This article focus on a traditional issue of measurement error, the questionaire and, more specifically, the question position. One will see that identical questions, or very similiar ones, can lead to totally different results, depending on the question position on the questionaire and, in addition, on the issue. This article shows that the same question about the voting memory in Benedita da Silva, a black woman, politician of the Worker’s Party, has obtained very different results in very similar surveys, except for the position question on their questionnaires.