A key step for analysing the mechanochemistry of mitosis would be the isolation of a functional spindle capable of anaphase chromosome movement in vitro. Although Mazia and Dan first isolated spindles in 1952, with one or two possible exceptions, isolated spindles are non-functional. An alternative approach has used permeabilized cells to study anaphase chromosome movement, but these preparations are biochemically and morphologically complex, and hence difficult to analyse. We describe here a simple procedure for isolating diatom spindles which are capable of anaphase spindle elongation in vitro. With addition of ATP, the two half-spindles slide completely apart, with concomitant decrease in the zone of overlap. Electron microscopy reveals decreased numbers of microtubules throughout the spindle after ATP addition and confirms the complete absence of structures beyond the spindle poles. These results are inconsistent with theoretical models of mitosis which suggest that spindle poles are pushed apart by microtubule growth, are pulled apart by external forces applied to the poles, or are released from tension generated during spindle formation. The results are consitent with models that postulate mechanical interactions in the zone of microtubule overlap as a factor in spindle elongation.