Split-panel classroom testing was conducted with 515 participants recruited from four regional Census offices. Participants were randomly assigned to complete either the DC-2-U(F) long form or an alternative census long form, followed by a debriefing questionnaire. Results indicated that: 1) item nonresponse rates were lower on the alternative questionnaire than on the 1986 form, but the difference was not statistically significant; 2) there were few differences in substantive responses between the two forms; and 3) alternative placement, wording and formats of particular questions decreased item nonresponse rates to some questions while increasing item nonresponse rates for others. Differences in responses to each question are also discussed.