When banking on meaning is not (yet) money in the bank: Explorations in connectionist modeling.
Steve Joordens, Derek Besner · Journal of Experimental Psychology Learning Memory and Cognition · 1994
A number of reports claim that humans perform lexical decisions faster to words with many meanings than to words with only one meaning. It is a challenge to simulate this ambiguity effect with a parallel distributed processing model because activation of ambiguous words produces competition at the semantic level; this ought to lead to less rather than more efficient processing of ambiguous words. Despite this problem, the present simulations show that it is possible to produce an ambiguity effect when the network settles into a learned semantic pattern through use of a proximity factor. However, competition is not completely countered by proximity as the network fails to settle into a learned semantic pattern for ambiguous words on more than 50% of the trials. Further simulations show that manipulations that increase the proportion of correct settling result in the loss of thi ambiguity effect