Revisiting Performance Models of Distal Pointing Tasks in Virtual Reality

TL;DR

The paper addresses the lack of consensus on distal pointing performance models in VR/AR and questions whether traditional $ID$ formulations suffice for modern hardware. It introduces a new data-collection methodology on a hemispherical grid to broaden the range of $(\alpha,\omega)$ values and tests multiple models, including $ID_{\text{ANG}}$, $ID_{\text{ANG}^3}$, and $ID_{\text{DP}}$. The main finding is that a simple, angular, Fitts'-law–style index of difficulty, $ID_{\text{ANG}}$, provides the best overall fit across easy and hard distal pointing tasks, making the two-part model unnecessary. The methodology and results have practical implications for predicting distal pointing performance in VR/AR interfaces and offer a scalable framework for evaluating future 3D interaction models across diverse hardware.

Abstract

Performance models of interaction, such as Fitts Law, are important tools for predicting and explaining human motor performance and for designing high-performance user interfaces. Extensive prior work has proposed such models for the 3D interaction task of distal pointing, in which the user points their hand or a device at a distant target in order to select it. However, there is no consensus on how to compute the index of difficulty for distal pointing tasks. We present a preliminary study suggesting that existing models may not be sufficient to model distal pointing performance with current virtual reality technologies. Based on these results, we hypothesized that both the form of the model and the standard method for collecting empirical data for pointing tasks might need to change in order to achieve a more accurate and valid distal pointing model. In our main study, we used a new methodology to collect distal pointing data and evaluated traditional models, purely ballistic models, and two-part models. Ultimately, we found that the best model used a simple Fitts-Law-style index of difficulty with angular measures of amplitude and width.

Figures (13)

• Figure 1: The selection task used in the preliminary study.
• Figure 2: Linear regression plot using the data from the preliminary study for the $ID_{\text{DP}}$ model.
• Figure 3: Linear regression plot that shows the left (blue) and right (red) regression models for ID = 0.32
• Figure 4: Linear regression plot that shows the left (blue) and right (red) regression models for ID = 2.807, the best fitting two-part model from the preliminary data.
• Figure 5: Linear regression plot using the data from the preliminary study for the $ID_{\text{ANG}}$ model.