Prior research indicates that retirees with less cognitive ability are at greater financial risk because they have lower incomes yet higher medical expenditures. Linking HRS data to administrative records, we evaluate two hypotheses about why this group spends more on health: (1) they are in worse health; (2) they receive more expensive or less effective care for the same conditions. We find that the bulk, but not all, of the cross-sectional relationship can be attributed to the poorer health of those with lower cognitive functioning. Much of this relationship appears to be driven by coincident declines in cognitive ability and health. While, in this respect, the data have important limitations, we find no evidence of substantial differences in care, conditional on observable health.