A digital video tape recorder has interpolation and crossfading circuits and records audio and video signals in independent frames, the audio frames being offset with respect to the video frames. For a four-channel audio signal, the even and odd samples are recorded near different edges of the tape, enabling burst errors near one edge to be recovered efficiently by Interpolating the audio signal from the other edge. When audio dubbing is performed, the first and last dubbed frames contain the even samples of the new signal and the odd samples of the old signal, or vice versa, and audio-dubbing flags are recorded in the these frames. When these frames are played back, the old and new signals are regenerated by interpolation and crossfaded to create a smooth transition between them. The same apparatus can also be used to record a two-channel signal, by recording each sample twice.