Several theoretical models identify perceived competence as a mediator of future choice of participation and persistence in activity (Eccles et al., 1983; Harter, 1978). Research with children has been conducted primarily in academic and formal sports settings where actual choice may be somewhat influenced by parental or contextual restrictions in which the activity takes place. Such research may not accurately reflect the variable of free choice. The purpose of this study was to determine the relationship between children's perceptions of competence and importance and actual engagement on a variety of playground activities during recess. A secondary purpose was to determine the stability of perceptions of competence. Gender differences were also investigated. Seven males and 7 females in grades 3 and 4 participated in the study. Perceptions of competence and importance were assessed for 54 potential activities three times over a 3-week period. Activities actually engaged in were assessed twice daily through the use of a self-report form completed immediately after each of 24 recess periods using procedures described by Watkinson and Causgrove-Dunn (2000). The validity of self-reports were checked against actual observations. Repeated measures ANOVAs with follow-up dependent and independent t-tests were used to analyse relationships between perceptions of competence and degree of activity engagement. Results indicated that children took part in activities in which they held higher perceptions of competence (p< .05) and which they considered more important (p< .01). This relationship was even stronger when activity behavior was measured over longer time periods, when perceptions of competence also discriminated between various levels of engagement (frequent vs. infrequent). It was also found that, for both genders, perceptions of competence were stable over the 3-week period, indicating a lack of serious fluctuation for this construct. This is important information that has implications for measuring such perceptions during research. Analyses of gender differences revealed that generally boys indicated higher mean perceptions of competence (p< .05) than girls and that these differences were particularly evident for activities that were not done. In summary, the findings provided support for the theories of Eccles and others concerning the relationship of behavior to perceptions of competence among children in a free choice setting. This has major implications for understanding, and perhaps intervening to expand, the activity choices of children.Keyword(s): . NA