Another avenue for progress in GOMS modeling has been to implement
GOMS models within more complex computational cognitive architectures,
such as Soar and ACT-RPM (Anderson and Lebiere, 1998; Pew and Gluck, 2004).
These architectures are embodied in software systems that model cognitive,
motor, and perceptual processes, but these models inherit the limitations of
GOMS models, predicting only skilled user execution time for familiar tasks. A
current goal is to employ these architectures to predict a broader spectrum of
human performance, including learning time, errors, performance under stress,
and retention over time.