Symbolic vs. Connectionist Accounts of SLA

Abstract

Generative Linguists following (Chomsky 1965, 1982, and 1995) have argued that grammar is innate, exist in brain as domain-specific module, and is transmitted by genetic inheritance. They also argued for rule-governed nature of language and language acquisition. They have resorted to many arguments to justify these claims among which, the complexity of language, the poverty of stimulus and the lack of negative evidence can be mentioned (Cook and Newson, 1996). For some decades these theories have been widely accepted as being not controversial and even undeniable. But within the last two decades these ideas have been strongly disputed by emergentists, construction grammarians, associationists, and connectionists. These approaches differ strikingly from other accounts of language learning. They do not believe that language acquisition is the result of internalizing language rules. Instead, in these approaches, the importance is put on construction of associative patterns (Mitchell & Myles 2004). Among these approaches to language the last one, connectionism, is greatly distinguished by others in its research techniques. The development of neural network computer simulations or what has come to be known as Artificial Neural Networks (ANN) has helped researchers in this approach to make stronger claims about the nature of language and language acquisition. This has helped them to move from making abstract and obscurant theories toward entangling with concrete and physical realities. The present paper is an attempt to compare and contrast the symbolic and connectionist approaches to second language acquisition.