We reported previously that the disappearance of cyclin a at the end of mitosis in early Drosophila embryos starts at centrosomes and spreads into the spindle [1], Here, we used a novel mutation, centrosome fall off (cfo), to investigate whether centrosomes are required to initiate the disappearance of cyclin a from the spindle. In embryos laid by homozygous cfo mutant mothers, the centrosomes coordinately detached from the mitotic spindle during mitosis, and the centrosomeless spindles arrested at anaphase. Cyclin a levels decreased on the detached centrosomes, but not on the arrested centrosomeless spindles, presumably explaining why the spindles arrest in anaphase in these embryos. We found that the expression of a non-degradable cyclin a in embryos also caused an anaphase arrest, but most centrosomes remained attached to the arrested spindles, and nondegradable cyclin a levels remained high on both the centrosomes and spindles, These findings suggest that the disappearance of cyclin a from centrosomes and spindles is closely linked to its destruction, and that a connection between centrosomes and spindles is required for the proper destruction of the spindle-associated cyclin a in early Drosophila embryos. These results may have important implications for the mechanism of the spindle-assembly checkpoint, as they suggest that unattached kinetochores may arrest cells in mitosis, at least in part, by signalling to centrosomes to block the initiation of cyclin a destruction.