How fast can f-VEP BCIs ever be?

VEP-based BCIs offer a number of advantages that make them a promising candidate for applications in everyday environments: They do not require user training, the high signal-to-noise ratio of VEPs allows reliable classification, and high information transfer rates (ITRs) of up to ~100 bits/min have been achieved during recent years. In this article we estimate an upper bound of the ITR for VEP BCIs that use frequency and phase coding for classification (f-VEP BCIs). The estimate is based on an idealized classification process that operates on real EEG data of the steady-state (SSVEP) from naive subjects. Our study yields subject-specific upper bounds in the range of approx. 200 to 500 bits/min. We identify causes for the significantly lower ITR of existing f-VEP BCIs and suggest solutions that can narrow the gap to the upper bound.