Case alternation in argument structure constructions with prepositional verbs : a case study in corpus-based constructional analysis

In present-day German, nine “two-way prepositions” can take either the accusative (ACC) or dative (DAT), viz. an, auf, hinter, in, neben, über, unter, vor and zwischen. In traditional explanations, ACC is associated with ‘motion’/‘directionality’ and DAT with ‘location’/‘state’. A similar conceptual contrast can be found in many modern accounts, which oppose ‘focus on the path’ and ‘endpoint focus’. While such a contrast may be satisfactory with respect to verbs such as gehen ‘go’, laufen ‘run’, wandern ‘wander’ etc., the ACC/DAT alternation with more complex verbal constructions cannot be accounted for in this way. This article provides an alternative explanation, which is corpus-based and couched in a constructional framework. The case study focuses on two prepositional verbs: aufsetzen auf ‘set down on, land on, base on’ and aufnehmen in ‘allow (as a member), incorporate, shelter, assimilate mentally’. These prepositional verbs were selected because they figure prominently in the literature on case alternation with two-way prepositions and exhibit interesting formal and semantic differences, which are potentially relevant from a constructional point of view: aufsetzen auf occurs in transitive and intransitive sentences, aufnehmen in only in transitive sentences, and they both have a range of different senses. The data is drawn from the Mannheim German Reference Corpus (DeReKo). The sentences (aufsetzen auf: N = 644; aufnehmen in: N = 454) are annotated for a set of morpho¬syntactic and lexical-semantic factors, viz. transitivity, voice, perfect tense and conventionalized senses. Classification trees are used to gauge the effects of the factors on the case alternation. The quantitative findings are interpreted according to a “three-level” approach to meaning based on the distinction between i) the general systemic meaning of a prepositional verb as part of the lexicon, ii) its conventionalized or default senses which pertain to “normal language use” and iii) its unique, referential readings in particular contextualized occurrences. The analysis shows that a constructional account makes it possible to determine the extent of the effects of the various factors on the case alternation in a quantitatively satisfactory and qualitatively innovative way. At the same time, the analysis leaves room to incorporate assumptions about the meaning of ACC and DAT proposed in previous accounts, in particular Paul (1920)