Against Token Fetishism: Replication Package for a Corpus-Sociolinguistic Study of ekki/eigi in the Icelandic Saga Corpus

This Zenodo record contains the article 'Against Token Fetishism: Cluster Robustness and Categorial Disambiguation in Historical Corpus Sociolinguistics' together with its accompanying dataset and replication package. The article examines the methodological problem of treating raw token counts as stable sociolinguistic evidence in historical corpus research. Using the distribution of ekki and eigi in direct speech from the Icelandic saga corpus, it argues that apparent gendered patterns become substantially weaker once cluster structure, attribution categories, and manual disambiguation are taken seriously. The study therefore proposes a cautionary approach to historical corpus sociolinguistics: tokens should not be fetishized as self-interpreting evidence, but must be situated within textual, philological, and categorial constraints. The accompanying dataset/code archive contains the extraction scripts, derived datasets, manual classification files, robustness checks, final result tables, appendices, reproducibility notes, and a bundled copy of the Saga Corpus used for replication. It allows users to reproduce the main analyses concerning ekki/eigi, direct speech, gendered attribution, and cluster robustness. The article PDF and the dataset/code ZIP are deposited together in this record so that the argument and its empirical basis remain permanently linked.