> But I got 34 sub-bricks. Didn't know what's wrong.
It sounds correct to me with 34 sub-bricks. You can look at the sub-brick labels to see what exactly they are.
Again 3dMVM is a better choice in your case.
> There are two groups(Chinese and English), each subject read and listen the story in their own language.
OK, you have one within- and one between-subjects factor. However, 3dLME shows the following
Subj vs Task:
Listen Reading
sub01 1 0
sub02 1 0
sub03 1 0
...
which means that you didn't arrange the data table correctly. In other words, the above contingency table should look like this:
Subj vs Task:
Listen Reading
sub01 1 1
sub02 1 1
sub03 1 1
...
Gang