The episodic context account (Gaskell et al., 2019) proposes that the act of language comprehension gives rise to an episodic discourse representation, and that this representation is prone to sleep-related memory effects. In three experiments, we tested this prediction by asking participants to read/listen to naturalistic stories before their memory was tested after a 12-hr interval, which included either daytime wakefulness or overnight sleep. To assess discourse memory, we used sentence recognition (Experiment 1; N = 386), free story recall (Experiment 2; N = 96), and cued recall (Experiments 2 and 3; N = 192). We found no evidence of sleep-related effects in sentence recognition or free recall, but cued recall (aka fill-in-the-blank) showed that the degree of time-dependent distortion, as indexed by both a subjective categorisation measure and Latent Semantic Analysis, was lower after sleep than after wake. Our experiments suggest that the effect of sleep on discourse memory may [1] be constrained by the retrieval processes (recollection vs. familiarity & associative vs. item), [2] lie on a qualitative level that is difficult to detect in an all-or-nothing scoring metric, and [3] primarily situate in the textbase level of the tripartite model of discourse processing. Overall, in line with the episodic context account, our findings highlight a specific role of declarative memory and sleep in day-to-day language comprehension, but with limitations on their impact depending on the nature of the retrieval processes.