Submitted to 27th Annual Meeting, Sciety of Neuroscience, 1997.

MULTIPLE REPRESENTATIONS IN THE BASAL GANGLIA LOOPS FOR ACQUISITION
AND EXECUTION OF SEQUENTIAL MOTOR CONTROL.  H. Nakahara^1, K. Doya*^2,
O. Hikosaka^3, and S. Nagano^4. 1:Frontier Res. Prgm., RIKEN, Saitama,
351-01, Japan. 2:Kawato Dynamic Brain Project, JST. 3:Dept. of Physiol.,
Juntendo Univ. Sch. of Med. 4:Dept. of General Sys. Studies, Univ. of
Tokyo.
  The basal ganglia (BG) are known to play an important role in
sequential motor control but their computational roles are poorly
understood. In recent studies using "2x5 task", in which a monkey
learned to press five sets of two LED buttons in a predetermined
order, functional differentiation was observed between anterior and
posterior BG (Miyachi et al., Soc Neurosci Abstr 20: 357, 1994), as
well as presupplementary motor area (pre-SMA) and supplementary motor
area (SMA) (Miyashita et al., Soc Neurosci Abstr 22: 1862, 1996), in
acquisition and execution of the sequential movement. Here, we propose
a computational model on the functions of those areas based on a
theory of reinforcement learning (Nakahara, Ph.D. thesis, 1997). The
central idea of the model is that a visuomotor sequence is easier to
learn in spatial representation (e.g. visual coordinate) but is easier
to control in body-based representation (e.g. joint angle coordinate). 
Based on anatomical and physiological evidence, we propose that the
loop between anterior BG and the prefrontal cortex, perhaps together
with pre-SMA, learns the sequence using visual coordinate. The loop
between posterior BG and SMA learns the sequence using body
coordinate.  Although both loops concurrently learn the sequence,
because of the difference in the speed of learning, the anterior loop
is more critical in learning of new sequences and the posterior loop
is more involved in performance of well learned sequences. We then
performed computer simulations of the proposed model. The results
replicated both behavioral and neurophysiological data, and further
provided several predictions which are experimentally testable using
the 2x5 task. Supported by JSPS, Grant-in-Aid from The Ministry of
Edu. and Sci, and SPRF of RIKEN to HN, HFSP to KD, Uehara Memorial
Foundation and JSPS Research for the Future Program to OH.