system has stopped on the representation of the ex-
act number of ob jects in the scene. The cooperation
0 20 40 60 80 100 120 140 160 180
0 20 40 60 80 100 120 140 160 180
0 20 40 60 80 100 120 140 160 180
Figure 5: Activity of four neurons in the number map.
From top to bottom, these neurons respond preferentially
for the words ’zero’ ’one’ ’two’ and ’three’.
between the two systems has been tested and Figure
5 shows the evolution of the activity of four neurons
on the number map when the reward activity is pro-
duced by the attention-switching mechanism. These
neurons activate sequentially at each dopamine de-
pletion so that at end of the visual search the only
neuron remaining active is the neuron representing
“three”. This works with one, two or three visual
objects, but for more objects we would just need
more neurons in the number map thanks to the dis-
tributed architecture of the system.
6. Conclusion
We presented here two computational models that
seem in a first view totally independent as they deal
with different modalities (vision and reproduction of
phonetical sequences), but that can cooperate to pro-
duce a new behaviour, namely a cardinal counting
task. Our hypothesis was that counting objects in a
scene needs to sequentially focus these ob jects (what
is a motor ability) and to associate this sequence
with the remembered ordinal sequence of numbers.
These two models have two separate basal ganglia
channels that communicate via a unique dopamin-
ergic unit, what is coherent with the diffuse inner-
vation of dopamine throughout the Striatum. The
first model works in real-time on a robot but for rea-
sons of computational cost the merging of the two
systems has only been tested in simulation. Never-
theless, there are some problems: the coding of num-
bers from phonological inputs is not enough compact
and stands for numbers up to ten maybe. We would
need another architecture to deal with greater num-
bers. The sequence-learning system also learns the
sequence offline (without the attention mechanism),
it would be an interesting feature if the learning oc-
cured online.
References
Amari, S.-I. (1977). Dynamical study of formation of
cortical maps. Biological Cybernetics, 27:77-87.
Berns, G. and Sejnowski, T. (1998). A computational
model of how the basal ganglia produce sequences.
Journal of Cognitive Neuroscience, 10:108-121.
Brannon, E. M. and Van de Walle, G. A. (2001). The de-
velopment of ordinal numerical competence in young
children. Cognitive Psychology, 43:53-81.
Gurney, K., Prescott, T. J., Wickens, J. R., and Red-
grave, P. (2004). Computational models of the basal
ganglia: from robots to membranes. Trends in Neu-
rosciences, 27(8):453-459.
Hikosaka, O., Takikawa, Y., and Kawagoe, R. (2000).
Role of the basal ganglia in the control of purpo-
sive saccadic eye movements. Physiological Reviews,
80(3):953-978.
O’Regan, K. and Noe, A. (2001). A sensorimotor ac-
count of vision and visual consciousness. Behavioral
and Brain Sciences, 24:939-1031.
Piaget, J. (1972). The psychology of the child. New
York: Basic Books, New York.
Rizzolatti, G. and Arbib, M. A. (1998). Lan-
guage within our grasp. Trends in Neurosciences,
21(5):188-194.
Rizzolatti, G., Fadiga, L., Gallese, V., and Fogassi, L.
(1996). Premotor cortex and the recognition of mo-
tor actions. Cognitive Brain Research, 3:131-141.
Rougier, N. and Vitay, J. (2005). Emergence of atten-
tion within a neural population. Accepted to Neural
Networks.
Schaal, S., Ijspeert, A. J., and Billard, A. (2003). Com-
putational approaches to motor learning by imita-
tion. Philosophical Transactions: Biological Sciences
(The Royal Society), 358(1431):537-554.
Schultz, W., Apicella, P., Scarnati, E., and Ljungberg,
T. (1992). Neuronal activity in monkey ventral stria-
tum related to the expectation of reward. Journal
of Neuroscience, 12:4595-4610.
Steels, L. (2003). Evolving grounded communication for
robots. Trends in Cognitive Science, 7(7):308-312.
Sugita, Y. and Tani, J. (2002). A connectionist model
which unifies the behavioral and the linguistic pro-
cesses: Results from robot learning experiments. In
Stamenov, M. I. and Gallese, V., (Eds.), Mirror
Neurons and the Evolution of Brain and Language.
John Benjamins.
Taylor, J. G. (1999). Neural bubble dynamics in two
dimensions: foundations. Biological Cybernetics,
80:5167-5174.
Vitay, J., Rougier, N. P., and Alexandre, F. (2005).
A distributed model of spatial visual attention. In
Wermter, S. and Palm, G., (Eds.), Neural Learning
for Intelligent Robotics. Springer-Verlag.
Vygotsky, L. (1986). Thought and language. MIT Press,
Boston.
128