The UC diagram technique described in the companion paper (Part 1), is applied to nine sets of Lg phases recorded through the CEDAR system in southern California, and two sets of Lg phases recorded along the northwestern margin of the Sierra Nevada. A clear image of the signal is obtained in time-frequency-wavenumber space, and we discuss in particular observations at 2.5-sec period, for events 200 to 300 km outside the profiles. From the gross features of UC diagrams we conclude that a representation of Lg as a single coherent multi-mode wave train is oversimplified in the case of southern California but is more appropriate for the Sierra block. In southern California, peaks observed at group velocities smaller than 3.2 km/sec are not predicted by realistic crustal models of the area, and are probably due to lateral heterogeneities effects such as mode conversion and multipathing. On the other hand, for group velocities between 3.2 and 3.6 km/sec, peaks observed in either area can generally be interpreted in terms of overtones excited at the source and propagating through spatially averaged structures, although care must be taken to monitor the stability of the algorithm on actual short-period records.