Goal
Support menus with up to 12 items per level.
Twelve evenly spaced items are 30 degrees apart, so the current 45-degree limit blocks them. That limit comes from generated strokes, not human measurements.
Evidence
Kurtenbach and Buxton tested expert 4, 8, and 12-item marking menus. Their study found that breadth and depth raise errors, with menus of eight or more items becoming error-prone beyond two levels. It does not show whether this library, its input devices, or a different recognizer can reliably support 12 items per level. The Limits of Expert Performance Using Hierarchic Marking Menus
An earlier nonhierarchical experiment gives a reason to pursue it. A 12-slice clock layout was easier for participants to reconstruct from memory than an 11-slice menu, and more accurate in both hidden and marking conditions. The result depends on its clock-face layout and ordered labels, so it does not justify every 12-item arrangement. An empirical evaluation of some articulatory and cognitive aspects of marking menus
Work
- Collect human strokes in a versioned corpus with device, breadth, depth, direction, and target metadata. Keep a held-out portion to measure accuracy, false selections, and latency.
- Investigate recognition methods for segmented directional gestures and prototype the strongest practical alternatives alongside the current algorithm.
- Compare the methods against the same corpus, including error recovery costs and implementation complexity.
- Use the results to implement and validate support for 12-item menus.
Done when
The library supports up to 12 items per level. A documented human-stroke corpus and reproducible benchmark show the supported depth and device limits, and the library documents those limits.
Goal
Support menus with up to 12 items per level.
Twelve evenly spaced items are 30 degrees apart, so the current 45-degree limit blocks them. That limit comes from generated strokes, not human measurements.
Evidence
Kurtenbach and Buxton tested expert 4, 8, and 12-item marking menus. Their study found that breadth and depth raise errors, with menus of eight or more items becoming error-prone beyond two levels. It does not show whether this library, its input devices, or a different recognizer can reliably support 12 items per level. The Limits of Expert Performance Using Hierarchic Marking Menus
An earlier nonhierarchical experiment gives a reason to pursue it. A 12-slice clock layout was easier for participants to reconstruct from memory than an 11-slice menu, and more accurate in both hidden and marking conditions. The result depends on its clock-face layout and ordered labels, so it does not justify every 12-item arrangement. An empirical evaluation of some articulatory and cognitive aspects of marking menus
Work
Done when
The library supports up to 12 items per level. A documented human-stroke corpus and reproducible benchmark show the supported depth and device limits, and the library documents those limits.