Abstract

Motor tremor is among the most common barriers to effective pointer-based computer interaction for users with Parkinson’s disease, essential tremor, and related conditions. Dwell-based selection, where a user holds the cursor within a target zone for a fixed duration to trigger a click, is a widely used accommodation. But fixed dwell zones and fixed dwell times fail to serve a population whose tremor characteristics vary by time of day, medication cycle, and fatigue. This article examines adaptive dwell tolerance as an HCI design problem and surveys current approaches in research prototypes and production accessibility tools.

The Variability Problem

Tremor-affected users do not exhibit a single, stable pointing profile. Parkinson’s resting tremor, for example, oscillates typically between 3 and 6 Hz with amplitudes that can range from under 1 cm to over 3 cm depending on the individual and the moment. Essential tremor, by contrast, is an action tremor that intensifies during intentional movement. A fixed dwell zone of 32x32 pixels, which is the default in most accessibility settings panels, will be well inside the tremor envelope for some users and provide no accommodation for others. Microsoft’s Windows Accessibility settings and the macOS Switch Control panel both expose dwell time and zone size as manual sliders, but they treat these as one-time configuration choices rather than runtime-adaptive parameters.

Adaptive Zone Sizing via Motion Sensing

Research into real-time tremor characterization has explored using the cursor motion trace itself to estimate current tremor amplitude and frequency, then scaling the dwell zone accordingly. Work presented at ASSETS 2024 demonstrated a spectral analysis approach that computes a rolling FFT over the last 1.5 seconds of cursor movement, identifies the dominant tremor frequency band, and adjusts the acceptance radius to cover 95% of the user’s current tremor envelope. The key insight is that this adjustment can happen within a single interaction session without any explicit calibration step, because the tremor signal is always present in the motion data.

Temporal Filtering vs. Zone Expansion

Two distinct architectural approaches appear in the literature. Temporal filtering stabilizes the cursor’s displayed position using a low-pass filter, presenting a visually smoothed cursor while tracking the raw position for intent inference. Zone expansion instead leaves the cursor display unmodified but widens the target acceptance boundary. Temporal filtering creates a perceptual lag that some users find disorienting, particularly when they intentionally move the cursor quickly. Zone expansion avoids this perceptual artifact but risks unintended activations when targets are closely spaced, as in dense toolbars or text. Some prototypes, including the TremorTamer research tool from UCL, combine both, applying temporal filtering only when spectral analysis detects a tremor episode and falling back to direct mapping otherwise.

Production Integration Barriers

Despite solid research results, production operating systems have been slow to adopt adaptive dwell tolerance. The barriers are partly organizational, since accessibility settings are maintained by different teams than input subsystems, and partly technical. Estimating tremor in real time requires low-latency access to raw cursor samples at 100 Hz or higher, which conflicts with how most accessibility APIs are structured. Web-based approaches face additional constraints: the Pointer Events specification exposes pointer coordinates at display refresh rate, but applications cannot request higher-rate input streams. Closing this gap would require either OS-level middleware or browser API extensions, neither of which has a clear standards home as of mid-2025.