A signal processing method for recording and playing back recorded data with decreased detection errors originating from residual marks characteristic of an overwrite medium is performed by 8/9 modulating input data, and by inverting the polarity of the modulated data string at each "1" bit to produce a DC-free recording pulse. The thus-recorded signal is played back through a playback equalizer 8 before being converted by an A/D converter 9 into digitized amplitude data. The digitized data are processed to determine an identification level, which identification level is stored as well as used to determine displacement quantities for quantized bits derived from the analog playback signal. The number of "1" bits is counted by a counter 20, and an error in the number of "1" bits is detected by a comparator 21. An error bit decision circuit 22 determines which bit is most likely to be erroneous on the basis of displacement quantity, and the likely erroneous bit is corrected prior to outputting the data for 8/9 demodulation.