Vertical stacking of Vibroseis® sweeps for the purposes of reflection data enhancement and data compression often smears sporadic noise bursts into summed traces with degrading effects. One noisy record can easily nullify the signal enhancement benefits of stacking the remaining good sweeps (e.g., Gimlin and Smith, 1980). This paper compares different methods of noise editing before vertical stacking, as applied to deep seismic data collected by COCORP (Consortium for Continental Reflection Profiling) in conditions of heavy traffic noise. Noise editing is of particular concern in crustal seismic work where weak reflections from great depths are often critical to interpretations. In our test, individual sweeps rather than the field sum of these sweeps were recorded separately on tape. This procedure, single‐sweep recording, allowed four computer simulations: (A) recording with no noise editing; (B) recording with a field noise‐rejection (NRJ) system; (C) using a more optimal NRJ system which was computed using the characteristics of the recorded sweeps and applied after recording; and (D) recording using mantissa‐only (MO) format. In this initial, limited experiment, we found significantly greater noise reduction with both tailored, postrecording NRJ systems and MO format than with the simple field NRJ currently employed by COCORP. Normally, these alternative recording methods are not used, either because they entail excessive tape handling and preprocessing, as in the case of single‐sweep recording and application of postrecording NRJ, or because they entail loss of dynamic range, as with MO format recording. However, this study suggests that in the presence of severe traffic noise the benefits of these alternative techniques may outweigh their drawbacks.