This study investigated the common practice in infant research of eliminating from reported data large numbers of subjects who prove uncooperative (sleepy, fatigued, fussy) during the experiment. It was suggested that these excluded infants constitute a special class of subjects and that the inclusion of their data would greatly alter the research results. Subjects compared were 37 3- to 6-month-old infants, 15 of whom were unable to finish the experiment sessions. Two visual and two auditory tests were administered and measured according to fixation time and cardiac deceleration. Activity data were collected by use of an infant stabilimeter. Results demonstrate that infants unable to complete the experiment show different attentive patterns from those able to complete the sessions. The findings suggest that elimination of large numbers of infants may result in serious biasing of obtained data. (Author/AJ)