An emerging technique for the recovery of microbial secondary metabolites, such as antibiotics, is the use of colloidal liquid aphrons (CLAs) in pre-dispersed solvent extraction (PDSE) processes. Knowledge of the extraction kinetics and the limiting mass transfer resistances will be vital for efficient process design and operation. In this work, the rates of erythromycin extraction using CLAs and conventional aqueous-organic, two-phase systems have been investigated. The CLAs used were formulated from 1% w/v Softanol 120 in decanol and 0.5% w/v SDS in water. The rate of erythromycin extraction with CLAs dispersed in well-mixed systems was found to be extremely rapid with equilibrium being achieved within 15 s or less. Overall erythromycin mass transfer coefficients, K-o, were typically 6.3 x 10(-6) m s(-1) for extraction experiments, and 1.0 x 10(-6) m s(-1) for stripping experiments. The rapid rates of erythromycin transfer were attributed to the small size of the dispersed CLAs, typically 5 mum diameter, and hence the large interfacial area available for mass transfer of around 15 x 10(3) m(2) m(-3). Experiments over a range of Reynolds numbers indicated that both the extraction and stripping processes were probably under mixed control of an interfacial resistance and boundary-layer diffusion. To investigate the influence of the surfactants used for aphron formulation on erythromycin extraction, further experiments were performed using aqueous-organic, two-phase systems in a non-dispersive stirred cell. The measured K-o values were of the same order of magnitude as in the case of experiments with dispersed CLAs. When present individually, or together with SDS, the non-ionic Softanol surfactant was seen to retard erythromycin extraction rates in all cases. In contrast, the presence of SDS was found to enhance K-o values by a factor of 2-4 compared to surfactant-free systems. This may be attributed to a specific interaction between individual SDS acid erythromycin molecules and/or the generation of interfacial turbulence at a microscopic level. Further investigations are underway to elucidate the mechanism responsible. The concentration of either surfactant was also found to significantly affect the measured K-o values. The prediction of K-o values using existing correlations, and the implications of the experimental results on contactor design and operation are also discussed.