Feature 3. Position of oblique elements

found in question(s): 6a, 6b, 6c, 15a, 15b, 15c, 15d, 15e, 15f

Definition and illustration

Oblique elements are defined as adpositional phrases or noun phrases marked with peripheral cases that are used in adverbial function (Dryer & Gensler 2013, Hawkins 2008). Whereas previous research did not clearly differentiate between different kinds of oblique elements, this database makes a distinction between three different kinds, i.e. instruments (question 6a), places (question 6b), and directions (question 6c). Drawing more fine-grained distinctions allows a higher resolution and more qualified predictions concerning potential cross-linguistic correlations. As shown in the background description of this database, some languages such as Oromo exhibit different word order properties depending on the type of oblique. Whereas question 6 addresses the neutral word order in a wide-focus context, question 15 additionally tests for the influence of information structure.

A methodological problem pointed out by Sposato (2014: 118) for the rare XVO languages is that many of the prepositions that are part of the oblique element derive from verbs. In example (1) from Mandarin, yòng is analyzed as a preposition meaning 'with' but is formally still identical to the verb yòng 'to use'.

(1) Mandarin (Sinitic, Trans-Himalayan)

Therefore, example (1a) could potentially be interpreted as an instance of verb serialization. However, the postposition in (1b) lacks a verbal origin and demonstrates that Mandarin does have XVO order. To rule out potential other explanations of this sort, the project collects extensive background information for every language under investigation.

Correlations

Oblique elements, although adverbial in function, exhibit a syntactic behavior that is slightly different from adverbs as described in feature 2 (Hölzl 2025). Table 1 shows the data on the placement of oblique elements provided by Dryer & Gensler (2013). The data need to be treated with caution since the relative order of O and X is often unknown or inferred from information on OV/VO and XV/VX in grammatical descriptions, which leads to a biased sample with OVX being overrepresented (here indicated with an asterisk). Another problem is the category of “no dominant order”, which does not allow for further conclusions.

Table 1: Order of verb, object, and oblique (X) in a global sample

The dataset in Table 1 exhibits different patterns (Hawkins 2008). There are general tendencies for oblique elements to follow both V and O (VOX, OVX) and for obliques and objects to occur on the same side of the verb (VOX, XOV, OXV). If V stands between X and O, the order is usually OVX. XVO is rare and mostly found in Sinitic and some surrounding languages, such as Hmong-Mien (Sposato 2014). Most importantly, OV allows X to intervene between O and V (OXV), whereas intervening X is not found in languages with VO order (absence of VXO). When VXO order is attested, as for example in (2), it is not the only or dominant order; in Czech, VOX as in (1a) is more common and perceived as less marked.

(2) Czech (Slavic, Indo-European)

One type of pattern can be observed when focusing on the positioning of the oblique element: OV shows much more variability than VO (Table 2). While XOV, OVX, and OXV are all relatively widely attested, VO languages usually have VOX order. A hyphen indicates the position of O and V relative to X.

Table 2: Matrix showing the order of OV/VO and obliques (X) (excluding languages with no dominant order)

This allows the establishment of unidirectional correlations:

(3) if VO then -X (VOX)

(4) if -X- or X- then OV (OXV, XOV)

However, the oblique element following both O and V cannot accurately predict the relative order of O and V because both VOX and OVX are widely attested. Similarly, OV cannot predict the positioning of X because all three possibilities are roughly equally common (XOV, OXV, OVX).

Another pattern is found when considering the relative order of X and O, exclusively, which compresses the dataset to four types as above (Table 3). This obscures the effects of intervening X, but puts into focus the relative order of two dependent elements, X and O.

Table 3: Matrix showing the order of OV/VO and OX/XO (excluding languages with no dominant order)

This illustrates an overlapping but slightly different unidirectional correlation, shown in (9) and (10).

(5) if VO then OX (VOX)

(6) if XO then OV (XOV)

In contrast, OX and OV do not allow for clear predictions.

References