In the classical 'slope-intercept' method of determining the zeta potential and the surface conductance, the relationship between ΔP and E(s) is measured experimentally at a number of different channel sizes (e.g., the height of a slit channel, h). The parameter (ε(r)ε₀ΔP/μE(s)λ(b)) is then plotted as a function of 1/h and linear regression is performed. The y-intercept of the regressed line is then related to the ζ-potential and its slope to the surface conductance. However, in this classical method, the electrical double layer effect or the electrokinetic effects on the liquid flow are not considered. Consequently, this technique is valid or accurate only when the following conditions are met: (1) relatively large channels are used; (2) the electrical double layer is sufficiently thin; and (3) the streaming potential is sufficiently small that the electroosmotic body force on the mobile ions in the double layer region can be ignored. In this paper a more general or improved slope-intercept method is developed to account for cases where the above three conditions are not met. Additionally a general least-squares analysis is described which accounts for uncertainty in the measured channel height as well as unequal variance in the streaming potential measurements. In this paper, both the classical and the improved slope-intercept techniques have been applied to streaming potential data measured with slit glass channels, ranging in height from 3 μm to 66 μm, for several aqueous electrolyte solutions. The comparison shows that the classical method will always overestimate both the ζ-potential and the surface conductance. Significant errors will occur when the classical method is applied to systems with small channel heights and low ionic concentrations. Furthermore, it is demonstrated that traditional regression techniques where the uncertainty is confined only to the dependent variable and each measurement is given equal weight may produce physically inconsistent results. (C) 2000 Academic Press.