Research in the field of written procedures (in an industrial context) has moved along two lines: regulating the performance of operators, especially in nuclear power plants, and creating effective technical manuals. The quality of such texts has remained a problem, perhaps because research has generally concentrated on superficial text features and task analysis, thereby neglecting the actual use (or non-use) of documents in working situations and the wider interests the organisation has in writing down procedures. A more promising strategy, along the lines of a distinction made by Rasmussen, is to concentrate on the relation between written procedures and the appropriate levels of human performance: skill based, rule based or knowledge based. In a case study (procedures for railway-track maintenance with uninterrupted train service), users described their problems with the written procedures as "lack of clarity", alluding to deficiencies in content and presentation. Further analysis revealed the following more specific problems: - unclear presentation: not designed as a job performance aid, awkward phraseology, unfit for easy reference; - too many detailed procedures; - unclear working situation: users often felt compelled to choose between following the procedures to the letter and accomplishing the jobs on schedule; - resistance to the predominating "control directing" function of the procedures, to the detriment of their "action directing" function; - lack of motivation, as a result of not having been involved in the drawing up of the procedures. The procedures had been conceived by the organisation on a rule based level. Due to the modus operandi of the human cognitive system, however, actual application of the procedures often takes place on a skill based level. Accordingly, it was decided to rewrite the procedures to a skill based level, i.e. to write very explicit instructions, specifying the actions to be taken by workers at every level in the organisation. From the present study it can be concluded that the usability of safety-related written instructions is not determined predominantly by their contents and their presentation, but also by at least the following factors: organisational context (working conditions, administrative control, safety policies); task involved; risk perception; feasible level of human functioning; frequency of error types; ergonomic possibilities; motivation of the users; knowledge and experience of the users.