ANSWERING THE CONNECTIONIST CHALLENGE - A SYMBOLIC MODEL OF LEARNING THE PAST TENSES OF ENGLISH VERBS

Supporters of eliminative connectionism have argued for a pattern asso ciation-based explanation of language learning and language processing . They deny that explicit rules and symbolic representations play any role in language processing and cognition in general. Their argument i s based to a large extent on two artificial neural network (ANN) model s that are claimed to be able to learn the past tenses of English verb s (Rumelhart & McClelland, 1986, Parallel distributed processing, Vol. 2, Cambridge, MA: MIT Press; MacWhinney and Leinbach, 1991, Cognition , 40, 121-157). In this article we critically review Rumelhart and McC lelland's as well as MacWhinney and Leinbach's ANN models and conclude that they do not succeed in the assigned task of learning the past te nses of English verbs. In order to answer their challenge to the symbo lic processing approach, we present our symbolic pattern associator (S PA) - a general-purpose pattern associator that can learn to associate arbitrary discrete patterns. We carried out several experiments with the SPA using the same set of verbs that was used in MacWhinney and Le inbach's simulation with more realistic training and testing procedure s. The SPA outperformed the connectionist models by a wide margin in t he accuracy of learning, and successful inductive generalizations to u nseen verbs. Our SPA has very natural and psychologically realistic ex planations to many psychological effects such as U-shaped learning cur ve, and is much closer to human subjects in predicting past tenses of the pseudo-verbs. In contrast to ANNs, whose internal representations are entirely opaque, the SPA can represent the acquired knowledge in t he form of production rules that allow for further higher-level proces sing and integration, resulting in linguistically realistic associativ e templates for irregular verbs and production rules for regular verbs . In the light of these findings, we conclude that eliminative connect ionists' vision of cognition as simple pattern association and pattern recognition without symbolic representation is inadequate. Pattern as sociation as such does nor imply rule-less or cue-based models of lang uage acquisition or of human learning in general.