We implement a methodology that identifies respondents who are reticent in answering sensitive questions on surveys. The methodology entails asking a series of randomized response questions and identifies the reticent as those who give a set of answers that can arise only with an implausibly low probability. In a sample of Romanian company officials, we identify a specific 10% of respondents as reticent. The identifiably reticent admit to corruption interactions significantly less often than others. On values questions, they claim to be of higher moral virtue than the average respondent. We show that reticence is related to the respondent's age, the reticence of others in the region, and region itself. Our results suggest a method of improving the accuracy of survey data on corruption and values, as well as caution in interpreting the causes of cross-country differences in such data. (c) 2009 by The University of Chicago. All rights reserved..