Rick,
Yes, handling a regressor with all 0's is more a practical issue of scripting and convenience than a statistical concern.
> Michael may be referring to the case of subjects not having data
> at all for some regressors, so an ANOVA might not be appropriate.
> Gang, I guess you still suggest 3dLME.R for that case, yes?
Well, I still prefer 3dMEMA if possible. Admittedly 3dANOVAx/3dLME is advantageous in the sense that you have all the tests done in one shot. However, that convenience aside, most of those tests essentially boil down to t-tests, which can be easily taken care of with 3dMEMA in a piecemeal fashion.
Gang