Clinical dextran was subjected to varying degrees of acid and enzymatic hydrolysis and fractionation with alcohol and ether. The number average molecular weights (Mn) of the different fractions varied from 17,000 to 850. Cross-precipitation with Type II antipneumococcal rabbit serum decreased with the molecular size of the fractions. Precipitation occurred with the Mn 5,000 but not with the Mn 2,200 fraction. Precipitation was inhibited about equally by fractions Mn 14,000 to 950. There was much less inhibition by a smaller fraction (Mn 850.) Native and clinical dextrans elicited fatal anaphylactic shock regularly in guinea pigs which had been passively sensitized with Type II antipneumococcal rabbit serum. The Mn 14,000 dextran fraction elicited a moderate reaction in only one of 10 similarly sensitized animals. Smaller fractions, even in large doses, did not provoke anaphylaxis at all. Varying degrees of desensitization against native dextran followed large challenging doses of several of the fractions. Desensitization was complete for 24 hours with the Mn 14,000 fraction, complete but transient (less than 3 hours) with the Mn 1,500, and negligible with the Mn 850 fraction. Studies of these fractions showed that progressive molecular cleavage transformed dextran from (1) an anaphylactogen to anaphylactogenically inactive fractions, (2) an anaphylactogen to specific inhibitors of anaphylaxis, and (3) a specific precipitator to specific inhibitors of precipitation. But there was no parallelism between the progressive reduction in molecular weights of the fractions and the loss of precipitating, anaphylactogenic inhibitory and desensitizing powers. Dextran is an allergen in man. This limits its usefulness as a plasma expander. These findings suggest the need to investigate the possibility that reactions to infusions of clinical dextran could be prevented by prior desensitizing infusions of dextran fractions, with molecular weights larger than 1,500 and smaller than 14,000.