We examined whether estradiol and norethindrone hormone therapy (HT) prevented decline in delayed verbal recall in older women with normal to mildly impaired memory functioning. This was a 2-year, randomized, double-blind, placebo-controlled trial of 142 women aged 61-87, randomly assigned to receive 1 mg 17-beta estradiol daily and 0.35 mg norethindrone 3 days/week or daily placebo for 2 years. The primary outcome was short-delay verbal recall of the California Verbal Learning Test (CVLT). To look for differences in response to HT by baseline short-delay recall, we examined the primary outcome in participants grouped according to whether their baseline scores were below average for the age group or greater than or equal to this score and according to whether they met criteria for Mild Cognitive Impairment (MCI) or not. 133 women completed 1 year of the trial and 128 completed 2 years. Prespecified covariates in all repeated measures analyses of covariance (RANCOVA) included age, education, APOE epsilon4, and prior HT use. RANCOVA showed no overall significant treatment effects at year 1 or year 2. After testing for an interaction, which was significant (p=0.02), we found that women in the HT group who scored at or above the average showed significantly less decline than the placebo group in short-delay verbal recall after 1 year, p=0.007 and 2 years, p=0.01. No treatment effects were found in women below the average in either year. When grouped according to whether the participant met criteria for MCI, the interaction between treatment group and MCI subgroup was not significant. These results suggest that benefits of estrogen exposure may be limited to those with average to above average scores on the delayed verbal recall. HT dose and formulation may have contributed to these beneficial outcomes. Replication is warranted before recommendations can be made in the clinical setting.