Thank you for your reply,
Please tell me if I am off the mark, but I believe the logic goes...
If I am contrasting two variables in an event related design, and I do not want to use a wavered stim. file, then I would want to compare as much of the area under the curves as possible. In this case I would want to use as many time lags in my glt as multicollinearity will allow. Right?

If I am just doing a variable to base line "contrast," where I want to sum the coefficents for each time lag to give the area under the curve, then I would need to use a glt with the 0's for linear drift and 1's corresponding to the # of time lags. (ex. 4 concatenated trials, binary input, max lag of 4, with a glt: 0000000011111). Using a glt that looks like 000000001111100000 will only add unnecessary parameters to my hits vs base line analysis, decreasing its power. Right?

So in theory, the two methods of using wavered input and binary input for comparing variables to base line, can be comparable if we use a glt with the binary input. A glt which gives output that sums the coefficients to give one coefficient(where wavered just has one coefficient to begin with). The wavered is limited to assuming that each BOLD response peaks at the same lag, and the binary with glt will look at every possible time lags permitable, thus compensating for variable BOLD peakage. Right?

I apologize if my questions seem basic and are lengthy. They will help to answer my poorly phrased previous question though, at a theoretical level.

Thank you again soo much for your time and patience!
Jeremy