It is known that the saccadic system shows adaptive changes when the comman
d sent to the extraocular muscles is inappropriate. Despite an abundance of
supportive psychophysical investigations, the neurophysiological substrate
of this process is still debated. The present study addresses this issue u
sing (H2O)-O-15 positron emission tomography (PET). We contrasted three con
ditions in which healthy human subjects were required to perform saccadic e
ye movements toward peripheral visual targets. Two conditions involved a mo
dification of the target location during the course of the initial saccade,
when there is suppression of visual perception. In the RAND condition, int
ra-saccadic target displacement was random from trial-to-trial, precluding
any systematic modification of the primary saccade amplitude. In the ADAPT
condition, intra-saccadic target displacement was uniform, causing adaptive
modification of the primary saccade amplitude. In the third condition (sta
tionary, STAT), the target remained at the same location during the entire
trial. Difference images reflecting regional cerebral-blood-flow changes at
tributable to the process of saccadic adaptation (ADAPT minus RAND; ADAPT m
inus STAT) showed a selective activation in the oculomotor cerebellar vermi
s (OCV; lobules VI and VII). This finding is consistent with neurophysiolog
ical studies in monkeys. Additional analyses indicated that the cerebellar
activation was not related to kinematic factors, and that the absence of si
gnificant activation within the frontal eye fields (FEF) or the superior co
lliculus (SC) did not represent a false negative inference. Besides the con
tribution of the OCV to saccadic adaptation, we also observed, in the RAND
condition, that the saccade amplitude was significantly larger when the pre
vious trial involved a forward jump than when the previous trial involved a
backward jump. This observation indicates that saccade accuracy is constan
tly monitored on a trial-to-trial basis. Behavioral measurements and PET ob
servations (RAND minus STAT) suggest that this single-trial control of sacc
ade amplitude may be functionally distinct from the process of saccadic ada
ptation.