The aim of this paper is to characterize more precisely the condition when the Japanese focus particle sae is felicitously used in terms of ”unexpectedness” indicated by sae. Paying close attention to the acceptability of sentences of form ”X sae P” (”X” indicates a focused element and ”P” a predicate) in various situations, we reached the following generalization.(i) Sae can be used only when P(X) was unexpected in the sense that the speaker originally believed P(X) to be false, though actually true.When P(X) was the furthest thing from the speaker's mind, sae cannot be used in spite that P(X) was unexpected for the speaker. Thus, in identifying the condition of felicitous usage of sae, it is important to differentiate two types of unexpectedness. We therefore conclude that the function of sae is to declare that the speaker's state of knowledge has changed from ¬P(X) to P(X). Consequently, we propose that the Semantic Representation of ”X sae P” consists of two elements: One is an assertion ”P(X) holds” and the other is a presupposition ”¬P(X) has been deleted”.