Batches of Sepharose–lysine, which varied in lysine content from 35 to 430 μmol/g of dry gel, were prepared by varying the quantity of CNBr in the activation reaction. The batches were tested for heparin binding by using a controlled chromatographic procedure. Sepharose–lysine, containing < 150 μmol of lysine/g, did not significantly bind heparin whereas conjugates with > 400 μmol/g retained the entire heparin load. For intermediate batches of Sepharose–lysine (150–400 μmol/g) the quantity of heparin bound largely paralleled the lysine content. Thus, Sepharose–lysine of an intermediate lysine content separated heparin into an unretained fraction and a bound fraction which was recovered from the column by eluting with 1 M NaCl. On testing for anticoagulant activity by factor Xa inhibition assay, no significant difference in specific anticoagulant activity was observed between these heparin fractions and the heparin load. However, from gel filtration studies, a substantial difference in molecular size was noted. An unretained heparin fraction from Sepharose–lysine was of a lower average molecular weight than the parent heparin. In contrast, a retained heparin peak was of a higher average molecular weight compared with the parent heparin. These observations were confirmed by studying the chromatographic properties of low (10 000) and high (23 000) molecular weight heparin samples on various Sepharose–lysine batches. A model is proposed to explain this discriminating property of Sepharose–lysine. For a conjugate containing 400 μmol/g, the mean lysine spacing is calculated at 47 Å (1 Å = 0.1 nm), which is approximately equivalent to five to six disaccharide units in heparin.The property of Sepharose–lysine to bind heparin was compared with the affinities of the mucopolysaccharide for both thrombin and antithrombin III. Evidence has been proposed for the involvement of lysine residues of antithrombin III in this process. Our investigations suggest that lysine, in addition to arginine, groups of thrombin are also involved in heparin binding. By specifically modifying two to four lysine residues using nitrous acid, the heparin-binding capacity of the enzyme and its plasma clotting activity were largely destroyed, although the esterase activity was retained.