condition. Both the normal and rotated settings involved sequence learning, whereas
the follow condition did not involve any learning. Hence, we associate any activity
observed in the subtractions reflected in rotated > normal contrasts with rotational
transformations, but not with sequence learning. Activity observed in the superior
occipital gyrus and superior parietal cortex in the visual > normal contrast (Table 1b)
may support processes involved in target rotation and the activity in the middle
temporal gyrus and inferior parietal cortex in the motor > normal contrast (Table 2b)
may be related to the processes involved in cue rotation.
Brain areas supporting the formation of representations during sequence learning
Our results suggest that visuo-spatial sequence representation engages cortical and
subcortical network involving the left anterior striatum, hippocampus, extrastriate
visual areas, dorsal and ventral premotor, and parietal cortical areas (see Table 1a, 3).
The activation in the extrastriate visual areas may be related to the visuo-motor
processes required for synchronizing the motor actions to visual cues (Bower, 1995).
The activation in the extrastriate visual areas and hippocampus may form part of the
ventral stream (Mishkin, Ungerleider, & Macko, 1983) that encodes information in
visual coordinates and conveys the information to ventral premotor to enable the
formation of visual stimulus-to-response associations (Caminiti, Ferraina, & Mayer,
1998). The activity in the parietal cortex may be part of the dorsal stream (Mishkin,
Ungerleider, & Macko, 1983) conveying information in spatial coordinates to the
dorsal premotor to enable formation of spatial cue-to-response associations (Wise et
al., 1997). Anterior striatum may be in the best position to combine the information
from the ventral and dorsal streams to formulate goal-directed action sequences based
on abstract information. Our results also revealed that effector-specific sequence
representation is subserved by the dorsal premotor and SMA (Table 3, 4). A summary
of our findings on various representations acquired during the process of visuomotor
sequence learning is given in the supplementary figure (Fig. S4).
Cortico-subcortical networks subserving visuo-spatial and somato-motor sequence
representations
Our hypothesis in sequence learning is that the early stage involves abstract (visuo-
spatial) representation and the late stage involves effector-specific (somato-motor)
representation. This is true regardless of whether subjects are performing the sequence
task in the normal, motor or visual settings in our experiments or they are learning
some other visuo-motor skill such as cycling or boxing. The usual acquisition of
visuo-spatial sequence being very rapid, we designed the visual-normal experiment so
that subjects were required to use the sequence of visual cues over an extended period
and eventually acquired a second motor sequence for the visual setting. The design
strategy of the motor-normal experiments was such that subjects could focus on motor
movement sequence right from the early stage and progress to performance of an
over-learned somato-motor sequence. Ours is an explicit sequence learning task and
the instructions and pretraining given to the subjects ensured that subjects took
advantage of the representations that were being learned at various stages of the task
(see Experiments subsection in Materials and Methods). However, it is possible that
there are alternative explanations for the results we obtained.
First, subjects might simply be performing the display and keypad rotations to retrieve
the sequence from the normal setting in order to perform the rotated settings. This is
23