Yes, I think a T-contast would highlight voxels with a common task-related response. At the second level, a T-test would constrast whether (common) mean activation is significantly greater than zero. That is, you are testing for task-related similarities in your population. On the other side, a F-test at the second level (as you defined) would test if exist task-related differences in your constrast. As you pointed, a F-contrast give common 'differences' in both directions (positive and negative, or task > control/baseline and control > task).

But you can't extent these approximations to F and T tests, because they are different mathematical/statistical tools. If you are new in the matter, just see presentations with experimental fMRI examples to familiarize with it.

Hope it helps ;D,
Martin

2015-06-15 16:16 GMT+02:00 Joelle Zimmermann <[log in to unmask]>:

Hi Martin,

Thanks a lot for your input - that's very helpful I didn't know that! So the F-contrast tests for differences in activations between subjects while T-contrast with '1' would test for similarities in activations between subjects? So you think that a T-contrast would highlight voxels where many subjects show an activation?

I thought that the F-contrast tests for activation and deactivation similarities across subjects. But I'm totally new to this and just getting this from scraping info from various sites.

Thanks,
Joelle

On Mon, Jun 15, 2015 at 4:03 PM, Martín Martínez <[log in to unmask]> wrote:
Hi Joelle,

in the second level, you have modeled a F-contrast (testing for between subjects-related differences).
If you want to test for common-related activations on constrasts, I think you just need to model a T-contrast with a '1'.

Wellcome,
Martin

2015-06-15 10:48 GMT+02:00 Joelle Zimmermann <[log in to unmask]>:
Hi - First of all I'd like to thank everyone who has given me great advice on this mailing list - it's really such a helpful resource!

Now, to my question. I'm setting up a second-level fMRI analysis in SPM. I've previously done 5 first-level fMRI analysis for 5 individual subjects. I am attaching the results of one subject for viewing (results across subjects are quite similar to this). First-level analysis tested for task-related activations.

I now ran a second-level group analysis, forwarding the 5 individual subject con.nii's (ie the con.nii's from the first-level analyses) into the second-level model, specified a one-sample t-test, and set up a simple contrast (an F-test, and just put '1'). However, I didn't get any significant voxels, which I am surprised by, considering all of my first-level single subject analysis look very similar to what I am attaching here.

I think there must be something that I am doing wrong. Any pointers would be helpful.

Thanks,
Joelle