2. Recall how Example 8 (page 304) shows how wording can affect survey results. You can explore this by doing an experiment disguised as a survey. Choose a topic, then design two questions with a key difference in wording.

Use randomization to choose which version of the question you ask each person. Don’t reveal the design of the experiment to participants until after they have provided their answers. After you have collected roughly 20 or more responses to each version of the question, compare and interpret your results. If you’re interested in reading about an example of such an experiment, see John Rubin’s article “Weighing Anchors” in the June 1990 issue of Omni.