Near the eclipse contacts in 1951, radial velocities of 31 Cygni varied in a complicated manner impossible to explain in terms of rotation of the limb of the giant star. Before second contact, most elements departed from the orbital velocity about 5 km/sec negatively. Mter third contact, they also deviated -5 km/sec for the first 45 days, then changed to +5 to +10 km/sec for several days, before finally returning to the orbital velocity. Lines of Ca ii showed similar trends, but with greater amplitude and slower change. From 6 to 2 weeks before second contact, residuals were about -12 km/sec, but fell close to zero for several days before immersion. Luring egress, Can deviated - 10 to - 15 km/sec for 6 days, then changed to +10 to t 15 km/sec a few days after most other elements had changed. After several days more, the lines returned approximately to the orbital velocity. Again, from 5 to 9 weeks after emersion, Ca ii departed as much as + 20 km/sec. These fluctuations were due in part to variation of the relative intensities of two components of a blend. The behavior of Ca II is tentatively assigned to large-scale prominence activity. No interpretation is suggested, at this time, for the fluctuations of the lines formed in lower atmospheric layers.