Compensation requires both input filtering and target enlargement

What it is

Interaction compensation for pathological tremor usually needs two techniques applied together: filtering the raw input signal (suppressing the high-frequency oscillation component in the trajectory) and enlarging the hittable target area (expanding the effective click/tap zone). Either one alone is usually insufficient — pathological tremor's amplitude is too large for target enlargement alone to cover the full range of displacement, while its frequency and amplitude are too irregular for filtering alone to handle without sacrificing a large amount of usable responsiveness.

Why it happens

Filtering can suppress the oscillation in the signal, but at the cost of latency — and because pathological tremor's amplitude is far larger than physiological tremor's, averaging out an oscillation that big requires a longer filter window, which means a larger latency penalty. Relying on deeper filtering alone to cover the full amplitude makes the interaction noticeably laggy or even unusable. Target enlargement works on an independent dimension: it doesn't change the oscillation amplitude in the raw signal at all, but loosens the criterion for "landing inside the target," so that even the oscillation left over after filtering is likely to still fall within the enlarged zone. The two techniques address the signal-level and geometry-level parts of the problem separately and don't conflict — combined, each can be dialed to a moderate, acceptable strength instead of either one being pushed toward its own limit.

Studying it

Accessibility interaction research typically uses two paradigms to validate compensation: one collects real motion traces from people with pathological tremor (or simulates them with synthetic noise of known amplitude and frequency) and compares hit rate and completion time before and after filtering and enlargement; the other is a usability study where patients try different parameter combinations (filter strength × target enlargement factor), with hit rate, mis-tap rate, and subjective ease of use combined to find the best operating point.

Methodological note: artificial jitter simulated by having healthy participants add noise on purpose has a noticeably different spectral signature from real pathological tremor (the latter tends to be more concentrated and regular in frequency) — evaluation results based only on healthy participants with artificial jitter cannot be taken directly as evidence of effectiveness for the pathological-tremor population.

Where it stops holding

Neither filter strength nor enlargement factor is "the more the better." Excessive filtering makes the interaction feel sluggish and detached from the user's actual intended motion; excessive target enlargement causes adjacent targets to overlap, creating new mis-taps, especially where the interface already has high information density and limited spacing between targets. The right parameters for each technique are person-specific and layout-specific, and need to be tuned to the actual tremor amplitude and the interface's own layout — there is no single set of values that works across all contexts.

Applying it

• Expose filtering and target enlargement as two independently adjustable parameters in accessibility settings, rather than bundling them into a single, non-separable "tremor assist" toggle — this lets users tune each to their own tremor characteristics.
• For operations that need to preserve some real-time feel — dragging, drawing — keep filtering strength lower than for simple tap-confirmation operations, and shift more of the compensation burden onto target enlargement; single-tap confirmation actions can tolerate stronger filtering.
• To verify: using the same set of real or high-fidelity simulated pathological-tremor traces, test hit rate and completion time under filtering-only, enlargement-only, and combined conditions, and confirm the combination is significantly better than either technique alone rather than assuming they simply add up.

Related

• Same group: A8.16.1 Intention tremor grows stronger as the hand nears the target · A8.16.2 Pathological tremor amplitude far exceeds ordinary tolerance design · A8.16.4 This population has a higher mis-tap rate, and undo beats confirmation
• Nearby: A8.15.5 Input smoothing introduces latency, conflicting with directness · B1.02 Deviations from Fitts' Law on touchscreens
• Search terms: tremor compensation · target enlargement · input filtering · motor-impaired pointing