Noise in marine magnetic anoamlies becomes significant at wavelengths below about 3 km. Stacking (or averaging) together reasonable numbers of marine profiles in order to enhance two-dimensional anomalies at the expense of noise does not appreciably reduce this critical wave-length. It is shown, however, that such short wavelengths are of negligible importance to the waveform of the anomaly and that, if the noise is random, stacking substantially improves the anomaly-to-noise ratio for longer wavelengths. For non-random noise, such as two-dimensional topographic features, it is essential to compare data from widely separated regions to discern actual paleofield changes recorded by the ocean crust.