A new dayside source of O+ ions for the polar magnetosphere is described, and a statistical survey presented of upward flows of O+ ions using 2 years of data from the retarding ion mass spectrometer (RIMS) experiment on board DE 1, at geocentric distances below 3 RE and invariant latitudes above 40°. The flows are classified according to their spin angle distributions. It is believed that the spacecraft potential near perigee is generally less than +2 V, in which case the entire O+ population at energies below about 60 eV is sampled. Examples are given of field-aligned flow and of transversely accelerated “core” O+ ions; in the latter events a large fraction of the total O+ ion population has been transversely accelerated, and in some extreme cases all the observed ions (of all ion species) have been accelerated, and no residual cold population is observed (“toroidal” distributions). However, by far the most common type of O+ upflow seen by DE RIMS lies near the dayside polar cap boundary (particularly in the prenoon sector) and displays an asymmetric spin angle distribution. In such events the ions carry an upward heat flux, and strong upflow of all species is present (H+, He+, O+, O++, and N+ have all been observed with energies up to about 30 eV, but with the majority of ions below about 2 eV); hence, these have been termed upwelling ion events. The upwelling ions are embedded in larger regions of classical light ion polar wind and are persistently found under the following conditions: at geocentric distances greater than 1.4 RE; at all Kp in summer, but only at high Kp in winter. Low-energy conical ions (<30 eV) are only found near the equatorial edge of the events, the latitude of which moves equatorward with increasing Kp and is highly correlated with the location of field-aligned currents. The RIMS data are fully consistent with a “mass spectrometer effect,” whereby light ions and the more energetic O+ ions flow into the lobes and mantle and hence the far-tail plasma sheet, but lower-energy O+ is swept across the polar cap by the convection electric field, potentially acting as a source for the nightside auroral acceleration regions. The occurrence probability of upwelling ion events, as compared to those of low-altitude transversely accelerated core ions and of field-aligned flow, suggests this could be the dominant mechanism for supplying the nightside auroral acceleration region, and subsequently the ring current and near-earth plasma sheet, with ionospheric O+ ions. It is shown that the total rate of O+ outflow in upwelling ion events (greater than 1025 s−1) is sufficient for the region near the dayside polar cap boundary to be an important ionospheric heavy ion source.