This paper explores how change propagation analysis can be affected by the way system boundaries are defined. This is an important issue as engineering change can in reality propagate out of the system modelled and back through components that were not considered. The work builds on a diesel engine case study to examine the difference in analysis results generated based on a full system model (i.e. entire engine) and those generated based on a set of partial system models (e.g. sub-assemblies). It was found that partial system models with boundaries defined by physical sub-assemblies can produce analysis results that are highly correlated with the one produced using a full system model. It was also revealed that modelling more components (i.e. a more complete system model) does not necessarily increase the level of correlation. The findings can be used to support system boundary decisions in change propagation analysis