Daejeon, July 27–31
The nonhuman has been widely studied in the humanities and social sciences (Cohen & Lemenager, 2016). While its meaning varies among disciplines and contexts, in literary studies the term nonhuman generally encompasses “plants, animals, spirits, gods, and, more recently, machines” (Thomsen, 2021, p. 909). Within the digital environmental humanities—an emerging field that fosters interdisciplinary collaboration between digital and environmental humanities—the nonhuman world, as represented in fiction, media, archives, folklore, poetry, and narratives, has been widely explored through digital technologies (Cohen & Lemenager, 2016; Gould, 2017; Linley, 2016; Ryan et al., 2023; Travis et al., 2022; Zampaki & Karpouzou, 2025). Some DH studies have examined the narrative agency of human fictional characters at scale (Piper, 2024; Soni et al., 2023), analyzing human behavior and movement in literary texts to reveal how agency and mobility are represented in fiction published from 1800 to 2020. By extending this examination of agency to nonhuman figures, we decenter the human and arrive at a more ecologically aware understanding of narrative in ecofiction. Examining how representations of nonhuman animals and plants, and of their agency, span creative works can, in turn, illuminate how various human perceptions of the nonhuman world are constructed.
The corpus for the present study comprises 75 ecofiction books published from 1846 to 2002. The corpus is based on Dwyer’s introduction to ecofiction, Where the Wild Books Are, which includes an appendix of “100 Best [Ecofiction] Books” (Dwyer, 2010). The corpus consists entirely of English-language works, including translations from other languages. Seventy-five of Dwyer’s 100 works were in the HathiTrust Digital Library. The corpus was analyzed in a data capsule provided by the HathiTrust Research Center (HTRC), a secure, sandboxed computing environment that supports non-consumptive analytics—allowing researchers to run algorithms on in-copyright works while preventing viewing, downloading, or exporting the underlying text itself.
To identify nonhuman living entities—nonhuman animals and plants—we used BookNLP’s supersense tagging, which provides semantic information by assigning tokens to 41 WordNet lexical-semantic categories, including person, animal, and plant. Building on Ash et al. (2023), we categorize nonhuman actions into agentic (e.g., deciding, planning) and experiencer (e.g., feeling, sensing) verbs. To identify agentic and experiencer verbs, we used Word2Vec embeddings Codes are available at: https://anonymous.4open.science/r/dh2026-0DB0/
First, we extracted seed words for agentic and experiencer verbs from existing lists of agency and experience words in Ash et al. (2023). Using Word2Vec’s cosine similarity scores computed on the seed words, we labeled additional verbs in the corpus as having agency, experience, or both based on their semantic similarity to the seed words We used the following parameters: vector_size: 100; window: 5; min_count: 5; workers: 4. The similarity threshold was set to 0.5 in order to catch tokens within both agency and experience dimensions.
| lemma | supersense | verb | verb_type | sentence |
| tree | noun.plant | produce | agency | Each tree produced only one seed at a time , and when it reached maturity it exploded like a cannon shot . |
| grape | noun.plant | cure | agency | I d teach you what grapes cure the bite of deaf adders . |
| animal | noun.animal | understand | both | He said the animals understood what was happening , how they were dwindling . |
| rabbit | noun.animal | suffer | experience | Rabbits , like most wild animals , suffer hardship . |
Table 1. Nonhuman action examples.
We analyzed the collocations of nonhuman actors with agentic and experiencer verbs and compared them with those of human actors. Both human and nonhuman actors collocate more with agentic verbs than with experiencer verbs. Figure 1 suggests that nonhuman actors are depicted as agentic to a similar extent as human actors. However, human and nonhuman actions fall into distinct categories. Among the top 20 most frequent agentic and experiencer verbs, human actions are more likely to involve cognition and intellect (e.g., see, know, think, remember, understand, believe) (Rice & Newman, 2018). Nonhuman actions, on the other hand, are more embodied and physical (e.g., eat, leave, kill, survive, join, exist) (see Figure 1 and Figure 2).
Figure 1. Counts of the top 20 most frequent verbs with human and nonhuman subjects.
Figure 2. Proportions of the most frequent verbs with both human and nonhuman subjects.
This contribution explores the representation of various nonhuman living entities and their actions in ecofiction. It suggests differences in the representation of human and nonhuman actors and in their engagement in fictional narratives across agency and experiential roles. Examining how creative works depict nonhuman actions can reveal how human authors conceptualize and narratively construct the nonhuman world.