This paper presents a quantitative analysis of the use of the respect language of Javanese in a selection of short stories. Variation in this usage is shown to be conditioned by factors within three domains: interactional, sentential, and lexical. Because the effects from the different domains are at least partially independent, they may sometimes conflict, in which case one or another may take precedence. The relationship among the factors is thus quantitative, not categorical. The tendencies observed are, however, strong enough to have a significant effect. A program for further research along these lines is outlined.