Experimental design and analysis code for "Gaze Behavior Reveals Expectations of Potential Scene Changes"
Please refer to the manuscript Gaze Behavior Reveals Expectations of Potential Scene Changes (under review) or the Anonymized OSF Project for details about this project. The experimental code is based on Psychtoolbox-3 and is written in MATLAB. The parsing of the eye tracking data (recorded with an Eyelink 1000+) and the analysis code is written in Python.
Download the Stimuli data from the OSF Storage and place both the video and image files in the Stimuli folder (no subdirectories). It might be necessary to specify the full/absolute path to this folder in runLPA.m, line 32. The recorded eye-tracking data will be stored in the EDF folder, and the participant and trial information in the Data folder.
We used EyeLink Software to convert the EDF files to the ASC files uploaded to the OSF Storage. The parse_LPA_5s.py script takes the raw data as input, parses the EyeLink messages, and generates writes the eye-tracking data of the first 5 seconds of stimulus presentation in the LPA_5s_SUBJID_eye.csv.gz and the corresponding trial information in the LPA_5s_SUBJID_data.csv.gz flies for all participants (SUBJIDs). The parsed data is then evaluated (ideally on a high-performance computing cluster since the NSS calculations and the tolerance radius to determine which object was foveated can lead to a high computational load) first with the evaluate_metrics_psychsci_hpc.py script, whose outputs are subsequently used by evaluate_metrics_psychsci_nss_hpc.py. The resulting LPA_5s_SUBJID_eval_rad05_all_hpc.csv.gz files are then used to analyze the effect and produce the figures reported in the manuscript in the psysci_figures.ipynb notebook. The object-based analysis of the fixations is partially performed in the fix_obj_summary_psycsci.py script, which also creates summary statistics used in psysci_figures.ipynb (also uploaded to OSF).
All figures and statistical results in the manuscript can easily be reproduced directly in the browser by running the psysci_figures.ipynb notebook in the interactive interface of Google Colab. The notebook is also available in the Analysis folder of this repository.
If you want to reproduce the results on your local machine, it is sufficient to install the required Python packages with pip install -r requirements_min.txt and run the psysci_figures.ipynb notebook in a Jupyter environment.
If you want to use the full capabilities of the analysis scripts, including the PfC rating script (requires OpenCV), the parsing of the data, and the eye movement event detection (requires REMoDNaV) you will need to install the additional packages listed in requirements_all.txt.
Will be added after the double-blind review process.