IBL behaves comparably to neural network approach and does better than it for unseen signers. The scale of the problem is bounded, however and the system does not have to be built in a scalar manner, since there will always be 31 handshapes. C4.5 is not competitive.
The difference in performance between the PowerGlove and CyberGlove is significant, as would be expected. This becomes an issue if GRASP was to be expanded to things such as finger-spelling.