PURPOSE: This study aimed to assess the influence of anatomical placement of an accelerometer on physical activity energy expenditure prediction in manual wheelchair users. METHODS: Ten accelerometer units (ActiGraph GT3X+) were attached to a multiaxis shaker table and subjected to a sinusoidal oscillation procedure to assess mechanical validity and reliability. Fifteen manual wheelchair users (mean ± SD: age, 36 ± 11 yr; body mass, 70 ± 12 kg) then completed five activities, including desk work and wheelchair propulsion (2, 4, 6, and 8 km·h). Expired gases were collected throughout. GT3X+ accelerometers were worn on the right wrist, upper arm, and waist. The relations between physical activity counts and metabolic rate were subsequently assessed, and bias ± 95% limits of agreement was calculated. RESULTS: During mechanical testing, coefficients of variation ranged from 0.2% to 4.7% (intraunit) and 0.9% to 5.2% (interunit) in all axes. During human exercise testing, physical activity counts at each anatomical location was significantly (P <0.01) correlated with metabolic rate (wrist, r = 0.93; upper arm, r = 0.87; waist, r = 0.73). The SEE for each correlation were 3.34, 4.38, and 6.07 kJ·min for the wrist, upper arm, and waist, respectively. The absolute bias ± 95% limits of agreement values were 0.0 ± 6.5 kJ·min, 0.0 ± 8.5 kJ·min, and 0.0 ± 11.8 kJ·min for the wrist, upper arm, and waist, respectively. CONCLUSIONS: The ActiGraph GT3X+ is a reliable tool for determining mechanical movements within the physiological range of human movement. Of the three anatomical locations considered, a wrist-mounted accelerometer explains more of the variance and results in the lowest random error when predicting physical activity energy expenditure in manual wheelchair users.