Typological asymmetries and agreement alternations in ditransitive constructions

Ditransitive constructions across languages differ in how the two objects are expressed (case alignment) and whether they control agreement on the verb (agreement alignment). Interestingly, not all combinations of case and agreement alignment are attested in the languages of the world. No language seems to exclusively have a type of ditransitive in which the recipient is expressed like the typical single object of a monotransitive while the theme argument controls agreement on the verb. More technically, there is no language with only secundative or neutral case alignment and indirective agreement alignment. This observation is supported by a genetically and geographically broad sample of more than 120 languages from 99 genera.  

I propose a morphosyntactic explanation for this observation based on independently motivated evidence, namely locality, the syntactic structure of ditransitive constructions, and the effects of the case hierarchy. In brief, the unattested type of language is ruled out because the agreement relation between the predicate and the theme cannot be established across a recipient which must bear case that is accessible for agreement. I also discuss agreement alternations which seem to give rise to the otherwise unattested patterns and show how person and information structural properties can account for them.