We report studies of 20-MHz sound propagation in the A and B phases above the polycritical point, and in the B phase below the polycritical point. Some striking features are observed which were not seen heretofore.The propagation of ultrasound is a powerful probe of the low-temperature superfluid phases of 3 He. 1 Earlier studies of sound propagation along the melting curve 2 and at lower pressures 3 showed a large peak in the attenuation at the normal-superfluid transition accompanied by a sharp, but monotonic, decrease in the sound velocity 3 which appeared to extrapolate to the hydrodynamic first-sound value at still lower temperatures. Sound propagation at the B-A transition has not been studied before at pressures between the polycritical point (PCP) and the melting curve; the phase diagram was established by different experimental probes. 1 ' 4 In this Letter we report the observation of (1) a discontinuity in both the velocity, c, and attenuation, a, at the B-A transition for pressures well above the PCP; (2) a change in slope of the temperature dependence of the velocity just above and below the discontinuity at the B-A transition for pressures just above the PCP; and (3) a change in slope of the temperature dependence of the velocity (suggestive of a divergence with a sign reversal) just above and below the attenuation peak which occurs below the B -to-normal 711 liquid transition at pressures below the PCP.Our measurements were performed in an epoxy adiabatic-demagnetization cell. The cell was precooled through a specially designed mechanical heat switch by the epoxy mixing chamber of our 3 He-4 He dilution refrigerator. The cell and heat switch are described in detail elsewhere. 5 The heat switch was actuated by a bellows which could be expanded (switch closed) using pressurized liquid 4 He. The demagnetization cell contained 23.2 g of cerium magnesium nitrate (CMN) together with a measured volume of 2.4 cm 3 of 3 He (including the sonic chamber). The sonic components were contained in an epoxy plug which was screwed into the bottom of the demagnetization cell; the threads were sealed with Nonaq stopcock grease. The sound path was defined by an 0 o 6-cm-long fused quartz annulus with an outside diameter of 1.27 cm and an inside diameter of 0.95 cm; the diameter of the sound path was further reduced to 0.6 cm by an epoxy annulus of nearly the same length in order to reduce the volume of 3 He in the cell. The ends of the quartz spacer were castellated (to minimize spurious transmission through the spacer and to