The results from a study of linear regression of the estimated boundary layer potential on a set of solar wind variables is displayed in Table 1. The variables included in the study were the solar wind velocity, v, number density, n, pressure, P, the magnetic field components, BX, BY, BZ, and BT = BY2 + BZ2), and the electric field in the transverse plane, E = vBT. Among the simple variables tested, the solar wind pressure shows the best correlation with the derived boundary layer potentials, with the ratio of explained variance to total variance R2 = 16.5%. If we add an E-dependence to the model, i.e. ΦBL = c0 + c1P + c2E, where c0, c1, and c2 are regression coefficients, the R2 value only increases with 1%, a fairly insignificant increase. These results indicate that most of the correlation between ΦBL and E is likely due to a co-dependence on v and that no significant direct relation exists between the two. It is of importance to note that the results above do not necessarily indicate a direct (causal) dependence between ΦBL and P. It should rather be considered as an interplay of the different viscous parameters, i.e., the solar wind velocity, density and/or the pressure. Other viscous-like functions give similar or better results, such as the v2 function provided by Newell et al. , which gives an R2 value of 27% (corresponding to a correlation coefficient of 0.52), although the direct physical significance of this function is unclear. It is important to note that even though the R2 values reported are rather low, they are clearly statistically significant. These findings thus bring credibility to the method used for deriving the boundary layer potential, which leads us to believe that the potentials are on average reliable estimates and that the method is adequate for this type of analysis. When individual cases are considered however, they are still to be treated only as rough estimates. The analysis assumes that no dayside merging cells are present, as in the model by Burch et al. . Such cells are not likely to form for the IMF Bz conditions imposed on the data set [Reiff and Burch, 1985], and any larger contribution from such cells would be reflected in a stronger E or By dependence in the regression analysis.
 With the general acceptance of the boundary layer potential estimates, we can proceed to investigate their properties and distribution, shown in a histogram in Figure 2. Both the mean and the median values of the distribution are around 9–10 kV. If instead the ratio of the boundary layer potential to that of the total dynamo related potential is considered, i.e., ϕBL/(ϕBL + ϕRC), the mean and median values average to around 30–35%.