Bob,
Thanks for the quick response. Basically, I'm basing my last email off of a paper I recently came across by Gary Glover. It is called "Deconvoloution of Impulse Response in Event-Related BOLD fMRI" and it is from NeuroImage (1999). Anyway, I am not completely sure I understand the principle he was getting at, so mostly what I would like to do is improve the model I'm using to account for noise. When I look at the raw data as a graph in AFNI, the BOLD response doesn't just go up and stay at a constant level as one would hope for and expect in a block design. Rather it goes up and it fluctuates up and down a lot until it reaches its max response and then it starts to slowly decrease until the next stimulus block arrives. There's never a plateau of 20 seconds as I had hoped for. Normally I could just smooth the data and that would help with some of the multiple peaks that I'm seeing during one block period, but due to the small area of interest and the lack of a template (I'm not looking at the brain) I can't normalize or smooth my data. I know part of the reason for the fluctionas is the type of task I'm using, and I'm sure part of it really is just noise. So, I'd like to account for the multiple peaks a little bit better (without over-fitting) and was looking to do that through waver. Any suggestions? Thanks

-Nicolas