The hypothesis was tested of whether neurofeedback training (NFT)--applied in order to increase upper alpha but decrease theta power--is capable of increasing cognitive performance. A mental rotation task was performed before and after upper alpha and theta NFT. Only those subjects who were able to increase their upper alpha power (responders) performed better on mental rotations after NFT. Training success (extent of NFT-induced increase in upper alpha power) was positively correlated with the improvement in cognitive performance. Furthermore, the EEG of NFT responders showed a significant increase in reference upper alpha power (i.e. in a time interval preceding mental rotation). This is in line with studies showing that increased upper alpha power in a prestimulus (reference) interval is related to good cognitive performance.