In the initial mandate for the Armed Services Editions, no mention of gender was made. Initially, the purpose of the ASEs was to improve morale, to remind soldiers of why the US was fighting this war and to encourage them in that effort, and to help soldiers learn about the enemy. In 1944, however, with the passage of the Soldiers’ Voting Act, gender became an issue of contention in the composition of the ASEs. Several of the new criteria included specific mention of masculinity and the avoidance of all things feminine:
“The criteria for book selection includes the following definite aims:
a. Stress readability and the masculine viewpoint. […]
g. Provide recent fiction, avoiding only the mediocre, trashy, mawkish and books with a decided feminine interest.”
This emphasis on the “masculine viewpoint” is complicated by what historians of the ASEs are quick to report: the book most frequently requested by soldiers was Betty Smith’s A Tree Grows in Brooklyn, the story of a young girl’s coming-of-age. How does this institutional emphasis on masculinity square with the popularity of Smith’s novel? How is gender represented in this corpus?
There are several very easy ways to begin to operationalize and test this issue of gender in regard to the ASEs. Starting with the metadata, we can easily analyze the relative representation of female authors vs. male authors in the corpus. Of course, this is not an analog for gender representation in the books, but it does speak to a selection bias– and age-old gendered theories of gender, genre, and representation– that influenced the ultimate shape of the corpus.
Note: “NULL” values represent multiple authors or situations in which author gender was indeterminable (i.e., use of initials).
The results are hardly surprising. Books published by women make up only 12.6% of the entire corpus. One interesting course for further study might be to consider the percentage of books published by women nationally during the war years (if such data were available) to consider if this figure is, in fact, representative of national publishing trends. For now, however, suffice it to say that books written by women were not often chosen by the CBW Advisory Board when selecting books to send to soldiers, and books by women make up a tiny fraction of the total corpus. This makes the frequency of requests for A Tree Grows in Brooklyn all the more remarkable. It also suggests a disconnect between soldiers’ desires and institutional (or governmental) priorities.
Another way that we might consider the representation of gender in the books themselves is to consider gender pronoun usage and the relative frequency of gendered pronouns. Here, I calculated the relative frequency of masculine pronouns “he”, “his”, and “him”, and feminine pronouns “she”, “her”, and “hers.” This method would exclude books written in the first person (hardboiled detective fiction, perhaps), that might directly assume a “masculine [or feminine] viewpoint.” (The narrative implications of those SVA criteria are fascinating, aren’t they?) The results of this analysis are also quite clear: yes, it appears that the ASEs do, by and large, assume a “masculine viewpoint” and avoid “feminine interest.”
Feminine pronouns make up only 26.5% of the total gendered pronouns used in the corpus.
Most surprising to me in the relative pronoun frequency is how infrequently the pronoun “hers” is used: that is, a pronoun indicating possession. “Hers” represents just .09% of the total gendered pronouns.
OK, but is a pronoun really an accurate measure of gender representation across the corpus? It only speaks to the presence or absence of certain pronouns, which may or may not speak to the presence or absence of men relative to women– it does not address issues of form or style, or even the ideas that might be associated with masculinity or femininity (to say nothing of the problematic reduction of gender to a simple binary). So, I conducted a very very basic LDA topic model using MALLET, to consider how well these topics might map onto stereotypes of “masculinity” or “femininity.”
Working in broad, literary-historical stereotypes here, some of these topics might be considered more “masculine” than “feminine.” Topic 4 might be called a “naval” topic (“ship”, “sea”, “captain”, “boat”, “island”– and, yes, “men”). Topic 11 looks to be industrial, or work-related– (“use”, “work”, “import”, “increase”, “develop”, “power”). Topic 12 is concerned with issues of diplomacy (“new”, “state”, “war”, “general”, “american”, “king”, “govern”), and Topic 17 looks like a battle topic, maybe drawing from the high occurrence of Westerns in the corpus (“horse”, “run”, “dead”, “turn”, “fire”, “shot”, “kill”, “follow”).
Obviously, there are problems with topic models and I don’t want to generalize too much. The frequency of militaristic words or the language of statecraft might not be said to apply to issues of gender at all, but might rather meet the criteria of “Authentic military books of interest to any soldiers and the military books pertinent to his special branch of the service.” Regardless, only Topic 16 addresses issues of emotion, affect, or feeling– “love”, “heart”, “feel”, “thought”, “spirit”, “beauty”, “hope”, “soul”, “seem”. In his Macroanalysis, Matt Jockers has addressed the literary-historical stereotype that men write about war and women write about emotion– and a number of smart scholars have issued an important corrective to that analysis. But at least at this stage, this very simple list of topics would seem to align with Jockers’ analysis. However, unlike that large-scale study, these results would speak less to a sweeping literary history than to, perhaps, a selection bias, giving us more of an indication of what the CBW considers to be “masculine interest.” That is to say, what we see in these topics is less what men and women may write about, than what the CBW thinks men are more interested in reading about.