Adposition borrowing at SLE 2013

The Societas Linguistica Europae's annual conference finished today. The plethora of parallel sessions forced me to miss a lot of potentially interesting talks, but here are some highlights from the workshop I was participating in: adposition borrowing. This workshop was organised around a generalisation proposed by Edith Moravcsik 25 years ago, which has held up remarkably well (better than probably any other structure-based generalisation proposed about language contact):

"A lexical item that is of the 'grammatical' type (which type includes at least conjunctions and adpositions) cannot be included in the set of properties borrowed from a language unless the rule that determines its linear order with respect to its head is also borrowed." (source)

Eitan Grossman presented a number of apparent counterexamples – in fact, he reported that fully one-third of his sample of languages with borrowed adpositions displayed counterexamples. His effort to systematically test the hypothesis is laudable. However, the results cannot be taken at face value. Many examples, on closer examination, turn out to be amenable to one of three alternative explanations:

1. The adposition was originally borrowed as a preposition, and turned into a postposition in the course of a more general typological realignment of the language. (This applies to Sri Lanka Malay dative nang, ultimately from a Javanese preposition; Authier et al. presented a new example, an apudlocative preposition possibly borrowed from Tatic into Georgian: Tatic (b)-tan N > old Georgian tan-a N > modern Georgian N-tan.)
2. The source language order is not necessarily as postulated. (Thus the Khorasan Turkish postposition tā is assumed to be from Persian, in which it is a preposition, but could also derive from neighbouring Mazanderani, in which it is a postposition.)
3. The "adposition" is also used without a complement in the source language (eg as a noun or adverb), and hence was not necessarily borrowed as an adposition. (This applies, for instance, to the Brahui postposition savā "without", connected at some remove to Persian سوا sevā "separate, other", or to the Manambu postposition wantaim "with", from Tok Pisin wantaim "together (adv.) / with (prep.)". In some cases the adverb is unambiguously the source, for instance Turkish raǧmen "despite", from the Arabic adverb raghman رغما rather than the preposition raghma رغم.)

1 and 2 merely illustrate the need for in-depth historical linguistic investigation of each case, which should go without saying. 3, however, is more interesting in principle. If an adposition can readily occur as a noun or adverb, are we justified in classing it as "of the 'grammatical' type"? The answer I gave in my presentation, before discussing the borrowing of adpositions in Northern Songhay, was: no. Not all adpositions are functional, as various authors have been pointing out since at least 1990, and we should not expect the generalisation to apply to lexical adpositions. In fact, we need at least a three-way distinction (cp. Littlefield 2005): purely functional adpositions such as of, in, to; purely lexical items used in complex adpositions such as front, back, middle (Svenonius's (2010) "axial parts"); and mixed items which simultaneously express the meanings of both a nominal/adverbial stem and a functional adposition governing it, such as beside (by the side of), inside (on the inside of). Functional adpositions should be subject to Moravcsik's generalisation; mixed items should be able to go both ways; and lexical items should be subject to the recipient language grammar alone. This proposal appears to eliminate all the few genuine exceptions to Moravcsik's generalisation so far proposed; however, it remains to be seen whether this criterion can be defined unambiguously for all adpositions in all languages.

Petros Karatsareas gave a nice summary of the situation in Cappadocian Greek (cf. Dawkins 1916), which has taken advantage of Greek's word order flexibility to move a long way towards developing postpositions; relational nouns which in Medieval Greek normally preceded their complement came to obligatorily follow it, yielding circumpositions (governing the genitive) whose prepositional component then became optional. This strategy was in turn used for borrowing Turkish adpositions.

Riho Grünthal pointed out the striking rarity of borrowed prepositions on Finno-Ugric, even in languages such as Finnish or Saami that (as a result of contact) have developed prepositions. This seems to confirm a point that I had also made in regard to Northern Songhay: that it's much easier to borrow adpositions when they have the same syntax in the source and target languages. He did find one or two cases, though, notably Livonian pa, from Latvian. Brigitte Pakendorf showed that Even borrows a fair number of Yakut postpositions (with varying degrees of acceptance among speakers), but no Russian prepositions, which at first sight seems to confirm the role of congruence even more. However, it's also true that Yakut has influenced Even much longer than Russian has.

Edith Moravcsik herself finally gave a summing-up address, in the form of an outline of relevant factors that need to be considered in the typology of adposition/case marker borrowing, with allusions to the talks given; she didn't focus particularly on her original generalisation, and she gave the impression of seeing it as being only statistically true in light of the proposed counterexamples.

I won't go into detail on the contributions that did not directly address Moravcsik's generalisation here, since this is already getting too long for a blog post, but some were also very interesting. Notably, Bakker and Hekking revealed that, whereas Quechua and Guarani make little use of Spanish adpositions, Otomí has massively adopted them – probably because Otomí, unlike the other two, had no morphemes serving such a function before contact, leaving it to context.

Much work remains to be done on the topic. Do you know any prepositions that have been borrowed as postpositions, or vice versa?