The Multi-function Control and Display Unit (MCDU) has been identified as a source of issues pilots have transitioning to glass cockpits. Several aircraft manufacturers and avionics vendors have committed to replace the MCDU with graphical user-interfaces in the next generation of commercial aircraft. A cognitive task analysis of pilot-MCDU interaction, described in this paper, has identified that pilot failure to complete mission tasks using the MCDU is not a sole consequence of the physical dimensions or layout of the device. Instead, the MCDU interface works adequately when a given pilot task: (1) is supported directly by a function provided by the automation, and (2) the access of MCDU pages, and format and entry of data, are prompted by labels and other visual cues (and not by memorized actions sequences). Pilot tasks not supported directly by automation, and/or pilots tasks that rely on memorized action sequences are difficult to learn and likely not to be used effectively in the field.