Figure 3. Manual sampling is labor-intensive, so our samples are not huge, and our error bars leave considerable wiggle room. But it is safe to say that we have found no evidence that the broad trends in HathiTrust are produced merely by library purchasing patterns. Minor fluctuations, such as the suspiciously sharp drop at , for instance, may well be library-specific. But the broad decline in the proportion of fiction written by women is, if anything, even more dramatic in this larger sample.
There is one final source of skepticism to address. Some critics of distant reading complain that library-based samples are too narrow.
But others object that they are too broad. Do we really need to care about every novel ever published?
Shouldn't we care more about novels with cultural significance? So we checked the yearly fiction bestsellers reported by Publishers Weekly.
This sample covers a shorter timeline and displays slightly different patterns, but here too, the prominence of women in fiction broadly declines from the first half of the twentieth century to the period To tell the truth, checking bestsellers is probably overkill. A massive change in academic libraries would be important in itself.
But readers who have any doubts about the significance of libraries can set their doubts aside: this was a broad trend, and it had some impact on every aspect of literary life. Figure 4. The authors of yearly bestsellers from Publishers Weekly.
So why did it happen? Tuchman and Fortin offer several hypotheses. First, they emphasize that early-nineteenth-century fiction was dominated by women because novel-writing was not yet a high-status career. Additionally, after the job conditions of novelists improved" 9. Gentrification stories of this sort are familiar. Ellen Miller Casey's close analysis of The Athenaeum comes to an opposite conclusion, at least about reviews in one periodical .
Finally, Tuchman and Fortin also acknowledge that intellectual careers other than "novelist" were opening up to women. To fully map this expansion would require many different kinds of social evidence.
But if we stick with the evidence of authorship, fiction is one of the few parts of the library where representation of women seems to have declined. If we consider all other categories, collectively, we see an enormous expansion interrupted by the much briefer sort of pause in the middle of the twentieth century that we might have expected. It is not hard to see how expanding opportunities on this scale might have lured women away from the novel.
Books by women, as a fraction of all books in HathiTrust. The category "other" is mostly nonfiction, although it also includes small amounts of poetry and drama. Social causality is never easy to untangle. Tuchman, Fortin, and Casey, for instance, come to opposite conclusions about the impact of nineteenth-century reviewing.
In this paper, we won't completely resolve the tension between competing explanations either. But the occupational shifts that Tuchman and Fortin have described — male "gentrification" of the novel, and the opening of other careers for women — are very plausible explanations of the changes we see down to, say, After that point, a different pattern appears: participation in fiction and nonfiction move together, down and then spectacularly up, possibly in response to the broader fortunes of feminism.
From forward, it appears that authorship of fiction is shaped less by social factors that guided women toward or away from fiction in particular than by broader attitudes toward femininity and work.
But this is not to imply that ideological pressures of that broad kind were absent before Evidence gathered from BookNLP does hint that the decline from to also had an imaginative or ideological component. The change wasn't simply that more men decided to become novelists, or that women found other opportunities — but that fiction itself became more attentive to men. We can illustrate this by considering the space on the page that writers allot to fictional men or women.
A shy web designer thinks he has found the man of his dreams online. When we compare models closely, it turns out that women and men surprise! Any society needs a leader, but how does that leader cope with absolute power? In some modeling runs we made that distinction, but not in the final version represented above. Some of these archetypes are seemingly gender-specific, which is perhaps a result of gender roles that span back millennia. To address this question, it may be useful to juxtapose another model trained on the same data.
Women certainly write about women more than men do. In books written by men, women occupy on average only quarter to a third of the character-space. This gap between the genders is depressingly stable across two hundred years and may be, for some readers, the biggest story coming out of this inquiry.
Words devoted to description of women, broken out by author gender. HathiTrust and the Chicago Novel Corpus are folded together here. But there is also another, subtler story in this image, because the declining prominence of women as characters between and remains visible even after we separate volumes written by men and women.
This suggests that the underrepresentation of fictive women inside books cannot be completely explained by the underrepresentation of women writers on the shelves of libraries. When we separate authors by gender, we find that women were becoming less prominent even in books by women across this century. As we have said, this is only a subtle shift, and we are not yet sure how much should be inferred from it.
But in reality, more research is needed to understand what happened. For instance, how much of this change was due to shifting literary ideals the masculinity of high modernism , and how much to the displacement of domestic fiction by adventure, detective, or Western genres? It is also important to remember that any "masculinization of fiction" between and took place against the backdrop of a broader social change that gave women a vastly larger proportion of the nonfiction book market than they had held in the nineteenth century. We are in the habit of reading literary change as a direct reflection of deep cultural shifts.
But the relation between literature and other aspects of culture can also be indirect or even reversed: for instance, if women abandon the novel because they are finding wider opportunities elsewhere, fiction and nonfiction can move in opposite directions. Leaving the sheer allocation of space in books aside for now, let's move to consider the characters themselves. Up to this point we have taken public signs of gender at face value, as if their meaning remained constant. But we know, in fact, that characters are gendered to different extents and in different ways.
This is a complex topic, and we won't be able to capture all its nuances on a scale of centuries. But we can ask, at least, how strongly a public, binary conception of gender shaped other aspects of characterization. Were fictive men quite different from fictive women, or were the differences between characters mostly unrelated to conventional signs of gender?
And how did the answer to that question vary across the timeline? Questions about the relative strength of different relationships are a place where quantitative methods shine. The only thing that makes this problem tricky is that we can't define in advance which aspects of character might be surreptitiously gendered.
Instead, we need a capacious way of representing lots of different aspects of character at once, in order to ask how well gender explains them collectively. A bag-of-words representation has worked well for many similar problems. We represent each character by the adjectives that modify them, the verbs they govern, and so on — excluding words that are names of gendered roles like girlhood or boyhood, wife or husband, and personal names.
Then we show a model some of the characters, labeled with grammatical gender. The model will learn what it means to be "masculine" or "feminine" purely by looking at the tacit assumptions expressed in words associated with the character. It can then use those patterns to make predictions about other characters it hasn't yet seen. If the model turns out to be accurate, we can conclude that practices of characterization were powerfully organized by a binary conception of gender: even apparently innocuous words like say "smile" or "grin" appear to tacitly predict gender.
When we attempt this comparison, a clear long-term pattern emerges: the implicit differences between masculine and feminine characters get steadily harder to discern from the middle of the nineteenth century to the beginning of the twenty-first.
Figure 7. How easily can grammatical gender be predicted from ostensibly ungendered evidence? The blue dots are scattered in columns because we ran fifteen different models in each decade, randomly selecting characters each time, and classifying them using the most common words in that group of characters. That's not a lot of words many characters are described fairly briefly , so these models are never as accurate as a model of genre which uses a whole volume for each inference.
Find helpful customer reviews and review ratings for That's So Cliché (Gender Transformation Erotica Book 12) at tecnichenpulchper.ml Read honest and unbiased. Caroline is a computer programmer for WCI, a local software company. She has an intense loathing for men and hates the fact that her world is dominated by the .
This picture may align somewhat better with expectations about progress than the first half of this article did. As we move from forward, characters are less and less clearly sorted along a masculine-feminine axis. That sounds like the kind of convergence we might have expected to see as the notion of separate spheres was challenged.
But if there were a slight increase in accuracy from to , it would fit received narratives about domestic ideology reasonably well. Virginia Woolf, at any rate, believed that "the sexes drew further and further apart" as the nineteenth century began.