Nominal arguments, such as the subject and the object are not grammatically required to be overt in Japanese, and are frequently unexpressed, approximately 50% of the time in written narrative texts. Despite this in high frequency of ellipsis, Japanese is not equipped with such familiar devices as the cross-referencing systems and verbal inflections commonly found in pro-drop languages for referent identification. Yet the mechanisms governing argument ellipsis have been little explicated. This thesis elucidates the linguistic mechanisms with which to identify the referents of ellipted arguments. / These mechanisms stem from three tiers of linguistic system. Each sentence is structured in such a way as to anchor the subject., (using Sentence devices following the principle of direct alignment), with argument inferring cues on the verbal predicate (using Predicate devices). These subject oriented sentences are cohesively sequenced with the topic as a pivot (using Discourse devices). These subject oriented sentences are cohesively sequenced with the topic as a pivot (using Discourse devices). It is this topicalised subject which is most prone to ellipsis. I develop an algorithm summing up these mechanisms, using naturally occurring texts. I demonstrate how it can detect the existence of ellipsis in sentences and track the referential identity of it. / A generalisation for ellipsis resolution and the way in which the algorithm is constituted is as follows. Sentence devices formulate sentences to make the subject most prone to ellipsis, discourse devices enable the interaction of wa (the topic maker) and ga (the nominative marker), which mark the majority of subjects, to provide the default reading for referent identification of ellipsis, and predicate devices furnish additional cues to verify that reading. Since Japanese is an SOV language, it is intuitively tenable from the perspective of language processing that the interplay of wa/ga representing subjects gives initial cues from predicate devices. This multiple layering of mechanisms, therefore, can determine referents for ellipted arguments more accurately.

Amongst the multiple and diverse meanings and functions passive constructions hold, this study considers that the primary function of passives in Japanese is to portray an event from the point of view of an affected entity. The thesis identifies three types of affectedness in Japanese passive constructions: emotive affectedness, direct / physical affectedness, and objective affectedness. Emotive affectedness, often referred to as �adversative� meaning, has drawn attention from many researchers. It has been strongly associated in the past with the syntactic category called the �indirect passive�, but is actually also observed in many instances of the �direct passive�. Direct / physical affectedness is detected mainly in the construction here referred to as the �direct sentient passive�. This meaning is common in passives in many other languages, including English. The last type � objective affectedness � is primarily associated with �non-sentient passives�, more specifically with what is here called the �plain passive�. Many previous researchers have claimed a complete and apparently transparent correlation between syntactic and semantic distinctions of the Japanese passive. The present study rejects these direct correlations. In analysing authentic data, it becomes evident that the correlation is much more subtle than has generally been recognised, and that is a matter of degree or continuum, rather than a discrete, black and white issue. To reflect this view, this study proposes separate sets of categories for syntactic and semantic distinctions. The ultimate aim of this study is to reveal how Japanese passives are actually used in real contexts. In order to achieve this aim, detailed examination of authentic written and spoken data is conducted. Some findings of the data analysis in the present study contradict previous claims, such as the finding of a large proportion of passives with a non-sentient subject and very low frequency of occurrence of indirect passives. This research also finds that, although more than half of the propositional meanings in the passive data examined are �negative�, a considerable number of passives still appear in a proposition with a neutral or positive meaning. Another prominent finding regarding propositional meaning is that it seems to be related to the degree of centrality of the passive subject to the event. With regard to the syntactic classification of passive, in particular, it is observed that the lower the degree of the centrality of the subject of the passive to the event, the greater the likelihood that the passive clause involves a negative proposition.

Thesis (Ph. D.)--University of Washington, 1988. / Vita. Bibliography: leaves [211]-214.

Thesis (Ph. D.)--University of Washington. / Bibliography: l. [265]-285.

Ill., Univ., Diss--Chicago.

Thesis (M.S.)--Rochester Institute of Technology, 1992. / Typescript. Includes bibliographical references (leaves 70-73).

Thesis (M.S.)--Villanova University, 2008. / Biology Dept. Includes bibliographical references.

Thesis--Bonn. / Includes bibliographical references (p. 318-324).

Profschrift--Utrecht. / "Stellingen" p. [3] p. inserted. Bibliography: p. [220]-221.

Thesis (Ph. D.)--University of California, Los Angeles, 2004. / Vita. Includes bibliographical references (leaves 214-224).