Human platelet-rich plasma clots, formed with Reptilase, do not retract. Electron microscopy of such a clot reveals platelets of normal shape and ultrastructural features surrounded by fibrin. When ADP is added before Reptilase, the clot strongly retracts. A sequential ultrastructural study of this process shows that, before visible fibrin formation, the platelets form small clusters and show shape change, granule centralization and pseudopod formation. Upon immobilization of the platelets by fibrin, the pseudopods develop into large cytoplasmatic protrusions, which allow cell contact. During the retraction process, the platelet granules remain intact; some granule fusion only becomes apparent in fully retracted clots. When Thrombofax is added to platelet-rich plasma before Reptilase, the clots formed also retract. With this aggregation inducer, granule fusion occurs earlier and in more platelets. Platelet pseudopod enlargement and fibrin concentration around the platelet mass are similar to that in ADP-treated samples. Inhibitors of granule fusion and secretion (suprofen, indomethacin) do not modify ADP-Reptilase nor Thrombofax-Reptilase clot retraction but reduce the incidence of granule fusion in Thrombofax-Reptilase clots. Retraction of Thrombofax-Reptilase clots is unaffected by concentrations of apyrase which completely block ADP-Reptilase clot retraction. Prostaglandin E1, papaverine and amitryptiline inhibit both ADP-and Thrombofax-Reptilase clot retraction, platelet pseudopod formation and cell-to-cell adhesion. These findings suggest that interaction of fibrin with the mass of fused granules (granulomere) or platelet secretion are not responsible for clot retraction. Rather, contraction around adhesion sites formed by cytoplasmic protrusions from adjacent platelets would seem to be involved.