OBJECTIVE: Bilateral transcranial direct current stimulation (TDCS) is superior to unilateral TDCS when targeting motor learning. This effect could be related to either the current flow direction or additive polarity-specific effects on each hemisphere. METHODS: This sham-controlled randomized study included fifty right-handed healthy subjects in a parallel-group design who performed an exoskeleton-based motor task of the proximal left arm on three consecutive days. Prior to training, we applied either sham, right anodal (a-TDCS), left cathodal (c-TDCS), concurrent a-TDCS and c-TDCS with two independent current sources and return electrodes (double source (ds)-TDCS) or classical bilateral stimulation (bi-TDCS). RESULTS: Motor performance improved over time for both unilateral (a-TDCS, c-TDCS) and bilateral (bi-TDCS, ds-TDCS) TDCS montages. However, only the two bilateral paradigms led to an improvement of the final motor performance at the end of the training period as compared to the sham condition. There was no difference between the two bilateral stimulation conditions (bi-TDCS, ds-TDCS). CONCLUSION: Bilateral TDCS is more effective than unilateral stimulation due to its polarity-specific effects on each hemisphere rather than due to its current flow direction. SIGNIFICANCE: This study is the first systematic evaluation of stimulation polarity and current flow direction of bi-hemispheric motor cortex TDCS on motor learning of proximal upper limb muscles.