Person-Role/Case constraints in a cross-linguistic perspective

In the generative literature examples like Agnès me le présentera/*Agnès me lui présentera (Agnes will present it to me/ me to him) was first discussed by Perlmutter (1971) as the me lui/I-II Constraint:. (i) In a combination of a direct object and an indirect object, the direct object has to be third person; (ii) Both the direct object and the indirect object are phonologically waek (= clitics, affixal pronominal markers, weak pronouns) Given such a formulation in the following questions arise: What is the reason that person marking and role/case marking interact? Why is the interaction linited to phonologically weak elements (affixes, clitics); To what extent is this restriction cross-linguistically valid? This project is aimed at (i) further developing an analysis along the lines of Anagnostopolou (2001), in which person restrictions are derived from split phi-feature checking, while at the same time (ii) expanding the emprirical domain by sysematic crosslinguistic research, and see whether the cross-lingistics differences can be accounted for in such an approach.

The analysis could be extended in certain areas:

1. The *me/lui/I-II constraint and nominative objects in Icelandic. Could we establish a correlation between a constraint on agreement/clitics, The *me/lui/I-II constraint, and a restriction on nominative objects, the Prohibition against (agreeing) 1st and 2nd person nominative objects in Icelandic?
2. The *me/lui/I-II constraint extended to SE-reflexives. Should we link the *me/lui/I-II constraint to the observation that double object constructions also show a restriction on reflexive accusative clitics?: * Elle se lui est donnée entièrement (She gave herself to him entirely)
3. Person constraints in simple transitive constructions. It has been noted in the literature (Rosen 1990, Aissen 1999) that some languages show person restrictions in 'simple' transitive constructions. Are there similarities with the *me/lui/I-II constraint.

Languages seem to have ways to avoid the *me/lui/I-II constraint: French avoids taking a clitic, Georgian employs a strategy that Harris (1981) calls object camouflage. In this project we want to get a clear picture of which avoidance strategies are available cross-linguistically.

It has been argued that some languages which might be expected to exhibit the constraint, in fact do not seem to obey it. Polish would be such a language, and other examples mentioned are Kabardian, Lakhota, Noon and Haya. What could be reasons for a designated language not having t the *me/lui/I-II constraint constraint?