Thanks for this idea.
I used the mean RT of all subjects in all conditions. Do you have an idea whether it is reasonable that following this manipulation the HDR was reversed and from positive it became negative? I have a hard time understanding this.
In the modulation sub-brick [#1] the betas are usually highly positive, but in the modulated sub-brick [#0] they are negative in regions I expect them to be positive (e.g., intra-parietal sulcus). These regions showed positive HDRs before using RT as modulator. I understand that strong RT differences could be responsible for the non-modulated effect, but because these regions usually show positive HDRs, it is strange that the effect is totally reversed.
In addition to the reversed effects (from positive to negative), my GLTs show more activation for easy conditions than for harder conditions. All this together make me susspect that by subtracting the mean RT I changed more than I wanted.

Here is my 3dDeconvolve script (I also tried :0:2.8576 after the AM2 stim, thinking that the stim-time should also be taken into account, but the results were the same):

3dDeconvolve -input pb04_${subj}_r*_scale+tlrc.HEAD \
-censor motion_${subj}_censor.1D \
-polort 6 \
-local_times \
-num_stimts 14 \
-stim_times_AM2 1 stimuli/${subj}-stim1-E-1-cor.txt 'GAM' :2.8576 \
-stim_label 1 e1_cor \
-stim_times_AM2 2 stimuli/${subj}-stim2-E-DN-DD-cor.txt 'GAM' :2.8576 \
-stim_label 2 e-diffcomp_cor \
-stim_times_AM2 3 stimuli/${subj}-stim3-U-DN-DD-cor.txt 'GAM' :2.8576 \
-stim_label 3 u-diffcomp_cor \
-stim_times_AM2 4 stimuli/${subj}-stim4-U-SN-SD-cor.txt 'GAM' :2.8576 \
-stim_label 4 u-samecomp_cor \
-stim_times 5 stimuli/${subj}-stim5-E-1-inc.txt 'GAM' \
-stim_label 5 e1_inc \
-stim_times 6 stimuli/${subj}-stim6-E-DN-DD-inc.txt 'GAM' \
-stim_label 6 e-diffcomp_inc \
-stim_times 7 stimuli/${subj}-stim7-U-DN-DD-inc.txt 'GAM' \
-stim_label 7 u-diffcomp_inc \
-stim_times 8 stimuli/${subj}-stim8-U-SN-SD-inc.txt 'GAM' \
-stim_label 8 u-samecomp_inc \
-stim_file 9 motion_demean.1D'[0]' -stim_base 9 -stim_label 9 roll \
-stim_file 10 motion_demean.1D'[1]' -stim_base 10 -stim_label 10 pitch \
-stim_file 11 motion_demean.1D'[2]' -stim_base 11 -stim_label 11 yaw \
-stim_file 12 motion_demean.1D'[3]' -stim_base 12 -stim_label 12 dS \
-stim_file 13 motion_demean.1D'[4]' -stim_base 13 -stim_label 13 dL \
-stim_file 14 motion_demean.1D'[5]' -stim_base 14 -stim_label 14 dP \
-iresp 1 iresp_e1.${subj} \
-iresp 2 iresp_e-diffcomp.${subj} \
-iresp 3 iresp_u-diffcomp.${subj} \
-iresp 4 iresp_u-samecomp.${subj} \
-xjpeg Xmat \
-x1D Xmat \
-gltsym 'SYM: +e1_cor[0] -e-diffcomp_cor[0]' \
-glt_label 1 e1VSediffcomp \
-gltsym 'SYM: +e1_cor[0] -u-samecomp_cor[0]' \
-glt_label 2 e1VSusamecomp \
-gltsym 'SYM: +e1_cor[0] -u-diffcomp_cor[0]' \
-glt_label 3 e1VSudiffcomp \
-gltsym 'SYM: +u-diffcomp_cor[0] -u-samecomp_cor[0]' \
-glt_label 4 usamecompVSudiffcomp \
-gltsym 'SYM: +e-diffcomp_cor[0] -u-diffcomp_cor[0]' \
-glt_label 5 diffcomp-eVSu \
-gltsym 'SYM: +e-diffcomp_cor[0] -u-samecomp_cor[0]' \
-glt_label 6 ediffcompVSusamecomp \
-gltsym 'SYM: +e1_cor[0] +e-diffcomp_cor[0] -u-diffcomp_cor[0] -u-samecomp_cor[0]' \
-glt_label 7 eVSu \
-gltsym 'SYM: +e1_cor[0] -e-diffcomp_cor[0] -u-diffcomp_cor[0] +u-samecomp_cor[0]' \
-glt_label 8 nocomputVScomput \
-GOFORIT 11 \
-fout -tout -x1D X.xma.1D -xjpeg X.jpg \
-x1D_uncensored X.nocensor.xmat.1D \
-errts errts.${subj} \
-bucket stats.$subj

(* The GOFORIT 11 is due to some participants have no errors and thus their "inc" stim_times contain only *s.)

And here is an example of the stim_time file
0.0172*3.2776 66.6238*2.7509 85.6543*3.8671 123.7152*3.8376 171.2914*2.1536 190.3218*2.4149 209.3523*2.4671 352.0807*2.5893 390.1417*3.1968 399.6569*3.2214 447.233*2.3245 485.2939*1.9039 580.4462*3.1831 637.5376*3.9568 656.5681*1.686 666.0833*1.7806 732.6899*2.232 751.7203*1.8183 780.266*4.098 827.8422*1.885
19.0626*1.8128 76.154*2.5595 95.1845*2.0647 104.6997*1.8003 161.7911*2.114 171.3063*1.9676 199.852*1.6225 209.3672*2.4131 285.489*2.1711 323.5499*1.9785 361.6108*4.538 390.1565*1.9388 399.6718*2.6504 542.4002*2.0696 561.4306*2.7489 609.0068*1.8009 713.6743*2.2307 732.7047*1.841 742.22*2.5016 751.7352*2.1682
0.0248*2.6799 9.54*3.2256 38.0857*2.2275 85.6618*3.1206 123.7227*2.8561 133.2379*2.0857 152.2684*2.0729 161.7836*3.7085 209.3597*2.5606 218.875*1.6022 237.9054*2.3795 266.4511*2.7423 275.9663*1.728 380.6338*1.9138 409.1795*2.4306 551.9079*1.5829 637.545*2.4504 675.6059*1.5239 713.6668*1.3673 780.2734*3.2205

Thanks!
kallai