Three experiments show that, contrary to the current view, comprehenders do not unconditionally deactivate information marked by negation. Instead, they discard negated information when it is functionally motivated. In Experiment 1, comprehenders discarded negated concepts when cued by a topic shift to dampen recently processed information. However, in the presence of a global cue suggesting topic continuity, they retained it, despite a local negation marker that might prompt it. Specifically, when negative statements (The train to Boston was no rocket; Hasson & Glucksberg, 2006) were furnished with relevant (compared to irrelevant) subsequent contexts (The trip to the city was fast though), incompatible meanings ("fast"), related to the affirmative sense of the negative metaphor (rocket), were not suppressed. Instead they were retained and primed related targets (fast) appearing in the late context. Experiment 2 showed that preceding contexts had similar effects, inducing retention of probes related to the affirmative meaning of a negated target. Such effects, however, waned after a lengthy delay (Experiment 3).