Equilibrium isotherms have been determined for the adsorption of two single-component and one binary acid dye solutions on carbon. The ideal adsorbed solute theory (IAST) has been used to predict isotherm data for the binary system using the single-component dye sorption data alone. A detailed analysis has been carried out to investigate the effect of different error criteria for determining single-component isotherm parameters and their influence on the results of IAS model simulations using the Langmuir and Freundlich isotherm equations. It was found that the "best-fit" single-component isotherm based on the minimum error criteria did not result in the "best-fit" IAS model predictions. Furthermore, the IAS model producing the closest fit to the binary experimental data did not use the same isotherm model as the single-component data for one dye. Reasons for this apparently anomalous behaviour have been discussed and the most likely cause appears to be dye-dye interactions in the binary system.