To quantitatively evaluate the correlations between the amount of initial burst release and the surface-associated protein, and between the onset time for the second burst release and the matrix polymer degradation. Methods. Human serum albumin (HSA) was microencapsulated in polylactide (PLA) and poly-dl-lactide-poly(ethylene glycol) (PELA) with PEG contents of 5, 10, 20, and 30%, respectively, using the solvent extraction procedure based on formation of double emulsion w/o/w. Microspheres with similar particle size (1.7-2.0mm), similar protein entrapment (2.1-2.8%) but different surface-associated proteins (9.3-53.6%) were used to evaluate the in vitro matrix degradation and protein release profiles. Degradation was characterized by studying the intrinsic viscosity decrease, medium pH change, and weight loss of the microspheres. Results. The matrix degradation and protein release profiles were highly dependent on the polymer composition of the microspheres. Faster decreases in the intrinsic viscosity of recovered matrix polymer, the microspheres weight, and the pH of degradation medium, and earlier onsets for the break in intrinsic viscosity reduction and the mass loss were detected for PELA microspheres with higher PEG content. The hydration and swelling of microspheres matrix contributed greatly to the degradation of matrix polymer. The HSA release showed triphasic profile and involved two mechanisms for all the microsphere samples. Smaller amount of initial burst release, larger gradual release rate, and earlier onset for the second burst release were observed for HSA from matrix polymer with higher PEG content. The extent of the initial burst release was quantitatively related with the surface-associated protein. The second burst release of HSA was observed to occur within 1 week after the onset for mass loss, which was also the break in the intrinsic viscosity reduction rate. Conclusion. Protein release profiles could be rationalized by optimizing the matrix polymer degradation and microsphere characteristics.