Hi MB,

Functional MRI does not have great temporal
resolution to estimate baseline on a per-trial
basis, particularly in the typical case of a
fast event design. Even with a slow or block
design, one would need long rest periods (at
least 30 seconds) to even have a shot at doing
it robustly. And 30 seconds intervals would
only give maybe 15 seconds after BOLD recovery
to estimate the baseline with, which seems very
short and prone to be noisy, making the BOLD
estimates similarly noisy.

Since that would take a lot of extra "rest"
time, it means opportunity for stimulus time
would be similarly lost. And researchers tend
to covet time for stimulus events.

- rick