Neutral Drift and Cumulative Memory: A Wright–Fisher Model of Thematic Evolution in Online Horror Fiction

Cultural evolutionary models offer a principled framework for understanding how digital genres stabilise over time. In this paper, we apply a modified Wright–Fisher model to the thematic evolution of creepypasta narratives published on the Creepypasta Fandom wiki between 2010 and 2024. Drawing on a corpus of 13,128 stories tagged across 176 moderated thematic categories, we test whether the observed power-law distribution of themes can be reproduced by a neutral drift process augmented with cumulative archiving. Standard Wright–Fisher simulations fail to replicate the empirical long tail, as minor categories extinguish too rapidly. When a mechanism of cumulative memory is introduced—whereby each generation contributes a sampled archive that persists alongside the living population—the model successfully reproduces both the dominance of major themes and the stable persistence of rare ones. Maximum-likelihood fitting confirms that both the empirical distribution and the simulated archive follow a power law with comparable exponents, supporting the qualitative agreement between model and data. These results suggest that the thematic structure of the Fandom corpus is best understood not as a snapshot of active production, but as a sedimented historical record whose composition reflects processes of neutral drift operating over a cumulative substrate. We discuss implications for platform studies and the modelling of digital literary evolution.