Abstract 1. 1. An analysis of the problems encountered in obtaining reproducible electrophoretic separation of proteins in starch gel has shown that there are critical limits for the preparation of soluble starch, buffer, and proportions of these used in making the gels. These limits are easily obtained with ordinary laboratory equipment, and an awareness of those steps in which accuracy is needed will permit considerable modification without loss of reproducibility. 2. 2. Conditions of electrophoresis which employ temperatures lower than room temperature but above freezing and an 18-hour application of voltage resulted in the separation of more protein fractions than were described previously in normal serum. 3. 3. Semiautomatic staining procedures using continuously decolorized solvent for elimination of excess dye provide reproducible staining of starch gel electrophoretograms. 4. 4. An analysis of the reflectance curves obtained for normal serum electrophoretograms by use of a specially designed reflection densitometer demonstrated that a high degree of reproducibility can be obtained. 5. 5. A survey of 100 normal and 400 pathologic sera indicates that while there are many normal serum types, each with complex patterns, the changes in disease are sufficiently marked that this method holds considerable promise for further studies. Because of the marked variation in normal serum protein patterns, accurate information regarding the changes specific to a given disease will require a quantitative method of gel strip analysis and adequate data processing procedures. 6. 6. When employed in testing the homogeneity of several purified proteins, zone electrophoresis in starch gel gave rapid and sensitive results comparable with or better than those obtained by other methods.