Reading Usable Help
@UsableHelp on Twitter
Gordon R. Meyer
A good system can't fix bad Help
Robert Krull of RPI and Angela Eaton of Texas Tech University write Problems In Navigating Online Help: Clues from Users Search Patterns at the WritersUA site. This study focuses on help system organization and the difficulties users might have in finding information--due to poor terminology--and difficulties in navigating the structure of the help, because of its information design.
I think it's important to take Krull and Eaton's results in context, they were clearly studying a system with general language and naming problems; as evidenced by their recommendation that the help system provide definitions for similar terms (name, caption, and label). It sounds like the application could have benefited from a pro-active "word scrub" by the help writers, which would lessen the need to burden the help with attempting to clarify a flawed interface and nomenclature.
I am skeptical of the recommendations regarding an exposed hierarchy of information, and support for extensive horizontal and vertical movement within that structure. There were clearly issues with a poor search engine that returned hundreds of hits, the naming problems already discussed, and topic titles so poorly written that users couldn't discern between them or know what type of information they contained. Adding more links, so that users could transverse the five structural levels, and apparently thousands of pages, is not likely to make this particular help system easier to use. Although it would certainly provide a lot of opportunity for users to click things; which might at least make them feel like they were getting closer to an answer.
Most of the other recommendations are sound, such as incorporating more help into the user interface and presenting hits so that users know if it's a procedural or declarative piece. But I do think that more weight should have been given to the fact that the one help system being studied was for a programming language, and thus presumably more reference orientated and for a technical audience. Additionally, offering a general conclusion that searching in help is only successful 50% of the time excuses the technical writers from addressing what seems to be the core of the problem - "garbage in/garbage out."