Dear John This should be no problem, just choose to specify your onsets in seconds and enter the actual values. SPM will then insert a stick function at the corresponding timebin (the default setting is 16 timebins pr. TR), convolve the timeseries with stick functions with a square with a width of the stimulus duration, and finally convolve the result with your choice of basis set e.g. canonical hrf. In this way SPM can account for differences between stimulus onsets and TR's. Best Torben Torben Ellegaard Lund Associate Professor, PhD The Danish National Research Foundation's Center of Functionally Integrative Neuroscience (CFIN) Aarhus University Aarhus University Hospital Building 10G, 5th floor, room 31 Noerrebrogade 44 8000 Aarhus C Denmark Phone: +4589494380 Fax: +4589494400 http://www.cfin.au.dk [log in to unmask] Den Uge:45 09/11/2011 kl. 08.22 skrev John Gelburg: > Hi all, > > Suppose, my trial duration is 4 seconds and the TR=2.5 seconds. How bad is it that some of my trials starts at the middle of a TR? Clearly, the SPM onset matrix can coop with it, but how bad it is from model estimation point of view? > > Thanks, > John