I have placed 10 subjects in the scanner and asked them to carry out a tone-paced joystick movement task during two different conditions 'on' and 'off'. I am interested in the change in movement-related activity between on and off. I initially set up my design matrix with each scan in its own session with its own head movement multiple regressors. I then used a simple t-contrast to see the areas that change between the two conditions. I have now also produced a design matrix where I have concatenated the scans into one session with different conditions, one of which paramatrically modulates the boxcar so that the movement blocks occurring during 'on' have a vector +1 and blocks occurring during 'off 'have vector -1.
The second design has more degrees of freedom, but not sure which design is better? Which should I take to the second level?