Hello everyone,

I'm running an fMRI study with participants watching short audio-visual clips (4.5 s) and making button-press responses. Videos are divided in four categories/conditions A B C or D.

Pre-processing is quite standard: realign and reslice, slice time correction, normalization, and SMOOTH at 8 mm. 

My concern is that the “estimated smoothness” values - both in the first level and the second level analyses - are VERY LOW: i.e. around 5-7 mm in the first levels and 8 mm in the second level. 

I have nine participants and all show the same pattern in the first level analyses. Also, low smoothness values do not seem to be model dependent (I tried several models, including pooling different conditions and using HRF or FIR).

Data are acquired on a 3T Phillips Achieva, with the following parameters: TR=2100ms TE=30ms Flip Angle = 90 FOV = 192 Matrix = 64x64 voxel Size = 3 x 3 x (2.5+1.25 gap) mm, with a SENSE reduction factor of 2.

I've also noted that when comparing the different conditions with the baseline (rest) the results “look smooth”, but when comparing between conditions the results are much more “grainy/noisy”. Aside the low smoothness, the patterns of activation - also for the direct comparisons between conditions – appear reasonable.

Has anyone had this problem before or have any suggestion on where the problem may be?

Thanks a lot in advance for your help,

Luis.