Thanks for the previous reply. I ran 3dvolreg twice on the "same" data.
Just recall what I did:

Step 1: use 3dvolreg on subject1_run1+orig to output motion parameters (dfile named motion.1D)

Step 2: use 3drotate with -1Dfile option to read the motion parameters, motion.1D (say, the output named rotated_subject1_run1+orig )

Step 3: use 3dvolreg on rotated_subject1_run1+orig (dfile named rotated_motion.1D)


If I don't use Step 2, then I can see that running the 3dvolreg twice has no benefits whatsoever; however, earlier Bob Cox mentioned that 3drotate can be used along with the -1Dfile option to read in motion parameters.

Correct me if I am wrong. The purpose of using
"" 3drotate -1Dfile <motion parameters>""
is to correct motion by rotating and shifting each scan to a reference scan, which is the same reference scan that has been used in Step 1.
The motion parameters indicate how much each scan should rotate and shift to match the reference scan. As a result, each scan will have the same position as the reference scan. Therefore, motion is corrected.

Please feel free for any comments and correct any of my misunderstanding.

Thanks again.

Cheers,
Ricky