Some recent seismic studies have suggested the presence of a thin ultra-low P-wave velocity layer (ULVZ) at the base of the mantle, which is interpreted to be due to presence of partial melting. Partial melting would lead to a strong decrease of the S-wave velocity for which there is no seismic evidence. Such a decrease in the S-wave velocity would produce a strong precursor to SKS phase from the conversion of S to P at the upper boundary of the layer. We analyze records of events from the subduction zones in the south-west Pacific region obtained at stations in North America. At the source side, the converted phases propagate in the region, where the ultra-low P-wave velocity has been found earlier. Our analysis demonstrates that either the S-wave velocity drop in this layer is much smaller than predicted by the hypothesis of melting, or the layer is so thin (less than about 10–15 km) that it can not be detected with our technique.