In order to represent my data in a more complex and informative way-- and influenced by Joanna Drucker's warnings that graphs too easily give a false impression of objective truth-- I devised procedures for various graphic design decisions which generate the aesthetics of each graph from the data it describes.

Size

Each book is represented by a width of 3/100ths of an inch (in Excel's calculation); the number itself is chosen because it seemed to produce graphs of the appropriate size. The field is thus a square 6.24 inches in diameter, to represent the 208 books under consideration, and the diameter of each pie chart is determined by the number of books associated with the motif in question. Because my goal is to indicate the relative presence of each motif in the field as a whole, no particular attention has been paid to maintaining exact measurements as the images have been extracted from Excel or resized in Scalar, so long as all graphs are subject to the same processes and thus all relationships of scale are intact.

Color

My colour choices are also meant to qualitatively suggest further information about the relationship depicted. I wished to avoid graphs whose labels said that a motif featured a 50/50 split, but whose starkly-constrasted colours nonetheless implied that male and female books could be easily distinguished from each other. Instead, I wanted my reader's ease in distinguishing between male and female authors to be directly connected to the ease with which a reader in the 1790s might have been able to accomplish the same task if given only the information that a certain motif was present. 

Accordingly, each slice's proportions of cyan and yellow are based on the percentages in the graph. A 50/50 split is just a green circle, with each half 50% cyan and 50% yellow. Or, in a pie chart with a 73/27 split dominated by men, the 'male' slice will have its colour defined as 73% cyan and 27% yellow. The 'female' slice will have those proportions reversed, 73% yellow and 27% cyan.

Cyan and yellow, specifically, were chosen partly for convenience and partly for gendered political reasons. First, the choice of CMYK over RGB: given that these graphs were being generated for the web, it would have been preferable to use RGB colourspace, rather than CMYK; indeed, some graphs aren't 'perfectly' the appropriate percentages of cyan and yellow precisely because Excel must approximate those values in RGB. However, because RGB is defined on a scale from 0 to 255, rather than 0 to 100, generating colour values from percentages was a complicated and messy process. Because the resulting graphs were also less attractive than the cyan and yellow versions, I decided in favour of CMYK.

Second: an astute reader may have noticed that this decision has been framed as cyan-and-yellow vs blue-and-green, with no mention of magenta or red. The gendered association of blue with boys and pink with girls is anachronistic in the 1790s and antiquated in the present, so it has been avoided. Sample graphs with blue for female authors (and yellow for male) proved excessively confusing to test readers, and were discarded.