ipm and resulted in the curves shown in Fig. 11; similar results were obtained at 400 ipm. All specimens were compressed to approximately 55 percent reduction in height which increased the final D/H ratio to 3.3 times its initial value. For tall specimens of D 0 /H 0 -0.5, end-face expansion was negligible; ii became significant only at values of D o /H 0 = 1 (D/H = 3.3) for both lubricated and unlubricated conditions.Although end-face expansion began when the final diameter/ height ratio (D/H) was approximately 3, it did not approach the theoretical maximum even for lubricated conditions until the initial value, D a /H 0 , was equal to 3. In the latter circumstances, end-face expansion was about 80 percent of the theoretical maximum (Fig. 11) and gave a surface appearance which indicated very little folding over of the sides. Thus deviation from homogeneous deformation under these conditions is due to barreling only and not to folding over of the sides.Lubrication, therefore, plays a significant part, in increasing the expansion of the end faces and the homogeneity of deformation, although expansion still begins only after a diameter/height ratio of approximately 3 is reached.Although not shown here, it is worth mentioning that the glass was sarprisingly inefficient in promoting end-face expansion. Apparently, the squeeze-film that formed on impact became sealed in at the periphery of the billet end face and acted essentially as an immobile mass. Pheripheral breakdown of squeeze-films was previously noted [13] and was probably accentuated in the present work by cooling of the glass on the colder anvils.