Work

Normative Reference Values for Saccadic and Fixation Eye Movements in Healthy Young Adults: Centile Norms, Split-Half Calibration and Test–Retest Reliability from GazeBase

Year: 2026
Type: preprint
Abstract: Saccadic and fixation measures are widely used in psychiatry and neurology, yet no published centile norms exist for expressing a new recording as a centile or deviation score. We derived norms from GazeBase v2.0 (EyeLink 1000, 1000 Hz, left eye), using Round 1 recordings from 322 healthy young adults aged 18–46, most (284/322) aged 18–25. Twenty-nine features were extracted from three tasks—horizontal saccades (HSS), random oblique saccades (RAN), and central fixation (FXS). Norms are empirical centiles (5th–95th). We tested age and sex effects with generalized additive models for location, scale and shape (GAMLSS), assessed calibration with split-half resampling (50 random splits), and estimated 4-month test–retest reliability in 136 returners (ICC(2,1), absolute agreement). Median main-saccade latency was 121 ms in predictable HSS and 199 ms in unpredictable RAN, with median peak velocity 527 °/s and 440 °/s, respectively. Six of the 29 features differed by sex after Benjamini–Hochberg correction (q < .05), notably fixation small-saccade rate (female median 0.27/s vs. male 0.13/s). Split-half calibration was near-nominal except for ceiling and floor variables. Four features had good reliability (ICC ≥ .75), 8 moderate, and 17 poor (ICC < .50); fixation duration and small-saccade amplitude were least reliable (ICC = .11). At retest, HSS latency increased systematically (+53.0 ms), whereas RAN latency was unchanged. The HSS norms are therefore first-exposure norms. These frozen v1 norms allow new recordings on the same EyeLink tasks to be expressed as centiles and deviation scores, but their clinical validity requires prospective validation in patient samples.
Author: Bowen Fu
Language: English
Open Access status: green
OpenAlex ID: W7226018889
Cites:
Cited by:
Related to: 10