The answer to the above question may seem obvious: whichever of the two is the bottleneck, and if there is none, whichever of the two has longer ‘service times” and/or larger variation in service times. As we empirically show in this study, the answer is not quite so obvious. In fact, our simulation results suggest that, as long as there are no bottlenecks in the system, the material handling system should almost never be responsible for work-in-process even if it has service time parameters that are comparable to processing times at the workstations. In the paper we first show the simulation results, and subsequently we provide insight based on well-known analytical results in queueing theory. We also discuss certain properties of the workstations and the handling system that affect the work-in-process levels