The aim of the study was to validate the self-report Multimedia Activity Recall for Children and Adolescents (MARCA) against accelerometry for the assessment of physical activity in New Zealand children. Participants (n = 716, 10-18 years) recalled 3-4 days of activity using the MARCA and underwent a partially overlapping 7-day accelerometry protocol during a national survey. Spearman correlation coefficients (ρ) assessed the association between accelerometer-derived counts per minute and MARCA-derived physical activity level and time in locomotion. Both data sources estimated time spent in light and moderate-vigorous physical activity. Association and agreement between methods for light physical activity and moderate-vigorous physical activity was assessed using correlations and Bland-Altman plots respectively, and paired t-tests conducted. Accelerometer-derived activity counts were moderately correlated with both MARCA-derived physical activity level and locomotion (ρ = 0.38, P < 0.0001). The correlation between methods was -0.14 for light physical activity and 0.28 for moderate-vigorous physical activity (P < 0.0001). The MARCA overestimated moderate-vigorous physical activity compared with accelerometry (120 min, P < 0.0001), which increased as moderate-vigorous physical activity time increased. Some sex and ethnicity (Māori [indigenous] versus non-Māori) differences were observed. Overall, the MARCA indicated moderate validity for assessment of physical activity level, locomotion and moderate-vigorous physical activity and poor validity for assessment of light physical activity. This was comparable to other self-report tools. The MARCA has utility for future large-scale research.