Background: Mirror therapy (MT) was found to improve motor function after stroke. However, there is high variability between patients regarding motor recovery. Objectives: The following pilot study was designed to identify potential factors determining this variability between patients with severe upper limb paresis, receiving MT. Methods: Eleven sub-acute stroke patients with severe upper limb paresis participated, receiving in-patient rehabilitation. After a set of pre-assessments (including measurement of brain activity at the primary motor cortex and precuneus during the mirror illusion, using near-infrared spectroscopy as described previously), four weeks of MT were applied, followed by a set of post-assessments. Discriminant group analysis for MT responders and non-responders was performed. Results: Six out of eleven patients were defined as responders and five as non-responders on the basis of their functional motor improvement. The initial motor function and the activity shift in both precunei (mirror index) were found to discriminate significantly between responders and non-responders. Conclusions: In line with earlier results, initial motor function was confirmed as crucial determinant of motor recovery. Additionally, activity response to the mirror illusion in both precunei was found to be a candidate for determination of the efficacy of MT.