Oh. Ok, well, actually the same thing applies.

+ If you have multiple DW factors (for example, using b=500 for some and b=1000 for others), then you have to use the `-bmatr' form/options.
+ If you have just a single b=0 image and a single, DW factor, you can use the `-grad' option/form.
+ If you have multiple b=0 images, you will also have to use the `-bmatr' form, and each one has to get its own row in the file (as in the previously posted example).

Most DWI protocols take multiple b=0 reference images; if you're not going to do any other processing, one commonnon way of going is to average the b=0 images together to form a single reference, and then attach the other DWIs afterward in a 4D file. Thus, while scanning you might have 3 b=0 images and 30 DWIs (making a total, 4D file with 33 subbricks); you could average the first 3 b=0 slices together and then reattach the DWIs, making a new 4D file of 31 subbricks (with the b=0 as the first subbrick or subbrick [0], in AFNI notation). The latter file form with 31 subbricks could then be utilized with `-grad' format of gradients, while the former with 33 subbricks could only be used with `-bmatr'.

--pt