Objectives This study sought to clarify changes in both Sense of Coherence (SOC) and social support during one year and causal relationships among upper-grade elementary school children.<br/>Methods A one-year longitudinal survey using a self-rating questionnaire containing measures regarding individual attributes, SOC, and social support was conducted for all 403 pupils from 4th to 6th grade at a public elementary school in a suburban area of Kanagawa prefecture. Latent growth curve analysis was carried out on 237 pupils who completed all three surveys, conducted once in each semester. Firstly, to observe both overall and individual changes in SOC and social support scores, a model was constructed for each score, using the score for each semester as an observable variable, intercept (signifying the score of the first semester) and slope (signifying change of score during a year) as latent variables, and grade and sex as independent variables. Then, models of both were combined to examine causal relationships.<br/>Results For change in SOC score, the mean slope was 0.01 (n.s.) with a variance of 2.85 (P<.05). Thus the score had no change overall, but there were pupils whose scores gradually rose or fell during the year. For change in social support score, the mean slope was −1.25 (P<.05) and its variance was 8.47 (P<.01). Therefore, the score showed an overall decrease, but there were pupils whose scores gradually rose. Grade and sex were not related to change in either score. The intercept of the social support score contributed to the slope of the SOC score (0.44, P<.001), and the intercept of the SOC score contributed to the slope of the social support score (0.34 P<0.05). Thus, a high score of social support in the first semester raised the SOC scores in the following second and third semesters, and a high score of SOC in the first semester raised subsequent social support scores. Grade and sex were not related to these associations.<br/>Conclusion Among 237 upper-grade elementary school children, it was found that SOC overall demonstrated no change and social support decreased over one year. Pupils whose social support was abundant in the first semester had heightened SOC subsequently, and high levels of SOC in the first semester increased social support thereafter. Therefore, it was suggested that an interactive causal relationship between SOC and social support could be demonstrated by latent growth curve analysis.