Neuroimaging activation studies typically observe signals during two o
r more periods of differing cognitive activity which are then analyzed
by a subtraction to test for localized neuroanatomical dissociations
between cognitive tasks. Significant activity found between task condi
tions is frequently assumed to reflect a novel cognitive process prese
nt in one task and not the other. We present a conceptual framework th
at considers the neural mechanisms underlying such observed neuroimagi
ng changes. We propose that neuroimaging experiments which present sti
muli at a fixed pace (where each trial takes the same amount of time)
will be sensitive to changes in both duration and intensity of neural
processing. In contrast, the signal observed during a self-paced desig
n is derived from neural processing averaged over the reaction time an
d hence could be less sensitive to differences in duration of neural p
rocessing. As an empirical demonstration of these ideas, we studied no
rmal subjects using echoplanar functional MRI during two visuospatial
tasks (matching of either ROTATED or NONROTATED stimuli) performed usi
ng FIXED and SELF-PACED designs. In both pacing designs, reaction time
s were greater in the ROTATED than NONROTATED task, interpreted as a g
reater duration of neural processing during the ROTATED compared to th
e NONROTATED task. In the FIXED-PACED design, significantly greater si
gnal was present within a parieto-occipital cortical region during the
ROTATED task compared to the NONROTATED task. This difference was not
observed during the SELB-PACED design. This result illustrates the im
portance of considering trial pacing in the interpretation of function
al neuroimaging activation studies. (C) 1997 Academic Press.