We study queue abandonment from a hospital emergency department. We show that abandonment is not only in uenced by wait time, but also by the queue length and the observable queue ows during the waiting exposure. For example, observing an additional person in the queue or an additional arrival to the queue leads to an increase in abandonment probability equivalent to a fteen minute or nine minute increase in wait time respectively. We also show that patients are sensitive to being "jumped" in the line and that patients respond di erently to people more sick and less sick moving through the system. This customer response to visual queue elements is not currently accounted for in most queuing models. Additionally, to the extent the visual queue information is misleading or does not lead to the desired behavior, managers have an opportunity to intervene by altering what information is available to waiting customers.