>> If I tell the program that each condition is in a seperate run, are
>> the conditons from let's say column 8 still having an impact on column 1?
>
> Yes, it still does. As column 8 is not orthogonal to baseline+drifting for
> that run of data, that would change the fitting of the whole model.

Hmmm... after a second thought, let me clarify a bit: If all the 8 columns are mutually orthogonal with each other, then removing one column would not have effect on the estimation of others. However, this might not be the case in your situation. For example, although column 8 is orthogonal with column 1, column 1 might not be orthogonal with some other columns that are also not orthogonal with column 1, and this would end up with a situation of column 8 having impact on column 1. Is this consistent with your result?

Gang