Hi Hedvig,

> So the difficulty, then, would be to distinguish between a
> lack of activation and a deactivation, unless I've
> misunderstood something.

Would you be happy with this (assuming both a_threshold and b_threshold > 0):

expr '(a-a_threshold) + 2*(b-b_threshold) + 4*step (-a_threshold-a) + 8*step (-b_threshold-b)'

And this is how you interpret the results:

No activation and no deactivation only if value at a voxel is 0;
Activation in group 1 iff value at a voxel is 1;
Activation in group 2 iff value at a voxel is 2;
Activation in both group 1 and group 2 iff value at a voxel is 3;
Deactivation in group 1 iff value at a voxel is 4;
Activation in group 2 and deactivation in group 1 iff value at a voxel is 6;
Deactivation in group 2 iff value at a voxel is 8;
Activation in group 1 and deactivation in group 2 iff value at a voxel is 9.

Gang