The high thermal anisotropy of O+5 in coronal holes, as observed by the Ultraviolet Coronagraph Spectrometer (UVCS) on the Solar and Heliospheric Observatory (SOHO), suggests that these ions are being heated by the cyclotron resonance. The observations indicate that the O+5 temperature steadily increases from r ≈ 1.5 r S (where the ions become almost collisionless) out to r ≈ 3.5 r S (the outer limit of the observations). Previous models have not been able to reproduce even the qualitative result of a steady temperature rise. We suggest that the problem has been that previous models have considered O+5 resonating only with outward propagating waves. Once the ions are heated perpendicularly to the background magnetic field, they are accelerated to high outward flow speeds by the mirror force. As a result, they resonate with outgoing waves having higher (normalized) wave numbers, where there is presumably less power; they may even drop out of resonance altogether. We suggest here that resonances with inward (i.e., sunward) propagating waves may be the key to explaining the observed O+5 temperature rise. In that case, as the ions are accelerated by the mirror force, they never drop out of resonance, and they resonate with ingoing waves having lower (normalized) wave numbers where there is presumably more power. We offer a very simple strawman model to illustrate the differences between oxygen resonances with ingoing and outgoing waves and to show that the UVCS/SOHO results can be approximately reproduced if the ingoing wave power spectrum in the resonant range varies as k -γ , with γ ≈ 5/3. We point out that it is really necessary to take into account the fact that O+5 (and other heavy ions) can resonate with ingoing and outgoing waves simultaneously, which can only be studied via full kinetic solutions for the ion distribution functions; however, it is possible that once the ions are accelerated by the mirror force, the resonances with sunward propagating waves will be dominant.