Surveys of student populations provide important data to policymakers on underage alcohol and drug usage, risky student behavior, and school safety, among many other topics. Sensitive behaviors are, however, under-reported, and survey researchers continue to work to maximize data quality by using carefully-designed implementation protocols that create a more private environment and yield more accurate reporting. Our paper reports the results of a natural experiment to improve the data quality In student surveys in an ongoing study in township schools in Pietermarilzburg, Soufh Africa. As part of our research, a survey was administered to the same group of students at three time points. At time 1, all students received an identical paper and pencil version of the survey instrument. To reduce shoring of answers by students, at Time 2, we administered multiple PAPI versions of the survey with different question orders. Prior to survey administration we shared with students the fact that there were various survey orderings. At Time 3 we administered the survey using audiopersonal digital assistants [APDAs) that allowed participants to choose between two languages (Zulu or English) in both the screen text and sound components. Preliminary results indicate that survey administration conditions were improved and missing data was reduced in both the multiple PAPI version condition and in the APDA condition. Although having multiple versions of the survey increased students' sense of privacy, it did not increase self-reports of how honest they were in completing the survey. Using APDAs increased both privacy and self reported honesty. We focus on our preliminary results and analyze the impact of these mode experiments on other substantive variables in the survey. The paper connects these findings to recommendations for best practices in implementing student surveys in challenging conditions. Our results reemphasize how new technologies can help improve data quality in student surveys.