The aim of this paper is to find a way to incorporate word semantics into formal semantics. Formal semantics, or truth-conditional semantics as it has often been called, has concerned itself with sentence meaning. Little attention has been paid to word meaning because, as the theory says, it is secondary to sentence meaning in the sense that word meaning, whatever it is, is the word's contribution to the sentence's truth condition. And, as it turned out, it has nothing to say about word semantics.
Truth-conditional semantics has another shortcoming : it is incapable of treating conceptual sentences. As we talk/think not only about real things and real situations but also about imaginary matters and generalizations, this counts as a serious flaw in the theory. Sohmiya (1996) suggested that the theory of formal semantics be expanded so as to be more flexible and more linguistic.
Linguistic formal semantics should then be a theory that is equipped with word semantics and with a way of interpreting conceptual sentences, while maintaining all the good things about truth-conditional formal semantics. This paper concentrates on the first problem, although we come to be aware in the course of discussion that the second problem is closely related to the first.
As I have in mind a large-scale theme "What has man been thinking about language?", I start with a survey of some thinkers' views on word meaning, followed with a brief critique of some present-day approaches to word semantics. The thinkers are : G. Frege, F. de Saussure, C. K. Ogden & I. A. Richards, and L. Wittgenstein. The word theories are : the analysis of semantic relations, componential analysis, prototype theory, and frame theory.
The thinkers' views overlap and diverge. Some of them reappear in present-day word theories. Frege's interest in intension, sense, his claim for the autonomy of linguistic meaning, and his rejection of subjective idea have a close relative in componential analysis. The analysis of semantic relation, including the theory of semantic field, can be traced back to Saussure's structuralism and his indifference to material world. Ogden & Richards emphasized that the thinker/speaker is indispensable for word to have any meaning. Wittgenstein concluded that the meaning of a word is its use in the language community : meaning is private while use is public. O & R's and Wittgenstein's emphasis on user/use and Wittgenstein's "family resemblance" share a permissive attitude toward word meaning and its definition with prototype theory and frame theory.
In conclusion, I take the view that some form or another of componential analysis precedes or underlies the other methods. An important point is that semantic components, some of which I assume are universal, are hierarchically ordered. The primary components are either universal or perceptual. The secondary ones are abstract and conceptual, derived from the primary. The tertiary components signal that the word has first to combine with another word before receiving any kind of interpretation.
Words with primary components undergo real-world semantic interpretation. They are then fed into sentence semantics. Words with secondary components need conceptual semantic interpretation, whose details are yet to be developed in another paper.
By recursively defining semantic components in terms of hierarchy and by relating word semantics to sentence semantics, I succeed in retaining the good things about formal semantics : its formalism and its correspondence to extralinguistic world.
The present suggestion has a further advantage in that the components help predict which thematic role a word should have in a sentence. The primary components predict that the word is very likely to have Event, Agent and other thematic roles, whereas the secondary components signal the impossibility of Event.
Words are "stars" in our cognitive structure. They are connected by our mind's eyes into a constellation, where they are given a name indicating their role in the constellation. Word semantics as is suggested in the present paper thus makes a contribution toward the constellation theory of meaning, which I have been developing since 1995.