Evaluating fitts' law performance with a non-ISO task

We used a target-selection task to evaluate head-tracking as an input method on a mobile device. The procedure used a non-ISO Fitts' law task since targets were randomly positioned from trial to trial. Due to a non-constant amplitude within each sequence of trials, throughput was calculated using two methods of data aggregation: by sequence of trials using the mean amplitude and by common A-W conditions. For each data set, we used four methods for calculating throughput. The grand mean for throughput calculated using the division of means and the adjustment for accuracy was 0.74 bps, which is 45% lower than the value obtained using an ISO task. We recommend calculating throughput using the division of means (and not the slope reciprocal from the regression model) and with the adjustment for accuracy. We present design recommendation for non-ISO tasks: Keep amplitude and target width constant within each sequence of trials and use strategies to avoid or remove reaction time.