1. In essence this may be a diachronic universal: HeadNoun – Genitive constructions are a source of Adposition – Complement constructions, with head nouns reanalysed (“grammaticalized”) as adpositions. If the ordering remains unchanged in the course of such categorial reanalyses, then the order of constituents in the target construction would mirror that in source constructions.
In Greenberg 1963 itself, this way of dynamicizing word order typology is not highlighted; but it is in other work of Greenberg’s.
2a. All adpositions (in all their uses)? Most? An implicit claim, at any rate, is that adpositions tend to harmonize AMONG EACH OTHER. In no language, therefore, would half of them be expected to precede and the other half to follow their NPs.
2b. If there are exceptions to the rule of adposition ordering, especially when the rule is to precede, then it is probably not entirely unpredictable which particular adpositions will be the odd men out. For example, temporal adpositions meaning ‘ago’ seem to have a tendency to be postpositions – although I can really only quote a few examples like English ago and Italian fa. Their positional divergence is to do with their history: their sources are participles which were final in absolute constructions (two years (a)gone, due anni fatti ‘made’). Cf. Philippe Bourdin [still unpublished?].
1. In essence this may be a diachronic universal:
HeadNoun – Genitive constructions are a source of Adposition – Complement constructions, with head nouns reanalysed (“grammaticalized”) as adpositions. If the ordering remains unchanged in the course of such categorial reanalyses, then the order of constituents in the target construction would mirror that in source constructions.
In Greenberg 1963 itself, this way of dynamicizing word order typology is not highlighted; but it is in other work of Greenberg’s.
2a. All adpositions (in all their uses)? Most? An implicit claim, at any rate, is that adpositions tend to harmonize AMONG EACH OTHER. In no language, therefore, would half of them be expected to precede and the other half to follow their NPs.
2b. If there are exceptions to the rule of adposition ordering, especially when the rule is to precede, then it is probably not entirely unpredictable which particular adpositions will be the odd men out. For example, temporal adpositions meaning ‘ago’ seem to have a tendency to be postpositions – although I can really only quote a few examples like English ago and Italian fa. Their positional divergence is to do with their history: their sources are participles which were final in absolute constructions (two years (a)gone, due anni fatti ‘made’). Cf. Philippe Bourdin [still unpublished?].