Juan's RLAC Notes

---

On Gender, Aliens, and their Sentiments

written with Aryamaan Dholakia

In this piece, we analyse the differences between the choices made in two sets of works of science fiction set in similar contexts from two different authors. We ask ourselves whether the differences in the personal circumstances of our authors is reflected in how they reinterpret the core concept that spans their works, and whether such a conclusion can even be reached by looking at limited variables such as their gender and their historical period. On our exploration, we dialogue with Mandell's "Gender and Cultural Analytics: Finding or Making Stereotypes?" and discuss whether analyses like ours can giv eus a glimpse on how gender is historically defined, or if gender is itself a valid category through which to analyse texts.

Leigh Brackett (l) and Ray Cummings (r)

We chose the common topic of the encounter between humans and aliens in the books Enchantress of Venus (1949) and Black Amazon of Mars (1951), from female writer Leigh Brackett, and compared them to Brigands of the Moon (1931) and War Nymphs of Venus (1941), from male writer Ray Cummings. A surface reading of the synopsis of the books lets us infer that they share the common motif of human explorers finding themselves outside of Earth battling groups of aliens, however, given the different gender identities of the authors and the almost twenty-year difference between their works, we can ask ourselves whether the way they characterise their heroes, and their violent encounters with extraterrestrials differs similarly.

To explore this, we looked into how the books express the common topic of war and whether this can be related to the gender of the authors, then we looked at how books’ characters are gendered and what qualities are usually associated with these genders in both authors, and we performed sentiment analysis with the intent of having a sense of the development of the plot and try to establish whether it is significantly different between the authors.

The first question that arose was whether the gender of the authors resulted in different depictions of the characters and plot in their novels. We began analyzing the author’s writing styles in terms of their word choices. Since our texts consisted largely of negative words around the topic of inter-species conflict in sci-fi, we assumed wars and battles to be likely recurring scenarios.

Concordance plots of “war,” “battle” and “death” used by both authors. Brackett has a much higher frequency of these words as compared to Cummings.

When we looked at concordance plots of words such as “war” and “battle” some interesting patterns were brought forth. Brackett tended to use words like “war,” “battle,” “death” and “gun” more frequently than Cummings, expressing war in more explicit terms.

Concordance plots of “gun,” “dark” and “light” used by both authors. Brackett uses “dark” more than Cummings, while Cummings uses “light” more.

Additionally, the word “dark” was incorporated several times by both authors but “light,” in contrast to dark, was used by Cummings several more times than Brackett, highlighting his tone to be lighter and relatively uplifting as compared to Brackett. However, analyzing such words may not provide us any insight into how their gender played a role as their jargon could simply be a result of their writing style, which may be different for every author, regardless of their gender. As well, our preliminary reading (that Brackett, a female, is predominantly more "dark" and explicit in her writing), is contrary to the "stereotypes" Mandell describes when addressing stylometry - and might altogether be too broad a category to divide a corpus on. However, we then asked ourselves whether the gender and descriptions of the characters are related to the gender of the author, as character choice, we hypothesised, might be a choice closer to the author's own identity.

At first, we hypothesized that Cummings, a male, would prefer male protagonists, while Brackett, a female, would show off powerful female protagonists. However, when we analyzed the texts of each author individually, we discovered a fascinating observation: Cummings used words like “she” and “her” significantly more than Brackett, conveying the prevalence of female characters in his novel; exactly the opposite of what was expected. Is, then, the description of the character more related to the historical period and genre of the authors than their gender?

Selected collocates of Cummings' main male (Gregg, left) and female (Anita, right) characters. Complete screenshots here.

To describe his primary male character “Gregg,” Cummings incorporated words like “suitable,” “sharpest,” (collocates sorted by stats) “prince,” “safe” and “right” (collocates sorted by frequency). However, to describe an essential female character “Anita” in his novel, he used “darling,” “withdrew,” “weakling,” “squirmed,” (collocates sorted by stats) and “whispered” (sorted by frequency) clearly highlighting a certain gender-based hierarchy in Cummings’ characters.

Selected collocates of Brackett's main male (Stark, left) and female (Ciara, right) characters. Complete screenshots here.

Similarly, Brackett used words like “reckoned,” “wrestle,” “whipped,” (collocates sorted by stats), and “death” to describe a male character “Stark” in her novel, while she used incorporated “restrained,” “filth,” “cheered” (collocates sorted by stats) and “dying” (sorted by frequency) as collocates of a female character “Ciara.” A notable exception for Ciara could be the collocate “murderous”, however, these findings make us think that instead of being related to gender, the description of the characters is related to the conventions of the sci-fi genre at the time.

Concordance plots on AntConc showing the reference to male and female characters in both author’s works

When analyzed with all texts together, we learned another commonality: that both authors relied heavily on male characters, even with the titles including words like “Enchantress,” “Amazon” and “Nymphs.”

Selected collocates of him/his and she/her on a corpus of both authors. Complete screenshots here.

On further analysis of the gender of characters, we found that “him” and “his” had collocates, when sorted by frequency, like “voice,” “sword” and “struck” in contrast with words like “she” and “her” having collocates “whispered,” “shook,” “against” and “will;” words that signify weakness and fragility. Moreover, if the collocates were sorted by stats, then the male collocates become “slamming,” “disarming,” “hating” and “unruly,” while the feminine collocates were “unwillingly,” “loyally,” “wept,” “ratlike” and “scared”, which supports the hypothesis that the gender of the authors does not impact the descriptions associated to gendered pronouns. These observations, instead, could depict a pattern for how males are treated in science fiction novels of the early 1900’s compared to females: assigning themes of strength, violence and courage to men but weakness, loyalty and submission to women.

Lastly, we asked ourselves if the genders of the authors made a difference on how the plots of their books are developed. In order to explore this, we performed two kinds of sentiment analysis on the corpus: we first used the afinn lexicon to assign a sentiment score ranging from -5 (very negative), to 5 (very positive) and added these scores across 100-line sectors of each one of the books. The results can be seen in this plot:

Plot of sentiment score (positive-negative) per sector of the books. [This is an interactive graph. Hover over it to see the values.]

We can observe that all of the plots show an overwhelming concentration of negative words throughout their length, which to us might be a consequence of the high frequency of words used to describe the violent encounters between humans and aliens, and that words describing combat are ranked negatively in the afinn lexicon. However, we can notice that all the books seem to be organised in a series of negative spikes, which we think can mean that the books are composed of many short but rapidly intensifying violent segments. Therefore, we could associate these similarities in plot development to the common topic of conflict than to gender. However, as we didn’t find any clear differences between the two authors, we decided to use the nrc lexicon to pair every word in our corpus with an emotion (anger, anticipation, disgust, fear, joy, sadness, surprise, trust) and count these emotions per 100-line sectors in order to obtain a plot summary in terms of emotions.

Number of words per sentiment across 100-line sectors of "Brigands of the Moon" (above) and "War Nymphs of Venus" (below) by Ray Cummings.
[This is an interactive graph. Hover over it to see the values.]

Using this visualisation, we can try to confirm our suspicions about the nature of the negative words in the corpus. In Cummings, words relating to “disgust”, “joy”, and “surprise” are not common, instead being words related to “fear” the most frequent throughout the length of the books. We can identify again the spikes in conflict in the books through increased frequencies of words related to “anger”, “anticipation”, and “sadness” and the lulls between these spikes through larger amounts of “trust” words.

Number of words per sentiment across 100-line sectors of "Black Amazon of Mars" (above) and "Enchantress of Venus" (below) by Leigh Brackett.
[This is an interactive graph. Hover over it to see the values.]

In Brackett, similarly, “fear” is a prevalent sentiment in both books, and “joy”, “surprise”, and “disgust” are, too, rather infrequent. However, the spikes in conflict are represented only by an increase in “anger” and “sadness”, not “anticipation”, and words related to “trust” have a seemingly constant frequency throughout the books.

We can first observe that “trust” is constant in Brackett, while only appearing outside of conflict in Cummings. We could infer that in Brackett there is a partnership in combat between multiple characters, which can explain why “trust” appears even in the violent spikes of her books, while, on the contrary, we could infer that Cummings prefers to place a lone “hero” character in violent sections of his books, given the lack of words pertaining to “trust” during the peaks in “anticipation” and “anger”. Another difference is the lack of words related to “anticipation” in Brackett’s peaks of “anger” compared to Cummings’, which could mean a difference in how they portray conflict. Given that, as seen before, most of the words related to violence are collocates of male characters of masculine pronouns, we cannot associate these differences to the authors' own identities in relation to their characters.

Our last observation is about the endings of the works. Given that not all words are counted as pertaining to an emotion by the nrc lexicon, we can assume that the thicker parts of our streamgraphs correspond to sectors of the story with a larger emotional load. We can notice that Cummings’ endings are noticeably more emotionally charged than the rest of the books, while Brackett’s books end in levels of emotion similar to the other parts. We could infer from this that Cummings prefers grand endings, while the climax of Brackett’s books happens slightly before the ending. This coincides with the "stereotypes" Mandell criticises in her article, however our corpus is not large enough to make a generalisation about "grand endings" being a characteristic of male authors.

Looking at the concordance plots, collocates, and the sentiment analysis of the corpus, we can clearly infer that both authors adhere to the perception of gender and the general style of writing that defined science fiction at the time. Our analysis provides a glimpse into the perception of gender prevalent at the time the books were written: male characters as the physical heroes of intense combat scenes against aliens, and females performing submissive and loyal supporting roles. Nonetheless, we also found interesting contrasts in the styles of our authors, as Brackett used “darker” language than Cummings, contrary to the stereotypes criticised by Mandell, however the size of our corpus is not large enough to make a general statement about the stylistic differences between male and female authors. That literary context seems to be the primary factor influencing style makes us wonder whether gender is a valid variable to analyse corpora with: although distant reading can give us a sense of how gender is defined in a particular time, there always exists the risk of over-reducing the identity of an author to a pronoun or a binary.

See the R notebook for this assignment.
Download the AntConc screenshots.

ready for grading, november 2nd, 2021