Predicting functional recovery is important for the rehabilitation of stroke patients. This study aimed to examine the predictive validity of motor items of Functional Independence Measure (motor-FIM) using the prediction method based on a logarithmic model. This prospective observational cohort study included 111 patients with subacute stroke (mean age: 67.0±13.8 years). The motor-FIM score was assessed at admission, 1 month after admission, and at discharge. We calculated the predictive value of motor-FIM scores at discharge using a logarithmic model, which used only motor-FIM scores at admission and 1 month after admission. The relationship between the actual motor-FIM score at discharge and the predictive value was assessed with the intraclass correlation coefficient. The motor-FIM score at admission, 1 month after admission, and at discharge, and the predictive value were 41.9±15.0, 52.3±18.4, 66.0±18.8, and 65.1±21.2, respectively. The predictive value was highly correlated with the actual motor-FIM score at discharge (intraclass correlation coefficient=0.89). Our results suggest that we can use the prediction method based on a logarithmic model for predicting the motor-FIM score at discharge. This may be useful for designing rehabilitation programs for patients with subacute stroke.