Attributing Beliefs and Attitudes when the Experiencer is Implicit: The Difference between Discourse and Attitude Adjectives

This paper introduces and provides a linguistic analysis of Discourse Adjectives (Taranto, 2003), a natural class whose members include apparent, evident, clear, and obvious, and contrasts them with another class of proposition-taking adjectives; a class that will be referred to here as “Attitude Adjectives” (absurd, intriguing, ridiculous). The identification Discourse Adjectives, and their contrast with Attitude Adjectives is relevant to the exploration of attitude and affect in text as both classes license implicit experiencer arguments. However, the implicit experiencers licensed are not interpreted in the same way, and the distinction in their interpretations is subtle: the default interpretation for implicit experiencers of Discourse Adjective is the discourse participants, while the default for the implicit experiencer of Attitude Adjectives is the speaker only. This paper presents data illustrating the distinct, yet partially overlapping syntactic distribution of these adjective types, and identifies the characteristics to which any computational model will need to be sensitive in order to accurately attribute attitudes to appropriate discourse entities. While no formal model is proposed at this time, an approximation of one such model is suggested.