Skip to content

SightLab Eye-Tracking Data: Best Practices for Fixations, Saccades, and Dwell Time

SightLab produces four complementary datasets, each meant for a different level of analysis. The recommended workflow depends on whether the researcher needs high-level summary statistics, time-resolved gaze behavior, or object-level event reconstruction.

Below is the recommended way to work with each file, when to use which, and how to integrate them.

📁 1. Understanding the Files & Their Intended Use

A. Trial_Data (high-frequency sample-level data)

  • Logged at headset update rate (e.g., 90 Hz) or user-set rate.
  • Contains every sample: gaze intersection, combined/individual eye gaze vectors, fixation state, saccade state, head position, pupil diameter, etc.
  • Best for:
  • Precision analysis
  • Custom fixation/saccade algorithms
  • Reconstructing gaze paths
  • Machine learning or time-series modeling

→ This is your rawest stream-level dataset.

B. Trial_Timeline_Dwell (object-level dwell timeline)

  • Time-stamped events showing:
  • When gaze entered an object
  • Duration of dwell
  • Number of fixations during the dwell interval
  • Also supports custom flags/events.

Best for:

  • Object-level engagement
  • Dwell/onset timing
  • AOI/ROI studies
  • Time-to-first-fixation validation

→ “What objects were viewed, when, and for how long?”

C. Trial_Timeline_Fixation_Saccade (per-event fixation/saccade table)

Introduced in SightLab 2.5+.

Includes per-event:

  • Fixation start/end time
  • Fixation duration
  • Fixation dispersion angle
  • Saccade amplitude
  • Saccade peak & average velocity
  • Δ angles between fixations/saccades

Best for:

  • Fixation sequence analysis
  • Scanpath reconstruction
  • Saccade dynamics
  • Micro-behavior analysis

→ This is the highest-quality event-by-event file, already segmented by SightLab’s internal detection algorithms.

D. Experiment_Summary (per-trial summary metrics)

Contains:

  • Fixation count
  • Dwell count
  • Total dwell time
  • Average view time
  • Saccade averages
  • Time to first fixation
  • Trial length
  • Custom trial-level metrics

Best for:

  • Quick comparisons across trials/conditions
  • Statistical summaries
  • ANOVA/GLM-type analysis
  • High-level experiment reporting

✅ 2. Which File To Use? (Best Practices)

Here is the general rule:

Goal Recommended File
Compare conditions, summarize behavior Experiment_Summary
Object-level engagement or AOI analysis Trial_Timeline_Dwell
Fixation/saccade sequence analysis Trial_Timeline_Fixation_Saccade
Low-level gaze reconstruction, custom event detection, ML models Trial_Data

🔍 3. Should You Use Trial Summary Metrics or Detailed Timelines?

Use Trial Summary WHEN:

  • You want per-trial averages
  • Fixation counts per trial are enough
  • You don’t need to customize the detection algorithm
  • Fast analysis/visualization matters

The summary metrics are computed from the same underlying events, but SightLab already:

  • handles noise thresholds,
  • applies dispersion/duration filters,
  • merges micro-fixations,
  • handles missing samples gracefully.

These summaries are validated against the event timeline files.

Use Detailed Timelines WHEN:

  • You need the exact sequence of fixations and saccades
  • You want to visualize or analyze scanpaths
  • You require per-fixation metrics (dispersion, amplitude, saccade velocity, etc.)
  • You want to build custom metrics (e.g., revisits, transitions, Markov modeling)
  • You need to reconstruct time-aligned object viewing sequences

Use Raw Trial_Data WHEN:

  • You want to:
  • Recompute fixations (e.g., using your own IDT/I-VT/I-HMM algorithm)
  • Apply your own smoothing filters
  • Analyze per-sample gaze noise
  • Combine gaze sampling with external devices

This file is not typically used for standard experiments unless the lab has custom algorithms.

🧩 4. How to Integrate the Data Sources (Recommended Pipelines)

  1. Use Experiment_Summary for group-level analysis.
  2. Use Trial_Timeline_Fixation_Saccade to check or extend fixation analytics.
  3. Use Trial_Timeline_Dwell to connect fixations to semantic objects.
  4. (Optional) Look at Trial_Data only for troubleshooting or custom work.

Pipeline B — Object-Centric Behavior Analysis

  1. Start from Trial_Timeline_Dwell.
  2. For each dwell period:
  3. Link to fixation events via timestamps (the files are time-aligned).
  4. Analyze fixation duration distribution per object.

This is ideal for:

  • advertising research
  • UX/UI object engagement
  • product placement studies
  • medical image ROIs
  • educational content analysis

Pipeline C — Scanpath & Eye-Movement Dynamics

  1. Use Trial_Timeline_Fixation_Saccade exclusively.
  2. Construct:
  3. fixation–fixation transition matrices
  4. saccade amplitude/velocity distributions
  5. scanpath entropy
  6. sequence clustering

Pipeline D — ML / Deep Learning / Custom Algorithms

  1. Start from Trial_Data (raw samples).
  2. Apply:
  3. noise filtering
  4. clustering-based fixation segmentation
  5. custom dwell models
  6. GLM/HMM/Transformer sequence analysis
  7. Optionally compare results to:
  8. SightLab’s precomputed fixation/saccade results.
Was this page helpful?