This discussion of the design of user-initiated help systems in computers focuses on the information that users actively seek to help them with their tasks, with emphasis on how to help users ask the questions that will bridge the gap between the initial internal (mental) form of the query and their information need as expressed by the system. Ways to help users ask questions are suggested, and examples are provided of information structuring on the Ceefax teletext system and the UNIX and Rabbit computer systems. The process of formulating a question is discussed, including the different types of questions that can crop up when going from an initial vague question to the final answer. It is argued that, although several different methods of getting help should be made available to the user, it is important that all of these systems form a well-integrated whole that enables the user to go from each subsystem to the next in a smooth and uninterrupted fashion. The need for dynamic structures that will enable users to structure or restructure the information to suit their own purposes is then discussed and "hypertext" is suggested as one method of providing such a structure. It is concluded that current systems tend to handle internally generated questions with natural language or database-query kinds of questions, and externally generated questions with "direct manipulation" interfaces. Designers are urged to consider both types of information gathering mechanisms more systematically for both kinds of questions. Seven figures, and eight footnotes are included. (EW)