GazeBase: 12,334 Longitudinal Eye-Movement Recordings from 322 Subjects

A 37-month longitudinal study collected 12,334 monocular eye-movement recordings from 322 college-aged subjects using an EyeLink 1000 tracker at 1,000 Hz. Subjects completed seven tasks per session, including fixation, saccade, reading, video viewing, and gaming tasks. The dataset was created by Henry Griffith and is described in a corresponding data descriptor manuscript.

Use Cases

• Training gaze estimation models based on high-frequency (1,000 Hz) monocular recordings.
• Studying longitudinal changes in eye-movement patterns based on data collected over 37 months.
• Analyzing saccade dynamics based on horizontal and random oblique saccade tasks.
• Developing attention models for reading or free-viewing scenarios based on the described task battery.
• Benchmarking biometric identification algorithms based on longitudinal gaze data from a consistent subject pool.

Strengths

• Large scale with 12,334 recordings from 322 subjects.
• Longitudinal design with data collected over a 37-month period across nine recording rounds.
• High temporal resolution from a 1,000 Hz sampling rate eye tracker.
• Structured protocol with seven distinct tasks and pre-task calibration to ensure data quality.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.
• Data is monocular (left eye only), which may limit certain binocular vision analyses.

Provenance

Source

Henry Griffith

Collection Method

Data was collected using an EyeLink 1000 eye tracker in a controlled experimental setting with a calibration and validation protocol before each task.

Time Range

Data collection spanned 37 months across nine recording rounds.