Randomisation is the most important means of controlling bias in estimating the effects of behavior change interventions in randomized controlled trials.
The aim is to explore the mechanisms by which bias may be introduced, and thus the possible limits of randomisation in experimenting with people.
Existing evidence on assessment and other forms of research reactivity will be reviewed from brief intervention trials and elsewhere.
Plausible hypotheses concerning the possible limits of randomisation have not been systematically investigated in brief intervention trials, nor in other areas of behaviour change trials.
The reasons for the lack of development of study of this crucial asperct of behaviour change trials are discussed. Hypotheses and study designs are proposed to guide new research which seeks to quantify problems with existing practice in the design and conduct of trials.