thanks, Vladimir. I tried COH,which is worse and showed activation everywhere. another related question is the source results are written in MNI space by default. how SPM converts from sMRI space to MNI space. SPM didn't ask for AC_PC alignment. Just worry about inaccurate alignment might affect my source reconstruction.

thanks
Jun

On Wed, Oct 3, 2012 at 2:20 PM, Vladimir Litvak <[log in to unmask]> wrote:
Dear Jun,

First, EEG source localisation is imprecise in general and if you want
precise spatial localisation you should rather do fMRI or perhaps
dipole fitting. The purpose of EEG source analysis, especially imaging
approach is to separate effects taking place in different brain areas.
With this said, your result seems quite close to A1 to me, at least
much better than many things I've seen myself.

Another issue is that the MSP approach uses discrete patches and your
solution will be at the nearest patch rather than at the true
location. In some cases this can result in some weird looking
solutions even for very obvious dipolar scalp patterns. This is a
known problem that some people are working to resolve. You can try COH
or MMN option which will give you a much less focal solution but the
peak could be where you expect.

Best,

Vladimir

On Wed, Oct 3, 2012 at 7:57 PM, Jun Wang <[log in to unmask]> wrote:
> Dear Vladimir
>     This is Jun. I asked you some questions about EEG source reconstruction
> a while ago. but I have to approach you again for some problems which have
> puzzled me for one week. I am trying to do  source reconstruction on ERP N1
> for people listening speech sound "ah". What I got is not in primary
> auditory cortex. here is what did.
>
> (1) Coregistration individual polhemus sfp file with individual MRI. the
> fiducials(Nas,Lpa,Rpa) are selected by typing the MNI coordinates found in
> individual MRI ( using display images). also using headshape as recommended
> in SPM manual. see F1.jpg
> (2) In invert step, choose imaging.
> Custom Model with GS inversion.
> Time window(80-120ms) this is the significant window got from sensory
> statistics.
> PST Hanning: NO.
> High pass: 1Hz, Low pass:128 Hz.
> Source priors: No.
> Restrict solutions: No.
>
> the  results model log-evidence is -390.9.
>
> (3)Source results window:80-120 ms, Frequency of interest: 1-50Hz.
>
> Finally I got source reconstruction in F2.jpg. The activation in temporal
> cortex is anterior away from primary auditory cortex.
>
> I want to is there anything I did wrong, why I couldn't get maximum
> activation in primary auditory cortex. really appreciate your help.
>
> thanks
>
> Jun
>
>