A little follow up:
I have now realized that all my results are off because when conn has moved the conditions to 1st level covariates it has actually transferred those values only to the first subject, while for all the other subjects the covariates relative to the task are actually empty. In which format should I specify those covariates ? Is a subjects x 1 vector (for each condition in the task) enough, or should I also provide durations ? And if this is the case, what is the format in which I should provide it ?

Alain