We conducted tests of a three-component broadband ocean-bottom seismograph (OBS), including side-by-side comparisons with the broadband Global Seismic Network/U.S. National Seismic Network station HKT at Hockley, Texas, and a 28-day deployment in the Gulf of Mexico. Our goals were to evaluate seismometer performance and determine whether our seafloor deployment strategy allows useful earthquake data to be collected. The seismometer generally performed well, but showed unexpectedly high intrinsic noise at frequencies above 1 Hz and produced occasional spikes that were confined to a single component at a given time. We identified 32 earthquakes from the Gulf of Mexico data; only five additional events were observed at the nearby, highly sensitive station HKT on land. While noise levels were higher throughout the 0.01-20 Hz frequency band in the Gulf of Mexico compared with HKT, site amplification effects in the gulf were only significant at frequencies above 1 Hz and power in most earthquake signals peaked at frequencies below 0.1 Hz, nearly coinciding with a minimum in background noise. We found the seismometer and OBS package encouraging for far-regional and teleseismic studies, given the low cost of the OBS and inexpensive means for deployment and recovery, but less encouraging for detecting and locating small-magnitude local and near-regional events.