Abstract: A method of correcting errors in a digital VCR in which the error correcting capability is maximized and probabilities that no error is detected and data decoding fails are minimized, by using a relatively small length of redundancy. The method comprises the first step of detecting errors from vertical and horizontal sections of a code of data played back from a tape and setting cross-points of the detected errors to flags, the second step of detecting errors from the horizontal code sections of the played back data, correcting the detected errors and erasing only the errors positioned at the cross-points set to the flags at the first steep, in ones of the horizontal code sections from which detectable but non-correctable errors are detected, and the third step of correcting the errors erased at the second in a vertical direction.