Information theory measures of grammatical goodness-of-fit

Reasons for the declining use of information-theory-based statistics were reviewed, and were found to reflect overly grandiose or overly constrained interpretations of distributional uncertainty measures. The range of definitions of H as a characterization of dispersion was extended to cases in which the scope, mode or specification, and unit of count were not limited to single terminal units. Information statistics may be generally described as reflecting the fit between a given grammar and a given language sample.Preparation of the article was partially supported by a grant from the U.S. Office of Education to the Early Education Research Center, University of Chicago.This article is based on papers given by the senior author at the Biennial Meeting of the Society for Research in Child Development, Spring 1969, and at the University of Iowa Conference on Language, October, 1970.