On the History of Models in American Linguistics

Abstract

In this chapter, I investigate how the term ‘model’ appeared in American linguistics in the 1940–50s, essentially in the works of Z.S. Harris, C.F. Hockett and N. Chomsky, and how it took a mathematical turn gradually contributing to the mathematization of linguistics and the development of generative grammars. Several steps are examined: early uses of ‘model’ evolved from Bloomfield’s grammatical paradigms and Sapir’s patterning to Harris’s diagrammatic representations and process models. Doing so, Harris developed an axiomatic and algebraic type of mathematization of linguistics. With Shannon and Weaver’s work and Information theory, Markov chains and probabilistic methods started to be used for modeling linguistic structures. They were taken up by Hockett to build his General Headquarter as a generative grammar with a neuropsychological aspect. Finally, Chomsky’s criticism of Markov’s chains led him to a logico-mathematical conception of models and a computerized view of linguistics modeling. I examine to what extent the use of these different meanings of ‘model’ was a controversial issue, and how it evolved into the emergence of a common approach to models and modeling in linguistics in the 1960s.