BACKGROUND:Somatosensory loss occurs often following stroke. A proportional recovery model is proposed for spontaneous motor recovery, with implication for treatment planning. It is currently unknown if initial severity of sensory impairment influences stroke survivors' response to treatment to improve sensation. OBJECTIVE:To examine if initial (pre-treatment) severity of upper limb somatosensory impairment is related to sensation outcomes following treatment. METHODS:Regression analysis was used to investigate the relationship between initial and post-treatment sensation performance. Data were pooled from two randomized controlled trials of somatosensory discrimination retraining (N = 80). Upper limb somatosensation was measured using standardized tests of sensory discrimination: Fabric Matching Test, Wrist Position Sense Test, and functional Tactile Object Recognition Test. RESULTS:Post-treatment somatosensory improvement patterns were proportional to the extent of initial pre-treatment somatosensory impairment (Texture discrimination: B = 0.74, 95% CIs [0.52, 0.96]; Proprioception: B = 0.35, 95% CIs [0.24, 0.47]; Object recognition: B = 0.85, 95% CIs [0.75, 0.95]). CONCLUSIONS:The effect of somatosensory retraining on post-treatment sensation was proportional to the extent of upper limb initial somatosensory impairment. Findings suggest sensory retraining can benefit stroke survivors of varying severity of sensory impairment, including those with more severe somatosensory loss.