The ability of diazepam, a benzodiazepine full agonist, and imidazenil, a benzodiazepine partial agonist, to protect hippocampal area CA1 neurons from death for at least 35 days after cerebral ischemia was investigated. Diazepam (10 mg/kg) administered to gerbils 30 and 90 minutes after forebrain ischemia produced significant protection of hippocampal area CA1 pyramidal neurons 7 days later. In gerbils surviving for 35 days, diazepam produced the same degree of neuroprotection (70% +/- 30%) in the hippocampus compared with 7 days after ischemia. The therapeutic window for diazepam was short; there was no significant neuroprotection when the administration of diazepam was delayed to 4 hours after ischemia. The neuroprotective dose of diazepam also produced hypothermia (approximately 32 degrees C) for several hours after injection. To assess the role of hypothermia in neuroprotection by diazepam, hypothermia depth and duration was simulated using a cold-water spray in separate gerbils. Seven days after ischemia, neuroprotection by hypothermia was similar to that produced by diazepam. However, 35 days after ischemia, there was no significant protection by hypothermia, suggesting that hypothermia does not play a significant role in long-term diazepam neuroprotection. Imidazenil (3 mg/kg), which produced only minimal hypothermia, protected area CA1 of hippocampus to the same degree as that by diazepam 7 days after ischemia. At 35 days after ischemia, significant protection remained, but it was considerably reduced compared with 7 days. Like diazepam, the therapeutic window for imidazenil was short. Imidazenil neuroprotection was lost when the drug was administered as early as 2 hours after ischemia. The ability of ischemia to produce deficits in working memory and of benzodiazepines to prevent the deficits also was investigated. Gerbils trained on an eight-arm radial maze before ischemia demonstrated a significant increase in the number of working errors 1 month after ischemia. The ischemia-induced deficits in working memory were completely prevented by diazepam but not by imidazenil. There was a significant, but weak, negative correlation between the degree of CA1 pyramidal cell survival and the number of working errors in both the diazepam and imidazenil groups. Thus, if given early enough during reperfusion, both benzodiazepine full and partial agonists are neuroprotective for at least 35 days, but the lack of sedating side effects of imidazenil must be weighed against its reduced efficacy.