Detecting and monitoring change in models

System dynamics is often brought into connection with a double‐loop learning process. Learning has been the object of an increasing number of studies. However, inquiry has focused on using models rather than modeling, and there are huge differences in assessment approaches. If learning changes models then it can be inferred from comparing models. Here it is argued that monitoring learning from modeling is feasible and desirable. One possibility is to conceive of a model as a series of versions and compare their structure. One possible method for comparing model versions is presented. Copyright © 2006 John Wiley & Sons, Ltd.