Discussion
The aim of the study was to investigate different representations of visuomotor
sequences as learning progressed from the early to the late stage. We demonstrated
that response times were shorter when subjects used effector-specific information (as
in the motor setting) than when they utilized abstract sequential information based on
visual cues (in the visual setting). We identified the subcortical and cortical brain
areas that mediated the two sequence representations. Firstly, left anterior putamen in
the dorsal striatum was found to be selectively active in the visual setting. Secondly, it
appears that when sequences are learned utilizing visuo-spatial representation, the
focus of activation moved from parietal in the early stage to parietal-premotor areas in
the late stage. In contrast, when sequences were performed with an emphasis on the
somato-motor representation, the transition was in the opposite direction, that is, from
parietal-premotor in the early to premotor areas in the late stage.
It is known that visuo-spatial sequence representation can be acquired quite quickly
but somato-motor optimization takes time (Bapi et al., 2000; Nakahara et al., 2001).
In the visual setting, although subjects attained similar level of accuracy to that of
motor setting by the late stage, performance speed was significantly slower (Fig. 2).
Behavioral results further indicated that the subjects utilized abstract sequential
information provided by the visual cues over an extended period of time and
eventually learned a new somato-motor sequence in the visual setting. Superior
performance speed in the motor setting by the early stage itself (Fig. 2) underlines the
advantage of using effector-specific representation. We further demonstrated that
chunking patterns of the sequences were identical between the motor and normal
settings (Fig. 3). Thus, while the onset of learning of visuo-spatial and somato-motor
sequence representations is the same, the activity associated with each setting at
various stages pointed out how learning of the representations unfolded over time.
The unique experimental design we used in this study enabled us to tap into the
representations that facilitate the learning of motor sequences and their unfolding
process.
Subcortical structures
While activation in the right anterior cerebellum sustained from the early to late stage
of visual setting (Table 1a), the activity decreased by the late stage in the motor
setting (Table 2a). Anterior cerebellum might possibly be involved in the optimization
of movement parameter and timing information in both the sequence representations
(Jueptner & Weiller, 1998; Sakai et al., 2000). Based on previous studies involving
trial and error learning (Jenkins et al., 1994; Jueptner et al., 1997b), sustained activity
observed in the ventral striatum in both early and late stages of visual setting may be
attributed to the trial and error process adopted for cue selection. Interestingly, when
there was no emphasis on cue selection process as in the motor setting, ventral striatal
activity was absent. Further evidence comes from the activation of ventral striatum
observed in the direct comparison contrast of early visual > motor (Table 3). Left
putamen activation in the dorsal striatum extended into both anterior and posterior
regions in the visual setting (Table 1a), whereas in the motor setting activation was
found exclusively within the posterior putamen (Table 2a). Further, left anterior
putamen in the dorsal striatum was found to be selectively active in the visual setting
(Table 3). Additionally, while activation in the putamen became stronger from the
early to late stage in the visual setting, it decreased by the late stage in the motor
21