Testing the guidance hypothesis for location learning in GUIs

Knowing command locations is an important part of expert performance in GUIs, but building this knowledge requires laborious visual search. Many UI tutorials provide visual guidance during training (e.g. icon highlighting) to make visual search easier – but the “guidance hypothesis” suggests that users will rely on the guidance rather than learning locations, compromising performance in the real UI where guidance is unavailable. To test the guidance hypothesis for GUI location learning, we developed a design framework that identifies two key dimensions of guidance: visual style and temporal delay. We then performed three studies with two GUIs (a CommandMap and the Blender 3D modeling application), investigating three styles of guidance (highlighting, animated trails, and stencils) and three presentation delays (immediate, static, and increasing). Results showed that delay had a substantial effect on location learning: immediate presentation impaired learning of icon locations compared to unguided training, but static and increasing delays showed no significant differences in learning compared to the baseline. Though the effects of guidance style were less obvious, animated trails reduced location learning on some measures. Our work provides new understanding of the guidance hypothesis and new opportunities for tutorial designers to build effective and usable training systems for GUIs.