Liquefaction and paleoliquefaction studies are being used increasingly to interpret ground motion parameters ( i.e. , acceleration distribution, earthquake magnitude, and epicentral location) in regions of the world that experience infrequent, but damaging, earthquakes. These studies are especially applicable when the most recent damaging earthquakes occurred prior to the development of ground-motion instrumentation. For example, Martin and Clough (1994) examined liquefaction features formed during the 1886 Charleston, South Carolina earthquake; Pond (1996) and Obermeier and Pond (1999) examined paleoliquefaction features formed during a number of paleoearthquakes in the Wabash Valley region along the Indiana–Illinois border; Dickenson and Obermeier (1998) and Obermeier and Dickenson (2000) examined paleoliquefaction features formed during the 1700 Cascadia earthquake of the Pacific Northwest; and Ellis and de Alba (1999) recently examined liquefaction features formed during the 1755 Cape Ann, Massachusetts earthquake. Many of these studies back-calculate seismic parameters using the results of penetration tests conducted long after the earthquake. The most common method to back-calculate a lower-bound peak ground surface acceleration is to use the lowest value of penetration resistance obtained at a historic or paleoliquefaction site in conjunction with a correlation for liquefaction resistance of sandy soils, e.g. , Seed et al. (1985) or Stark and Olson (1995). In some studies, the current value of penetration resistance is assumed to be the “representative” value for back-analysis ( e.g. , Pond, 1996), while in other studies corrections are made to the current value of penetration resistance prior to back-analysis ( e.g. , Ellis and de Alba, 1999). This paper examines some of the factors that affect the interpretation of a “representative” penetration resistance for back-analysis of seismic parameters. The interpretation of these factors is illustrated using an example historic liquefaction site. The back-calculated values of a max from the example case history show the sensitivity of the back-analysis …