By the way, is your data scaled to a mean of 100? It seems to be...

So your regression is presumably for a single run with a quadratic baseline,
and 3 regressors of interest. That suggests that you are running waver but
not using the '-peak 1' option, which you should consider using to scale the
regressors (and therefore the betas) to mean a percent change.

As a separate thought, you stated that regressor C is baseline. Are you
actually modeling the baseline with that third regressor? If so, there may
be multicollinearity in your design matrix, since the basline is already
modeled as a quadratic polynomial (or you have 3 runs of constant baseline).

---

After all that babbling, let me try to address your initial question.
Your three betas of interest are very small in the precuneus (wherever that
is... :), relative to those from the ACC. None of the points that have been
brought up address that, even multicolinearity.

Since they are _relatively_ small, how you are scaling things really does
not matter (unless regressors were scaled very differently say, which does
not appear to be the case). And even if you were modeling the baseline in
multiple ways (such as if A+B+C = constant), that would still have to come
out with A, B and C canceling each other by being bigger than they should
be, perhaps with opposite signs. Since they are all small, that is not the
case either.

So it looks to me like that average response there is simply small. One way
that can happen is if your fixation stimulus also activates the region, and
so it is active for the entire run. Or perhaps it is a very small region (I
*could* look it up, but that would take seconds), and you have applied a large
blur which bleeds it away.

That's all that comes to mind at the moment.

- rick