Predictors of performance on Raven's Progressive Matrices (RPM) and We
chsler Logical Memory (LM) were examined in a longitudinal study of in
itially healthy old people in the HOPE (Healthy Old People in Edinburg
h) study. In total, 387 subjects were tested twice, first (wave I) whe
n their mean age was 75.1 years and again (wave 2) when the mean age w
as 79.3 years. Multivariate regression models were constructed for all
subjects and for only those subjects who remained healthy over the fo
ur year gap between the test sessions (n = 195). RPM and LM were signi
ficantly correlated (r = 0.44, P < 0.001). Demographic, premorbid inte
lligence and blood pressure data collected at wave 1 accounted for up
to 39% of the reliable variance in RPM at wave 2, but only 12% of LM v
ariance. Blood pressure was related prospectively to fluid intelligenc
e but not memory differences. It is suggested: (i) that the concept of
fluid intelligence might be given more emphasis in studies of cogniti
ve aging; and (ii) that there are likely to be individual differences
in brain reserve capacity for different mental abilities. (C) 1998 Els
evier Science Ltd. All rights reserved.