Skip to content

Reporting eye-tracking data quality

A manuscript-ready quality section should describe the measurement quantities that matter for the analysis rather than dumping every possible tracker statistic.

Minimum data-quality items

Report, when relevant:

  • the native/analysis coordinate system and unit;
  • how validation targets were presented and segmented;
  • target-referenced accuracy and its summary level;
  • the precision operationalization (RMS-S2S, SD, BCEA) and BCEA probability when used;
  • nominal sampling frequency and whether sampling was irregular;
  • empirically observed/effective frequency or interval statistics when needed;
  • when long-gap diagnostics are reported, distinguish the number of long intervals from the estimated number of nominal samples lost within those intervals;
  • the definition and amount of data loss;
  • whether loss causes such as blinks or tracker invalidity were distinguishable;
  • preprocessing that can change these metrics;
  • pre-specified review/exclusion criteria and the number of affected units;
  • sensitivity analyses when conclusions may depend on quality decisions.

Suggested wording

Gaze-position accuracy was quantified as target-referenced Euclidean error. Precision was characterized separately using RMS sample-to-sample displacement and spatial SD; 68% BCEA was additionally reported as an area-based dispersion measure. Sampling behavior was summarized from timestamps using effective frequency, median inter-sample interval, and interval jitter alongside the device's nominal sampling rate. Data loss was defined as unavailable gaze coordinates or explicit invalidity and was summarized with loss fraction and missing-run duration. Pre-specified thresholds triggered review rather than automatic deletion, and exclusions were reported separately.

What not to write

Avoid claims such as “data quality was good” without an operational definition, using nominal Hz as proof that the realized stream was regular, reporting a single unspecified “precision” number, or describing a vendor calibration score as if it were a universal measure of measurement uncertainty.

References

The package terminology follows the 2023 minimal reporting guideline for eye-tracking studies (Dunn et al.; published in Behavior Research Methods 2024) and Niehorster et al.'s 2026 Behavior Research Methods tutorial on determining eye-tracking data quality (doi:10.3758/s13428-026-03039-4).