For several months, nurses completed ratings of the degree to which certain events relevant to absence were present during each of their scheduled workdays. The event ratings for days when the nurses decided to be absent were then compared with those for days when the nurses attended. As expected, certain events, such as ill health and tiredness, tended to covary and proved to be consistently related to absenteeism across nurses. Also as expected, some events that were not especially relevant for the nurses as a whole, like having a sick family member or friend and concerns about previous poor attendance, nonetheless emerged as being relevant to the absence behavior of certain individuals. Finally, some events were consistently related to the nurses' expressed desire to be absent but not to actual absences. We discuss these differences from two perspectives, one emphasizing the role of attribution bias and the other, a two-stage process in which such bias has no major role.