Area and bubble cursors are fundamentally incompatible with dense layouts

What it is

Area and bubble cursors have a fundamental limitation in dense layouts: when several targets sit close together, expanding the pointer region quickly covers multiple candidates at once, so it cannot both substantially enlarge the effective target and keep a unique hit. Tiny distance differences between nearby targets also make the candidate switch frequently.

Why it happens

A bubble cursor relies on a "nearest and unique" geometric relationship. Dense objects shrink nearest-neighbour boundaries, driving the allowable expansion radius toward zero; any larger radius includes two candidates. Pointer noise, irregular object boundaries, and dynamic layout changes cross the narrow Voronoi boundary between them, flipping the algorithm's output between neighbours.

This limitation is not a matter of mistuned parameters — it is a mathematical dead end. The bubble cursor's core assumption — "there exists a radius that, when expanded to it, covers exactly one uniquely nearest target" — only holds when nearest-neighbour distance exceeds some usable threshold. In a dense layout, nearest-neighbour distance itself approaches zero, which means the range of radii satisfying that assumption also approaches zero: the algorithm degrades into an ordinary point cursor, while people still carry the expectation that it "should" expand. The result is often worse than using a point cursor from the start — people perceive the assistance as intermittently working, which is harder to build a stable expectation around than assistance that is simply absent.

Studying it

Vary spacing, object shape, overlap, density, and speed, measuring candidate stability, error, switching, and selection time. Report failures in the densest regions and near boundaries, not only averages, since sparse regions would otherwise mask the limitation. The study design should treat "nearest-neighbour distance approaching zero" as an explicit level of the independent variable rather than letting density increase smoothly, or the point at which degradation occurs is easy to miss.

Where it stops holding

Density does not mean pointing cannot be improved at all: actual hit regions can still be enlarged, layout can be reflowed, people can zoom or filter, and keyboard focus can help. The limitation is specific to unique-nearest-target geometric expansion, not to every assistance technique. When objects carry a clear hierarchical priority, semantics rather than pure geometry can also resolve the ambiguity.

Applying it

• Disable or shrink area expansion in high-density regions, relying on zoom, filtering, or explicit selection instead.
• Show candidate changes and never auto-commit a selection when the candidate is not unique.
• How to check: monitor layout via nearest-neighbour distance and switch to a more predictable input strategy past a density threshold; specifically test whether people still believe the assistance is active once it has degraded to an ordinary point cursor near the critical density.

Related

• Same group: C1.13.1 Area and bubble cursors enlarge activation without changing visual target size · C1.13.2 Benefits are substantial when targets are sparse
• Nearby: C1.12 Target snapping and sticky targets · C1.14 Side effects of pointing assistance
• Search terms: bubble cursor · dense targets · Voronoi boundary