It is POSSIBLE to do this analysis with the GAMMA (or BLOCK(1,1)) response function, but it probably won't give significantly different results -- if every event in the block is the same stimulus class. If the events are NOT in the same class, then your estimated amplitudes for the various closely-spaced classes may be highly correlated, since their responses will overlap strongly. You can judge this by looking at the correlation coefficient of the estimates, using program 1ddot on the -x1D matrix output (for example).