We present a new technique for imaging spatially distributed heterogeneities using information from combinations of source and receiver arrays. The method is based on the single scattering assumption and is closely related to the double beam method of Kr¿ger et al. <1993, 1995, 1996> in that it exploits amplitude, delay time, slowness, and azimuth information in two arrays simultaneously. A crucial step in the method is the application of static time corrections for a chosen reference phase (here PcP). One type of image is obtained as a spatial likelihood distribution for prescribed source and receiver array slowness and azimuth values and delay times with respect to a reference phase. This allows the determination of the spatial origin of coherent phases in the P coda. In a complementary approach, we perform a double beam stack migration according to the theoretical slowness and azimuth values for candidate scatterers distributed over a three-dimensional (3-D) grid. The scattering strength is expressed as the resulting beam power, beam amplitude, or semblance in a time window determined from the theoretical delay time with respect to the reference phase. While likelihood mapping provides an image reflecting the probability of a region to explain observed kinematic phase properties by single scattering, double beam migration provides information about the scattering strength. The method is applied to nuclear explosion sources in eastern Kazakhstan recorded at the Yellowknife array in northern Canada. We identify several scattering volumes within the lower mantle below the arctic producing individual anomalous lower mantle phases in the P coda. These anomalies range in depth from the core-mantle boundary (CMB) up to 500 km into the lowermost mantle and are mainly located under the Eurasian side of the arctic. The joint interpretation of these results with the results of previous studies suggests a connection of these anomalies to the Cenozoic and Mesozoic subduction of the Pacific plate and the Kula plate. The determination of the nature of such 3-D anomalies necessitates further array studies with better azimuthal coverage.¿ 1997 American Geophysical Union