This paper presents evidence which challenges the view that techniques which are designed to measure the social preferences of subjects can always be accomplished in a nonintrusive manner. We find evidence that such measurements can influence the preferences which they are designed to measure. Researchers often measure social preferences by posing a series of dictator game allocation decisions; we use a particular technique, Social Value Orientation (SVO). In our experiment we vary the order of the SVO measurement and a lager stakes dictator game. We find that subjects with prosocial preferences act even more prosocially when the SVO measurement is administered first, whereas those with selfish preferences are unaffected by the order of the measurement. Additionally, we find evidence that this difference is driven by the presence of choices involving the size of surplus.
