Languages that have clitic pronouns frequently prohibit certain combinations of these clitics (e.g., the Person-Case Constraint). Why do these constraints restrict just clitic pronouns, not arguments more generally? And, why are only some combinations of clitics prohibited and not others? We identify two patterns in the clitic combinations that are allowed across languages and across phi-domains (across person and gender). These patterns arise, we propose, from how clitics are licensed syntactically; certain asymmetries point, in particular, to the universal role played by a cyclic version of Agree in clitic licensing. The attested variation across languages in how they constrain clitic combinations can then be derived entirely from variation in their lexicons.
[Joint work with Steven Foley]