Sister chromatid separation creates a sudden loss of tension on kinetochores, which could, in principle, re-activate the spindle checkpoint in anaphase. This so-called "anaphase problem" is probably avoided by timely inactivation of cyclin B1-Cdk1, which may prevent the spindle tension sensing Aurora B kinase from destabilizing kinetochore-microtubule interactions as they lose tension in anaphase. However, exactly how spindle checkpoint re-activation is prevented remains unclear. Here, we investigated how different degrees of cyclin B1 stabilization affected the spindle checkpoint in metaphase and anaphase. Cells expressing a strongly stabilized (R42A) mutant of cyclin B1 degraded APC/C(Cdc20) substrates normally, showing that checkpoint release was not inhibited by high cyclin B1-Cdk1 activity. However, after this initial wave of APC/C(Cdc20) activity, the spindle checkpoint returned in cells with uncohesed sister chromatids. Expression of a lysine mutant of cyclin B1 that is degraded only slightly inefficiently allowed a normal metaphase-to-anaphase transition. Strikingly, however, the spindle checkpoint returned in cells that had not degraded the cyclin B1 mutant 10-15 minutes after anaphase onset. When cyclin B1 remained in late anaphase, cytokinesis stalled, and translocation of INCENP from separated sister chromatids to the spindle midzone was blocked. This late anaphase arrest required the activity of Aurora B and Mps1. In conclusion, our results reveal that complete removal of cyclin B1 is essential to prevent the return of the spindle checkpoint following sister chromatid disjunction. Speculatively, increasing activity of APC/C(Cdc20) in late anaphase helps to keep cyclin B1 levels low.