The manner in which nulls affect the drifting subpulses of PSE 0809+74 has been investigated. Immediately following a null, subpulse reappear at approximately the longitudes expected for subpulses in the first nulled pulse, as previously reported. The higher quality of the new data, however, has made it possible to demonstrate that the ``phase memory'' is not perfect; there is a small syetematic phase shift in the direction of normal drift. This result is probably inconsistent with the hypothesis that during a null, electron-positron pair production in the polar gap is absent and ``hot spots'' identify the regions over which discharges last occurred. If nulling is the manifestation of temporally steady discharge wihin discrete regions in the polar gap, on the other hand, a slight phase shift is expected, but the observations do not confirm the prediction that its magnitude should be proportional to null length. In addition, this scenario has difficulty explaining the abnormally slow subpulse drift following long nulls and the absence of radio radiaiton during nulls associated with large phase shifts.