OBJECTIVE Recovery of ambulatory function after stroke is routinely assessed using standardized subject- or observer-rated instruments that do not directly measure ambulatory activities in the home-community setting. Accuracy of conventional pedometers in stroke patients is not established, limiting their application in mobility outcomes monitoring. This study investigates the accuracy and reliability of a mechanical pedometer versus microprocessor-based step activity monitoring (SAM) in gait-impaired hemiparetic stroke patients. METHODS Accuracy and test-retest reliability of ankle-worn SAM and belt-worn pedometer were tested directly against hand tallied stride counts and cadence during a battery of timed walks in 16 chronic hemiparetic stroke patients. Patients performed replicate 1-min floor walks at self-selected and fastest comfortable paces, and two 6-min walks on separate days. RESULTS SAM cadence and total stride counts are more accurate than pedometers during 1-min walks at self-selected (99 +/- 1 vs 87 +/- 11.3%, mean +/- SD, P < 0.01); fast pace (98 +/- 2.3% vs 85 +/- 15%, P < 0.01); and repeated 6-min walks performed on separate days (99 +/- 1% vs 89 +/- 12%, P < 0.01). Although SAM is highly reliable (r = 0.97, P < 0.0001) and accurate in all patients under every walking condition tested, the mechanical pedometer demonstrates this high level of accuracy in only half of stroke patients and has poor test-retest reliability (r = 0.64, P < 0.05). CONCLUSIONS SAM, but not the conventional pedometer, provides accurate and reliable measures of cadence and total stride counts in hemiparetic stroke patients. Portable microprocessor-based gait monitoring offers potential to quantitatively measure home-community-based ambulatory activity levels in this population.