INTRODUCTION: The increasing number of restorative stroke trials has generated increased discussion about optimal endpoints to detect changes in behavior over time. Robotic devices are under study as a therapeutic intervention, and in some cases the design of these devices might also make them useful for assessing behavioral status. If true, similar devices might be useful for remote assessment of behavioral status after stroke, i.e., tele-outcome. As an initial step, the current study examined the validity of 3 robot-based behavioral assessments. METHODS: Data were collected at baseline for patients enrolled in a study of robotic therapy for arm weakness after stroke (clinicaltrials.gov # NCT01244243 ). Entry criteria include age >18 yr, arm motor deficits, and stroke 3-6 months prior. Pre-treatment assessments included the Fugl-Meyer Arm Motor Scale (FM-total), which has often been used as a behavioral endpoint in clinical trials, as well as 6 secondary, related tests: the Box/Blocks (B/B) Test, Action Research Arm Test, the hand/wrist subsection of the FM-total, hand motor subscale of Stroke Impact Scale-2, affected hand grasping force, and affected hand pinching force. These 7 tests were compared with score on robotic games that require targeting movements by wrist or by finger and that therefore emphasize movement speed and accuracy (using Bonferroni-adjusted p<0.007). In addition, score from the live version of the B/B test was compared with score from a robot-based version of the B/B test. RESULTS: The 28 subjects were age 59 +/- 14 yr (mean +/- SD), stroke affecting R side in 14 and L in 14, and had moderate-severe impairment (FM-total scores = 34 +/- 15, range 14-60). Robot-based assessments were rapidly and successfully obtained in all subjects. Performance on the robot-based test of wrist targeting correlated significantly with baseline FM-total score (r = 0.73, p<0.0001) as well as all 6 secondary tests (r = 0.54 - 0.83, B/B score being highest). Performance on the robot-based test of finger targeting also correlated significantly with baseline FM-total score (r = 0.69, p<0.0001) as well as all 6 secondary tests (r = 0.52 - 0.70). Values of the live version (real blocks) and the robot-based version (virtual blocks) of the B/B test were correlated (r=0.48, p<.02). While 10 subjects demonstrated a floor effect on the live B/B test, only 2 subjects had a floor effect with the robot-based B/B test. CONCLUSION: The current findings support the validity of robot-based methods for assessing upper extremity motor function in subjects with chronic stroke. Potential advantages of this approach as compared to traditional scales include simplicity, speed, and fewer subjects with a floor effect. In addition, the overall approach described here suggests the utility of tele-outcome assessments after stroke.