Objective: The evaluation of rehabilitation success as measured by different tools is becoming increasingly important in terms of time and money allocation. We wanted to know whether functional change in the first week predicts subsequent improvement in a geriatric inpatient rehabilitation clinic. Design: Observational longitudinal study. Setting: Geriatric inpatient rehabilitation clinic in Germany. Subjects: One hundred and sixty-one inpatients (117 women) with a median age of 82 years, capable of walking at baseline. Main measures: Weekly assessments of physical function were performed from admission until three weeks later. We used a self-rated tool (the function component of the Short Form — Late Life Function and Disability Index), a proxy-rated tool (the Barthel Index) and a performance-based tool (gait speed). We set up linear regression models to estimate the predictive capacity of change in physical function within the first week on change in physical function within the following two weeks. Results: Positive correlations were found between functional change within the first week and total change within three weeks. However, correlations of the same periods of change with subsequent change were negative. Correlations were highly significant for both analysis with P-values <0.0001 when the same measures for prediction and outcome were used. Correlations were inconsistent when prediction and outcome were different. Conclusions: Improvement within the first week of inpatient rehabilitation is negatively correlated with subsequent functional change.