An unsupported hand jitters several times more than a supported one

What it is

Aim a fingertip in free air at a virtual bead and the cursor draws a circle. Rest the wrist on a table edge and that circle collapses to a point. Unsupported pointing jitter is about amplitude: the sway of a free hand near the endpoint is typically several times the sway of a wrist, elbow, or fingertip that has somewhere to land. The question is not whether the hand trembles; it is whether the radius still fits the bead.

Why it happens

Without a distal rest, shoulder to fingertip is a cantilever. Postural tremor (roughly eight to twelve hertz) and the co-contraction used to correct it ride that beam; the longer the moment arm, the larger the endpoint. A wrist on a lip shortens the arm to the forearm; a fingertip on a surface shortens it to zero. The same neural noise, millimetres apart by a factor of several.

To hold a small mid-air target people freeze the shoulder and elbow with extra co-contraction. Amplitude drops for a few seconds, then the muscle puts the noise back. Filtering can smooth the cursor but not the visible hand; once hand and cursor split, the person chases, and the loop amplifies the shake. Jitter is first mechanical and neural gain at the endpoint, and only then a software lag.

Studying it

Same target, same distance, four supports: free, wrist rest, elbow rest, light fingertip contact. Take a position time series from the fingertip or controller. Report RMS or peak amplitude, not hit rate — hit rate scores a lucky shake as success.

Independent variables: support site (none / wrist / elbow / surface), moment-arm length, angular target size. Dependent variables: endpoint RMS, peak excursion, throughput, co-contraction if EMG is available.

Hold the point for ten to twenty seconds. The first two seconds of “looks steady” are not steady-state amplitude; the factor-of-several gap lives in the steady state.

Where it stops holding

A controller is heavier than a fingertip: inertia can damp high-frequency shake while growing a slow drift. The “several times” comparison has to be on the same device. Parkinsonian or essential tremor raises the factor and the value of a table edge. A tap that leaves immediately never sees steady-state amplitude; this leaf is about hover, drag, and alignment. A ray turns the same jitter into angular error and leverages it with distance — that is an angular lever, not these millimetres on a cantilever.

Applying it

• For millimetre alignment, give the wrist or elbow a landing, or grow the target to several times the unsupported RMS.
• Do not make dwell confirms and fine drags into sustained free-air holds. Use a trigger, a pinch, or an edge.
• Cursor filtering is fine until the hand and the cursor visibly disagree; then people chase the shake.
• How to check: one small target, a few seconds of free aim versus wrist-rested aim. If free RMS is still several times the rested value and the task must sit on the target, the precision budget was never left — grow the target or supply a rest, rather than another filter layer.

Related

• Same group: N2.12.2 A support surface collapses 3D pointing into 2D sliding · N2.12.3 Mid-air handwriting is far worse than writing on paper · N2.12.4 Borrowing a real desk or wall can buy the precision back
• Nearby: N2.08 Fatigue in Spatial Input · N2.03 Ray Pointing
• Search terms: unsupported pointing jitter · postural tremor · RMS tremor