The propagation mechanism of very low latitude (geomagnetic latitudes of less than 10--15¿) whistlers is poorly understood. There is a controversy on their propagation; some workers using the observational facts have suggested field-aligned propagation, but some theoretical (ray tracing) works have all indicated nonducted propagation. This paper reexamines the propagation characteristics of nonducted propagation, but we use three-dimensional ray tracing (different from previous works) for realistic ionosphere/magnetosphere models (the electron density profile with latitudinal and longitudinal gradients and the International Geomagnetic Reference Field (IGRF) magnetic field model instead of the conventional dipole model). By assuming small possible tilts (in the latitudinal and longitudinal direction) of the initial wave normal angle in the input southern hemisphere, we have found that it is possible for us to detect simultaneously, at a very low latitude position in the northern ionosphere, one-hop whistler rays started from slightly spaced locations in the south with different initial wave normal angles and that some of them can penetrate through the ionosphere, but some others cannot. On the basis of systematic analysis of important parameters, we come to the general conclusion that it is possible for us to find a closely spaced set of paths to reproduce the one-hop and three-hop whistlers in the north and to have the dispersion ratio of 1:3. The echo train whistlers, as were often observed by Hayakawa et al. [1990>, are realized also by this nonducted propagation without any serious requirements.¿ 1997 American Geophysical Union