Hi Dr. Chen, thank you for your reply.

yeah, there are 23 trials in each condition, and you wrote that "Whether you should use AM1 or AM2 depends on the type of questions you'd like to ask regarding your data. ", does it mean that:

If the item-wise regressors is actually the main effect which is of interest, should I use AM1 instead of AM2? Say three critical conditions are differ in semantic distance, and I used GLT to test the difference between conditions in the past. And now I'd like to find the areas that activation is modulated by semantic distance, should AM1 be used? Should I just need to put the three conditions in ONE "stim_times_AM1"? Does the activaiton map we get refelct the correlation between brain activation and semantic distance? Can we contrast the activation maps which are both resulted from AM1?

When the three conditions differ in semantic distance but semantic distance is not what we are interested in, is it better for us to use AM2 so that we can discriminate the areas which is sensitive and insensitive to semantic distance? In this case, should I define each condition in a separate "stim_times_AM2"?