This is the second of three maps that show the relationship between two features that are shown separately in earlier maps, namely the order of relative clause and noun (Map 90A) and the order of object and verb (Map 83A). The two orders of object and verb and the two orders of relative clause and noun intersect to define the first four values shown in the value table.
|Go to map|
|Object-verb and relative clause-noun (OV&RelN)||132|
|Object-verb and noun-relative clause (OV&NRel)||113|
|Verb-object and relative clause-noun (VO&RelN)||5|
|Verb-object and noun-relative clause (VO&NRel)||416|
|Languages not falling into one of the preceding four types||213|
This map illustrates a case where three of the four types are common and one type is uncommon: the first, second and fourth types are common, while the third is uncommon.
The first type consists of OV and RelN languages, languages in which the object precedes the verb and the relative clause precedes the head noun. An example of such a language is Kolyma Yukaghir (isolate; Siberia, Russia); (1a) illustrates the OV order while (1b) illustrates RelN order.
‘My father has killed your husband.’
‘the fish [that we have lost]’
‘The starlings build their nests in a tree.’
‘one man [who is standing up in the stern]’
Languages of the first type are common, though as the map shows and as is discussed below, they are more common in Asia than in other parts of the world.
The second type consists of OV and NRel languages. While such languages are sometimes thought to be uncommon, it is clear from the map that they are in fact very common. An example of such a language is Harar Oromo (Cushitic; Ethiopia); the OV order is illustrated in (3a), the NRel order in (3b).
‘I saw my camel.’
‘the girl [he gave a present to]’
‘Those authorities defended that woman.’
‘the man [that you want to see]’
The third type consists of languages which are VO and RelN. An example of such a language is Mandarin, illustrated in (5), (5a) illustrating the VO order, (5b) the RelN order.
‘They steal bicycles.’
‘the book [that I gave you]’
Languages of this third type are distinctly rare. Apart from Mandarin and other varieties of Chinese, the only other cases on the map are two languages in close geographical proximity to Chinese, namely Bai, a Tibeto-Burman language of China that has been heavily influenced by Chinese, and Amis, an Austronesian language of Taiwan (Joy Wu, personal communication), where there is also a possibility of influence from Chinese. The examples in (6) illustrate this for Bai, (6a) illustrating VO word order, (6b) the RelN order.
‘Grandpa add(ed/s) fertilizer (to the field).’
‘words [that are written tidily]’
the book [that I bought]
‘They are building a house.’
‘the pig [that my father bought for me]’
Most VO languages are of this type.
The fifth type shown on the map includes various types of languages not falling into one of the first four types. This includes languages lacking a dominant order of object and verb; languages with one of the four less frequent types of relative clauses shown on Map 90A, namely internally-headed relative clauses, correlative clauses, adjoined relative clauses, and double-headed relative clauses; and languages which have two or more of the relative clause types without one being dominant. Although not shown on the map, it is worth mentioning that these four less frequent types of relative clause are mostly found in OV languages. In fact, my data contains only one exception to this generalization, Kutenai (isolate; western North America), illustrated in (9), which is VO, in fact verb-initial, as in (9a), with internally-headed relative clauses, as in (9b).
(9) Kutenai (own data)
‘The old woman rolled up the teepee door.’
‘The woman that Mike saw died.’
One other VO language with internally-headed relative clauses is Tukang Besi (Austronesian; Sulawesi, Indonesia), but this is not the dominant type of relative clause in the language, since it co-exists with externally-headed postnominal relatives (Donohue 1999a: 304, 367).
OV and RelN languages are very common in much of Asia except in an area extending from the Middle East to around Afghanistan and in eastern China and Southeast Asia. Outside this area, they are not common, although there are multiple instances in (i) Ethiopia and Eritrea and (ii) New Guinea; in both of these areas, however, there are also many languages that are OV and NRel. OV and NRel languages are common in parts of North America, in northwestern South America, in those areas in Africa where OV languages are common, in an area in Asia stretching from Iraq to Afghanistan, and in Australia. In effect, the distribution of the two orders of relative clause and noun among OV languages is largely determined geographically: OV languages in Asia are generally RelN, while OV languages elsewhere in the world are much more commonly NRel, though there are small areas, as noted, in which RelN languages are common. It is often assumed, because of the high frequency of RelN order among OV languages in Asia, that it is normal for OV languages to be RelN; however, the evidence here shows that much of the rest of the world differs from Asia in this respect.
Since most VO languages are NRel, the distribution of VO and NRel languages on the map is similar to the distribution of VO languages on Map 83A: they are common in (i) Europe and around the Mediterranean; (ii) much of sub-Saharan Africa; (iii) an area stretching from southeast Asia eastward through the Pacific; and (iv) Mesoamerica. Languages which are VO and RelN are confined to mainland China and Taiwan. Note that the location of these VO and RelN languages is adjacent to the area in which OV and RelN is most common, in Asia, and it seems likely that the occurrence of such languages in Asia is related to the overall high frequency of RelN order in Asia.
This map resembles in some respects the distribution on the next map, Map 97A, which shows the relationship between the order of object and verb and the order of adjective and noun. The distribution of the two OV types on this map resembles the distribution of the two OV types on that map. On both maps, the type combining OV with AdjN or RelN is found in a large area covering most of Asia, and in roughly coterminous pockets elsewhere in the world. Apart from the fact that there are fewer languages shown on this map than on Map 97A (because grammatical descriptions more often give information on the order of adjective and noun), the most noticeable differences are (i) the fact that languages in the area around Burma are part of the large area in Asia in which OV and RelN order is common, but not part of the area in which OV and AdjN is common; and (ii) while there are many OV and AdjN languages in the southwestern United States, this is not a region in which OV and RelN is common. On the other hand, while the distribution of the two OV types is similar on the two maps, the distribution of the two VO types is quite different: on this map, almost all the VO languages are NRel, while on Map 97A both AdjN and NAdj are common among VO languages.
The distribution on this map can be described in terms of an implicational statement “If a language is VO, then it is usually NRel”, or by what is an essentially equivalent statement “If a language is RelN, then it is usually OV”. Note that this implication is unidirectional, in contrast to that discussed in Chapter 95, where the implication is bidirectional. In other words, one cannot reverse the above implicational statements; one cannot say “If a language is NRel, then it is usually VO” or “If a language is OV, then it is usually RelN”, both of which are false.
The fact that one can formulate implicational generalizations relating these two typological features means that there is an interaction or correlation between them. This correlation is one of many correlations between the order of object and verb and the order of other pairs of elements, and the attempts at explaining these correlations, discussed in Chapter 95, are relevant to the correlation discussed in this chapter as well, although it is not clear that there is any good account in terms of grammaticalization. In addition, since relative clauses are not complements of the noun, accounts in terms of consistent ordering of heads and complements will not account for the correlation found with relative clauses. Accounts in terms of heads and dependents will work, but will not account for the existence of a correlation with relative clauses but not with adjectives (see Chapter 97).
Any explanation for the correlation must also account for the unidirectionality of the implication. One type of approach is in terms of competing motivations; on this approach, there are explanatory forces that favour the type VO and NRel and the type OV and RelN, and other explanatory forces favouring NRel over RelN and thus favouring the type VO and NRel and the type OV and NRel. These forces would work together to favour NRel order among VO languages, but would be in competition with each other in OV languages, where one would favour OV and RelN, while the other would favour OV and NRel. Proposals by Hawkins (1990, 1993, 1994, 2002) are accounts of this sort.
The absence of a correlation between the order of object and verb and the order of adjective and noun, while there does exist a correlation between the order of object and verb and the order of relative clause and noun, might seem surprising in light of the fact that relative clauses resemble adjectives in terms of their relationship to the nouns they modify, as reflected by the fact that relative clauses are often referred to as adjective clauses. It is proposed in Dryer (1992) that the fact that we get a correlation only with relative clauses is due to the fact that they are phrasal, and in general contain more than one word, while adjectives are generally nonphrasal, consisting of a single word. Placing relative clauses before the noun in an OV language will tend to make the language consistently left-branching, while the order of adjective and noun will not affect the direction of branching in a language.