This systematic review describes the effect of exercise training on physical performance in frail older people. Randomized controlled trials were identified from searches in PubMed, EMBASE and CENTRAL from January 1995 through August 2007. Two reviewers independently screened the trials for eligibility, rated their quality, and extracted data. Randomized controlled trials that examined the effects on performance-based measures of physical function among frail older adults were included. The systematic search identified 20 studies, examining 23 different exercise programmes. The methodological quality score (0-9) of the trials ranged from 2 to 7 points. Sixteen of the studies were scored as high quality. There was a large variety in the studies concerning sample size, degree of frailty, types of interventions and types of assessments. The majority of the programmes were facility-based, group-exercise programmes that were performed three times a week for 45-60 minutes. The intervention programmes comprised resistance training (n = 9), Tai Chi training (n = 2), or multi-component training (n = 12). Six of the total selected 20 studies did not find a beneficial exercise effect on functional performance. This systematic review suggests that older adults with different levels of abilities can improve their functional performance by regular exercise training. To determine the most appropriate design of the exercise programme (type, intensity, frequency and duration of exercise) for functional improvement or prevention of loss of function, more high-quality trials are needed in which different training protocols are compared.