Hikosaka et al (2002) proposed that a sequence of movements is represented in two
ways - spatial sequence and motor sequence. In their hypothetical scheme spatial
sequence learning and representation are supported by parietal-prefrontal cortical
loops with the associative region of the basal ganglia (anterior striatum) and
cerebellum (posterior cerebellum). Motor sequence learning and representation are
mediated by the motor cortical loops with the motor region of the basal ganglia
(posterior striatum) and cerebellum (anterior cerebellum). Further, in their scheme
premotor area mediates the transformation of spatial to motor coordinates and pre-
supplementary motor area (pre-SMA) participates in coordination or switching
between the two representations. In this connection it is interesting to note that this
scheme is partly based on an earlier proposal of Alexander, DeLong, & Strick (1986)
of distinct cortico-basal ganglia-thalamus loops serving different functions. In this
scheme that stresses parallel information processing, the dorsolateral prefrontal cortex
(DLPFC)-caudate nucleus loop takes part in spatial sequencing whereas the
supplementary motor area (SMA)-putamen loop mediates motor sequencing.
Our hypothesis was that motor sequence learning involves two representations - an
early acquisition of effector independent (abstract) representation and a late
consolidation of effector dependent representation (Hikosaka et al., 1999, 2002; Bapi,
Doya, & Harner, 2000; Nakahra, Doya, & Hikosaka, 2001). In an earlier behavioral
study (Bapi et al., 2000), we used a sequential button-pressing task in which subjects
performed either the same visuo-spatial sequence with altered finger movements or a
different visuo-spatial sequence with the same finger movements. We found that the
response time was significantly shorter when the finger movements remained the
same compared to when the visuo-spatial sequence was the same. These results
suggest that an effector independent representation develops early in the learning
process and subsequently an effector dependent sequence representation is formed.
Using a whole-brain fMRI study, we set out to investigate the question of the brain
areas subserving such representations acquired at various stages of explicit learning of
motor sequences. In the current study, subjects learned a sequence of 12 finger
movements, using a 2x6 task (Fig. 1 a) modified from Hikosaka et al. (1995), in two
settings - normal setting where the visual display and keypad are arranged in the
usual position and a rotated setting. In the rotated (motor and visual) conditions,
subjects were required to rotate the visual cues by 180° and press the corresponding
keys. The display sequence was also rotated for the motor condition, requiring an
identical set of effector movements to be performed as in the normal condition. Thus
the display-to-keypad mapping was identical for both the motor and visual settings.
Further in the visual setting, the sequence remained the same as in normal in visuo-
spatial coordinates, whereas it was different from normal in somato-motor
coordinates. On the other hand, in the motor setting, the sequence in somato-motor
coordinates was the same as in normal, but it was different from normal in visuo-
spatial coordinates. This experimental design allowed us to explicitly tap into the
neural loci of abstract and effector-specific representations of motor sequences.