Gaze location is not attention or comprehension
Aliases: looking without seeing · covert attention · gaze is not understanding
What it is
An eye tracker reports where gaze landed: which patch of screen the fovea was aimed at. Attention is which stream of information is selected for processing; comprehension is whether that processing built a usable meaning. The three can come apart. Someone can stare at button copy while thinking about the previous error, or glance over a warning without encoding its constraint. Writing “was looked at” as “was noticed” or “was understood” treats a motor record as a cognitive conclusion.
Why it happens
Saccades send the fovea to candidate locations, but attention can shift before the saccade and can leave during a fixation (covert attention). A fixation may be a pass-through in search, capture by a high-contrast patch, or confirmation after the meaning is already known. Reading further needs dwell and a scan pattern; a short dwell can pick up only a glyph outline. A landing therefore proves optical alignment, at most an opportunity to have been selected—not that selection occurred, and not that a proposition was encoded. The busier the layout and the more practiced the task, the more pass-through and confirmatory fixations, and the wider the gap between landing and comprehension.
Studying it
Do not use total dwell on an area of interest as a comprehension index by itself. Combine time to first fixation, regressions, pupil or dwell thresholds, and a probe that can tell “seen” from “used”: can the person restate the constraint, or act on the warning. A control that replaces copy with unreadable text or removes the critical sentence tests whether the gaze pattern is almost unchanged—if it is, gaze was layout-driven rather than semantic. Concurrent or retrospective talk is only an aid, because it changes looking rhythm.
Where it stops holding
Under forced reading, proofreading, or search for a named target, gaze and attention couple more tightly, and landings are a safer attention index. Coupling loosens in infants, some clinical groups, and high-load dual tasks. Poor calibration, spectacle glare, and off-screen gaze create false landings. In advertising or games, gaze itself may be the commercial target and need not be promoted to “understanding”; claims that users were “informed” still need comprehension evidence.
Applying it
- Do not write a hot zone or fixation count as “users saw and understood”; the most it supports is “the fovea was aimed at this region.”
- For critical notices—fees, permissions, irreversible actions—add a behavioral or restatement check; treat gaze as side evidence of opportunity.
- If a control is fixated heavily and still skipped or mis-clicked, inspect wording and timing before making the control larger.
- Split “looked at” and “used” into two columns in review materials; do not conclude informed consent from the first column alone.