Lexical vs. structural case: a false dichotomy

This article takes issue with the two dichotomies of structural vs. lexical case and thematic vs. idiosyncratic case, on the basis of their predictions on: (a) synchronic productivity, (b) language change, and (c) language acquisition. It is shown here that these predictions are not borne out in Icelandic. In fact, productivity data from Icelandic suggest that accusative objects to new verbs are assigned lexically and not structurally. Another problem is presented by different changes in case marking in the history of the Germanic languages, changes that can only be captured by two complementary approaches to structural case, which in turn severely undermines the general explanatory power of this concept. It turns out, moreover, that the case preservation property of lexical case, as opposed to structural case, in passives and raising-to-object constructions, is a construction-specific property, not generalizable to the language as a whole. An alternative approach is sketched in terms of a usage-based Construction Grammar where all case marking of core arguments in Icelandic is regarded as lexical, i.e. word-bound, and modeled in terms of lexicality-schematicity hierarchies which capture verb-specific idiosyncrasies, higher-level generalizations, as well as the default status effect found for the Nom-Acc Construction.