Touch bias varies with finger angle and screen position

What it is

Touch bias is not a constant vector: it varies with finger angle, screen location, grip, and handedness. A thumb approaching an edge differs from an index finger tapping from above, changing the relation between contact centroid, occlusion, and subjective aim; one global offset cannot fit every region.

Why it happens

Wrist posture and device support change finger projection and contact ellipse, but the screen's edges and corners are the most easily overlooked special case in this variation: the closer to the device bezel, the narrower the range of approach angles the finger can actually take, constrained by the physical edge itself and by the holding palm, so it is often forced into a shallower, more edge-hugging angle than it would use at screen centre — and that angle change directly reshapes the contact ellipse and the relationship between the aim point and the centroid. Precisely because the bias itself varies continuously with position, the appropriate model treats the whole screen as a two-dimensional error field — each local region carrying its own offset vector — rather than assuming one direction and magnitude shared across the entire screen; approximating that continuously varying field with a single global vector necessarily leaves far larger residual error near the edges than at the centre.

Studying it

Collect intended and reported points over a screen grid, orientation, handedness, and common grips; map the error field and repeatability. Include portrait one-hand use, two-handed holding, and desktop tapping rather than extrapolating from one laboratory posture. A concrete test fits two models — a single screen-wide offset vector, and per-cell local offset vectors over a grid — and compares their residual error specifically at the edges and corners; a markedly smaller residual for the local model confirms the bias really is a position-dependent field that cannot be summarized by one global constant.

Where it stops holding

Complex real-time personal correction can be unstable, opaque, and harmful when posture changes. For most apps, large targets, spacing, and clear feedback are more reliable; fine correction belongs only where the system has sufficient device data.

Applying it

• Validate edge, corner, and thumb-reachable regions independently, rather than treating error measured at screen centre as representative of the whole screen.
• Put small targets in visible, approachable places or offer magnification and alternative entry.
• How to check: measure aiming bias separately at screen centre versus edges and corners, and compare whether direction and magnitude agree; a clear mismatch means one uniform offset assumption cannot cover the whole screen.

Related

• Same group: C2.06.1 The user’s aim point is systematically biased above the touch point · C2.06.3 Offset compensation belongs at the system layer and should not be repeated by applications
• Nearby: C2.05 Finger occlusion · C2.02 Minimum touch target size
• Search terms: touch offset · grip · screen position