setting (Table 1a, 2a). Another structure that can possibly be implicated with visuo-
spatial representation is the hippocampus, which was found to be activated in early
and late stages of visual > motor contrast (Table 3). The brain behaviour relationship
in the hippocampus was significantly correlated with the visual setting, but not the
motor setting (Table 4). In contrast to earlier studies where cerebellum (Doyon et al.
2003) and basal ganglia (Penhune & Doyon, 2002; Shadmehr & Holcomb, 1997)
activity was not observed during the recall stage, the activity in these regions persisted
till the more automatic late stage in the motor setting. One possible reason for this is
that in our experiment, recall was measured within the experimental session on the
same day as opposed to the earlier studies where recall was assessed after a delayed
consolidation period. The conclusions on cerebellum are provisional as the coverage
of cerebellum did not extend into posterior lobe in the fMRI scans at the chosen field
of view.
Cortical structures
There seems to be a trend of shift in activation from the parietal in early visual to
parietal-premotor areas in late visual (Table 1a) and early motor settings (Table 2a).
The activation in premotor areas becomes stronger by the late stage of motor setting
(Table 2a). The trend of shift in parietal areas is strengthened by its activation in the
late visual > motor comparison and early motor > visual contrasts (Table 3). The brain
behaviour relationships clearly demonstrate a decreasing trend of activation in the
parietal areas and an increasing trend of activation in the premotor areas for the motor
setting (Table 4). The activation in the frontal areas did not show any selectivity to
either of the rotated settings (Table 3 and 4). The rostral part of supplementary motor
area, pre-SMA, was selectively active in the late stage of visual setting (Table 2a). In
the visual-normal experiments subjects used the same visuo-spatial sequence but
learned two motor sequences, one corresponding to the normal and the other to the
visual setting. The brain-behavioral relationship result of pre-SMA (Fig. 6, Table 4)
combined with the selective activation found in the late but not in the early stage of
visual setting (see Table 1a) indicates its role in new somato-motor sequence learning
process. The activation of pre-SMA during new motor sequence learning was also
observed in earlier studies (Hikosaka et al., 1996; Sakai et al., 1998). Together with
the activity observed in the anterior cingulate (Table 3), we speculate that the pre-
SMA may have a dual role, one in sequence learning and the other in sequence
switching (Shima et al., 1996). Consistent with the explicit and implicit sequence
learning literature (Tanji & Shima, 1994; Grafton, Hazeltine, & Ivry, 1998;
Willingham, 1998; Keele et al., 2003; Tanji, 2001), we found SMA activity in the late
stage of motor setting (Table 3) and brain- behaviour correlation (Fig. 7, Table 4)
reflecting its role in somato-motor sequence representation. We observed activity in
M1 during somato-motor sequence learning and performance as evidenced by its
activation in early and late stages of both the rotated settings. These results suggest
that M1 participates in the learning of motor sequences but may not be the actual
locus of representation. These observations on M1 are consistent with most of the
earlier proposals for its role in motor learning (for example Karni et al., 1995; Grafton
et al., 1998; Sanes, 2003). We can speculate that activity in the right DLPFC observed
in both the rotated settings (decreasing trend as seen in Table 4) is related to the
optimization of the sequencing process.
It is possible that the rotated > normal contrasts point out learning related activations
in addition to rotational transformations. Our experimental design involved
interleaved blocks of normal and rotated settings alternating with the follow
22