Raising insulin acutely in the periphery and in brain improves verbal memory. Intranasal insulin administration, which raises insulin acutely in the CNS without raising plasma insulin levels, provides an opportunity to determine whether these effects are mediated by central insulin or peripheral processes. Based on prior research with intravenous insulin, we predicted that the treatment response would differ between subjects with (epsilon4+) and without (epsilon4-) the APOE-epsilon4 allele. On separate mornings, 26 memory-impaired subjects (13 with early Alzheimer's disease and 13 with amnestic mild cognitive impairment) and 35 normal controls each underwent three intranasal treatment conditions consisting of saline (placebo) or insulin (20 or 40 IU). Cognition was tested 15 min post-treatment, and blood was acquired at baseline and 45 min after treatment. Intranasal insulin treatment did not change plasma insulin or glucose levels. Insulin treatment facilitated recall on two measures of verbal memory in memory-impaired epsilon4- adults. These effects were stronger for memory-impaired epsilon4- subjects than for memory-impaired epsilon4+ subjects and normal adults. Unexpectedly, memory-impaired epsilon4+ subjects showed poorer recall following insulin administration on one test of memory. These findings suggest that intranasal insulin administration may have therapeutic benefit without the risk of peripheral hypoglycemia and provide further evidence for apolipoprotein E (APOE) related differences in insulin metabolism.