We present evidence for precursors to the SCS and SSCS (SDS and SSDS) phases observed at epicentral distances smaller than 30¿. These precursors are intermittently observed in broadband recordings from the six Incorporated Research Institutions for Seismology stations used (western Pacific region and South America). They appear approximately 35--50 s ahead of the SCS arrival on the transverse component and are independent of hypocentral depth. The observed precursors are characterized by relatively large amplitudes in the frequency band from 0.05 to 0.2 Hz. A more detailed analysis is restricted to a subsidiary data set from station SNZO (New Zealand). The coherency, frequency dependence, slowness, polarity, and polarization of the precursors are discussed. Many explanations for SDS can be rejected, and we conclude that a reflector 180 km above the core-mantle boundary causes these occasionally strong precursors. The large SDS/SCS amplitudes at low frequencies require an unrealistically large impedance contrast for a one-dimensional model. We test the possibility of focusing the SDS phase by a discontinuity with topography and show that structure with scale lengths of ~19¿--25¿ (1200--1600 km, i.e., larger than the Fresnel zone) can account for large intermittently observed SDS amplitudes with their geometrical reflection points within the same Fresnel zone. This is surprising, since it is often assumed that scale lengths smaller than the Fresnel zone must be responsible for variations within a Fresnel zone. The limited data set permits no conclusions about the global properties of this phase or its implied reflector. ¿ American Geophysical Union 1996