Two different approaches are considered to the problem of determining seismic velocity ratios from P and S arrival data from a group of earthquakes. Each method arises naturally from a different assumption about the sources of error. If the S reading error is the predominant source of error, S versus P regression, an extension of the Wadati Plot method to pooled data, should be used. If, however, S reading errors do not dominate, a symmetric regression should be used. A natural extension of the latter method allows the determination of station-by-station variation of α/β.
The methods are illustrated with data from an aftershock sequence which occurred off the coast of northern California. The average velocity ratio is not as well determined as the station-by-station variation about the average. However, the results show that the method has some utility in the study of regional variation of seismic velocities.