Rick:
It was the 3dcalc help that had me puzzled in the first place, because "nth order statistic" sounds like something more complicated than simply the "nth largest value".
Second, if pursuing the issue of rotating a label volume via splitting into separate single-label volumes then rotating each... the recombining function needed is *index* of maximim value (from which one can translate to "winning label value"), not the actual *value* of the max value.
But maybe that's all moot, as you and Ziad have convinced me to investigate the top secret "NN" feature.
Thanks,
Graham