Abstract
An empirical study examined the effectiveness of four different versions of a self-instruction manual for a word processing system: a Skeletal version that explicitly states only the essential information, an Inferential version that has the users infer some of the essential information, a Rehearsal version that is like the Skeletal manual, but adds opportunities to rehearse the explicitly stated information, and a Lengthy version that adds nonessential explanatory and descriptive information to the Skeletal version. The best learning performance was obtained with the inferential approach, particularly for more realistic tasks.
