Dear Daniel,

If the objective is to replicate the FSL activation-based analyses as close as possible then I would not select any effect there (leave the 'confounding effects' list empty). Then, in the first-level analysis tab enter there any effect that you have entered in your FSL design matrix (e.g. task effects and motion covariates; if you had entered ART scrubbing regressors then select those as well). In that way, the first-level analyses performed in CONN will be most similar to the ones performed in your activation-based FSL analyses. Last, regarding filtering, a 100s high-pass filtered in FSL would be equivalent to a [.01 inf] filter in CONN (note: [0 0.01] would be a low-pass filter instead)

Also, and just reiterating to make sure we are all in the same page, all of this is meant to estimate task-related activation effects (i.e. look at the change in BOLD signal associated with the presence/absence of a given task condition) using CONN in a way that mimics similar task-related activation analyses in FSL. If, instead, you want to estimate task-related connectivity effects (e.g. PPI) in CONN in a way that mimics similar task-related connectivity analyses in FSL, that would be much simpler and the Denoising and other steps options would be very similar to the ones implemented using the default CONN pipeline. 

Hope this helps
Alfonso
Originally posted by Daniel Yang: