01 — Attention and gaze
Attention and eye movements are tightly coupled.
Attention can move without the eyes. A saccade — the rapid movement between two gaze locations — is nevertheless usually preceded by an attentional shift toward its target. Eye movements therefore make part of attentional orientation observable.
Eye tracking records these movements. A fixation indicates where the fovea points; the fovea is the central retinal region that provides the sharpest vision. Interpretation then connects the signal to the task, stimulus and an explicit cognitive hypothesis.
02 — Saliency and goals
Gaze combines image properties with the observer’s objectives.
Exogenous attention is captured by physical scene properties such as contrast, color, orientation or shape. This stimulus-driven influence is called bottom-up processing. Saliency models attempt to represent the visual priority it creates.
Task and prior knowledge exert a goal-driven influence called top-down processing. Semantic context, faces, age, expertise and personality also guide the next saccade. Modern priority maps combine both families of factors. The same painting produces different scanpaths — ordered gaze trajectories — when viewed for memorization or to locate a detail.
03 — Two complementary modes
First grasp the whole, then inspect the detail.
Ambient mode is classically associated with fixations below 180 ms followed by saccades above 5 degrees. It quickly captures spatial organization and the global impression of a scene, sometimes called its gist. Focal mode shows the reverse pattern: longer fixations, smaller saccades and deeper local processing.
Exploration often begins with an ambient-to-focal transition, after which both modes alternate as needs change. The K coefficient combines fixation duration and saccade amplitude to place this dynamic on a continuum without relying only on fixed thresholds.
04 — What this changes
A heatmap summarizes space and reduces the temporal dimension.
Interface analysis combines the most fixated regions with their viewing time, order and task instruction. A heavily viewed area may be important, ambiguous or simply difficult to decode.
Ambient and focal modes have been linked to dorsal and ventral brain pathways. This proposal remains a hypothesis about their neural signatures and provides an interpretive framework to test with dedicated neurological measures.
05 — From concept to decision
A heavily viewed region calls for a question, rather than a verdict.
On an offer page, sustained viewing of the price may reflect useful decision information or difficulty understanding it. I would examine first viewing, returns to the conditions and the timing of the choice, then ask participants what they wanted to check.
A saliency prediction helps prepare this spatial inspection. Explaining the sequence or comprehension requires observations of gaze, behavior or users’ accounts. That distinction helps select the right tool for each research step.