Branches were produced by agglomerative clustering in R using cor() and hclust(), and the resulting tree cut to highlight major divisions. The central node (labeled '1of1') contains all topics; each subsequent branch point represents a point of convenience for considering groups of topics, with the first number being an arbitrary name and the second indicating the number of groups formed by that cut.
Seven topics are excluded from this figure because they were especially non-content-bearing, e.g. topics comprised of non-words formed by unsuccessful character recognition during the scanning process, or topics indicating merely a non-English language.
NB: These dissertations were completed, as best I could determine, for PhD programs within the Doctoral Consortium of Rhetoric and Composition.