Objective: To determine whether older cognitively impaired people benefit from physical training with regard to motor performance or fall risk reduction and to critically evaluate the methodologic approach in identified randomized controlled intervention trials. Design: Published randomized controlled intervention trials from 1966 through 2004 were identified in PubMed, CINAHL, Gerolit, and the Cochrane Central Register of Controlled Trials according to predefined inclusion criteria and evaluated by two independent reviewers using a modified rating system for randomized controlled intervention trials developed by the Cochrane Library. Results: There were 11 randomized controlled intervention trials that met the predefined inclusion criteria. There was a large heterogeneity regarding methodology, sample size, type of intervention, study outcomes, and analyses. We found conflicting evidence regarding the effect of physical training on motor performance and falls in older people with cognitive impairment. However, a considerable number of the studies had methodologic limitations, which hampered the evaluation of the effectiveness of training. Conclusions: The randomized controlled intervention trials showed only limited effectiveness of physical training in patients with cognitive impairment. More studies with adequate sample size, sensitive and validated measurements, and higher specificity for the types of intervention targeting subgroups of patients with different degrees of cognitive impairment are required to give evidence-based recommendations.