The use of the t-test in Shakespeare scholarship

Abstract

Abstract In a recent article, the New Oxford Shakespeare scholar Hugh Craig used the t-test in statistics to compare the uses of one hundred function words across plays, authors, genres, characters, and periods in Shakespeare’s era. Craig drew some strong conclusions; for example, that authorship is a greater differentiator between plays than genre. Using both theoretical arguments and the results of further experiments, this note challenges Craig’s analysis, showing that he applied the t-test even when it is not clear that its premises are satisfied, failed to apply necessary controls, and, moreover, that his reasoning contains a fundamental error which invalidates his interpretation of the results.