School consultants who rely on direct observation typically conduct observational samples (e.g., 1 30‐min observation per day) with the hopes that the sample is representative of performance during the remainder of the day, but the representativeness of these samples is unclear. In the current study, we recorded the problem behavior of 3 referred students for 4 consecutive school days between 9:30 a.m. and 2:30 p.m. using duration recording in consecutive 10‐min sessions. We then culled 10‐min, 20‐min, 30‐min, and 60‐min observations from the complete record and compared these observations to the true daily mean to assess their accuracy (i.e., how well individual observations represented the daily occurrence of target behaviors). The results indicated that when behavior occurred with low variability, the majority of brief observations were representative of the overall levels; however, when behavior occurred with greater variability, even 60‐min observations did not accurately capture the true levels of behavior.