Sorry, I already sent this mail, but last time, I forgot to add a subject and I reformulated one question. Dear SPM experts I’m a French doctoral student (forgive me for my poor English) and I’m trying to construct an event-related fMRI protocol: Without being too precise (I know you’re busy), this protocol might be the following : the first part (maybe one run) will be an auditory lexical decision task comprising about 80 items divided in 4 experimental conditions. The second part (maybe one or two runs) will be again an auditory lexical decision task and will comprise 160 items (divided in 8 conditions) which will be, for the half of them, repeated from the first part. My purpose is to evaluate only the differences of activation between these conditions (particularly within the second part) and also to make correlations or regressions (I don’t know for the moment) between theses differences of activation and behavioural measures indexed by the reaction time (RT). My naïve questions are the following: 1) In the case of many conditions (as for my protocol, I think) and with the aim to evaluate differences, is it better during the analyses to have a priori assumptions about the shape of hemodynamic response (HR) (as Friston, 1999) or no assumptions (as Dale, 1999)? In other words, will the efficiency of the design for testing differences differ when one have to estimate the form of the BOLD response or when one have to estimate the parameters modelling the BOLD response? 2) In the perspective of making correlations between activations (or deactivations) and RT, which of this choice (assumptions or not, about the hemodynamic response) is preferable? 3) If I choose to model the signal with a canonical HRF and its partial derivatives, can I make correlations between RT and time-to-peak for example? 4) In the perspective to detect differences in the shape of the HR for different conditions: Must I detect regions commonly activate by all the conditions (and so introduce null events) and after detect differences in the shape between the conditions within the activated regions? Or can I detect theses shape differences between 2 particular contrasts (without introducing null events). For example, is there a difference for time-to-peak between the contrasts [1 -1 0 0] and [0 0 1 -1]? 5) I know that, to evaluate differences of activation between conditions, short SOA are the best, but which distribution of SOA are the best? In the case of multiple trial types paradigm, is it always a random sequence which is optimal? I believe that the SOA must be short and randomly jittered, but is there a constraint on the interval of time for the events belonging to the same condition (TOA)? 6) How can I determine the optimal sequence for the different trial-types order and for the distribution of SOA and TOA? Can I do this with SPM99 or with another Tool before acquiring my data Thanks a lot for you helpful comments and pardon me for my poor English. Pierre Gagnepain GIP Cyceron Inserm E0218-Université de Caen Boulevard Henri Becquerel BP 5229 14074 CAEN France Tel +33(0)2.31.47.02.60