Hi Dr. McLaren,

    I have A, B, C, and D tasks/conditions in run1 and run2. They are named as A_run1, B_run1, C_run1, D_run1, A_run2, B_run2, C_run2, D_run2 for run1 and run2 respectively in the SPM. mat.
    When constructing P structure, I did as follows:

     P.task={'0','A_run1','B_run1','C_run1','D_run1', 'A_run2','B_run2','C_run2','D_run2', 'x_trans_run1', 'y_trans_run1', 'z_trans_run1', 'pitch_run1', 'roll_run1', 'yaw_run1', 'x_trans_run2', 'y_trans_run2', 'z_trans_run2', 'pitch_run2', 'roll_run2', 'yaw_run2'};

     Provided that I am interesting the PPI effects of A, so I did the contrast as follows:

    P.Contrasts(1).left={'A_run1' 'A_run2'};
    P.Contrasts(1).right={'none'};
    P.Contrasts(1).STAT='T';
    P.Contrasts(1).Weighted=0;
    P.Contrasts(1).MinEvents=1;
    P.Contrasts(1).name='A';

 My question is: 1). Will the contrast combine two runs for condition A? and 2). Are the motion regressors necessary for the P.task? and 3). Do I need to list all the task names in the SPM.mat to the P.task?


 Thank you in advance,

Best,
Chunliang

(1) You should only include actual "tasks" in the task field. The "tasks"
that you list need to match the "task" names in SPM.Sess.U.name fields.

(2) The contrast you specified will average the two tasks.

(3) You should list all the tasks as leaving tasks out will reduce the
model fit. Keep in mind if you don't have enough trials, then modeling all
the trial types may also be bad.

-Donald