Wave propagation of crustal phases, in particular Lg, near the ARCESS array in northern Norway is examined and compared to propagation characteristics near the NORESS array in southern Norway. Relying on array analysis, we show that ray-theory explanations of arrivals in local and regional seismograms are very useful in deducing the composition of the crustal wave trains. Frequency-wavenumber analysis is applied to the array data to identify the arrivals; then a composite of array beams is used to approximate each event. Record sections of composite seismograms are constructed for different directions from ARCESS to study propagational characteristics with distance and azimuth and then compared to composite-seismogram record sections from NORESS to study different regional characteristics. To model the observed character of Lg and Rg, synthetic record sections are constructed by wavenumber integration in velocity models representative of structure in the different regions. The results show that Rg propagation distance varies significantly between regions and that the velocity gradient in the lower crust is the dominant factor in specifying Lg characteristics in the Baltic shield region. Lg is dominated by several discrete arrivals representing multiple Moho reflections, or by multiple Moho reflections and turning waves, with each multiple confined to a small distance range so that only one multiple set dominates at each distance. Furthermore, we demonstrate that incomplete knowledge of the propagation characteristics leads to mislocations of local and regional events and inhibits source-depth discrimination of events. We identify three earthquakes in the ARCESS data set based on source depths. These are constrained by phase velocities and depth phases, none of which were properly identified by the automatic array-detection software.