This study has investigated the characteristics of P coda and Lg waves recorded at the NORSAR and Graefenberg arrays from presumed nuclear explosions in western Russia. Lg waves were found to be more energetic, relative to the preceding coda, in the mid-band frequency range (0.2 to 1.0 Hz) recorded on the broadband Graefenberg instruments than in the high-frequency band (0.6 to 3.0 Hz) at either array. Comparison of the log-rms coda envelopes at the two arrays revealed that the early high-frequency P coda at NORSAR is stronger, relative to Lg, than at Graefenberg. Also, the amplitudes of P waves and early P-coda waves, within 55 sec after P, recorded at the centers of the NORSAR subarrays, are positively correlated indicating the effect of focusing-defocusing on the P and coda waves. However, for subarrays which cluster spatially, there is an inverse correlation which seems to be due to scattering attenuation, suggesting that much of the early P coda is generated by local P-wave scattering beneath the NORSAR array. Also, the NORSAR coda envelopes flatten between about 219 and 352 sec after P, whereas the Graefenberg coda decays monitonically from the P wave out to the Sn and Lg arrival times. The standard deviation across the NORSAR array of the log-rms amplitudes of Lg and S-coda phases are higher by about 0.05 to 0.06 magnitude units than those of P-coda phases. These differences are probably related to differential site and propagation path effects on P, S, and Lg waves in the P coda. The flattening of the NORSAR coda envelopes appears to be a burst of energy not associated with any known regional phases. The burst may be caused by forward scattering of Lg to P waves along the path between the source and NORSAR. Based on a regional travel-time curve for western Russia, the start and end time of the coda flattening is consistent with Lg-to-P scattering in the distance range of 23° to 29° from NORSAR, which brackets the Ural Mountains. The Lg propagation path from the source region to Graefenberg does not appear to traverse the Ural Mountains which may explain why the coda flattening is not observed at Graefenberg.