Is the Ontological Commitment to "Rules" Really Preferable to Linguistic Science? : A Critical Assessment of the Symbolist-Connectionist Debate on the "Learning of the Past Tenses of English Verbs

Abstract

This paper aims to assess critically the debate on the "learning of the past tenses of English verbs" between those symbolists who are proponents ofsymbol manipulation model of the language and those connectionists who are proponents ofparallel distributed processing model of it, thereby suggesting that between such extreme positions there is still a realistic position that linguists can take to blur the debate. It is when linguistics is conceived as a phmomenology ofthe language, as much as chemistry, for example, is a phenomenology of the nature. If the position is adopted, theoretical constructs such as "rules, " "schemes, " "patterns, " can be defended only for a negative reason: without them we could not describe linguistic phenomena correctly and effectively enough to draw any scientifically significant generalizations. To take this position is to take a perspective from which rules are no more than a heuristic device. This is, I claim, the most realistic position in which scientific identification of linguistics, or linguists' looser talks about the language, can be maintained.