Optimal Linearization: Word‐Order Typology with Violable Constraints

Insofar as linearization is a postsyntactic phenomenon pertaining to phonological forms, it is desirable that it be modeled in a violable‐constraint framework, like other PF‐branch phenomena. This paper presents such a model, with the goal of capturing the Final‐over‐Final Condition, a typological generalization that says that head‐final phrases embed only other head‐final phrases, while head‐initial phrases may embed phrases of either headedness. The specific model proposed here, termed Optimal Linearization, is couched in Optimality Theory and comprises three interacting violable constraints on the mapping of syntactic structure to linear order. One constraint, A ntisymmetry , closely mimics the action of Kayne's familiar Linear‐Correspondence Axiom. An opposing constraint H ead F inality penalizes deviations from an idealized head‐final order. Lastly, a domain‐specific constraint H ead F inality ‐α enforces head‐final ordering only within one constituent, allowing for disharmonic orders that respect the Final‐over‐Final Condition. Optimal Linearization has several advantages beyond typology, including offering insight into the left positioning of specifiers, which can be seen as the emergence of an unmarked preference for head finality. Linearizing at PF also allows us to model cases where phonological or prosodic properties of words affect their order; this is illustrated with the case of Irish pronoun postposing.