Question: Are inpatients undergoing rehabilitation who appear able to count exercises able to quantify accurately the amount of exercise they undertake? Design: Observational study. Participants: Inpatients in an aged care rehabilitation unit and a neurological rehabilitation unit, who appeared able to count their exercises during a 1-2 min observation by their treating physiotherapist. Measurements: Participants were observed for 30 min by an external observer while they exercised in the physiotherapy gymnasium. Both the participants and the observer counted exercise repetitions with a hand-held tally counter and the two tallies were compared. Results: Of the 60 people admitted for aged care rehabilitation during the study period, 49 (82%) were judged by their treating therapist to be able to count their own exercise repetitions accurately. Of the 30 people admitted for neurological rehabilitation during the study period, 20 (67%) were judged by their treating therapist to be able to count their repetitions accurately. Of the 69 people judged to be accurate, 40 underwent observation while exercising. There was excellent agreement between these participants' counts of their exercise repetitions and the observers' counts, ICC (3,1) of 0.99 (95% CI 0.98 to 0.99). Eleven participants (28%) were in complete agreement with the observer. A further 19 participants (48%) varied from the observer by less than 10%. Conclusion: Therapists were able to identify a group of rehabilitation participants who were accurate in counting their exercise repetitions. Counting of exercise repetitions by therapist-selected patients is a valid means of quantifying exercise dosage during inpatient rehabilitation.