Restrictive eating is common across the lifespan and eating disorder (ED) severity levels, and associated with negative psychological outcomes. Little is known about functional processes that maintain restriction. Here, we extend research on four-function models (identifying automatic negative, automatic positive, social negative, and social positive reinforcement functions) that have previously been applied to nonsuicidal self-injury (NSSI), binge eating, and purging to restricting. We assessed restrictive eating functions in three samples: transdiagnostic adolescents (Study 1: N=457), transdiagnostic adults (Study 2: N=145), and adults with acute or weight-restored anorexia nervosa (Study 3: N=38). Study 1 indicated the four-function model was a good fit for restricting (RMSEA=0.06, TLI=0.88). This factor structure replicated across Studies 2 (CFI=0.97, RMSEA=0.07, TLI=0.97, SRMR=0.09) and 3 (CFI=0.99, RMSEA=0.06, TLI=0.99, SRMR=0.14). Unlike NSSI, binge eating, and purging, which have been found to primarily serve automatic negative reinforcement functions, all three present studies found automatic positive reinforcement was most highly endorsed (by up to 85% of participants). In Studies 1 and 3, automatic functions were associated with poorer emotion regulation (ps<.05). In Study 1, social functions were associated with less social support (ps<.001). Across studies, automatic functions were associated with greater restriction ps<.05). Functions varied slightly by ED diagnosis. Across ED presentation, severity, and developmental stage, restrictive eating may be largely maintained by automatic positive reinforcement, with some variability across presentations. Continued examination of restrictive eating functions will establish processes that maintain restriction, allowing more precise treatment targeting for these problematic behaviors.