Sign-bit digital recording means that only the sign of the analog signal is recorded with one bit. In conventional seismic recording, 16 to 20 binary bits are acquired per sample point. The economic advantages of sign-bit acquisition are immediately obvious. Complete amplitude recovery, comparable to full-gain recording, can be achieved by correct application of sign-bit techniques. We describe the amplitude recovery process in a semiintuitive manner to promote the understanding necessary for proper application of the technique. The dynamic range requirements in seismic applications are discussed.Sign-bit digitization is a completely viable technique for recording seismic data, provided that two conditions are fulfilled. First, in real time, the coherent-signal-to-random-noise-ratio must be < or = 1.0. Second, the data must be recorded with sufficient redundancy. Redundancy is achieved by source repetition, sweep correlation, and high-fold common-depth-point stacking, usually in combination. Failure to abide by these two restrictions results in (1) incomplete amplitude recovery, i.e., clipped data, and (2) insufficient dynamic range in the recovered signal.We derive the requirement that the signal-to-noise ratio be less than one; we also discuss the consequences of violating that requirement, namely clipping, at various points in the processing sequence. The amount of information lost is proportional to the degree of clipping; a small amount can be tolerated. Calculated expectation values show that (subject to the requirement that the signal-to-noise ratio be less than 1.0) an unbiased estimator can be chosen. The variance of these estimators is approximately the same as that for full-gain seismic techniques. With sufficient redundancy, the variance can be made as small as necessary to achieve the required dynamic range. With proper attention to these findings, sign-bit digitized data are found to be a totally viable tool.