For this feature, which is based on Dryer (2011e), an adposition is defined as a separate free word that stands before, inside or after a noun phrase and establishes a grammatical or semantic relationship between the noun phrase and a verb in the same clause. There are four types of adpositions: prepositions stand before, postpositions after, circumpositions both before and after a noun phrase, while inpositions are situated inside the noun phrase. The latter type does not occur in our sample.
Since our criterion is that an adposition indicates some kind of relationship between a noun phrase and a verb, as in speak of the devil, attributive possessive constructions like the tail of the dog are disregarded here because of only establishes a relationship between the nouns tail and dog but not between either of them and a verb. We also disregard case affixes on nouns, which in some languages have functions that are similar to adpositions in European languages. On the other hand, cliticized relation markers, which are phonologically integrated into the noun but whose position depends on the syntax, do count as adpositions for the purposes of this feature.
The type or token frequency of adpositions in a language is irrelevant for this feature. As long as there is at least one adposition, even if it has a low text frequency, its position relative to the noun phrase is what matters. For example, Gurindji Kriol has only two adpositions, langa and bo, which are quite rare (Meakins 2013), but since they always stand before the noun phrase whenever they occur, Gurindji Kriol is treated as a language that has only prepositions.
A number of APiCS languages use serialized verbs to encode meanings that are expressed by prepositions in other languages, e.g. come for ‘to’, go for ‘away’, take for ‘with’, pass for ‘through’, etc. (see also Chapters 84-86). Compare ba ‘give’ = ‘for’ in Guyanais:
As the degree of grammaticalization of such verbs is often hard to determine, they pose a potential problem for this feature and are disregarded. However, none of the APiCS languages relies on serialized verbs alone. For example, the instrumental meaning in Nengee can be expressed both through serialized tek ‘take’ and the preposition anga ‘with’:
The exclusion of serialized verbs from the present feature did therefore not result in a language being classified as having no adpositions.
Three different values are distinguished for this feature:
None of the languages in the APiCS sample has inpositions and there is no language that does not make use of adpositions at all.
Value 1. Most APiCS languages have only prepositions, like e.g. Mixed Ma’a/Mbugu:
Value 2. A handful of languages have only postpositions, like Yimas-Arafundi Pidgin:
Value 3. Circumpositions are illustrated by an example from Michif:
The APiCS languages Nengee, Early Sranan, Sranan and the student lect of Ghanaian Pidgin English have constructions that look like potential circumpositions. Compare na … tapu in the following example:
Such “circumpositions” consist of a general preposition (here: na) and a postposed locative or temporal nominal (here: tapu), whose degree of grammaticalization into a postposition is hard to determine (cf. Plag 1998). Since in all cases it is possible to interpret such constructions as a prepositional phrase containing a preposition and a possessive noun phrase (PrepP[Prepna NP[eksi tapu]] lit. ‘on the egg’s top’), these cases were not counted as circumpositions.
Feature 4 is a multiple choice feature and 9 languages in our sample allow adpositions to occur before (value 1) and after the noun (value 2). Compare e.g. Berbice Dutch, where prepositions (7a) and postpositions (7b) occur with about the same frequency:
All APiCS languages that have circumpositions also have both prepositions and postpositions. Compare the examples from Diu Indo-Portuguese (a. pə = preposition, b. jũt = postposition, c. də … jũt = circumposition):
Note that the corresponding WALS feature 85A “Order of adposition and noun phrase” (Dryer 2011e) records only the dominant value and disregards the less frequent alternatives. By contrast, APiCS maps all possible orders in a language, regardless of their frequency.
Numerical. Prepositions are by far the most common adpositions in the APiCS sample, with 62 languages relying exclusively on them and another nine having prepositions along with other adpositions. None of the three languages that make use of circumpositions – Afrikaans, Diu Indo-Portuguese and Michif – relies exclusively on them, and only in Afrikaans are they not marginal.
When the APiCS values are modified to match WALS, the following picture emerges: prepositions predominant 65, postpositions predominant 7, no dominant order 4. This is a stark contrast to the trend among the world’s languages: postpositions 577, prepositions 512, no dominant order 58 (of a total of 1158 languages). APiCS languages thus show a much stronger reliance on prepositions, but it has to be kept in mind that the APiCS sample is not typologically balanced.
By lexifier/substrate. For reasons of space, the following discussion will be restricted to postpositions, which are significantly less frequent in the APiCS sample than in the languages of the world. Dutch has pre-, circum- and postpositions (Koopman 2000: 206), and postpositions are also found in Afrikaans and Berbice Dutch. Interestingly, postpositions are marginal in Afrikaans, even though the contributing language Khoekhoe has postpositions. In Berbice Dutch, on the other hand, prepositions and postpositions occur with about the same frequency. The reason is possibly that the strong Ijo substrate (Smith 1999: 254f) has postpositions (Tepowa 1904: 122). Against this background it is surprising that the third Dutch-based language in our sample, Negerhollands, does not have postpositions at all, even though there was a strong postpositional input: Ewe (predominantly postpositions) and Akan (no dominant order; Dryer 2011e) were the major African substrates during the formative years of Negerhollands (van Sluijs 2013a). None of the five APiCS languages that rely exclusively on postpositions had a consistent postpositional input: either the lexifiers (predominantly) had prepositions and the substrates (predominantly) postpositions – as in Chinese Pidgin Russian, Sri Lanka Portuguese and Sri Lankan Malay, or the other way about – as in Pidgin Hindustani and Yimas-Arafundi Pidgin.