BACKGROUND This study aimed to establish the variability in the measurement of habitual physical activity using the ActiGraph® GT3X+ accelerometer in children with cerebral palsy (CP). METHODS Repeated measures: Independently ambulant children with unilateral CP (n = 30; age, 11 yr 3 months (2 yr 4 months)) completed standardized tasks over two consecutive days, wearing an ActiGraph® GT3X+ accelerometer and HR monitor. Testing protocol comprised 5 min of seated rest (REST), walking at light, moderate, and vigorous pace, and rapid stepping on/off a step. Agreement was calculated between days using intraclass correlation coefficients (ICC) (two-factor mixed agreement model). Minimum detectable difference was calculated (minimum detectable difference = [SD√1 - ICC] × 1.96√2). Performance variability: Participants (n = 102) wore an ActiGraph® GT3X+ accelerometer for 4 d in the community. Activity counts were converted into activity intensity using uniaxial-derived cut points to classify the time spent in moderate-to-vigorous physical activity (MVPA). Between-day intraclass reliability coefficients (R) and Spearman-Brown prophecy formula ([ICCdesired/(1 - ICCdesired)][(1 - ICCestimated)]/ICCestimated]) were calculated. RESULTS Agreement between repeated measures was strong for light physical activity and MVPA (ICC, 0.80). For MVPA, the minimum detectable difference was 1412 counts per minute. In the community, 345 d (87%) were recorded. Three days of monitoring produced acceptable variability estimates of MVPA (R = 0.63-0.73). Spearman-Brown prophecy analysis estimated that 3 d would achieve a reliability coefficient of 0.7 and 11 d would achieve 0.9. CONCLUSIONS Measurement of habitual physical activity using the ActiGraph® GT3X+ accelerometer is reliable under controlled walking and stepping conditions as well as in a community environment in independently ambulant children and adolescents with CP.