like I did here for here for another controversial topic), I’ll mention the high-level issues that make the studies irrelevant to me. All studies I could find had at least one of the issues listed below; if you have a link to a study that isn’t irrelevant for one of the following reasons, I’d love to hear about it!
- Age of study: it’s unclear how a study on interacting with computers from the mid-80s transfers to how people interact with computers today. Even ignoring differences in editing programs, there are large differences in the interface. Mice are more precise and a decent modern optical mouse can be moved as fast as a human can move it without the tracking becoming erratic, something that isn’t true of any mouse I’ve tried from the 80s and was only true of high quality mice from the 90s when the balls were recently cleaned and the mouse was on a decent quality mousepad. Keyboards haven’t improved as much, but even so, I can type substantially faster a modern, low-travel, keyboard than on any keyboard I’ve tried from the 80s.
- Narrow microbenchmarking: not all of these are as irrelevant as the
e -> | without search and replace task, but even in the case of tasks that aren’t obviously irrelevant, it’s not clear what the impact of the result is on actual work I might do.
- Not keyboard vs. mouse: a tiny fraction of published studies are on keyboard vs. mouse interaction. When a study is on device interaction, it’s often about some new kind of device or a new interaction model.
- Vague description: a lot of studies will say something like they found a 7.8% improvement, with results being significant with p < 0.005, without providing enough information to tell if the results are actually significant or merely statistically significant (recall that the practically insignificant scrolling result was a 0.08s difference, which could also be reported as a 16.3% improvement).
- Unskilled users: in one, typical, paper, they note that it can take users as long as two seconds to move the mouse from one side of the screen to a scrollbar on the other side of the screen. While there’s something to be said for doing studies on unskilled users in order to figure out what sorts of interfaces are easiest for users who have the hardest time, a study on users who take 2 seconds to get their mouse onto the scrollbar doesn’t appear to be relevant to my user experience. When I timed this for myself, it took 0.21s to get to the scrollbar from the other side of the screen and scroll a short distance, despite using an unfamiliar mouse with different sensitivity than I’m used to and running a recording program which made mousing more difficult than usual.
- Seemingly unreasonable results: some studies claim to show large improvements in overall productivity when switching from type of device to another (e.g., a 20% total productivity gain from switching types of mice).
It’s entirely possible that the mysterious studies Tog’s org spent $50M on prove that the mouse is faster than the keyboard for all tasks other than raw text input, but there doesn’t appear to be enough information to tell what the actual studies were. There are many public studies on user input, but I couldn’t find any that are relevant to whether or not I should use the mouse more or less at the margin.
When I look at various tasks myself, the results are mixed, and they’re mixed in the way that most programmers I polled predicted. This result is so boring that it would barely be worth mentioning if not for the large groups of people who believe that either the keyboard is always faster than the mouse or vice versa.
Please let me know if there are relevant studies on this topic that I should read! I’m not familiar with the relevant fields, so it’s possible that I’m searching with the wrong keywords and reading the wrong papers.
Appendix: note to self
I didn't realize that scrolling was so fast relative to searching (not explicitly mentioned in the blog post, but 1/2 of the text selection task). I tend to use search to scroll to things that are offscreen, but it appears that I should consider scrolling instead when I don't want to drop my cursor in a specific position.
Thanks to Leah Hanson, Quentin Pradet, Alex Wilson, and Gaxun for comments/corrections on this post and to Annie Cherkaev, Chris Ball, Stefan Lesser, and David Isaac Lee for related discussion.