Fault plane solutions are derived from systematic trial-and-error (“grid”) testing of three-component body waveform data from a single station. Modeling P and SH waveform data from five shallow events recorded teleseismically demonstrates that radiation pattern information contained within the interference of the direct wave and surface reflections and the overall relative amplitude between P and SH waveforms is sufficient to discriminate between fault type (e.g., strike-slip versus dip-slip) and often agrees with well-constrained first-motion studies. Events studied are the 9 April 1968 Borrego Mountain, California; 20 June 1978 Thessaloniki, Greece; 13 August 1978 Santa Barbara, California; 20 May 1979 Alaska; and 6 August 1979 Coyote Lake, California, earthquakes. It is also shown using data from the 27 July 1980 Sharpsburg, Kentucky, earthquake that inclusion of pP/P and sP/P polarity and amplitude information to an otherwise unconstrained first-motion study can significantly improve the quality of the fault plane solution. Although there are many potential problems (source multiplicity, directivity, etc.) which can prohibit finding a good model with these techniques and inclusion of data from many stations is clearly desirable, the results of this study suggest that sparse, high-quality waveform data sets may be as or more useful for obtaining source mechanisms than standard first-motion studies. At a minimum, they should be performed together as a consistency check. This procedure would be most useful in the common situation where only a few receivers are available for a particular event.