Direct democratic participation through referenda is often contested, because one faces the problem of determining referendum questions which avoid confusion or subjectivity. However, detailed knowledge concerning socalled ‘question wording effects’ is available within the domain of survey research. In this body of literature, several wording effects such as the use of suggestive wordings, the ambiguity of yes/no questions etc., have been well documented. Yet, despite the similarities between referendum and survey questions, knowledge from survey methodology is rarely employed within the literature on referenda. The present study discusses a number of question wording effects studied in survey research and shows their relevance in referendum settings. In addition this article explores these effects in twelve local referenda in Flanders, Belgium. Building on this empirical evidence, we conclude with a number of precise guidelines regarding the quality of referendum questions.