Docs / Studies / Studies.Dixon1982

Where Have All the Adjectives Gone? and Other Essays in Semantics and Syntax

Dixon, R. M. W. (1982). Mouton

Dixon argues that the adjectives of a language draw on seven semantic types, and that where a language's adjective class is small the types fall to its other parts of speech in a regular way. In a language with a closed adjective class, age, dimension, value and colour stay adjectives, human propensity goes to the nouns, and speed follows physical property. Table 2 records the parts of speech of the seven types in seventeen languages, Dixon1982.StructureDataset.table2, and each of Dixon's generalisations holds of it with exactly the exceptions stated here: those he lists, and three he does not.

Main definitions

Dixon1982.StructureDataset.PartOfSpeech.parts: the basic parts of speech a code of Table 2 involves.
Dixon1982.StructureDataset.Row.Takes, Dixon1982.StructureDataset.Row.Predominantly: some, or more than a few, of a type's items are of a part of speech.
Dixon1982.openClass, Dixon1982.noClass, Dixon1982.closedClass: the languages by their adjective class.

Implementation notes

A code's parts are the basic parts of speech it involves. A pair code such as A/V involves both members. A Sango adjunctive is an adjective when it modifies nouns, an adverb when it modifies verbs and a noun when it is a substantive, which is how Dixon reads the speed adjunctives for (d). A Hua verbalised noun is a verb on a noun root, and a Telugu N-A item a noun that also modifies.
A code Table 2 qualifies with some, one, two or few is minor, and a type is predominantly of a part of speech when a code that is not minor involves it.
The languages are classed by their adjective class as Dixon classes them in §1.4.3. He leaves Chinook out of that classification, and so do the theorems that depend on it.

TODO

(c) also says that the smaller the closed class the fewer its physical-property adjectives; Dixon gives the sizes only roughly, and this is not stated.
The commentary gives Kiriwinian some physical-property verbs ('be strong', 'be weak', p. 43) that Table 2 does not list; as a minority they would leave (d) unchanged.
The survey of twenty small adjective classes (§1.1) and the English criteria of Table 1 (§1.3) are not formalized.

References

Dixon 1982, Where Have All the Adjectives Gone? and Other Essays in Semantics and Syntax

Declarations

A language with an open adjective class, or none, has one part of speech predominant in all seven types, except Tzotzil (pp. 43–44).

theorem exists_predominantly_iff L119–125

(r : StructureDataset.Row) :

r ∈ StructureDataset.table2 → r.language ∈ openClass ++ noClass → ((∃ p ∈ [UD.UPOS.ADJ, UD.UPOS.VERB], ∀ (t : PropertyConcept.Class), r.Predominantly t p) ↔ r.language ≠ StructureDataset.tzotzil)

Built from , ,

By (a), in every closed adjective class age, dimension, value and colour have adjectives, except Bemba and Kiriwinian colour, Dixon's (3), and Hua value.

theorem takes_adjective_iff L127–133

(r : StructureDataset.Row) :

r ∈ StructureDataset.table2 → r.language ∈ closedClass → ∀ t ∈ [PropertyConcept.Class.age, PropertyConcept.Class.dimension, PropertyConcept.Class.value, PropertyConcept.Class.color], r.Takes t UD.UPOS.ADJ ↔ (r.language, t) ∉ [(StructureDataset.bemba, PropertyConcept.Class.color), (StructureDataset.kiriwinian, PropertyConcept.Class.color), (StructureDataset.hua, PropertyConcept.Class.value)]

Built from ,

(b) Human propensity is predominantly nouns in a language with a closed adjective class, except in Alamblak and Kiriwinian, the exceptions Dixon names.

theorem predominantly_noun_iff L135–140

(r : StructureDataset.Row) :

r ∈ StructureDataset.table2 → r.language ∈ closedClass → (r.Predominantly PropertyConcept.Class.humanPropensity UD.UPOS.NOUN ↔ r.language ∉ [StructureDataset.alamblak, StructureDataset.kiriwinian])

Built from ,

Every closed class with a human-propensity adjective has a physical-property one except in Sango, as (c) predicts (pp. 39, 48).

theorem takes_physicalProperty_of_humanPropensity L142–147

(r : StructureDataset.Row) :

r ∈ StructureDataset.table2 → r.language ∈ closedClass → r.Takes PropertyConcept.Class.humanPropensity UD.UPOS.ADJ → (r.Takes PropertyConcept.Class.physicalProperty UD.UPOS.ADJ ↔ r.language ≠ StructureDataset.sango)

Built from ,

By (d), in every language of Table 2 but Kiriwinian, speed shares physical property's predominant part of speech or is an adverb where that is verbal (p. 48).

theorem speed_iff L149–156

(r : StructureDataset.Row) :

r ∈ StructureDataset.table2 → ((∃ v ∈ r.speed, ∃ p ∈ v.code.parts, r.Predominantly PropertyConcept.Class.physicalProperty p) ∨ r.Takes PropertyConcept.Class.speed UD.UPOS.ADV ∧ r.Predominantly PropertyConcept.Class.physicalProperty UD.UPOS.VERB ↔ r.language ≠ StructureDataset.kiriwinian)

Built from , ,

In Table 2, A/V and N/V antonym pairs occur exactly in Hausa and Alamblak (p. 51).

theorem pair_iff L158–163

(r : StructureDataset.Row) :

r ∈ StructureDataset.table2 → ((∃ (t : PropertyConcept.Class), ∃ v ∈ r.cell t, v.code = StructureDataset.PartOfSpeech.adjectiveVerb ∨ v.code = StructureDataset.PartOfSpeech.nounVerb) ↔ r.language ∈ [StructureDataset.hausa, StructureDataset.alamblak])

Built from