Oh, I didn't think of TENT functions.

That suggests that for one stimulus class the events
are TR-locked, but for the other they are not, assuming
the TR is 3.0 s. That would probably apply on a 1.5 s
time grid, too.

With TR-locked events, the initial TENT regressor (for
a given class) will be all zero except for a 1 at each
event TR (and related regressors would be shifted
verisons).

With events being off the TR, each event will be
split proportionally across 2 TRs.

If events are never within 6 s of each other, the
off-TR count should double the on-TR count. If
they do get close, the ratio will be somewhere
between 1 and 2, depending on how frequently
they are close.

- rick