The action of strong dc electric fields at high latitudes combined with a predominance of ion-neutral collisions has for some time been expected to create appreciable departures of the ion velocity distribution from the equilibrium Maxwellian configuration. We present new evidence gathered with the retarding potential analyzer on the AE-C satellite that shows that the ion velocity distribution departs significantly from the Maxwellian shape at ion temperatures greater than 1500¿K. From the analysis of more than 50 volt-ampere characteristics we have determined the shape of the ion velocity distribution in the velocity plane perpendicular to the magnetic field direction. Theoretical analysis shows that the use of a relaxation model for the description of ion-neutral collisions gives velocity distributions for both O+ and molecular ions that are qualitatively correct. However, the departures from the Maxwellian configuration are always exaggerated by the theory. An immediate consequence is that the threshold for the onset of the Ott and Farley microinstability caused by the double-hump character of the velocity distribution must be increased to at least 80 mV/m. Finally, if the ion temperature exceeds 2000¿K, we find that the interpretation of the retarding potential analyzer data can be seriously affected by the use of the normally assumed Maxwellian ion velocity distribution in the data inversion process. For very strong electric fields, ion temperatures can be underestimated by more than 2000¿K, and composition ratios can be affected by more than a factor of 2. Other ground-based or in situ measurements should be similarly affected.