Feature 1. Position of objects (O arguments)

found in question(s): 1a, 1b, 2a, 2b, 3, 4, 12a, 12b, 12c, 12d, 12e, 12f, 22a, 22b, 22c

Definition and illustration

Traditionally, the basic word order of a language is described in terms of the six possible relative positions of subject, object, and verb (i.e., SOV, SVO, OSV, OVS, VSO, VOS). However, Dryer (2013a) and others have shown that in many cases the relative position of object and verb (i.e., OV, VO) has more predictive power than the traditional six-way tyopology. The focus within this database, therefore, lies on the OV / VO distinction.

This database aimed to include different types of languages with respect to this typology. First, there are languages where either OV or VO is the dominant, unmarked, or only possible word order.

(1) Akan

(2) Ika

Second, the two languages Jula and Tunen exhibit the typical and rigid West African (S)AuxOV(X) word order. The language Bwamu, which otherwise has dominant VO order, also exhibits one auxiliary that allows this AuxOV word order (next to AuxVO).

(3) Bwamu

Two languages have the characteristic V2+OV order: Standard German and Bernese German. In both languages, the finite verb is usually in the second position within the clause while OV order is the only possibility in subordinate clauses or if an auxiliary is present. They are classified as having (underlying) OV order in this database.

(4) Bernese German

A few languages allow both VO and OV order. In these cases, one of the two orders is usually dominant or unmarked in terms of information structure and/or morphosyntactic marking. For instance, the Sinitic languages in the sample usually allow both word orders although one of them is additionally marked (so-called positional differential argument marking, see Seržant et al. preprint). In Kangle Chinese, the dominant and unmarked order is VO, whereas OV is possible if a definite object is additionally marked with a prepositional marker (cf. the well-known ba-construction in Mandarin). See also question 66 on differential object marking.

(5) Kangle Chinese

Finally, the sample contains one language that arguably lacks a clear basic word order (i.e. Hungarian) as well as one language with verb initial VO(S) order (Mopan Maya).

Questions 1a, 1b, 2a, 2b, 3 and 4 test for the relative order of object and verb in a neutral, wide-focus context, taking into account some of the typological variation described above (presence and absence of an auxiliary, definiteness, embedding). Questions 12 and 34 additionally tests for information structure (i.e., information focus, constrastive focus/topic), which can have an impact of the possible word orders or the preference. Questions 22, 41, 42, and 47 additionally test for the internal structure of the verb phrase, e.g. whether the verb can be fronted with or without the object or whether the object is omissible (with or without the verb) under certain conditions.

Correlations

The majority of the questions within the questionnaire used to construct this database are known or suspected to correlate with basic word order (for details, see the other feature descriptions).

Previous research concerning linear word order identified two different types of correlations: bidirectional correlations, represented by adpositions (Adp), and unidirectional ones, represented by relative clauses (Rel) (Dryer 2007). These are two well-known cases that can be described in terms of cross-categorial harmony of head and dependent. First, languages with OV order usually have adpositions that follow the noun phrase, while those with VO have prepositions (see question 77). In this bidirectional correlation, both features predict each other (Dryer 2007: 89):

(6) if OV then NPAdp and if NPAdp then OV

(7) if VO then AdpPN and if AdpNP then VO

Table 1 shows the relevant numbers in a global sample. There is a strong tendency towards harmony of head-initial (VO & AdpN) and head-final types (OV & NAdp). The few exceptions from the correlations indicate that these should be considered statistical tendencies rather than absolute universals.

Table 1: Matrix showing the order of OV/VO and AdpNP/NPAdp (Dryer 2013b, 2013c), excluding other types

A certain number of languages is not classifiable according to this simple typology. For instance, many languages with SVO order and prenominal genitives, such as Mandarin, have both pre- and postpositions (Dryer 2019).

Second, as opposed to adpositions, adnominal relative clauses only exhibit a unidirectional correlation with the order of the transitive verb and its object (see question 81). In this case, only the following predictions are cross-linguistically valid (Dryer 2007: 97):

(8) if VO then NRel

(9) if RelN then OV

In contrast, OV cannot predict the relative order of relative clauses and noun phrases, and, knowing that a language has NRel order, one cannot accurately anticipate the order of objects and transitive verbs, although VO is much more common (Table 2).

Table 2: Matrix showing the order of OV/VO and RelN/NRel (Dryer 2013b, 2013d), excluding other types

Apart from a tendency towards harmony (VO & NRel, OV & RelN), there is a cross-linguistic preference for postnominal relative clauses (NRel) that is unrelated to the order of object and verb (Dryer 2011: 342). Therefore, the 113 languages displaying OV & NRel in Table 2 break the head-dependent harmony but are in line with the preference for relative clauses to appear after the noun.

There are only a few exceptions to these correlations, such as RelN & VO order in Sinitic and some surrounding languages (Comrie 2008; Sposato 2014).

(10) Mandarin (Sinitic, Trans-Himalayan)

Nevertheless, the correlations remain cross-linguistically valid as very strong tendencies. Further examples of head-dependent order and potential correlations with basic word order are described in features 30 and 32. However, most of the other other features in this database concern more complex patterns that cannot be reduced to head-dependent order. Correlations that go beyond the order of head and dependent can be illustrated with the dominant position of oblique elements (X) with respect to O and V, where X is defined as a noun phrase or adpositional phrase used as adverbial modifier or adjunct (e.g., with a key). While the linear relation of X and V would instantiate just another case of head-dependent ordering, the relative position of X and O in this triplet cannot be broken down to a head-dependent relationship. As such, cross-categorial harmony alone would not suffice to explain potential correlations in this domain (see feature 3).

References