Mark Liberman at the Log has a very interesting post about the common origin of prepositions (and postpositions), verbal prefixes, and adverbs in the Indo-European languages, quoting Virginia Anne Goetz’s 2006 dissertation, “The development of Proto-Indo-European local adverbs into Germanic prepositions and verbal elements” (O = object; see Mark’s post for other explanations):
In the initial stage of its development, PIE x was a free constituent in a functional language. Its role was to add a place adverb (xv) modality to a sentence. Often it related a case-bearing object to a verb. In the sequence OxV, for example, O ( – village) might be in the locative case and x would provide additional place adverb information to relate O to V (= go):
the village – toward, within, into, through, around, etc. – go.
From this earliest stage, there were innovations which related x to the object or to the verb. Sanskrit and Hittite are considered to be the most conservative in terms of these developments. There are in these languages some recurring expressions in which x appears to have an attachment to a case-bearing object as an O-x, so that Sanskrit and Hittite may be seen to be on the cusp of developing x as a case assigner. [...] In these languages, x was mostly xv (a free adverb of place) or part of an OxV. In the latter, the role of x is ambiguous in terms of [verb proclivity] versus [object proclivity].
The most innovative in terms of x development are Latin and Greek. While there is still some relic structure, especially in older Greek, the “classical” stages of these languages have b.xv (bound verbal prefix), x-O (“preposition”) and xv (“adverb”).
The early Germanic languages, and hence reconstructed Proto-Germanic, fall between the extremes of the conservative (Hittite, Sanskrit) and the innovative (Latin, Greek). While Germanic has a group of b.xv’s and sets of x-O, it still retains x’s that are ambiguous.
There’s plenty of other meaty historical stuff there; Mark’s conclusion:
From PIE to the present day, the consistent driver of change in this arena has been re-analysis — typically, syntactic re-interpretation of a functional relationship. Sometimes this is simply re-parsing of an ambiguous sequence, as when O x V was interpreted as O x-V. And sometimes it’s a simplification of a more complex structure, as when by cause that S becomes simply because S.
I don’t have strong opinions about whether the different functional and structural relations involved should be terminologically split or lumped or both — but I think that Geoff [Pullum] makes a good case for seeing the complex usage patterns of words like in, from, and because as variations on a single grammatical theme.
Makes sense to me.