Objective: Recovery of the paretic arm post-stroke can be assessed using observational and self-reported measures. The aim of this study was to determine whether the correspondence (match) or non-correspondence (mismatch) between observational and self-reported improvements in upper limb capacity are significantly different at 0-3 months compared with 3-6 months post-stroke. Methods: A total of 159 patients with ischaemic stroke with upper limb paresis were included in the study. Recovery of arm capacity was measured with observational (Action Research Arm Test; ARAT) and self-reported measures (Motor Activity Log Quality of Movement; MAL-QOM and Stroke Impact Scale Hand; SIS-Hand) at 0-3 and 3-6 months post-stroke. The proportion of matches was defined (contingency tables and Fisher's exact test) and compared across the different time-windows using McNemar's test. Results: The proportion of matches was not significantly different at 0-3 months compared with 3-6 months post-stroke for the ARAT vs MAL-QOM and SIS-Hand (all p > 0.05). In case of mismatches, patients' self-reports were more often pessimistic (86%) in the first 3 months post-stroke compared with the subsequent 3 months (39%). Conclusion: The match between observational and self-reported measures of upper limb capacity is not dependent on the timing of assessment post-stroke. Assessment of both observational and self-reported measures may help to recognize possible over- or under-estimation of improvement in upper limb capacity post-stroke.