From Social Withdrawal to Depression: A Quasireplication and Extension of Boivin, Hymel, and Bukowski (1995)

Tenets of the Boivin et al. (1995) social process model were reexamined with two longitudinal samples using both the original and contemporary analytic strategies. Study goals included reconstructing (e.g., quasireplicating) Boivin et al.'s (1995) original findings and evaluating hypothesized relations across both comparable and longer developmental epochs. Samples included 491 children (245 girls, M-age = 10.0; 80.1% White;19.1% low-, 43.1% middle- or higher-income) followed from grades 4 to 12 and 272 children (148 girls, M-age = 9.61; 84.2% White; 8.2% low-, 17.1% middle-, 74.7% upper middle to higher-income) followed from grades 4 to 5. The assumption that social withdrawal instigates a cascade of within-person changes in the quality of peer relationships, sense of loneliness and social dissatisfaction, and depression was evaluated using Boivin et al.'s (1995) original regression strategy plus two variants of cross-lagged panel models (classical CLPM; Latent Curve Model with Structured Residuals [LCM-SR]). Unlike classical CLPM, LCM-SR allowed for isolating within-person changes and testing hypothesized predictors of within-person increases and decreases. Results differed by type of analysis. Regression and classical CLPM yielded greater substantiation for some of the processes stipulated by Boivin et al. (1995). LCM-SR results, however, called into question the assumption of a cascade effect of early social withdrawal and the reliance on traditional regression and CLPM analyses to test for presumed predictors of within person change.