Is it ever possible to separate two signals (events) that have never have any gap between them? For example, the onset of a face stimulus (event 1, 700ms) followed by an apparent motion facial feature change (event 2, 1000ms). I'm aware that it makes the two events highly collinear in the design matrix (so the answer to my question is probably no) but are there any 'tricks' that can be used to allow a good/valid estimation of the two signals? For example, jittering the duration of one or both of the events to try and introduce more variance between them?
Note that, the intended ER design has more factors than described here and I have attempted to optimize it for these other factors using a pseudo-randomized ordering and varying ISI. I am actually only (more) interested in event 2 and will never contrast events 1 and 2. However, based on ERP data, combining 1&2 into a single event seems like a bad idea as, for example, event 1 (remember I'm more interested in event 2) can have larger peak amplitudes. I think in the scanner they will display different BOLD and I would like to be able to get at the BOLD of event 2 with as little 'contamination' from event 1 as possible.
All insights greatly appreciated.