In 1952 it was shown by one of us that the sources of galactic nonthermal radio emission form a nearly spherical system concentrating toward the plane and the center of the galaxy. It was pointed out independently by the other author that the field between the clouds must be sufficiently strong to retain the cosmic rays in the galaxy. The density of the kinetic energy of the gas between the clouds could be taken as equal to that of the magnetic energy. Thus the velocity dispersion of the rarefied gas in the space between the clouds should be large and form a spherical, not a flat, subsystem. The spherical distribution of the radio emission supports this suggestion. The wide H and K absorption lines appearing in the spectra of early supergiants are also an argument in favor of the reality of fast movements of the rarefied gas. However, L. Spitzer points out that at least a part of these lines belong to the stars. Spectrograms taken by G. Münch show that some of the wide lines consist of a few faint narrow lines. These phenomena may be explained by the density fluctuations of rarefied gas (kv ∝ n2), but some other interpretations are also possible. This phenomenon supports the hypothesis that the more rarefied gas possesses a higher velocity dispersion. The existence of wide H and K lines is not the principal argument of this theory. It will be shown below that the gas of the halo is too rarefied to give observable lines.