Mizobuchi et al. [13] analyzed the effects of walking and keyboard sizes when performing text input. A stylus was used to interact with a virtual keyboard displayed by a PDA. Their results show that the walking condition had no significant effect on text input speed, but lead to a higher error rate. Furthermore, they showed that the text input speed increases and the error rate decrease the larger the keys are. The difference to our research was again that they used two-handed input using a mobile device and a stylus. Moreover, no effect on text input speed was shown which stays in contrast to our findings.
Kane et al. [8] examined the effect of walking and the adaption of the user interface on the performance when using two hands and the thumb to interact with soft-buttons on a mobile. The first reported experiment showed a decrease in error rate when increasing the target size while the other results are not conclusive (as discussed by the authors) due to the low number of participants and the high variance in the study results. Their second experiment focused on a user interface that automatically increases the target size once the user is walking. Their study did not provide any evidence for the advantages of this approach when compared to a non-adaptive user interface with large targets. The authors argue that this was due to some issues with the prototype design (e.g. some buttons in the adaptive interface were just too small) and due to the tradeoff between screen size and button size.
Parhi et al. [16] performed a user study comparing different target sizes for thumb-based one-handed interaction with a touch screen while standing. Their results show that task times and error rates decreased with larger key sizes. The authors argue that a target size of 9.2 to 9.6 mm is an optimal tradeoff between target selection time and error rate. Unfortunately, those rather large buttons are relatively big when compared with current screen sizes of mobile devices. Consequently, the iPhone, for example, has a minimal target size of 6.74 mm leading according to Parhi et al. [16] to an error rate of circa 8-12% and according to our research even to 23% (see Figure 7).