Chapter Order of Adjective and Noun

by Matthew S. Dryer

1. Defining the values

This map shows the distribution of the two possible orders of modifying adjective and noun.  English is an example of a language which is AdjN, with the adjective preceding the noun (as in large dogs ). Another example is Mising (Tibeto-Burman; northeast India), as in (1).

(1) Mising (Prasad 1991: 69) 

azɔ́në 

dɔ́luŋ 

small 

village 

Adj 

‘a small village’ 

Examples of NAdj languages, with the adjective following the noun, are given in (2); Apatani, another Tibeto-Burman language spoken in northeast India, is illustrated in (2a), while Temiar (Aslian, Mon-Khmer; Malaysia) is illustrated in (2b).

(2) 

a. 

Apatani (Abraham 1985: 23) 

 

aki 

atu 

 

dog 

small 

 

Adj 

 

‘the small dog’ 

b. 

Temiar (Benjamin 1976: 155) 

 

dēk 

mənūʔ 

 

house 

big 

 

‘big house’ 

Values of Map 87A. Order of Adjective and Noun
Go to map
 ValueRepresentation
Modifying adjective precedes noun (AdjN) 373
Modifying adjective follows noun (NAdj) 878
Both orders of noun and modifying adjective occur, with neither dominant 110
Adjectives do not modify nouns, occurring as predicates in internally headed relative clauses 5
Total: 1366

In some languages, both orders of adjective and noun occur. In some of these, an argument can be given that one of the two orders is dominant (see “Determining Dominant Word Order”). For example, in Huasteca Nahuatl (Uto-Aztecan; Mexico) the words for ‘good’ and ‘big’ precede the noun, but other adjectives more frequently follow the noun (Beller and Beller 1977: 233). This is taken here as a basis for saying that NAdj order is dominant in Huasteca Nahuatl and it is thus shown on the map as NAdj. Tagalog (Austronesian; Philippines), in contrast, is shown on the map as an instance of a language of the third type, having both orders with neither order dominant, because there is no evidence from the source that one of the orders is dominant (Schachter and Otanes 1972: 118, 121-122).

It should be emphasized that this map shows the order of adjectives modifying  a noun. It does not show the order of noun and predicative adjective, when the the noun is subject and the adjective is functioning as the predicate, as in English the boy is tall  and in the example in (3) from Simeulue (Austronesian; Sumatra, Indonesia).

(3) Simeulue (Kähler 1963: 131) 

mexiao 

luan 

ere 

clean 

river 

this 

‘This river is clean.’ 

The adjective mexiao  ‘clean’ in (3) is not modifying the noun luan  ‘river’; rather, luan ere  ‘this river’ is the subject and mexiao  ‘clean’ is the predicate. The position of mexiao  ‘clean’ in (3) reflects the fact that it is functioning as the predicate and predicates precede their subjects in Simeulue. Adjectives modifying nouns, in contrast, follow  the noun in Simeulue, as illustrated by the adjective tu’a-tu’a  ‘very old’ following ata  ‘person’ in (4).

(4) Simeulue (Kähler 1963: 17) 

ŋaŋ 

sa’a 

bəsaŋ 

sara 

ata 

tu’a-tu’a 

already 

then 

come 

one 

person 

old-old 

‘Then a very old man came.’ 

For the purposes of this map, the term adjective  should be interpreted in a semantic sense, as a word denoting a descriptive property, with meanings such as ‘big’, ‘good’, or ‘red’. It does not include nondescriptive words that commonly modify nouns, such as demonstratives (like this  in this dog ) (see Map 88A), numerals (as in two dogs ) (see Map 89A), or words meaning ‘other’ (as in the other dog ). In some languages, like English, adjectives form a distinct word class. In other languages, however, adjectives do not form a distinct word class and are verbs or nouns (see Chapter 118). For example, in Eastern Ojibwa (Algonquian; eastern Canada and United States), words expressing adjectival meaning are just like verbs morphologically and syntactically. The example in (5a), involving a word meaning ‘tall’ being used predicatively, inflects for a first person singular subject with a prefix n-  in the same way as the inflection for the verb meaning ‘sing’ in (5b).

(5) Eastern Ojibwa (Rich Rhodes, p.c.) 

a. 

n-ginooz 

 

1sg-tall

 

‘I am tall.’ 

b. 

n-nagam 

 

1sg-sing

 

‘I am singing.’ 

Similarly, these two words inflect in the same way when they are used attributively to modify a noun, as in (6). 

(6) Eastern Ojibwa (Rich Rhodes, p.c.) 

a. 

nini 

e-gnoozi-d 

 

man 

rel-tall-3sg

 

‘a tall man’ 

b. 

nini 

e-ngamo-d 

 

man 

rel-sing-3sg

 

‘a man who is singing’ 

Both modifying words in (6) bear third person subject marking and a relativizing prefix e-. Because words expressing adjectival meaning are really verbs in Ojibwa, instances in which such words modify nouns, like (6a), are, strictly speaking, relative clauses. In other languages, words expressing adjectival meaning form a well-defined subclass of verbs, sharing certain grammatical properties with other verbs, but differing in other respects. For example, in Lealao Chinantec (Oto-Manguean; Mexico) there is a subclass of verbs that express adjectival meanings which occur with verbal inflections but which differ from other verbs in that they can directly modify nouns without a relative marker. Compare (7a), with a nonadjectival verb preceded by the relative marker ʔiM, with (7b), in which an adjectival verb immediately follows the noun without the relative marker.

(7) Lealao Chinantec (Rupp 1989: 86) 

a. 

mïʔM

[ʔiM

kaL-láH

miVH-liuʔH]

 

clothes 

rel 

pst-buy.3

clf-little

 

‘the clothes that the child bought’ 

b. 

VH-kuï:M

tia:M

   
 

clf-corn

white 

   
 

‘white corn’ 

For the purposes of this map, these distinctions in word class are ignored: a word is treated as an adjective, regardless of its word class in the language, as long as it denotes a descriptive property. The map also ignores the question of whether the adjectives are modifying nouns directly or whether they are the predicate of a relative clause which is modifying the noun. It is a matter for future research to determine whether any of these distinctions provide a basis for further patterns in the distribution of AdjN and NAdj order, either typologically or geographically. 

The fourth type shown on the map are languages in which the adjectives do not modify nouns, in which in the closest equivalent to such structures, the adjective is actually the predicate in an internally-headed relative clause (see Chapter 90), and the noun is serving as its subject. Internally-headed relative clauses in Mesa Grande Diegueño (Yuman; southern California and northwest Mexico) are illustrated in (3) in Chapter 90. The example in (8a) below illustrates the translation equivalent of an adjective modifying a noun, but in fact the word for ‘white’ in (8a) is the verbal predicate of an internally-headed relative clause and the word aq  ‘bone’ is functioning as the subject of that verb; its structure is exactly parallel to the structure in (8b), with a nonadjectival intransitive verb.

(8) Mesa Grande Diegueño (Couro and Langdon 1975: 224, 236) 

a. 

'iikwich=ve=ch 

[aq 

ku-nemshap]=vu 

aakwal 

 

man=def=subj

[bone 

rel.subj-white]=def

lick 

 

‘The man licked the white bone.’ 

b. 

kwenychekwii=ve=ch 

[hekwany 

ku-mii]=vu 

selyewelyuu 

 

old.woman=def=subj

[baby 

rel.subj-cry]=def

tickle 

 

‘The old woman tickled the baby that cried.’ 

While superficially it might not be obvious that the examples in (8) involve internally-headed relative clauses, the fact that they have exactly the same form as the examples in (3) in Chapter 90, which are clearly internally-headed, means that these examples apparently involve internally-headed relative clauses as well. Note that the order of the noun and adjective in (8a) simply reflects the normal order of subject and verb in Diegueño.  Languages in which adjectives do not really modify nouns, but are predicates in internally-headed relative clauses, are probably more common than the map suggests, both because grammarians have until recently often failed to recognize internally-headed relative clauses and because the simple structure of internally-headed relative clauses with just noun plus adjective is such that it may not be recognized that they are simple instances of internally-headed relative clauses. Some of the languages that are shown as AdjN or as NAdj may prove under more careful analysis to be better treated as languages in which the adjectives are predicates in internally-headed relative clauses.

2. Geographical distribution

Both AdjN and NAdj orders are common in the world, though there are more than twice as many NAdj languages on the map. There are also clear geographical patterns. NAdj order is overwhelmingly the dominant order in Africa, though there exist a few well-defined pockets of AdjN order.  This area of NAdj order in Africa can be seen as extending northward into southwest Europe and to the northeast into the Middle East. NAdj order is also the dominant type in a large region stretching from northeast India through Southeast Asia eastward among Austronesian languages into the Pacific, except in the Philippines. It is the dominant order in both New Guinea and Australia, though there are many exceptions. Both orders are common in North America, but NAdj order is noticeably more common in the eastern half of the United States and among the more centrally located languages of Mesoamerica. NAdj is the majority type in South America, again with many scattered exceptions.

By far the largest area in which AdjN is found is a large area covering much of Europe and Asia, except in southwest Europe, the Middle East and Southeast Asia. Within this area, there are relatively few exceptions to the dominance of AdjN order, though a number of Tibeto-Burman languages of the Himalayan region are NAdj. AdjN order is clearly a minority type in Africa, but there are some clear pockets, notably in Ethiopia, in central Africa, and among Khoisan languages in southern Africa.  Similarly, AdjN is a minority type in Australia, though there is a scattering of them, including pockets in the southeast and in the middle of the north coast. The situation is similar in New Guinea, with a couple of pockets of AdjN order in the eastern Highlands and in the lower Sepik valley. AdjN order is as common as NAdj order in North America, and is more common in the western regions of Canada and the United States. Both orders are found in Mesoamerica, though AdjN order is more common in the northeast and southeast. In South America, the AdjN languages are confined to the western half of the continent, except for a pocket of languages in the vicinity of Suriname.

While some of the geographical patterns shown on the map reflect areal phenomena that cross genealogical boundaries, there are cases in which knowing genealogical classification can explain instances where languages in the same area are of different types. For example, Romanian is a NAdj language surrounded by AdjN languages, but this reflects the fact that it is a Romance language separated from other Romance languages and like other Romance languages is NAdj. 

Languages lacking a dominant order of adjective and noun are widely scattered, but are noticeably more common in the Philippines, in an area in and around Burma, in Australia and in the Americas. The languages shown as lacking constructions with an adjective modifying the noun because the closest equivalent involves an internally-headed relative clause are all in the Americas.

3. Theoretical issues

The order of adjective and noun has been of most interest because it is often thought that it correlates with the order of object and verb. However, as shown by Dryer (1988a, 1992) and in Chapter 97 of this atlas, this is not the case: NAdj order is more common than AdjN order, both among OV languages and among VO languages. Greenberg (1963) and Hawkins (1983) discuss other crosslinguistic generalizations involving the order of adjective and noun.  As noted above, it remains to be investigated whether distinguishing among different sorts of languages on the basis of the extent to which adjectives are a distinct word class (or subclass) might lead to new generalizations relating to the order of adjective and noun.