We monitored in normal adult BALB/c mice the serum concentrations of four natural IgM antibodies, two of which show idiotypic complementarity in in vitro assays. In each individual, serum concentrations ofall four idiotypes were found to fluctuate in complex dynamical patterns with low correlation. The spectral power of some such patterns was found to be compatible with the existence of a chaotic regime. Groups of normal adult mice were injected intravenously with low' (10 ng) or moderate (10 ,ug) doses of either of the two complementary idiotypes in saline. This treatment resulted in a pronounced inhibition of the fluctuation in the serum concentration of both complementary idiotypes for periods up to 3 months. Such compensations were not detected for the two unrelated natural idiotypes and were specifically induced, for they did not occur following the injection of unrelated antibodies. These results indicate the functional operation of an idiotypic network among natural antibodies.Two alternative theoretical frameworks are prevalent in immunology. One, derived from the clonal selection theory, is based on the exclusion of self-reactivities from mature antibody repertoires (1, 2). The other, developed from the idiotypic network concept (3), considers (idiotypic) autoreactivities as the basis of immune repertoires and functional activities (4)(5)(6). In spite of extensive experimentation on idiotypic regulation of immune responses and the early demonstration of potential antibody networks (7, 8), evidence for widespread connectivity in the actual repertoire (9) of activated lymphocytes and natural antibodies of normal individuals has been obtained only recently (9,10). These observations, however, fall short of indicating the dynamical properties and functional significance of such idiotypic molecular interactions detected in vitro. Thus, serologic assays, usually carried out after immobilization of one of the two antibody partners on a solid phase, may well detect positive interactions with affinities that are too low to be relevant in the operation of an idiotypic network in vivo.This question has been extensively investigated by injection into normal mice of "artificially" prepared anti-idiotypic antibodies (11). These treatments have drastic consequences on the B-cell repertoire that is expressed upon subsequent antigenic challenges. Although such experiments have shown that idiotypic selection of available antibody repertoires takes place under these particular conditions, the high doses of antibodies injected, and the fact that these had not been isolated from normal individuals, could cast doubts upon the operation of a network under physiological conditions. Furthermore, none of those experiments addressed actual antibody repertoires and they provided no information on network dynamics.We have approached these questions by analyzing the dynamics of two naturally expressed idiotypes known to be connected. We find that their circulating levels normally vary in individually characteristic ways with...