Young and Scoville (1982) have argued that the CO distribution, and by inference the H2 distribution, follows the shape of the blue luminosity of the disk, and is thus exponential. However, full maps of bright-CO galaxies (Rickard and Palmer 1981) show considerable structure, with real peaks and depressions on scales as small as the telescope beam. This means that the noise in the determination of the underlying structure is dominated not by the instrumental contribution but by the intrinsic noise of the structure itself. A correct analysis of the axisymmetric structure requires the use of statistical tests comparing the observations with different hypothetical distributions. One finds that equally good fits can be obtained with exponential distributions, r-1 distributions, or even flat disks with central nuclei. However, profiles of the full map data averaged over azimuth, with “error” bars determined by the structural variation with azimuth, show a clear deviation from the optical luminosity profile in the outer disk of NGC 6946. Furthermore, if the profiles for NGC 6946 and IC 342 are fit with exponentials, the scale lengths are rather larger than previously suggested (8.9 and 6.6 kpc, respectively). By sampling only a few radii, one can miss much of the emission of the outer parts of the galaxies, and also underestimate the intrinsic noise of the structure.