If you have four stimuli that could be categorized as the same 'type' of stimulus, what is the theoretical difference, statistically speaking, between specifying four different stimulus vectors of 15 items as four different parameters in a deconvolution vs. specifying one one vector of 60 items as one parameter in the model?

For a more concrete example of my question, In our current study we have four stimulus items that are True statements, each repeated 15 times. We can treat them seperately, as four different 15-item stimulus 1D files in the deconvolution, or categorize them all as 'True stimuli' and build a single stimulus 1D file with 60 stimulus points to use as input in deconvolution.

I ask because the output of GLTs on these two different process methods appear to yield results that are slightly different.

In the first deconvolution, with four separate stim files, I conduct the glt to compare to baseline with a flag that looks something like this:

-gltsym 'SYM: +True1[2..4] +True2[2..4] +True3[2..4] +True4[2..4]' -glt_label 1 'TrueItems'

statistically testing TRs 2 through 4 across all four items.

In the second case, I would conduct the glt with this simpler flag:

-gltsym 'SYM: +True[2..4]' -glt_label 1 'TrueItems'

Since there is only one IRF output by the single-vector stimulus file.

The difference in the GLT results is the statistical significance of voxels in the output bucket datasets. In the results I have compared for one subject, the cluster regions are similar, but the former case yields clusters of voxels significant to 1e-9, whereas the second only around 1e-4.

I am curious about the theoretical underpinnings of this, whether this little test example makes sense statistically, and which approach would be advisable.

Thank you in advance for any input.

-George