Researchers studying the effects of diet on health sometimes use 'randomised intervention trials'. One group - the intervention group - is asked to change some aspect of diet, for example eat less fat or more fruit, while the other group - the control group - is told to eat as normal. At the end of the trial the number of cases of, say, colon cancer, in the two groups can be compared. Such results could be misleading, because as the study progresses, the 'intervention' subjects become less conscientious about sticking to their supposedly healthy diet, so the longer the study goes on, the less the diets of the two groups differ. Yet to get results that are statistically significant, the study must go on as long as possible.