not likely given that the accuracy was similar for the motor and normal settings in
both early and late stages. The visual setting had accuracy similar to normal by late
stage (Fig. 2). Further if subjects had continued to use rotational transformations till
the late stage, we would not observe improvements in response times from the early
to the late stage.
Second, subjects might develop novel representations for the sequence in rotated
settings independently from the sequence in the normal setting. This is not likely
given that subjects are slower in the visual setting compared to the motor setting in
both the early and late stages. We have manipulated the display and keypad rotations
such that display-to-keypad mapping is the same in visual and motor settings. If
subjects did not benefit from the sequence in the normal setting, then improvements
in performance must be similar for the two rotated settings and that is clearly not the
case (Fig. 2).
Third, it is possible that in both the rotated settings, subjects simply learned to apply
rotational transformations in the initial phase to learn the correct sequence but
eventually learned to replace this difficult operation with a novel sequence
representation. We suggest that this is a more plausible explanation for sequence
learning tasks that require performing a rotational transformation such as the current
experiment. Given that the visuo-spatial representation is acquired fairly quickly
during sequence learning (for example, in the normal setting), the visual setting
stretches this representation over a longer period of time as evidenced by its slower
performance indices. Similarly, given that a somato-motor representation takes quite
long to establish, the motor setting extends this late stage of sequence learning into
the early stage of the motor setting by using strategies of explicit sequence learning.
Thus the tasks appear to bias the sequence learning process in the two experiments the
way we intended.
The shift in activity from anterior putamen in early visual (Table 1a, 3) to posterior
putamen in late visual (Table 1a) and early motor settings and finally to the posterior
putamen in the late stage of motor setting (Table 2a) seems to have a cortical analog.
The transition of activity from the parietal cortex in the early visual setting to conjoint
activity in the pariteal-premotor areas in the late visual as well as early motor and
subsequently to the dominant activity in the premotor cortex in the late stage of motor
setting seems to mirror the type of transitions taking place in the basal ganglia
regions. Anterior striatum is part of the prefrontal and parietal cortex - basal ganglia
loops which may be involved in the visuo-spatial representation whereas posterior
striatum as part of the primary and secondary motor cortex - basal ganglia loops may
be involved in the somato-motor sequence representation (Alexander, DeLong, &
Strick, 1986; Hikosaka et al., 2002).
The novelty of our current findings is that the differential involvement of the cortico-
subcortical loops subserving various sequence representations could be demonstrated
as a direct outcome of our experimental design. The results suggest that during the
process of learning, an early acquisition of visuo-spatial representation is subserved
by frontal, parietal cortex - anterior striatal loop followed by an additional recruitment
of secondary motor areas (dorsal premotor cortex and SMA) - posterior striatal loop
during the acquisition of somato-motor representation. Finally, these results form the
first comprehensive and direct evidence for the model proposed by Nakahara, Doya,
24