A technique designed to reduce or eliminate evasive response bias in survey research on sensitive issues by preserving respondent confidentiality through randomization. For example, an investigator might instruct the respondents as follows: Toss a coin privately and then answer Yes if either your coin fell heads or you have smoked cannabis, otherwise answer No. With a sample size of 1,000, we would expect 500 Yes replies if none of the respondents had smoked cannabis; additional Yes responses would suggest that some respondents whose coins fell tails also replied Yes, and the proportion who had smoked cannabis could be estimated from the results without identifying the individuals concerned. If 580 answered Yes, then we might assume that approximately 80 of the 500 whose coins fell tails answered Yes, and we could therefore estimate the proportion of the population who have smoked cannabis to be 80/500 = 16 per cent, and this conclusion could be reached without having identified any individual cannabis smokers. The technique was first proposed by the US economist Stanley Leon Warner (1928–92) in an article in the Journal of the American Statistical Association in 1965. See also non-reactive measure, reactivity.