Research in schizophrenia and related severe mental illness (SMI) suggests that psychiatric rehabilitation facilitates recovery of psychosocial functioning although there is considerable heterogeneity in outcomes. The present study used growth mixture modeling (GMM), a longitudinal latent variable modeling strategy, to identify classes of psychosocial functioning recovery trajectories. Archival clinical assessment data representing the first 18 months of an inpatient psychiatric rehabilitation program were analyzed from a sample of participants with schizophrenia spectrum disorders (N=162). Based on a GMM analysis of monthly Nurse Observation Scale for Inpatient Evaluation (NOSIE) scores two classes were identified that differed in overall level of psychosocial functioning, rate and nonlinear shape of change. The identified lower functioning group was characterized by poorer neurocognitive functioning at admission, a history of negative symptoms, more previous inpatient psychiatric days, and a longer length of stay in the rehabilitation program. However, this group showed significantly greater positive change in the NOSIE domains directly targeted by behavioral treatment: daily schedule competence, neatness, and irritability. Methodological and theoretical implications of these modeling strategies are discussed in the context of understanding the rehabilitation process.