Some researchers worry that when reporting the frequency with which they have performed a category of events (e.g., “gone shopping”), survey respondents may fail to take into account all events in the category. One remedial strategy that some researchers have recommended is for a survey question to offer examples of events that might commonly be overlooked by respondents (Groves, et al., 2004). In our research, we examined the effects of offering two different types of examples that survey developers could use to help respondents recall commonly overlooked events when generating frequency estimates.

A total of N=1124 U.S. consumers completed an online survey in which they answered a general question about their average number of weekly shopping trips during the previous six months, and later in the survey reported their average weekly shopping trips to 15 different specific store types. Some respondents answered a No Examples version of the general question which read, “During the last 6 months, how many times per week, on average, did you visit a store to go shopping?” Other respondents answered an Object Examples version (defined as references to people, places, or things related to a target category of events) in which specific store types were added to the question as examples. The last group of respondents answered a Behavior Examples version (defined as references to activities related to a target category) in which the phrase “…to go shopping” was expanded to include other shopping-related behaviors. In each condition, respondents entered a numerical value into an open-ended response option box that accepted a maximum of two digits.

Our results indicate that questions offering either Object Examples or Behavior Examples generate larger frequency estimates than questions with No Examples . However, the effects of Object Examples on frequency estimates appear to be driven by enhanced recall of only the listed objects. In contrast, Behavior Examples appear to influence frequency estimates by enhancing recall of a wider array of events. This suggests that giving respondents examples of target behaviors, rather than locations where those behaviors might have occurred, yields more accurate reports of frequencies.