On touch, star precision is limited by finger width
Aliases: finger occlusion · half-star miss · touch rating
What it is
On a touchscreen, a row of stars cannot honestly collect more precision than a finger is wide. When the contact patch covers a star and a half — or two stars — the system cannot claim half-steps or adjacent whole stars. The finger-width limit is hit geometry, not how many bins the scale has on paper. A pointer can hit a half-star; the same five-plus-halves on a phone often turns “I meant four” into five. Grain must be stepped down by input device, not by how many stars the mock shows.
Why it happens
A finger’s contact ellipse is usually wider than the gap between stars. The system uses the ellipse centroid, while the user sees the stars hidden under the pad. Occlusion blocks a visual closed loop, so people tap from motor memory and overshoot into a neighbour. Half-stars halve the target width, Fitts difficulty steps up, and errors go from “off by a star” to “off by a half in a random direction.”
Enlarging stars helps until five large stars blow a narrow screen and spacing collapses, at which point the ellipse covers two again. “Bigger stars” and “more gap” fight on a narrow width. Drag-to-rate overshoots more than tap; the bin at lift-off is often not the bin that was fixated.
Studying it
Compare pointer versus touch, tap versus drag, whole versus half stars, and log intended bin versus landed bin.
Independent variables: device, target width, half-steps, tap versus drag. Dependent variables: hits on the intended bin, direction of neighbour bias, number of retries, whether people report “my hand was in the way.”
Use a real phone width. Occlusion is visual; a mouse has no pad.
Where it stops holding
A stylus hotspot is closer to a pointer, so half-stars can return. A tablet in landscape opens the gaps; the same component can fail on a phone and limp along on a tablet — branch on width, not on a single “is it touch” flag. Accessibility zoom that scales spacing as well as glyphs can help; scaling glyphs only makes the crush worse. Two large up/down buttons are outside this limit because they were sized for a finger; using stars to collect an attitude that only needed two bins is a precision crisis you chose.
Applying it
- Default to whole stars on touch, with each star’s hit width at least a finger edge; leave half-stars to pointers or genuinely large layouts.
- Do not make dragging the only input; provide tappable whole-star targets, and treat drag as acceleration at most.
- On a narrow screen, replace five squeezed stars with a stepper or two large buttons plus a numeral.
- How to check: with a real finger, hit a named bin (the fourth star) twenty times. Neighbour hits above chance mean widen the target or drop half-steps.