BACKGROUND: New tools are needed to examine physical activity and the contexts in which it occurs. Community parks contribute to physical activity, but measuring activity and associated variables in them is challenging because area contexts change and the numbers and characteristics of users are highly variable. METHODS: We developed SOPARC (System for Observing Play and Recreation in Communities) and tested its use by observing 16,244 individuals in 165 park areas. Reliabilities included 472 simultaneous measures by independent observers. RESULTS: Correlations between observers on number of area participants was 0.99 for female and male park users. Reliabilities (i.e., percent agreement) for age (89%, females; 85%, males), race/ethnic (80%, females; 82%, males), and activity level (80%, females; 88%, males) groupings met acceptable criteria. Reliabilities for area contexts (i.e., usable, accessible, supervised, organized, equipped) exceeded 94%. CONCLUSIONS: SOPARC is a reliable and feasible instrument for assessing physical activity and associated contextual data in community settings.