This repository contains the datasets used in a study investigating how participants visually assess harmony in horse–rider combinations using eye-tracking and verbal evaluation.
Participants viewed five video stimuli across equestrian disciplines and rated harmony on a 0–10 scale while their eye movements were recorded.
The dataset integrates:
eye-tracking metrics
harmony scores
participant characteristics
qualitative theme coding
Dataset structure
1. harmony_clean_long_dataset.csv
Long-format dataset at the level of participant × video × AOI.
Includes:
Participant ID (anonymised)
Video_label (discipline)
AOI_label (11 anatomical regions)
Fixation metrics:
NOF (number of fixations)
DOF (duration of fixation)
TFF (total fixation time)
OOF (order of fixation)
HarmonyScore (0–10 rating)
Participant characteristics (Category, Level)
Qualitative coding variables (horse, rider, connection themes)
This dataset contains 1,650 observations (30 participants × 5 videos × 11 AOIs).
-
harmony_fixation_analysis_dataset.csv
Aggregated dataset at the level of participant × video.
Includes:
Fixation proportions per AOI
Principal Component Analysis (PCA) scores:
FRC1–FRC5 (fixation-based gaze strategies)
HarmonyScore
Video_label and Level_label
Qualitative theme proportions (Horse, Rider, Connection)
-
harmony_duration_analysis_dataset.csv
Same structure as fixation dataset, but based on fixation duration.
Includes:
Duration proportions per AOI
PCA scores:
DRC1–DRC5 (duration-based gaze strategies)
Data processing
Eye-tracking data were processed in Python. Fixation counts and durations were converted into proportional measures per participant and video to control for variation in recording length.
Principal Component Analysis (PCA) with Varimax rotation was applied to identify broader gaze strategies, using standardized AOI variables and an eigenvalue > 1 criterion.
Ethical considerations
All data were anonymised prior to analysis. No personally identifiable information is included.
