OBJECTIVE: Longitudinal studies of neuropsychological changes in the preclinical phase of Alzheimer disease (AD) have yielded mixed results. Although some studies report tests of episodic memory, others report tests of attention and executive functions as reliable predictors of subsequent AD. Following theoretical models of neuropsychological processes before AD onset, the authors examined the predictive value of attention and executive function in the preclinical phase of AD in old age. METHODS: Authors studied the cognitive performance of 187 initially normal participants of the Berlin Aging Study, a community-based representative sample of Berlin citizens age 70 to 103, over a period of 4 years. Tests of attention and executive function (Digit Letter Test, Trailmaking Part B Test, Digit Symbol Substitution Test, and Identical Pictures Test) and of learning and recall functions (Activity Recall, Memory for Text, and Paired-Associate Learning) were administered at baseline. Diagnosis of AD was made according to NINCDS-ADRDA criteria (probable AD). Receiver operating characteristics curve analyses and Cox regression analyses were used to assess the diagnostic accuracy and predictive value of the neuropsychological tests at baseline for incident AD after 4 years. RESULTS: After 4 years, 15 participants had developed AD. Tests of attention and executive function discriminated best between nonconverters and incident AD cases. A similar pattern was found in survival analyses; attention and executive function tests, together with tests of learning and recall, significantly predicted incident AD over and above age, gender, and education. CONCLUSION: These results support theoretical models of attention and executive function in the preclinical phase of AD in old age.