Kumar el al. (1991, Chem. Engng Sci. 46, 2483-2489) have shown that in a stirred vessel, size of the largest stable drop, d(max), first increases with phi (holdup of the dispersed phase) at low phi, but decreases with phi at high phi. They have proposed two additional mechanisms of breakage-in shear and elongational flow regions in the front of the impeller blade-that operate along with the hitherto accepted mechanism due to turbulent fluctuations, and conclude that d(max) at high phi is controlled by breakage in shear flows in the range of parameters investigated by them. We show in this paper that their model is deficient on various counts. The new model proposed here overcomes these deficiencies. It predicts that at high phi, d(max) is controlled by breakage in the accelerating flow in the tip region of a rotating blade. The model predicts the data of Kumar er at. (1991) and Boye et al. (1996, Chem. Engng Commun. 143, 149-167). New experiments were also conducted to discriminate between the two proposed mechanisms. The experiments independently confirm that drop breakage at high phi is indeed controlled by accelerating flow. The model could predict the new experimental data also quite well.