Context
 Processor





 Meaning
 Processor








 Hidden units Hidden units








Orthographic Phonological
Processor Processor


Input
MAKE


Output
/mAk/


Figure 2-1. Seidenberg and McClelland's Word Recognition Model

 The model performed as in the simulations with 100 hidden units. Actually, the

"normal" model was configured with 200 hidden units, but for this simulation they

reduced a number of hidden units for the simulation. They reported that "with fewer

computational resources, the model was still able to learn; however, it performed poorly

on irregular words and non-words even after extensive training" (Seidenberg, 1992, p.