The purpose of this study was to investigate whether additional practice of either upper limb or mobility tasks improved functional outcome during inpatient stroke rehabilitation. This prospective, randomised, single blind clinical trial recruited 30 stroke subjects into either an Upper Limb or a Mobility Group. All subjects received their usual rehabilitation and an additional session of task-related practice using a circuit class format. Independent assessors, blinded to group allocation, tested all subjects. Outcome measures used were three items of the Jebsen Taylor Hand Function Test (JTHFT), two arm items of the Motor Assessment Scale (MAS), and three mobility measures, the Timed Up and Go Test (TUGT), Step Test, and Six Minute Walk Test (6MWT). Both groups improved significantly between pre- and post-tests on all of the mobility measures, however only the Upper Limb Group made a significant improvement on the JTHFT and MAS upper arm items. Following four weeks training, the Mobility Group had better locomotor ability than the Upper Limb Group (between-group differences in the 6MWT of 116.4 m, 95% CI 31.4 to 201.3 m, Step Test 2.6 repetitions, 95% CI -1.0 to 6.2 repetitions, and TUGT -7.6 sec, 95% CI -15.5 to 0.2 sec). The JTHFT dexterity scores in the Upper Limb Group were 6.5 sec (95% CI -7.4 to 20.4 sec) faster than the Mobility Group. Our findings support the use of additional task-related practice during inpatient stroke rehabilitation. The circuit class format was a practical and effective means to provide supervised additional practice that led to significant and meaningful functional gains.