Decomposing Pupil Dynamics to Reveal Attention-Dependent Processing of Emotional Faces

The aim of the dataset is to investigate how attention to emotional faces influences pupillary dynamics. The dataset includes both behavioural and pupillometry data collected from 35 participants who performed an emotion detection task involving happy, angry, and neutral facial expressions under different experimental conditions. In addition to the main task, a control condition was included to preserve visual input while minimizing emotional relevance. Trial-wise behavioural performance is provided in the file "Master_Accuracy_Trials.csv", which includes participant identifiers, emotion category, experimental condition, and response accuracy coded as 0 (incorrect) and 1 (correct).The file "Master_Pupil_Features.csv" contains detailed trial-level pupil data along with reaction times and six decomposed features capturing distinct phases of pupil dynamics. These features include constriction onset, constriction amplitude, latency of contriction amplitude, pupil value during dilation, constriction rate, and dilation rate. In addition, the file includes the full pupil time course for each trial, represented by columns "Pupil_Data_1" to "Pupil_Data_275", sampled at 8 ms intervals and preprocessed prior to analysis.The pupillary response was decomposed into biologically meaningful components to capture the temporal dynamics underlying autonomic and cognitive influences. All pupil measurements are baseline-corrected and reported in millimetres (mm), with temporal measures in milliseconds (ms) and rate-based measures expressed as change in pupil size over time (mm/ms). The code used for preprocessing and feature extraction is publicly available at: https://github.com/Sangramjit/Decomposing-Pupil-Dynamics-to-Reveal-Attention-Dependent-Processing-