An antecedent exercise (AE) intervention was conducted with two young children who demonstrated challenging behaviors during a large group circle time activity in a preschool classroom; outcomes were evaluated in the context of alternating treatments designs. Results from the initial visual analysis suggested no overlap between conditions, with small, consistent effects favoring the AE condition. However, plotting reliability data suggested observer bias was present for one participant. Subsequently (post hoc), a blind observer coded data from all conditions for that participant. Results suggested that no functional relation was present for that participant. Implications include the need for blind data collectors and additional, highly rigorous studies assessing the effectiveness of AE.