DIGITAL SCHOLARSHIP IN THE HUMANITIES, 2026 (AHCI, SSCI, Scopus)
This article proposes a reproducible mixed-methods framework for psychoanalytic film analysis that makes interpretive inference transparent, inspectable, and contestable while preserving a strict distinction between description and interpretation. Taking Poor Things (Yorgos Lanthimos, 2023) as its case study, it develops a locally regenerable workflow that produces a fixed-schema descriptive substrate of film form: visibility/access (mean luminance), activation (dense optical-flow intensity), temporal holding (segment duration), and contextual chroma (mean saturation). Distribution-calibrated cut points generate a regime map for navigation and anchor selection, while preregistered robustness diagnostics (+/- 10 per cent threshold perturbation and mean-mode agreement) define limits of generalization rather than validate interpretation. Across the analysed interval of thirty segments, eleven retain their regime assignment under perturbation, whereas within-segment mixture remains extensive, with mean-mode agreement at 20 per cent, suggesting that segment-level summaries often understate perceptual dominance. Interpretation therefore proceeds through an explicit warrant chain linking quantitative signature, formal mediation, and a bounded developmental-logic hypothesis, accompanied by clear scope conditions and evidence-compatible alternatives. Freud's stage terms are redeployed not as a linear model of character development, but as recurrent developmental logics that may overlap, interrupt, or reconfigure one another. Methodologically, the article offers a rerunnable descriptive workflow, a robustness-calibrated rule for anchor selection, and a vignette-based procedure for accountable interpretation. Analytically, it shows how Poor Things organizes access, containment, address, and agency in ways that can be interpreted with methodological discipline within digital film studies.