Perception bias in assessing rodent welfare during euthanasia and anaesthesia

Perception bias in assessing mouse welfare during euthanasia and anaesthesia This repository contains the anonymised data, preprocessing scripts and statistical analysis code for a mixed-methods pilot study on perception bias in the assessment of mouse welfare during anaesthesia and euthanasia. Study overview Professionals experienced in mouse experimentation watched four 75-second videos of mice undergoing either CO₂ euthanasia or isoflurane anaesthesia induction (one black C57BL/6 and one white CD-1 mouse per procedure). Each video was presented with a correct or incorrect framing label (“Anaesthesia” or “Euthanasia”), depending on the randomised group (A or B). While watching, participants wore eye-tracking glasses (SMI Eye Tracking Glasses) and verbalised their thoughts (think-aloud). After each video, they rated escape behaviour, fear, respiratory distress and overall distress on a 0–9 scale. A semi-structured interview followed. Repository contents Raw data Raw Data - Raw Data.txt: BeGaze sample-level export with pupil diameter (right eye, mm), gaze coordinates, trial timing, video identifiers and tracking quality, for all recorded participants.results_try_long_26.11.2025.xlsx: survey export with distress ratings and demographic and professional variables. Processed data long.xlsx: long-format rating dataset (one row per participant × video × question).aoi_use.xlsx: AOI-level eye-tracking metrics (net dwell time, revisits) per participant, video and AOI.pupil_long_5s_reduced.csv: pupil data aggregated into the five analysed 5-second bins (5–10, 20–25, 35–40, 50–55 and 65–70 s). Scripts (R Markdown, in the order they should be run) Survey_preprocessing.Rmd: recoding of demographic variables and creation of the long-format rating dataset.Pupil_Dilation_preprocessing.Rmd: participant exclusions, baseline correction (first 4.5 s), aggregation into 5-second bins and the reduced dataset.analysis_likert_final.Rmd: linear mixed-effects model of distress ratings (procedure × framing label), post-hoc comparisons and Table 1.AOI_DwellTime_Analysis.Rmd: median net dwell time per AOI and condition (Supplementary Table S1) and Figure 2.AOI_Revisits_Analysis.Rmd: median revisits per AOI and condition (Supplementary Table S2), the negative-binomial mixed model and Figure 3.Pupil_Dilation_descriptives.Rmd: descriptive pupil figures (Supplementary Figures S3–S8).Pupil_Dilation_modelfitting.Rmd: mixed-effects models of pupil diameter and distress ratings. Participants and exclusions Thirty-six volunteers were recruited; one was ineligible, leaving 35 for the rating analyses. Six participants were excluded from the eye-tracking analyses for insufficient calibration accuracy and one for a missing video trial, leaving 28 (Group A, n = 15; Group B, n = 13). The exclusions are applied in the scripts. Anonymisation Participants are identified only by codes (P01–P36). No directly identifying information is included. Interview transcripts are not shared, to protect participant confidentiality. Software R version 4.6.1, with the packages nlme, lme4, lmerTest, glmmTMB, emmeans, DescTools, dplyr, tidyr and ggplot2. Eye-tracking data were exported from BeGaze 3.7 (SensoMotoric Instruments).