Summary, in English
Unagi-sentences are, as defined in this thesis, sentences where there is a mismatch between the literal interpretation of the sentence and its meaning in context, often accompanying an apparent violation of selection restrictions of the predicate. This thesis argues that the phenomenon of unagi-sentences shows more variation than previously thought, as well as provides a much needed comparison of the use and underlying structures of unagi-sentences in Japanese and Korean. It also provides new insights into the understudied concept of frame setters. While unagi-sentences are traditionally interpreted as a copula construction, they are shown to be possible with verbal predicates as well. The comparison of unagi-sentences in Japanese and Korean is based on empirical data gathered from experimental acceptability judgment questionnaires targeted at a total of 100 native Japanese and Korean speakers. Interpreting the results, it is claimed that two types of unagi-sentences exist: topic-comment unagi-sentences and frame-setting unagi-sentences. Both copular and verbal predicates of topic-comment unagi-sentences have the property of being time-stable, which is hypothesized to be a result of their forming process. The predicate in a topic-comment unagi-sentence is formed through a process similar to the sentential predicate in multiple nominative constructions, existing in a separate nested IP. As for frame-setting unagi-sentences, their predicates are not restricted in terms of time-stability, and they do not exist in a separate IP. In Japanese, only temporal and locative frame setters can be marked using the topic marker alone, while experiencer frame setters can additionally be marked by the topic marker in Korean. Experiencer frame setters delimit the proposition to the frame which is experienced by its referent, i.e., its vicinity, opinion, etc. Furthermore, it is shown that unagi-sentences in Japanese and Korean can undergo right dislocation, as well as relativization.