Failure to generalize determinants of self-reported physical activity to a motion sensor.

We hypothesized that the prediction of physical activity from psychological determinants would differ when free-living physical activity was assessed by electromechanical or self-report methods. We assessed outcome-expectancy values and perceived barriers for physical activity (OE Values/P Barriers) and physical self-efficacy in 44 young (24 yr +/- 4.8) Caucasian men (N = 13) and women (N = 31). Two weeks later, subjects kept a daily diary and wore a Caltrac motion sensor during all ambulation for seven consecutive days. Reliability (PI) across days was high for MET.h of activity estimated by the diary (0.82) and for Caltrac counts (0.87). Canonical correlation analysis yielded one significant linear combination (Rc = 0.55, P less than 0.01; redundancy = 0.19 to 0.24) of the set of psychological variables [OE Values/P Barriers (beta = 0.76) and physical self-efficacy (B = 0.33)] and the set of physical activity estimates [7-d diary (B = 0.89) and Caltrac (beta = 0.24)]. Multiple correlations indicated that the linear combination of OE Values/P Barriers and physical self-efficacy explained 26% of the variation in MET.h reported in the 7-d diary (P less than 0.01) but was unrelated to Caltrac counts (P greater than 0.05). The strongest bivariate model included the 7-d diary and OE Values/P Barriers (r = 0.51, P less than 0.01). Caltrac counts were weakly related to OE Values/P Barriers (r = 0.34, P less than 0.05).(ABSTRACT TRUNCATED AT 250 WORDS)