Metacognitive Monitoring in Naturalistic Translation: a Signal Detection Approach
A novel signal detection approach was applied to translation behavioural data to assessmetacognitive sensitivity for complex, naturalistic tasks, using reformulation behaviour as animplicit proxy for the confidence response. Existing units of segmentation (Production units(PU)) were utilised to form trials made up of a Type-1 decision and a Type-2 response. Byusing AUROC2, metacognitive sensitivity was estimated for two language-pair corpora undera lower and upper bound on the response window: an English-Spanish group, where AUROC2= 0.618 (upper bound) and 0.567 (lower bound), and an English-German group, whereAUROC2 = 0.625 (upper bound) and 0.539 (lower bound). Under the upper bound, estimateswere reproducible across both corpora, agreeing to within 0.007. Individual differences inmetacognitive sensitivity were dissociable from translation accuracy and varied considerablybetween individual translators. Overall, reformulation behaviour during the drafting phasecarried information about the accuracy of individual productions, as judged against thetranslator’s own post-revision translation, and this information was not reducible to theiroverall accuracy.
Authors
- Aishvarya Raj (ORCID: https://orcid.org/0009-0003-7107-5261)
Publication Details
- Journal
- PsyArXiv (OSF Preprints)
- Published
- 2026-11-13
- DOI
- https://doi.org/10.31234/osf.io/tg2fy_v2
- Primary Topic
- Neurobiology of Language and Bilingualism
- Type
- preprint