We have previously established a linear relation between temperatures based on the silica content of groundwater and regional heat flow and used the relation to prepare a new heat flow map of the continental United States. We now examine the assumptions upon which the relation is based, the accuracy to which groundwater silica data can be used to estimate regional heat flow, and the limitation of the technique. By averaging silica geotemperatures and traditional heat flow values over 1¿¿1¿ blocks, the linear regression is TSiO2=mq+b, where m and b are constants determined to be 680¿67¿C m2 W-1 and 12.4¿5.1 ¿C. The physical significance of b is mean annual surface temperature, and the product of m times thermal conductivity reflects the minimum mean depth to which groundwaters may circulate. These values are not sufficiently different from our earlier values (m=670, b=13.2) to justify using the newer values. To illustrate the application of the linear regression in predicting regional heat flow, data sets are presented from upstate New York, south central New Mexico, and Egypt. In each case, the predicted heat flow is tectonically reasonable and consistent with whatever traditional data are available.