Survey researchers often use a check-all question format in Web and mail surveys, but revise it to a forced-choice format in telephone surveys. An underlying assumption is that format is unimportant. Recent experimental studies, however, have quite consistently demonstrated higher affirmative answers in the forced-choice question format than in the check-all question format in both mail and web surveys. Some authors have suggested that the forced-choice question may provide greater accuracy. However, none of the previous studies have included external validation measures. This is critical because without a benchmark one cannot know which format provides a more accurate answer. To examine these issues, a series of three experiments were conducted in 3 provincial park campgrounds in British Columbia in 2006. The experiments include manipulations of two experimental questions: an attribute question and a behavioral question. The experiments were designed to: 1) compare the two formats against an external validation measure for the attribute question; 2) examine the effect of reducing the length of response options for the two formats in both questions; and 3) determine the effect of “chunking” (changing from one column to two columns) for the check-all format. The external validation measure (i.e. type of camping equipment used) was obtained through observation at the time the questionnaire was distributed. Five different versions of the questionnaire were distributed to a random sample of campers (n= 133 to 199; response rates = 50% to 75%). While some of our results are consistent with previous studies, we found that differences in the two formats are unlikely to occur if the check-all question is highly salient and quite short. Surprisingly, our external validation measure indicates that both formats significantly overestimated.