To head off the question, I'm not using the -stdin option because I need to reference stdin twice in the call to 1dplot, drawing from a different columns of it.

[tjnycum@nx1 CatCondwise_fine_nomaint_2deriv]$ 3dhistog -mask 1006long_automask+orig.nii.gz -notit Decon_REML+orig.BRIK.gz'[Stim:RelNB-Nonrel#0_Tstat]' | 1dplot -X /dev/fd/0'[0]' /dev/fd/0'[1]'
++ 1dplot: AFNI version=AFNI_2009_12_31_1431 (Feb 25 2010) [64-bit]
++ Authored by: RWC et al.
++ 20576 voxels in mask
** FATAL ERROR: Can't read x-axis '-x /dev/fd/0[0]'
** Program compile date = Feb 25 2010

I thought perhaps the /dev/fd/0 form of referencing stdin wasn't compatible with piping, but that doesn't appear to be the problem:
[tjnycum@nx1 CatCondwise_fine_nomaint_2deriv]$ 3dhistog -mask 1006long_automask+orig.nii.gz -notit Decon_REML+orig.BRIK.gz'[Stim:RelNB-Nonrel#0_Tstat]' | cat /dev/fd/0
++ 20576 voxels in mask
-4.064107 1 1
-3.983314 0 1
-3.902521 0 1
-3.821727 0 1
[...]

So it appears 1dplot just doesn't like that file. Is this discrimination against special files really necessary? (You're thwarting my work-around! )

On a side note:
Before resorting to /dev/fd/0 I tried the '-' convention, but, alas, 1dplot doesn't accept it:
[tjnycum@nx1 CatCondwise_fine_nomaint_2deriv]$ 3dhistog -mask 1006long_automask+orig.nii.gz -notit Decon_REML+orig.BRIK.gz'[Stim:RelNB-Nonrel#0_Tstat]' | 1dplot -X -'[0]' -- -'[1]'
++ 1dplot: AFNI version=AFNI_2009_12_31_1431 (Feb 25 2010) [64-bit]
++ Authored by: RWC et al.
++ 20576 voxels in mask
** FATAL ERROR: Unknown option: --
** Program compile date = Feb 25 2010
Would you please consider this a feature request for AFNI as a whole (or at least as many of its binaries as possible)?

Thanks!