Lexical Information and Pragmatic Information : Reflexivity of an Event and Resultative Constructions in Japanese

This paper examines the interaction of semantic factors of reflexivity in the availability of result type te-ir and te-ar constructions in Japanese. The semantic factors of reflexivity have been examined in a number of studies, and they are known to be of relevance to the te-ir construction of result interpretation as well as to the availability of the te-ar construction. It has not been made explicit, however, whether the reflexivity is a lexical property of a predicate or it is part of the pragmatic information or the information provided in the actual event in context. I will demonstrate that reflexivity relevant in the two constructions derives from two separate sources; the reflexivity as lexically-encoded information and the reflexivity as contextually-supported. The interaction of lexicon and pragmatics has been studied from various points of view. The definition of lexical information and that of pragmatic information varies across theories. The present study suggests one way of discriminating the two. 1. LEXICAL INFORMATION AND PRAGMATIC INFORMATION It has been of much concern these days how lexicon is to be structured. Lexicon was once thought of as a simple list of words with a limited set of information necessary for syntactic mapping. In the trend of lexlicalism which dates back to early 1980s, more emphasis came to be placed on lexicon. At the same time, syntax was simplified and generalized. The more simplified the syntax was, the more complicated the lexicon had to be, since the overall complexity of linguistic structures was never reduced. For instance, one of the essential parts of lexical information is argument structure. The argument structure used to be more closely tied to subcategorization, which is relevant to syntactic mapping. As argument structure was elaborated in a greater detail, it came to be more associated with event structure of verbs, examining more closely at the semantic side of information than the syntactic side. The event structure is closely tied to the way the event denoted by the verb canonically evolves in the real world. This means that lexical information necessarily draws on extra-linguistic information. It is well-known that the interpretation of a sentence is greatly affected by pragmatic, contextual factors of various kind. However, it is not true that pragmatic factors can do anything. Many sentences are never grammatical in whatever kind of context they are used. In fact, the extent to which pragmatics can override syntactic restrictions is very limited. Pragmatic effect is usually of more modest kind: it, for instance, disambiguates the referent of a pronoun, or it saves a sentence which is grammatically correct but is uninterpretable due to semantic anomaly. There are still cases where pragmatic factors interact more closely with syntactic restrictions. Then the question is how pragmatic information interacts with syntactic and lexical information of constituent words.1 It has been observed that reflexivity of the event denoted by the predicate affects the interpretation of both te-ir and te-ar constructions in a contrastive way ([7], [10], [11], HO; roughly put, reflexivity of the predicate induces result interpretation in the te-ir construction ([10]), while reflexivity blocks result interpretation in the te-ar construction ([7]). What has escaped notice, however, is that those "reflexive predicates" concerned refer to overlapping but incongruent sets. Certain reflexive predicates that induce result interpretation of te-ir construction unexpectedly fail to block result interpretation of te-ar construction. The problem with the foregoing analyses is that they did not properly discriminate the types of reflexivity concerned. By examining the incongruency of the data, I will demonstrate that the semantic reflexivity discussed in the literature actually fall into two types, and that one type of semantic reflexivity resides in the specification of lexical items while the other one in the information of the context. Lexical information and pragmatic information ought to be thus distinguished, since their effects are independent and interactive. 2. ASPECTUAL INTERPRETATIONS OF TE-IR AND TE-AR AND REFLEXIVITY 2.1 Result Interpretation of Te-ir and Te-ar Te-ir construction is composed of a verb stem ending in te and an auxiliary verb ir, '(for an animate being) to exist.' Its meaning is mainly aspectual: it either denotes a progressive state of a process or a resultant state of a process, besides allowing for repetitive, perfect, or experiential interpretations. Of great concern up to present in the literature is the ambiguity between the progressive interpretation and the result interpretation. It has been pointed out that, very roughly speaking, progressive interpretation potentially obtains when the verb in the stem form denotes a durative process, while result interpretation potentially obtains when the process denoted by the verb entails some change of state or position of the theme argument. When both interpretations are potentially available, i.e., when the verb is [+durative, +change in theme], the progressive one prevails unless the verb is in passive voice or the verb is unaccusative ([6], [10]): (1) a. Taroo -ga miti -o hasitteiru (progressive) nom road acc run-ir 'Tom is running on the road.' b. Taroo -ga omotya -o kowasiteiru (progressive) nom toy acc break-ir 'Tom is breaking the toy.' c. omotya -ga kowasareteiru (result) toy nom break-PAS-ir 'The toy is broken.' d. Taroo -ga miti -ni taoreteiru (result) nom road loc collapse-ir 'Tom has collapsed and is lying on the road.' Te-ar construction, on the other hand, is composed of a verb stem ending in te and an auxiliary verb ar, '(for an inanimate thing) to exist.' It also has a mainly aspectual meaning, either result or perfect, besides experiential and repetitive meanings. The following are some examples of te-ar construction in