Heycock, Caroline (1993) Focus projection in Japanese. In Gonzàlez, Mercè, Eds. Proceedings North East Linguistic Society 24, pages 157-171, University of Massachusetts, Amherst.
Full text available as: Adobe PDF
In a Japanese matrix clause the subject of an individual-level predicate obligatorily receives a narrow-focus interpretation, while the subject of a stage-level predicate receives this interpretation only optionally (Kuroda 1965). In embedded contexts, however, subjects of these two types of predicate behave alike in not necessarily receiving the narrow focus interpretation. These phenomena, while well-known, have not yet received a satisfactory general explanation. Diesing 1989 proposes that ``ga'' is a focus-marker, and that the subjects of individual-level predicates are generated in Spec(IP), a position from which focus cannot project, hence the narrow focus interpretation; stage-level subjects are generated in Spec(VP), from which focus can project to yield the wide focus interpretation. This analysis, however, predicts no asymmetry between main and subordinate clauses; further, the proposal that ``ga'' is a focus marker makes its restriction to subject arguments mysterious.
In this paper I argue that nothing has to be specified about the focus properties of ``ga''. Rather, the distribution of the focus readings depends on the information structure of the entire clause. I adopt the proposal of Vallduvi 1990 that there is a level of representation---I[nformation] Structure---derived from S-Structure by applications of Move-alpha, at which sentences are articulated into a structure consisting of the FOCUS and the GROUND, the latter further subdivided into the LINK and the TAIL. The focus is the informative part of the sentence; the ground its complement-knowledge assumed to be part of the hearer's beliefs. Within the ground, the link is the initial constituent that acts as an address under which information is entered; the remainder of the ground is the tail. The only part of the structure that must always be overtly present is the focus. However, since the link provides the address under which new information is stored, only if it is recoverable is an apparently linkless sentence felicitous.
In Japanese, a sentence-initial ``wa''-phrase occupies Spec(CP), and this is a necessary and sufficient condition for an overt constituent to function as the link at I-Structure---the Japanese example is infelicitous because, given the context, the subject should be part of the focus, but cannot be because it is marked with ``wa''. Thus if Spec(CP) is empty, the sentence must either be entirely in focus, or consist of tail and focus, and the link must be recoverable from context. If the predicate is stage-level, the Davidsonian event argument referring to the slice of time and space at which the event takes place is always available as the link, and as a result an entire sentence can be interpreted as being in focus (cf. Erteschik-Shir 1992). If the predicate is individual-level, however, there is no such event argument. Consequently, it is harder to interpret the entire content of such a sentence as focus, giving rise to the preferred interpretation of narrow focus on the subject. Given an appropriate question, however, it is in fact possible to interpret such a sentence as being entirely focus, as I demonstrate. It is the relative rarity of this type of context that gives rise to the heavy favoring of the narrow focus interpretation when such sentences are taken out of context.