Travel times and amplitudes of long‐period SH, ScSH, sSH, and sScSH phases from several deep focus earthquakes in the northwest Pacific are analyzed for evidence of lithospheric slab penetration into the lower mantle. Inclusion of amplitude observations in the analysis provides constraints on lateral velocity gradients present in the deep slabs which are not resolvable using travel times alone. Travel time and amplitude residual spheres are presented for two deep focus events with good azimuthal coverage, but while interesting patterns are present, quantitative analysis is precluded by the lack of accurate methods for calculating synthetic long‐period seismograms for three‐dimensional slab models. Therefore, we focus our analysis on a two‐dimensional approximation to the downdip geometry of the Kurile and Japan slabs, which allows comparison of a much larger data set with accurate two‐dimensional synthetics, as well as the use of sS and sScS travel time and amplitude patterns as empirical corrections for deep mantle structure. While the limited azimuth range required for this approximation reduces the diagnostic capability of the data, it also increases our confidence that the corrected data are most sensitive to the near‐source region. The sS and sScS observations indicate that the downdip shear wave travel time pattern previously attributed to a deep slab extension is primarily caused by broad‐scale lower mantle heterogeneity. Once corrected for this structure, the S wave observations do not support a simple, undeformed slab steepening in dip and penetrating deep into the lower mantle beneath the Kurile and Japan Islands, as proposed by Jordan (1977) and Creager and Jordan (1984, 1986). Rather, the observations support shorter and/or broader slab models man those previously hypothesized: models which can probably be reconciled with P wave data. Further analysis of the complete three‐dimensional patterns will be required for more precise resolution of the penetration depths of these slabs, if, indeed, any deep slab heterogeneity is actually seismically detectable.