Training Wheels User Interface
(Sidebar to Jakob Nielsen's column "Progressive Disclosure.")In two pioneering studies, Dr. John M. Carroll and colleagues compared "training wheels" designs (where users were initially limited to using a few features) with corresponding unlimited designs (where users had continuous access to all features).
During early use, the training wheels design scored better than the unlimited system in both studies on each of the three usability metrics:
| Initial use | Study 1 | Study 2 |
|---|---|---|
| Time on task | 26% faster | 21% faster |
| Knowledge of system | 69% more facts learned | 21% more facts learned |
| Subjective satisfaction | 28% higher rating | (not measured) |
In Study 2, the test participants continued to use the two designs after the initial learning phase. In this phase of the study, the "training wheels" came off for the first design, giving users access to the full set of features. In other words, the two versions were identical during advanced use; the only difference was that users had been limited to a core features set during their initial use of the training wheels design.
On advanced tasks, people who’d been using the training wheels during their initial system exposure performed better than people who’d used the advanced features throughout. The following table shows how much better these users performed:
| Advanced use | Study 2 |
|---|---|
| Time on task | 52% faster |
| Knowledge of system | 10% more facts learned |
It's not so surprising that training wheels users performed better during initial use, but it’s striking that they also performed better during advanced use, as the second table indicates.
You might expect users to do better on the advanced features if they'd been using them during the entire study, including the initial use period when they first learned to use the system. You might expect people to construct a more complete mental model if they had all the features available all the time. Not so.
People performed better on the advanced features when they were initially limited to a smaller set of core features. The training wheels focused users' attention and helped them build a more complete and well-structured mental model. This better understanding of the core concepts gave them a framework that later helped them with the advanced features.
Carroll's training wheels studies provided early scientific support for the progressive disclosure design concept. Since then, much practical experience has extended this support: It's good to help users focus their initial learning and avoid confusing them with too many features.
(By the way, this very sidebar is an example of progressive disclosure as applied to Web content: the detailed research results are advanced information that readers don't need to understand the primary article, so I deferred them to a secondary page using a hyperlink. In my experience, only 10% of readers will follow the link, but that's okay because it's a deliberate editorial decision to remove the research studies from the primary information path. Otherwise, the primary article would be much too long; it's one of the guidelines for writing for the Web to use secondary pages for specialized information that most readers won't need.)