Scaling in the first level analysis


I have questions about scaling in the first level analysis by running

  1. Does ‘scaling’ mean percent signal change calculation?
  2. fMRI raw data value can have negative value? or all value 0 or positive value? When I see ‘proc.$subject.glt’ file, the ‘scale’ part is saying that

================================= scale ==================================

scale each voxel time series to have a mean of 100

(be sure no negatives creep in)

(subject to a range of [0,200])

foreach run ( $runs )
3dTstat -prefix rm.mean_r$run pb03.$subj.r$run.blur+tlrc
3dcalc -a pb03.$subj.r$run.blur+tlrc -b rm.mean_r$run+tlrc
-c mask_epi_extents+tlrc
-expr ‘c * min(200, a/b*100)*step(a)*step(b)’
-prefix pb04.$subj.r$run.scale

According to this, it looks like that fMRI raw data(or preprocessed data) can have negative value, and ‘proc.$subject.glt’ would ignore them by step(a) or step(b).
What is this meaning?

  1. As my understanding, scaling is performed normalization of data for group-level analysis before regression. This data should be preprocessed raw fMRI data.
    The preprocessing(tshift, align, space normalization, volreg, and blur) does not affect the raw FMRI data value changed so much (for example, positive value change to negative value or totally different value). But after scaling, some negative value can be changed to positive value by calculating percent signal change based on the average value through time series in voxel by voxel. Will it affect calculation of beta value to show if activate of deactivate?
  2. is ‘scaling’ necessary for all fMRI analysis? if so, what is clear reason?




Yes, scaling is done voxelwise. For more information, particularly on the arguments for scaling, and a description of interpretability, please see the “units and scaling section” here:
(or biorxiv version:
Actually, that whole article is relevant for your question(s).
See also section C-3 here:

The scaling is used to translate the unitless BOLD signal acquired by scanners to something interpretable as “BOLD percent signal change”. And yes, that can certainly have negative values (–> a negative change from baseline). This is something that seems useful quantitatively, as well as for interpretation, and also for how the BOLD signal works (modulating response from blood oxygenation).