01 — What color shows
A spatial density built through aggregation.
A heatmap adds a contribution around every fixation, often weighted by duration, then aggregates and smooths those contributions. Hot areas therefore represent greater density under specific choices of kernel, bandwidth and color scale.
It quickly reveals a shared center of attention, an ignored region or a contrast between conditions. Its pedagogical power is also its danger: the image feels direct even though it results from many analytical decisions.
02 — What it removes
Time disappears when every gaze is stacked.
The map does not say whether A came before B, whether gaze returned three times, or whether a long fixation occurred early or late. Two people may cover the same regions in reverse order and produce nearly identical heatmaps.
It also hides individual differences when a group is aggregated. A red area may reflect moderate consensus or a small subgroup dwelling intensely. Without distributions or trajectories, those scenarios are indistinguishable.
03 — Context changes gaze
Instructions transform exploration of the same image.
Since Yarbus’s classic work, task has been known to transform scanpaths. Mills et al. showed that instructions change spatial and temporal parameters, with some effects evolving across viewing time.
Schütt et al. separated bottom-up influence from image properties and top-down influence from goals and prior knowledge. A hot zone may therefore reflect visual saliency, task relevance, comprehension difficulty or proximity to screen center.
04 — A toolbox of views
Combine space, sequence and dynamics.
A scanpath, or ordered gaze path, restores fixation sequence through numbered points and the saccades connecting them. A transition matrix summarizes movement between areas of interest. Time-binned heatmaps reveal how exploration shifts. The K coefficient combines fixation duration and saccade amplitude to place exploration between a global ambient mode and a local focal mode.
Together these views answer different questions: where information is sampled, how attention travels, when strategy changes and under which conditions that change appears.
05 — From concept to decision
Comparable maps require a shared protocol.
Before concluding that a variant attracts more gaze, I would check participant count, exposure duration, fixation weighting and color normalization. A map built from more data can look stronger without each person viewing the region more.
I would then report the proportion of people who viewed it, time to first fixation and transitions with neighboring regions. The metric would follow the question: visibility, discovery or comprehension.