See these papers for examples of dealing with
RT confounds.

Desai, R., Conant, L.L., Waldron, E., & Binder, J.R. (2006). fMRI of past tense processing: The effects of phonological complexity and task difficulty. Journal of Cognitive Neuroscience, 18, 278-297.

Binder J.R., Medler D.A., Desai R., Conant L.L., Liebenthal E. (2005). Some neurophysiological constraints on models of word naming. NeuroImage, 27, 677-693.

(they are availalbe at [www.neuro.mcw.edu]
under "publications").

Basically, the idea is to include the RT for each trial as a within-condition regressor. If you have two conditions A and B, then you have two binary regressors in 3dDeconvolve coding A and B. Include two additional regressors, A-RT and B-RT, that have the RT instead of the '1' in the A and B regressors respectively. You don't want a single regressor coding RT for both conditions, because that can also account for some non-RT related differences between the conditions. The you can compute a map of areas that are commonly activated by A-RT and B-RT to see the areas that are modulated by RT regardless of the condition. Hope this helps.