This study tested a variation of the random response surveying strategy designed for lengthy questionnaires to be used in group administrations or mail surveys. A questionnaire was mailed to a random sample of 3,000 company grade officers divided into three groups. One group received a conventional questionnaire. The second group received a random response questionnaire (FARR-50 technique) using a 50 percent likelihood of having to respond honestly. The last group received a random response questionnaire (FARR-83 technique) using a 83.3 percent likelihood of having to respond honestly. It was found that subjects respond more honestly, i.e., with greater estimated incidence of true positive response, to sensitive questions when given a low level of protection provided by the FARR-83 technique. Rates of response and estimate levels of prevalence were higher when subjects were given some protection than when they were forced to answer the questions honestly more frequently--even though the latter afforded greater protection. (Author)