The vertical directionality of ambient low‐frequency (shipping) noise at midlatitudes has been measured by Wales and Diachok [S. C. Wales and O. I. Diachok, J. Acoust. Soc. Am. 70, 577–582 (1981)] in the Atlantic and by Williams and Fisher (R. B. Williams, personal communication. The measurements were taken in October 1979) in the Pacific. At the sound axis the angular distribution in intensity is flat from 0° (horizontal) up to the inclination of the surface‐limited rays. The intensity falls off sharply for steeper rays (which are associated with lossy bottom reflections). This observed flat angular distribution is in contrast with the bimodal distribution (peaked near the inclinations of surface‐limited rays and zero for flat rays) which is expected from surface sources in a range‐independent environment. A flat angular distribution is equivalent to an equal partition of energy among modes; once such equipartition is attained, the distribution remains unaltered with propagation (or scattering) in a lossless channel. The simplest model consists of surface sources generating rays of steep axial inclination, which are then scattered into flat rays by internal waves throughout the ocean volume. The model fails because of the weakness of the scatterers, requiring 104 km for equipartition. There are two regions for the noise energy to couple directly into flat (near‐axial) rays: (i) at high latitudes where the rising sound axis intersects the surface, and (ii) at the continental slope where the sound axis intersects the rising bottom. The second model consists of sources in the high latitude duct (where the sound axis is essentially at the surface), with energy broadly partitioned among modes from the very start (dipole sources in a linear sound speed gradient give precisely equipartition). Subsequently, the energy is ducted downward along the axis into midlatitudes, preserving equipartition. This model fails because of the low density of shipping at high latitudes. In the third model, the steeply descending rays from coastal shipping are bounced into flat near‐axial rays over the continental slope, as proposed by Wales and Diachok. Although there is some mystery as to how this will lead to the observed high degree of mode equipartition, we conclude (reluctantly) that reflections at the continental slope are probably the dominant mechanism for getting energy into the flat rays.