I guess the author refer to the individual estimation for each trial here. Thus it results a beta-series for one condition in a run (vs. what we usually called time-series). Someone used this "beta-series" to do the correlation or connectivity analysis with ER design. I think "-stim_times_IM" option in 3dDeconvolve could do this.
But -Gang, Does it make sense, in statistics, to model individual trial in a ER design with fMRI data?
Qiuhai