Models of neutrino masses are discussed capable of explaining
in a natural way the maximal mixing between $\nu_{\mu}$ and
$\nu_{\tau}$ observed by the Super-Kamiokande collaboration.
For three generations of leptons two
classes of such models are found implying:
a) $\Delta {m_{23}}^2 \ll \Delta {m_{12}}^2 \approx \Delta {m_{13}}^2 $
and a small mixing between $\nu_{e}$ and the other two neutrinos,
b) $\Delta {m_{12}}^2 \ll \Delta {m_{13}}^2 \approx \Delta {m_{23}}^2$
and a nearly maximal mixing for solar neutrino oscillations in vacuum.