This paper investigates modal marking in conditionals with respect to two research questions: (1) How tightly integrated are conditional sentences, both relative to similar adverbial clause constructions, and with respect to different variants of conditional constructions. (2) What are the pragmatic biases and discursive patterns, if any, that motivate conditional constructions with modal marking. These issues are investigated with the help of a large corpus of Modern Japanese. The data suggest that (1) conditional sentences are relatively tightly integrated compared to causal, and probably also concessive constructions, but there are considerable differences between different types of conditional constructions, and (2) modally marked conditional sentences are overwhelmingly associated with deontic speech acts. A number of discursive patterns associated with spoken language can be identified, some of which are probably cross-linguistically replicable.