> I'm told there's a way to run new contrasts (GLTs)
> that were not specified in the initial 3dDeconvolve
> without having to re-run the whole thing.

Well, this is possible only if you added -xsave option in your previous 3dDeconvolve script. Here is Bob's old note for 3dDeconvolve new options:

# In combination with the old "-bucket bprefix" option, the new "-xsave" option saves the X matrix (and some other information) into file "bprefix.xsave". Use this option when you first run 3dDeconvolve, if you think you might want to run some extra GLTs later, using the "-xrestore" option (below) -- this is usually much faster than running the whole analysis over from scratch.

# The new "-xrestore filename.xsave" option will read the -xsave file and allow you to carry out extra GLTs after the first 3dDeconvolve run. When you use -xrestore, the only other options that have effect are "-glt", "-glt_label", "-gltsym", "-num_glt", "-fout", "-tout", "-rout", "-quiet", and "-bucket". All other options on the command line will be ignored (silently). The original time series dataset (from "-input") is named in the -xsave file, and must be present for -xrestore to work. If the parameter estimates were saved in the original -bucket or -cbucket dataset, they will also be read; otherwise, the estimates will be re-computed from the voxel time series as needed. The new output sub_bricks from the new -glt options will be stored as follows:

* no "-bucket" option given in the -xrestore run
==> will be stored at end of original -bucket dataset
* "-bucket bbb" option given in the -xrestore run
==> will be stored in dataset with prefix "bbb", which will be created if necessary; if "bbb" already exists, new sub-bricks will be appended to this dataset

Gang