I agree, although you should use a linear contrast of [ -3 -1 1 3 ] and
not [ -2 -1 1 2 ] as the latter is not linear (there's a gap of +2 between
-1 and 1 but +1 between the others).

All the best,
Mark

On 10 Jun 2011, at 09:48, Stephen Smith wrote:

I would probably recommend option 1, as any nonlinearity of response (which you can then test for with other contrasts) will not damage the quality of the GLM fitting.

Cheers.

On 9 Jun 2011, at 11:03, Annouchka Van Impe wrote:

Dear FSL users,

I'm new to using FEAT and I have a question concerning parametric modulation.

My subjects perform a mental rotation task according to 4 rotation angles (45°, 90°, 135°, 180°). I’m assuming that the BOLD-response goes up with higher rotation angles, seeing as the RTs increase linearly with rotation angle.

What would be the best way to model this parametric modulation?

1)      Using a separate EV for each rotation angle and then specifying a contrast -2 -1 1 2

2)      Using one EV for all rotation angles, and entering the RT in the 3rd column.

        If so, should I demean the RTs?

3)      Using one EV for all rotation angles, and entering 1,2,3 or 4 in the 3rd column

4)      Using one EV for all rotation angles, and entering -2, -1, 1 or 2 in the 3rd column

Thanks for the info.

Kind regards,

Annouchka

---------------------------------------------------------------------------

Stephen M. Smith, Professor of Biomedical Engineering

Associate Director,  Oxford University FMRIB Centre

FMRIB, JR Hospital, Headington, Oxford  OX3 9DU, UK

+44 (0) 1865 222726  (fax 222717)

[log in to unmask]    http://www.fmrib.ox.ac.uk/~steve

---------------------------------------------------------------------------