Objectives: Recent findings of better health outcomes in older caregivers than noncaregivers suggest a healthy caregiver hypothesis (HCH) model may be more appropriate than the stress process model for evaluating the health effects of caregiving. In a cross-sectional study, we tested the HCH on two cognitive domains: verbal memory and processing speed. Method: Participants from the Caregiver Study of Osteoporotic Fractures who had a 2-year follow-up interview were categorized as continuous caregivers (n = 194), former caregivers (n = 148), or continuous noncaregivers (n = 574). The Hopkins Verbal Learning Test (HVLT; memory) and Digit Symbol Substitution Task (DSST; processing speed) were administered at the follow-up interview. Results: Continuous caregivers had better memory performance and processing speed than continuous noncaregivers: adjusted mean scores for HVLT were 18.38 versus 15.80 (p < .0001), and for DSST were 35.91 versus 34.38 (p = .09). Discussion: Results support the HCH model for cognitive outcomes in older women caregivers; however, the relationship may be domain specific.