We administered hyperbaric oxygen or air in a double-blind prospective protocol to 39 patients with ischemic cerebral infarction. We interrupted the study when we noticed what appeared to be a trend favoring the air-treated patients, whose neurological deficits were less severe (mean±SEM score on graded neurological examination: air, 25.6±4.9; oxygen, 34.5±7.5) and whose infarcts were smaller (air, 29.0±12.2 cm 3 ; oxygen, 49.2 ±11.7 cm 3 ) at 4 months. The trend, we decided, was probably an artifact of the randomization process. Nevertheless, we chose not to resume the trial because the treatment was difficult to administer by schedule (for various reasons the treatment protocol was broken in 15 of the 39 patients), was poorly tolerated (eight of the 39 patients refused to continue treatments), and did not produce dramatic improvement (Stroke 1991^2:1137-1142)