Background Research has demonstrated that physical activity serves an
important preventive function against the development of cardiovascula
r disease. The recognition that U.S. children are often sedentary, cou
pled with the observation that physical activity habits tend to persis
t into adulthood, has prompted the investigation of exercise determina
nts consistent with social learning theory. The purposes of the presen
t study were to identify social learning variables relevant to childre
n's exercise and to explore the longitudinal predictive value of the d
eterminants. Methods. Data were collected from 111 families (N = 54 gi
rls, N = 57 boys) who were interviewed in both Phase 1 (fifth and sixt
h grades) and Phase 2 (eighth and ninth grades) of this study. Data fr
om mothers (N = 111) were collected during both phases; data from 80 f
athers were collected at Phase 2 only. Results. The results of simulta
neous stepwise regression analyses indicated that child's enjoyment of
physical activity was the only consistent predictor of physical activ
ity during Phase 1. At Phase 2, child's exercise knowledge, mother's p
hysical activity, and child's and mother's friend modeling/support eme
rged as predictors for girls. For boys, child's self-efficacy for phys
ical activity, exercise knowledge, parental modeling, and interest in
sports media were important. Longitudinally, mother's self-efficacy, b
arriers to exercise, enjoyment of physical activity, and child's self-
efficacy for physical activity were important for girls. Only child's
exercise knowledge predicted boys' physical activity. The addition of
information from fathers nearly doubled the explanatory power of the p
redictors for both genders. Conclusions. Socialization in the family u
nit exerts a tremendous influence on health-related behaviors such as
exercise. The relative importance of determinants seems to differ for
girls and boys and the pattern of these determinants appears to change
over time. (C) 1998 American Health Foundation and Academic Press.