Summary: Frontiers in Computational Neuroscience www.frontiersin.org October 2010 | Volume 4 | Article 24 | 1
Original research article
published: 04 October 2010
previous theoretical works that have shown that specific forms of
mixed selectivity can be harnessed to perform computation such
as complex sensorimotor transformations (Zipser and Andersen,
1988;Pouget and Sejnowski,1997;Pouget and Snyder,2000;Salinas
andAbbott,2001) and to model serial working memory (Botvinick
and Watanabe, 2007) and visuomotor remapping (Salinas, 2004a)
(see Discussion for more details).
Rules are prescribed guides for problem solving and flex-
ible decision making and they vary in the degree of abstraction.
Examples include conditional (arbitrary) sensorimotor associa-
tions (if red light, then stop), task rules (respond if two stimuli
match),strategies for decision making (if win,stay; if lose,switch).
We assumed that the rule in effect is actively maintained by a recur-
rent neural circuit. In particular we hypothesized that the neural
correlate of a rule is a self-sustained persistent pattern of activity