We present the first spaceborne observation of a Global Positioning System (GPS) signal reflected from the Earth's surface, specifically from the Pacific Ocean. This result is scaled to obtain the expected voltage signal-to-noise ratio (SNR) and altimetric accuracy for a generic GPS reflections altimetry mission and the current SAC-C and CHAMP missions. Cross-correlating a three-parameter phase model with both a 1-s and 4-s segment of spaceborne imaging radar-C (SIR-C) calibration data, recorded before and after a Galapagos Islands imaging pass, results in beam-limited signals having voltage SNRs of 10 and 334, respectively. Evidence for these results being reflected GPS signals includes: (1) The signals' temporal shapes agree closely with that predicted using a detailed scattering model, at two different observation geometries, and differ significantly from the expected direct signal shapes. (2) The signal in the 4-s data has a measured coherence time of 1.0 ms, which agrees closely with that expected for a reflected signal and is completely inconsistent with the direct signal's coherence properties. (3) The 1- and 4-s signals' voltage SNR is maximized by shifting the model frequency -2740 Hz and 497 Hz, respectively from that expected from their respective specular reflection points, or -2875 Hz and 690 Hz from the expected direct signal frequencies. These values agree with the -2900 Hz and 510 Hz Doppler frequency shifts expected from those points on the surface corresponding to the antenna's pointing direction, thus illustrating beam-steering effects on the surface. (4) Plausible hypotheses for the detected waveform being a corrupted direct signal, including second-order mismodeling effects, shuttle multipath effects, or a band-pass cutoff of the GPS spread spectrum, are shown to be inconsistent with the data. Space-based observations of reflected GPS signals, like the ones presented here, may enable a new class of ocean topography measurements unavailable from traditional altimeters, such as TOPEX/Poseidon, and perhaps surface wind vector measurements. Making such observations with sufficient SNR will require unusually large, high-gain antennas. The measurement presented here is scaled to assess the expected SNR for those applications. Because this result lies in a nonlinear scaling regime, the correct scaling equations are presented, and the expected signal strength from a generic GPS reflections altimetry mission is derived to illustrate the most important contributions to the signal SNR. An SNR estimate is also derived for the SAC-C and CHAMP missions, which are expected to make GPS reflection measurements in the near future. Finally, a qualitative wind speed determination is extracted from the observed signal.