Longitudinal validity of Brief Symptom Inventory subscales was examined in a sample (N = 318) with schizophrenia-related illness measured at baseline and every 6 months for 3 years. Nonlinear factor analysis of items was used to test graded response models (GRMs) for subscales in isolation. The models varied in their within-time and between-times parameter constraints, with the homogeneous model being the least constrained, followed by the 2-parameter GRM and 1-parameter GRM. Results show that 4 subscales (Interpersonal Sensitivity, Hostility, Paranoid Ideation, Psychoticism) were consistent with the 1-parameter GRM, and 5 subscales (Somatization, Obsessive-Compulsive, Depression, Anxiety, Phobic Anxiety) were consistent with the 2-parameter GRM. There is evidence that the 9 subscales may be validly used to study change in single constructs over time.