Fieldwork provides maximum flexibility in allowing analysts to test a questionnaire under conditions similar to those intended for its actual use. In this case, it was not the usual demographic characteristics of people, but rather a specific experience that defined the survey population-that of having lived through a natural disaster. We chose to conduct our cognitive interviews in central Florida, where the region had experienced a record number of hurricanes in the 2004 season.

With the help of a Federal Emergency Management Agency (FEMA) representative who was familiar with the communities most devastated by hurricanes Charlie, Frances, and Jeanne, we decided to locate in Orlando, near a suburb that had sustained extensive damage. Our contact agreed to lend us a private room in a government facility temporarily housed by FEMA. This was ideal from a recruitment standpoint because it would allow us instant access to people inquiring about or applying for disaster relief aid. However, two weeks before our departure, we were informed that, because of the extensive hurricane damage, these facilities were in (unexpected) continued use by FEMA, and no spare rooms would be available.

Forced to alter our plans, we decided on a hotel in Orlando that bordered the suburban towns. Because this hotel also served as an evacuation shelter for its own employees, it provided a word-of-mouth recruitment opportunity that would not have been afforded by the government facility. Moreover, the building itself was structurally unaffected by the hurricanes, and was able to provide a reliable source of power (for videotaping, holding conference calls with survey sponsors, and copying or faxing), as well as adequate shelter for the interviewing team.

The proximity of the hotel to towns bordering Orlando was strategic as well. In most fieldwork projects, conducting interviews in respondents' homes is a good choice. It affords a maximum amount of privacy to respondents in an atmosphere that is most comfortable to them. In this case, however, home interviews were not deemed the best option. Many-if not all-homes had sustained either wind damage to their roofs (many had missing tiles and blue tarps covering all or a portion of the roof) or water damage inside the house. On the other hand, the public infrastructure was in reasonably good shape two months after the hurricane: Lifelines (water, electricity, sewage service) were functional, roads were passable, and public transportation was operational. Therefore, we determined it would be easier to have respondents travel to a conveniently located facility than to have staff conduct interviews inside respondents' homes.

Besides offering test conditions closer to real life than those of a lab, another advantage to fieldwork is that it can be tailored to any time-sensitive issues related to cognitive testing. The Disaster Needs Assessment survey had a temporal component dictating that respondents' hurricane experiences be recent. The goal was to conduct interviews within about a month of the disaster. However, because multiple hurricanes placed an unusual strain on FEMA, our fieldwork could not begin in the preferred timeframe. Instead, we faced a two-month lapse.

At first we did not judge the delay to be critically detrimental to the project, particularly as Florida residents were still in the recovery phase of the disaster. As the cognitive testing unfolded, however, it became apparent that the temporal aspect of the survey was more important than initially assumed. For example, one series of questions was designed to measure feelings of well-being and general mental health in the recovery phase of a disaster. While intended to be administered within a month of the disaster, these questions were not necessarily intended to measure respondents' feelings about the event itself. Instead, they were crafted to measure a person's general outlook on life.

One question in this series originally read, "In the past thirty days, how often have you felt that everything was an effort?" Because respondents were being asked the question two months after the disaster, we changed the beginning clause to reflect this fact. The question became, "Since the hurricane, how often have you felt that everything was an effort?" During testing we found that respondents were very confused by this question, asking "Everything was an effort? Like what?" or "I don't understand the question." Upon probing we found that respondents had no way to link "everything was an effort" to their experience with the hurricane-which is what they thought the question was asking them to do. Most of the people we interviewed considered "an effort" to mean, as one respondent put it, "Did people come together after the storm" to help in cleanup?

Other questions in this series exhibited similar problems. One read, "Since the hurricane, how often were you on guard, watchful, or easily startled?" Still confused, respondents asked, "About the disaster?" or "Do you mean after it left? Right after it left?" In other words, most respondents interpreted " since the hurricane" as " because of the hurricane" or even " during the hurricane." The original frame of reference, "In the past thirty days," might not have linked the hurricane experience so tightly to the question.