Résumé : An experiment using a representative survey of the German population shows that letting respondents report a number rather than asking them to choose from a list of predefined ranges lowers the response rate for both perceived past and expected inflation and decreases (increases) reported past (expected) inflation. Income, education, gender, objective and subjective knowledge about monetary policy, and political affiliation affect the effect’s size but not its sign. East and West German respondents who were 15 or older when the Berlin Wall fell have reactions different from those who were younger at that time, which supports the ‘impressionable years’ hypothesis based on different inflation experiences.