Cognitive models based on rules and symbols typically require a high degree of handwiring.
One way of avoiding the hand-wiring of rules is to rely on connectionist
networks for modeling some of the core processes in language acquisition. However,
connectionist models have problems with scalability and generativity. Ideally, we want a
model that will keep the strengths of both symbolic and connectionist models, while
overcoming their respective shortcomings. One model that may be able to do this is a
connectionist extension of the Competition Model which relies on the lexical item and
lexical categories as ways of coordinating processing and learning on the levels of syntax,
audition, articulation, and semantics.