Yonelinas et al. (2002) found that hypoxic patients exhibited deficits in recollection that left familiarity relatively unaffected. In contrast, Manns, Hopkins, Reed, Kitchener, and Squire (2003) studied a group of hypoxic patients who suffered severe and equivalent deficits in recollection and familiarity. We reexamine those studies and argue that the discrepancy in results is likely due to differences in the hypoxic groups that were tested (i.e., differences in amnestic severity, subject sampling methods, and patient etiology). Yonelinas et al. examined memory in 56 cardiac arrest patients who suffered a brief hypoxic event, whereas Manns et al. examined a group of severely amnesic patients that consisted of 2 cardiac arrest patients, 2 heroin overdose patients, 1 carbon monoxide poisoning patient, and 2 patients with unknown etiologies. We also consider an alternative explanation proposed by Wixted and Squire (2004), who argued that the two patient groups suffered similar deficits, but that statistical or methodological artifacts distorted the results of each of Yonelinas et al.'s experiments. A consideration of those results, however, indicates that such an explanation does not account for the existing data. All of the existing evidence indicates that recollection, but not familiarity, is disrupted in mild hypoxic patients. In more severe cases of hypoxia, or those with more complex etiologies such as heroin overdose, more profound deficits may be observed.