To test the applicability of Gibbs-Donnan diffusion equilibrium formulae to ultrafiltration, we have examined sodium (Na) and chloride (Cl) concentrations in ultrafiltrates collected directly from cellophane and cuprophane coils perfused with albumin-electrolyte solutions in a closed reservoir system. The ratio of ultrafiltrate to perfusate water Na concentration (UF/P Na) and the similar ratio UF/P Cl were 1.0 when albumin concentration was zero. UF/P Na decreased and UF/P Cl increased as albumin concentration was increased and pH maintained at 8.0–8.4. By linear regression, UF/P Cl = 0.017 [albumin (g/l00 ml)] +0.97 and UF/P Na = 0.99–0.011 [albumin]. With a nearly constant circulating albumin concentration (7.4–8.5 g/l00 ml) UF/P Na increased curvilinearly from 0.85 to 1.2 as circulating pH was decreased from 8.7. to 1.8 with HCl, and approached 1.0 near the isoelectric point of albumin (pH = 4.2). Simultaneously UF/P Cl decreased from 1.05 to 0.66 in a similar manner and approached 1.0 at a pH near 7.2. In all studies, measured UF/P Cl was usually less than the reciprocal of UF/P Na. Neither ultrafiltration rate nor the type of membrane used altered UF/P values. We conclude: 1. Protein charge does influence UF Na and Cl concentrations in directions predicted by formulae applicable to diffusion equilibrium; 2. Deviations of UF/P Cl from Gibbs-Donnan and isoelectric predictions are compatible with albumin chloride binding as reported in diffusion studies; 3. The electrolyte composition of filtrates from capillary beds such as glomeruli might be predicted to demonstrate similar deviations from Gibbs-Donnan formulae.