Although information system development methodologies supposedly improve development processes and end-products, information systems have continued to fail. In practice, methodologies have not been universally accepted, and even when accepted, not consistently used. Often development teams adapt methodologies to respond to perceived problems in past applications and/or specifics of the project under consideration. While it is possible that a combination of these factors contributed to less-than-effective system development processes or final end-products, we believe that there may be a subtler explanation for such failures. Although methodologies recommend best practices for system development, they rarely prescribe mechanisms to capture or evaluate problems encountered during system development. Specifically, methodologies do not include “learning” mechanisms that may be useful in improving the effectiveness of processes and end-products. We outline a meta-methodology for system development methodologies, which describes ways in which learning at different levels can be used to improve the effectiveness of system development methodologies.