'Reality orientation training' (ROT) is a well-established therapy used with the elderly, especially those with dementia. It aims to improve orientation and reduce negative behaviours, e.g. confusion. ROT has been recommended for non-demented patients with acquired neurological impairment. However, this suggestion has not been investigated further in controlled trials. This paper describes an informal 24 h ROT programme with an amnesic subject (H.J.). It combined single-case experimental designs derived from behaviour therapy. Target behaviours were items of temporal orientation—current time/time of day, year, season, month and day of week. Orientation regarding the current date was not trained, and thus served as control variable for non-specific effects (e.g. spontaneous recovery). In order to improve oriented behaviour in different relevant situations outside the clinic, we chose 24 h ROT instead of formal therapeutic sessions. The patients' spouse offered 24 h ROT at home 7 days a week. This consisted of reminding the patient of orientation information, e.g. the current day of the week in different situations. Furthermore, negative behaviours such as irrelevant questions were ignored. The patients' spouse was trained and supervised to carry out this task in 27 sessions each lasting 10 min. For 14 weeks supervision was provided twice a week in the clinic (i.e. 2×10 min). Afterwards the spouse continued to apply ROT at home without further supervision in two follow-up periods. As expected, learning was slow but reliable. Generalization of improvement to another setting was shown (home vs clinic). Stable therapeutic gains over a long follow-up period could be demonstrated. Temporal orientation neither improved nor worsened after cessation of supervision, suggesting that booster sessions would be crucial for further improvement.