This paper examines the generally accepted multipath model of wideband radio channels. We use a simple multipath model based on paths of unlimited spatial extent, and compare it to a model that allows a finite spatial extent to the paths (that may result, for example, from shadowing). The comparison shows that the introduction of a finite spatial extent improves the accuracy of the model. Both models are fitted with parameters and compared to actual measurements from an indoor campaign. Our analysis is based on measurements in the 2-18 GHz band; we made static impulse response measurements with the receiving antenna at different positions along a one meter long linear axis.