During analysis of pyruvate kinase distribution in developing guinea-pig liver it was observed that a substantial proportion of the activity remained associated with the microsomal membrane fraction ('microsomes'). Although some of this could be removed by washing with sucrose, the majority required detergent treatment for liberation, and even then at least one-half remained attached to the microsomes. Estimates of the contribution of this fraction to total cell pyruvate kinase activity indicated that it was more than 50% of the total, and this is likely to be an underestimate because of the continued latency of the enzyme even in the presence of detergent. The susceptibility of the microsomal enzyme, whether released by detergent or sucrose washing, to inactivation by Triton X-100 suggested it to be different from the cytosolic enzyme, which was stable under such conditions. (The microsomal enzyme required the presence of additional protein, such as bovine serum albumin, to maintain stability.) This view was confirmed by DEAE-cellulose chromatography and particularly isoelectric focusing, where the microsomal enzyme was shown to consist of at least four forms, which were distinctly different from those in the cytosol. Those data and the kinetic properties of the four forms in the membrane fraction indicate that the microsomal pyruvate kinase could consist of four counterparts to the cytosolic isoenzyme forms. These results are discussed in relation to the two possible explanations for the phenomenon (not mutually exclusive): that the more hydrophobic membrane forms are precursors of the cytosolic enzyme and that they may be part of functional glycolytic pathway in the microsomes of developing liver.