Seismic events along a geophysical fault-line are recorded in (space, time) pairs, where the time component represents the distance of the point along the fault. A small fraction of the pairs result from bursts of energy which propagate along the fault at approximately constant speeds, causing seismic events as they go. These bursts generate roughly linear clusters of points in the (space, time) plane, the gradient of the line being proportional to the speed at which energy travels along the fault. Identifying these clusters is of intrinsic geophysical interest. The fault line in question is a section of the San Andreas fault near Parkfield, California. We discuss statistical methodology for identifying approximately linear clusters of points. This problem has connections to that of ley-line analysis, addressed in the UK in the 1980s.