Searching for style in modern American poetry

This essay examines a corpus of Modern American Poetry using Zeta and Iota, two recently developed measures of textual difference (Burrows 2007), and some alternatives to them that move beyond the most frequent words of the text, which have been the traditional focus of computational stylistics. These measures concentrate our attention on moderately frequent or rare characteristic words, most of which are content words. These highly characteristic words lead us back to the text, back to questions of interpretation and style, and highlighting such words in texts emphasizes their dense concentration and helps us to visualize how and why intuitive perceptions of stylistic difference are possible.