In this paper, we argue, mainly on the basis of Japanese data, that we need a theory of focus in which a contiguous sequence of expressions that does not form a morphosyntactic constituent is allowed to serve as a single focus, rather than merely as a sequence of two or more foci, and then present such a theory within the framework of linearization-based HPSG. In the proposed theory, prosodic constituents ("domain objects" in the HPSG parlance), rather than morphosyntactic constituents ("signs"), are claimed to be the principal carriers of semantic information, and meaning assembly is carried out on the basis of prosodic, rather than syntactic, structure. This theory, if correct, means that there is a certain dissociation between the morphosyntactic structure of a sentence and the way its parts are semantically put together.
(S. Yatabe and Seiji Hayakawa, "A linearization-based theory of non-constituent focus," in Frank Richter and Manfred Sailer, eds., Proceedings of the ESSLLI '05 Workshop on Empirical Challenges and Analytical Alternatives to Strict Compositionality, Heriot-Watt University, Edinburgh, 2005, pp. 99-119.)