We propose a set of requirements for designing handheld computer systems for electronic collection of patient diary and questionnaire data in clinical trials: (1) the system should be suitable for use by all types of patient to be included in the clinical trial programme; (2) patients must be capable of using the system and be comfortable with it after a short period of training; (3) responses should always result from an action by the user--defaults should not be taken as data; (4) all information necessary to a given question should be simultaneously available on the screen. This applies to both the questions and the response options. We present guidelines as to how these requirements may be met in practice, so that bias may be avoided both in patient selection and in the responses made; so that electronic data collection may be as effective as possible, and so that study procedures are convenient and unobtrusive for the patients.