The use of the Flanagan and Barondes model(14) describing affinity partitioning as an aid in designing separation systems is discussed. Experimental studies are described for affinity partitioning of vancomycin, a glycopeptide antibiotic, in a water-methoxypolyethylene glycol-dextran system using methoxypolyethylene glycol-dextran system using methoxypolyethylene glycol bound D-alanyl-Dalanyl-D-alanine or D-alanyl-D-alanine as the reversible affinity ligand. Even for this ideal case of 1:1 binding interaction, the model only qualitatively predicts the affinity effect when all model parameters are measured independently. The discrepancy between measured and predicted values can be attributed to a difference in exposed surface of the free antibiotic and ligand compared to that in the bound state.The effect of experimentally varying model parameters is also described. It was determined that a polymers-ligand which partitions more strongly to the top phase would provide the most significant enhancement to this affinity partitioning system. Such an improvement can be made by increasing the molecular weight of the hydrophobicity of the polymer-ligand. A process for vancomycin recovery from fermentation broth using D--alanyl-D-alanine sepharose as affinity ligand is described.