There is a slightly complex way. Suppose you have a collection of 3D+time datasets that you want to 3dTcat together, then 3dDeconvolve-ize. The problem is that the intermediate 3dTcat-ed dataset is too big. The solution I can offer you is
(a) Use 3dZcutup to cut a 1 slice dataset out of the each of the input 3D+time datasets.
(b) 3dTcat these 1 slice+time datasets together
(c) Analyze the resulting long 1 slice+time dataset with 3dDeconvolve
(d) Do (a-c) repeatedly for each slice, resulting in N output datasets from 3dDeconvolve (1 for each slice)
(e) When done, use 3dZcat to put the 1 slice 3dDeconvolve buckets back together into a glorious 3D bucket dataset with all your results.
If you need help writing a script to do this, post a message. --
bob cox