The researchers concluded that the Hamburg study biased respondents by eliciting favorable attitudes toward retribution. They assumed that the Freiburg population was less favorable toward restitution than the Hamburg population, that the order of items in a specific response scale significantly affected the results, and that supplementing the incomplete range of answers in the same item would affect the analysis. To test these assumptions, they conducted a mail survey in 1994, using three versions of a 16-page questionnaire. Responses came from 378 of the 1,000 participants. Findings clearly demonstrated the effect that a question's formulation on the results and revealed that the Hamburg data do not permit conclusions regarding public attitudes toward victim compensation or nonpunitive approaches. Results provided clear support for the significance of instrument design in criminological surveys and other social science research. Tables, figure, and 54 references