Visual information is acquired mainly during fixations
Aliases: fixation · foveal encoding · gaze pause
What it is
Looking is an alternation of fixations and saccades. A fixation is a relatively stable pause with the fovea on a location, commonly about 200–400 ms in adult reading and UI search. High-spatial-frequency detail—letter shape, icon differences, status text, fine tick marks—is encoded almost only during those pauses. The eyes do not finish fine discrimination while traveling from A to B. “Having seen” a place is, mechanistically, closer to “having fixated there” than to “the scan path crossed that coordinate.”
Why it happens
Only the central 1–2° of the fovea has cone density high enough for fine spatial resolution. Seeing a new object clearly requires aiming the fovea at it and then holding still long enough for photoreceptor integration and early cortical extraction of orientation and spatial frequency. A saccade is ballistic; peak angular velocity can reach hundreds of degrees per second, smearing the retinal image. The visual system therefore reserves fine encoding for the following fixation. Fixations are not perfectly still: microsaccades, slow drift, and tremor continue, but their amplitude is far smaller than a targeting saccade. Eye-movement analysis clusters samples into a fixation on the basis of that relative stability.
Whether a UI region entered high-resolution processing therefore depends on whether a long enough, stable enough fixation landed there, not on whether a heatmap was brushed by a trajectory. A pause shorter than recognition needs leaves only a coarse peripheral impression.
Studying it
Reading, visual search, and free viewing of scenes are the classic ways to treat fixation as an encoding window. Video-based or head-mounted trackers record landing position, duration, and sequence; area-of-interest (AOI) hits ask whether a critical field was encoded. Typical independent variables: eccentricity, type size and contrast, crowding, and task instruction (read aloud, find an error, look freely). Typical dependent measures: first-fixation latency, fixation duration, number of fixations, skip rate, and later recognition or reading accuracy. Treating completion time as “they saw it” misses skips: people can hit the right button quickly without ever forming a readable fixation on its label. The detector itself rewrites the conclusion—velocity-threshold (I-VT), dispersion-window (I-DT), or hidden-Markov segmentation change when a fixation starts and ends. Sampling below about 60 Hz, or heavy filtering, swallows short fixations. The rhythm of forced laboratory viewing also does not transfer unchanged into a product’s saccade–fixation cycle.
Where it stops holding
Low vision, macular disease, foveal damage, or some intraocular lenses shift fine encoding toward the parafovea, loosening the mapping from landing point to “seeing clearly.” Uncalibrated trackers, spectacle glare, or dirty lenses can label a stable fixation as jitter. Infants and some neurological populations fixate less stably; adult reading duration cuts do not apply. In very dim light, integration time lengthens and fixations stretch on their own—that is photon statistics, not greater care. Treating fixation as attention, or a saccade path as already-read content, uses the encoding window for the wrong job.
Applying it
- Place text, status, and easily confused adjacent icons that must be discriminated at a size and location where people will pause, rather than assuming a passing gaze equals reading.
- When assessing readability, collect fixations, not only task time: a recognition-length fixation on a critical field is more informative than “the eyes went by.”
- Verify with AOI analysis on the target device and real vision conditions (corrective lenses, low contrast, outdoor glare): did the first fixation land in the region, and was it long enough to recognize?