consists of properties in the same way that the Dinka class does, including colours, physical properties, and speed (Sohn 2004: 237).
The question of whether Korean has adjectives, therefore, seems to be potentially unresolvable between these two perspectives, but only because they do not share criteria for adjectivehood.
I argue that the subclass laid out in this paper is the adjective class of the Dinka language-a subclass of intransitive verbs defined by two morphophonological features, syntactically indistinguishable from non-adjectival intransitive verbs.But if one is to approach the search for adjectives with the perspective that, for example, 'adjective' as an abstract concept that requires inflection for comparison, then the Dinka adjectives either cannot be considered adjectives or pose significant problems for the definition of adjectives.In fact, the Dinka adjectives do not exhibit several of the features that have been suggested as defining characteristics of 'adjective': they are not defined by an exceptional role in adnominal modification; they do not have a degree system; and they cannot be reliably predicted by semantic types.In contrast to this final point, however, the Dinka adjectives do provide some evidence that there is a correlation between properties and adjective, despite the fact that semantic type is not a defining feature of the lexical category.

Conclusion
Adjectives in Dinka, as explained in Section 3, are a subclass of intransitive verbs defined by two characteristics: a Low-toned finite form and an overlong (VVV) attributive construct state, which is found in the position after a noun and before another modifier.Importantly, this system is found across dialects and dialect clusters of Dinka.Dialects of Dinka can differ in number of tonemes, tonal behaviour, and morphosyntax, and while the presence of a binary voice quality contrast and a ternary vowel length contrast is consistent cross-dialectally, cognates can have differing vowel length and (less commonly) differing voice quality.Therefore, the fact that cross-dialectally, the same intransitive verbs satisfy Conditions 1 and 2-conditions based on suprasegmental features, not semantic types-supports the analysis of these verbs as a subclass.
The items in the subclass can differ slightly across dialects, which should perhaps be expected given the cross-dialect variability in suprasegmental features.However, it should be noted that semantic types that tend to appear in the smallest adjective classes (cf.Dixon 2004), such as size (big/small), color (red/white/black), and value (good/bad) are consistently adjectival across dialects of Dinka.

Condition 1 :
The tone of the finite form of an adjectival intransitive verb is low.Condition 2: The attributive construct state form of an adjectival intransitive verb has an overlong vowel.

Table 1 :
Examples of adjectival verbs and non-adjectival intransitive verbs.
Set B inflections are found following the clause-initial conjunction kṳ , which can precede both declarative and non-declarative clauses.The inflected verb must directly follow the conjunction.It also seems to be the case that Set B inflections only appear when they are utterance final, though it is unclear whether this is for syntactic or prosodic reasons.Example (8) displays Set B inflections for an adjectival intransitive verb, and (9) displays them for a non-adjectival ) contains examples for a non-adjectival intransitive verb.

Table 2 :
Tone alternations for nominalizations and construct states in intransitive verbs.based on the non-finite form of the verb, and is expressed by a tone change to high (irrespective of the tone on the original non-finite form), and sometimes nasalization of the coda consonant.This is true of both adjectival intransitive verbs and non-adjectival intransitive verbs, though a minority of non-adjectival intransitive verbs may retain the non-high-toned nominalized form, as seen in no. 8 of Table2, 'walk,' whose construct state also has a nasalized coda consonant.The construct states here are notably different from what Andersen (2020) refers to as the 'construct state' of an adjectival intransitive verb; I call that form the 'attributive construct state,' and it is explored in Section 2.4.Since the 'construct state' of an item in Dinka generally refers to a form of a noun, This lengthened form of adjectival intransitive verbs is what nouns and verbs, both transitive and intransitive, appear in this position, directly modifying nouns.Various examples of postnominal modification are seen in example (11).

Table 4 :
Remijsen & Ayoker (2020)ransitive verbs with a plural form.inDinka,vowelsthatarealreadyshortdonot shorten.This shortening phenomenon is also found in Shilluk, another member of the Luo sub-group.In Shilluk, seven adjectives shorten; two of these-'long' (sg.bʌʌr, pl.bʌr-ɔ)and 'bad' (sg.râac, pl.rɪ ̀c-ɔ)-have clear cognates in Dinka (ba̤ aar and ràac)(Remijsen & Ayoker 2020).Also notable is that the length of adjectival intransitive verbs (and all other verbs) can differ between Dinka dialects; for example, Andersen (2014) cites 'short' as an adjectival intransitive verb that can shorten in the plural, as doRemijsen & Ayoker (2020)for Shilluk.However, this verb, cìk, has a short vowel in its finite form in the Bor South dialect of Dinka, and is therefore unable to exhibit this vowel length alternation.Two non-adjectival intransitive verbs also display number agreement when used as both predicates and attributes: lɔ, 'go', and bɔ̤ , 'come.' Rather than vowel length, the exponent is tone, specifically a change from high tone in the singular to low tone in the plural, as seen in (17).It should be noted that these two verbs are exceptional in the sense that they do not have a coda consonant.However, this tonal alternation implies that number marking alone does not distinguish adjectival intransitive verbs from non-adjectival intransitive verbs.
show the negative context, in which the short stem verb has a high

Table 6 ,
'angry with somebody' is an adjectival intransitive verb, but 'proud (in a positive sense)' is a non-adjectival intransitive verb.Both are

Table 5 :
Finite and attributive construct states (ACS) of adjectival and non-adjectival intransitive verbs.

Table 6 :
Examples of a property that is not an adjectival verb and an adjectival verb that is not a property.ACS refers to the attributive construct state.The basic principle underlying the grade system, a version of which is displayed in Table7, is that each verb has an underlying 'root vowel' upon which various inflections alternate.For example, the unmarked form of the transitive verb màan, 'hate,' has a Grade 1 long vowel; the third person singular inflection, mɛɛɛn, has a Grade 2 long vowel.The unmarked form of every transitive verb has a Grade 1 vowel, and the third person singular inflection of every transitive verb has a Grade 2 vowel.Transitive verbs in Dinka also exhibit regular lexical quantity: they are either 'lexically short,' alternating between short (V) and long (VV) vowels, or 'lexically long,'

Table 8 ,
grades of the ACS of four adjectival intransitive verbs are clearly classifiable based on the root vowel.With a verb that has /e/ as its root vowel,

Table 7 :
Vowel quality grade alternations of both lexical lengths.This chart is based on the system in

Table 8 :
Vowel grades of attributive construct states (ACS) of adjectival intransitive verbs.The grade is in parentheses.

Table 9 :
Grade alternation patterns for adjectival intransitive verbs with a short vowel length in the finite form.

Table 11
displays the finite forms and attributive construct states (i.e.Conditions 1 and 2, respectively) of a number of adjectival intransitive verbs in Ngok Dinka.One sees that Table11follows the patterns found in other dialects of Dinka.That is, Condition 1 is satisfied because all items have low-toned finite forms, and Condition 2 is satisfied because all items have overlong attributive construct state forms.In the Ngok dialect, ŋɔɔɲ, 'rough,' does not have an ACS. 7

Table 10 :
Singular and plural adjectival verb forms in the Hol dialect of Dinka, compared with the corresponding forms in the Bor South dialect.One can also see that the length of the verb stem in (26b) has shortened, mirroring the behaviour of 'red' in other dialects, despite the differing inflectional form.A natural following question is how the dialect behaves with regards to the handful of adjectival intransitive verbs that shorten 8 Nos.11 and 12 are subject to some between-speaker variation; 'near' and 'far' were judged ungrammatical in the postnominal position by one speaker.

Table 11 :
Finite forms and attributive construct states of adjectival intransitive verbs in the Ngok dialect of Dinka.

Table 12 :
Singular and plural adjectival verb forms in the Ngok dialect of Dinka.