One of the first to present a working BCI based on EEG-signals was Emanuel Donchin from Beckman Institute for Advanced Science and Technology, Illinois, in 1988. Unlike most of his colleagues, he used an EEG-signal called p300. It belongs to the event-related potentials, that are also used by the Berlin group. Like the Berlin BCI, Donchin's BCI does not require the user to perform months of biofeedback training initially, as the p300 is "naturally" occuring as a brain reaction on unsuspected or "unnormal" events. Some people think, though, that the p300 is not independent from ocular motor activity, which would make it less suitable for severlely paralyse people.
The apparatus used differs a bit from the other EEG-based systems, as it has to incorporate "surprise". In fact, the user has to count how often a specific choice (a letter, say) is highlighted (more). Doing so he produces a prominent p300-wave, that can then be transformed into information. FIrst experiments seem to show that a p300-based BCI is quicker than most other EEG-based systems, but more research is needed (more).