An objection often raised against dip filtering is that it can give data a mixed appearance. Mixed means that adjacent channels appear to have been averaged and that they are no longer independent. This is indeed an effect of dip filtering, and it is inevitable at late times since the horizontal resolving power of reflection data decreases with time. There are two reasons for decreasing lateral resolution. First, dissipation causes high frequencies to disappear. Second, ray bending causes the angular aperture to decrease for deeper sources. It is unrealistic to ignore this fundamental limitation and imagine that adjacent channels should have an appearance of independence. If a mixed appearance is to be avoided for display purposes, then I advocate removing the low-velocity, coherent, signal-generated noise and replacing it by low-velocity, incoherent, Gaussian, random noise. Many plotters lose dynamic range at close trace spacing, and random noise can tend to restore it.