Hi Peter,
Thanks for your reply. I've used the same script for all subjects. This is the only one that had poor alignment. The subject had two runs within the same scan session and only one of the runs has poor alignment, so it doesn't seem to be specific to this subject, it seems specific to this particular EPI dataset. The anatomical and EPI in orig space are well aligned. It's only after volreg that the alignment looks poor.
I'll take a look at uber_align_test.py to see if I can understand the problem better. In the meantime, here's the script we used to generate the proc script. I can send the full script if it's helpful.
Josh
#!/usr/bin/env tcsh
# created by uber_subject.py: version 0.36 (April 5, 2013)
# creation date: Fri May 9 14:47:49 2014
# set data directories
set top_dir = /raid10/studies/varenicline
set anat_dir = $top_dir/subjects/exclude/jjl
set epi_dir = $top_dir/subjects/exclude/jjl
set stim_dir = $top_dir/files
# set subject and group identifiers
set subj = jjl
set group_id = Varenicline_Faces
# run afni_proc.py to create a single subject processing script
afni_proc.py -subj_id $subj \
-script proc.$subj -scr_overwrite \
-blocks tshift align tlrc volreg blur mask scale regress \
-copy_anat $anat_dir/mprage_series010+orig \
-tcat_remove_first_trs 4 \
-dsets $epi_dir/faces30slicespost_series011+orig.HEAD \
-volreg_align_to third \
-volreg_align_e2a \
-volreg_tlrc_warp \
-blur_size 4.0 \
-regress_stim_times \
$stim_dir/Fear.txt \
$stim_dir/Neutral.txt \
-regress_stim_labels \
Fear Neutral \
-regress_basis 'GAM' \
-regress_censor_motion 0.3 \
-regress_censor_outliers 0.15 \
-regress_opts_3dD \
-jobs 5 \
-gltsym 'SYM: Fear -Neutral' -glt_label 1 F-N \
-gltsym 'SYM: 0.5*Fear +0.5*Neutral' -glt_label 2 mean.FN \
-regress_make_ideal_sum sum_ideal.1D \
-regress_est_blur_epits \
-regress_est_blur_errts