Objective: To develop an adaptive measure, the Stroke Arm Ladder, to parsimoniously quantify upper extremity function post stroke. Setting: Various studies conducted in Montreal and elsewhere in Canada. Subjects: A total of 4058 persons with stroke who participated in various studies were assessed on different occasions post stroke. Assessments spanned from 3 days to 1 year post stroke. Main measures: A calibrated bank containing 49 items from tests and indices designed to assess global motor recovery of the upper extremity, upper extremity function and activities involving the upper extremity was used. Data including all testing occasions were analysed to test whether they adhered to the expectations of the Rasch partial credit model and whether item hierarchy remained stable across testing occasions. Results: Fifteen items did not meet the requirements of the Rasch model and were deleted. An adaptive measure of upper extremity function, the Stroke Arm Ladder, containing 34 items was created. The easiest item was partially executing the bilateral task of tying a scarf around one’s neck, while the hardest item was being able to transfer more than 60 blocks on the Box and Block Test. The items’ difficulty thresholds cover a wide range of difficulty levels from −7.4 to 6.2 and the global fit statistics (χ2: 331; probability: 0.18), confirmed content validity. Conclusion: The Stroke Arm Ladder demonstrates adequate initial psychometric properties. Further testing of the measure in its adaptive format is necessary before it can be used clinically.