Abstract
A neural network model for Stroop test is proposed. Subjects experience conflicts when they are presented colored letters, and
there are differences between color and letter presented. Although,
there are, so far, several neural network models have been proposed,
all of them did not represent this kind of conflicts. The model
proposed here is explicitly implemented Winner-Takes-all circuits for
solving these conflicts. The model was consist of four modules, those
are a word perception, a color perception, a task switch, and an
output. Each module has its own winner-takes-all circuit to select an
appropriate candidate. The word superiority effect was implemented as
the initial value in the task switch layer. The letter in the word
perception and the color in the color perception layers have also
winner-takes-all circuits. The combination of the winner-takes-all
circuit can explain the reaction time and error type.
