A session (also referred to as a 'run') is a period of data collection in the scanner. Between sessions, data acquisition is paused, typically to give the participant a rest and enable them to communicate with the experimenter. While the GLM-based analyses in SPM include provision for multi-session data, fMRI timeseries extraction and connectivity analyses such as DCM operate on a per-session basis. It is common, therefore, to concatenate the timeseries from multiple sessions prior to further analysis.
1. Specify a first level design matrix including the image volumes from all sessions. Don't forget to remove any dummy scans if required by your sequence. The tricky part is ensuring your onsets, collated across sessions, are correct. Do check them carefully!
2. Having run the batch to specify the GLM, run the following code in the main Matlab window, changing 'scans' to the number of volumes in each session in the original timeseries:
scans = [60 55 40]; spm_fmri_concatenate('SPM.mat', scans);
This will change your GLM in several respects. It will replace the usual mean column in the design matrix with regressors modelling each session - you can check this using the 'Review' button in the main SPM window. It will also correct the high-pass filter and temporal non-sphericity calculations to account for the original session lengths.
3. Estimate the GLM and add contrasts in the normal way.