The objective of the study was to examine determinants of midlife level and long-term changes in a general cog-nitive ability (g) factor. The data were from a Swedish sample of men (n=262;M=49.9years,SD=4.0)forwhich cognitive (conscript) test scores at age 18 were retrieved. In midlife the men completed a battery of cog-nitive tests that was re-administered atfive-year intervals up to 15 years after the baseline assessment. Second-order latent growth curve models were used to examine predictors of midlife level and longitudinal changes in agfactor reflecting four cognitive measures (WAIS-R Block Design, vocabulary, action recall, and wordfluency).The results showed education (years of schooling) to be related to ability level (intercept) before (β= 0.71),but not after (β= 0.09), adjustment of an early adult (age 18)gfactor (reflecting three different cognitive mea-sures)that washighly predictive of midlifeglevel (adjustedβ= 0.89). Neither education norgat age 18 (or mid-lifeglevel) was related to long-term changes ing, though. Conversely,baseline age, BMI, and pulse pressure wereunrelated to midlife ability level, but higher baseline age, higher BMI and higher pulse pressure in midlife werepredictive of cognitive decline. Thus, whereas higher levels of initial ability or educational attainment do not ap-pear to buffer against onset of age-related decline ingin midlife and young-old age, maintenance of lower levelsof pulse pressure and body weight could possibly have such an effect. However, further research is required toevaluate the mechanisms behind the observed relationships of the targeted variables and cognitive decline.