The performance of electroencephalogram (EEG)-based brain-computer interfaces (BCIs) still needs improvements for real world applications. An improvement on BCIs could be achieved by enhancing brain signals from the source via subject intention-based modulation. In this work, we aim to investigate the effects of task complexity on performance of motor imagery (MI) based BCIs. In specific, we studied the effects of motor imagery of a complex task versus a simple task on discriminability of brain activation patterns using EEG. The results show an increase of up to 7.25% in BCI classification accuracy for motor imagery of the complex task in comparison to the simple task. Furthermore, spectral power analysis in low frequency bands, alpha and beta, shows a significant decrease in power value for the complex task. However, high frequency gamma band analysis unveils a significant increase for the complex task. These findings may lead to designing better BCIs with high performance.