The seminal work of Le Douarin and colleagues (Ohki et al. 1987; Ohki et al. 1988; Salaun et al. 1990; Coutinho et al. 1993) first demonstrated that peripheral tissue-specific tolerance is centrally established in the thymus, by epithelial stromal cells (TEC). Subsequent experiments have shown that TEC-tolerance is dominant and mediated by CD4 regulatory T cells (Treg) that are generated intrathymically by recognition of antigens expressed on TECs (Modigliani et al. 1995; Modigliani et al. 1996a). From these and other observations, in 1996 Modigliani and colleagues derived a general model for the establishment and maintenance of natural tolerance (MM96) (Modigliani et al. 1996b), with two central propositions: (1) T cell receptor (TCR)-dependent sorting of emergent repertoires generates TEC-specific Treg displaying the highest TCR self-affinities below deletion thresholds, thus isolating repertoires undergoing positive and negative selection; (2) Treg are intrathymically committed (and activated) for a unique differentiative pathway with regulatory effector functions. The model explained the embryonic/perinatal time window of natural tolerance acquisition, by developmental programs determining (1) TCR multireactivity, (2) the cellular composition in the thymic stroma (relative abundance of epithelial vs hemopoietic cells), and (3) the dynamics of peripheral lymphocyte pools, built by accumulation of recent thymic emigrants (RTE) that remain recruitable to regulatory functions. We discuss here the MM96 in the light of recent results demonstrating the promiscuous expression of tissue-specific antigens by medullary TECs (Derbinski et al. 2001; Anderson et al. 2002; Gotter et al. 2004) and indicating that Treg represent a unique differentiative pathway (Fontenot et al. 2003; Hori et al. 2003; Khattri et al. 2003), which is adopted by CD4 T cells with high avidity for TEC-antigens (Bensinger et al. 2001; Jordan et al. 2001; Apostolou et al. 2002). In the likelihood that autoimmune diseases (AID) result from Treg deficits, some of which might have a thymic origin, we also speculate on therapeutic strategies aiming at selectively stimulating their de novo production or peripheral function, within recent findings on Treg responses to inflammation (Caramalho et al. 2003; Lopes-Carvalho et al., submitted, Caramalho et al., submitted). In short, the MM96 argued that natural tolerance is dominant, established and maintained by the activity of Treg, which are selected upon high-affinity recognition of self-ligands on TECs, and committed intrathymically to a unique differentiative pathway geared to anti-inflammatory and antiproliferative effector functions. By postulating the intrathymic deletion of self-reactivities on hemopoietic stromal cells (THC), together with the inability of peripheral resident lymphocytes to engage in the regulatory pathway, the MM96 simultaneously explained the maintenance of responsiveness to non-self in a context of suppression mediating dominant self-tolerance. The major difficulty of the MM96 is related to the apparent tissue specificity of Treg repertoires generated intrathymically. This difficulty has now been principally solved by the work of Hanahan, Kyewski and others (Jolicoeur et al. 1994; Derbinski et al. 2001; Anderson et al. 2002; Gotter et al. 2004), demonstrating the selective expression of a variety of tissue-specific antigens by TECs, in topological patterns that are compatible with the MM96, but difficult to conciliate with recessive tolerance models (Kappler et al. 1987; Kisielow et al. 1988). While the developmentally regulated multireactivity of TCR repertoires (Gavin and Bevan 1995), as well as the peripheral recruitment of Treg among RTE (Modigliani et al. 1996a) might add to this process, it would seem that the establishment of tissue-specific tolerance essentially stems from the "promiscuous expression of tissue antigens" by TEC. The findings of AID resulting from natural mutations (reviewed in Pitkanen and Peterson 2003) or the targeted inactivation (Anderson et al. 2002; Ramsey et al. 2002) of the AIRE transcription factor that regulates promiscuous gene expression on TECs support this conclusion. The observations on the correlation of natural or forced expression of the Foxp3 transcription factor in CD4 T cells with Treg phenotype and function (Fontenot et al. 2003; Hori et al. 2003; Khattri et al. 2003) provided support for the MM96 contention that Treg represent a unique differentiative pathway that is naturally established inside the thymus. Furthermore, Caton and colleagues (Jordan et al. 2001), as well as several other groups (Bensinger et al. 2001; Apostolou et al. 2002), have provided direct evidence for our postulate that Treg are selected among differentiating CD4 T cells with high affinity for ligands expressed on TECs (Modigliani et al. 1996b). Finally, the demonstration by Caramalho et al. that Treg express innate immunity receptors (Caramalho et al. 2003) and respond to pro-inflammatory signals and products of inflammation (Caramalho et al., submitted) brought about a new understanding on the peripheral regulation of Treg function. Together with the observation that Treg also respond to ongoing activities of "naïve/effector" T cells--possibly through the IL-2 produced in these conditions--these findings explain the participation of Treg in all immune responses (Onizuka et al. 1999; Shimizu et al. 1999; Annacker et al. 2001; Curotto de Lafaille et al. 2001; Almeida et al. 2002; Shevach 2002; Bach and Francois Bach 2003; Wood and Sakaguchi 2003; Mittrucker and Kaufmann 2004; Sakaguchi 2004), beyond their fundamental role in ensuring self-tolerance (e.g., Modigliani et al. 1996a; Shevach 2000; Hori et al. 2003; Sakaguchi 2004; Thompson and Powrie 2004). Thus, anti-inflammatory and anti-proliferative Treg are amplified by signals that promote or mediate inflammation and proliferation, accounting for the quality control of responses (Coutinho et al. 2001). In turn, such natural regulation of Treg by immune responses to non-self may well explain the alarming epidemiology of allergic and AID in wealthy societies (Wills-Karp et al. 2001; Bach 2002; Yazdanbakhsh et al. 2002), where a variety of childhood infections have become rare or absent. Thus, it is plausible that Treg were evolutionarily set by a given density of infectious agents in the environment. With hindsight, it is not too surprising that natural Treg performance falls once hygiene, vaccination, and antibiotics suddenly (i.e., 100 years) plunged infectious density to below some critical physiological threshold. As the immune system is not adapted to modern clean conditions of postnatal development, clinical immunologists must now deal with frequent Treg deficiencies (allergies and AID) for which they have no curative or rational treatments. It is essential, therefore, that basic immunologists concentrate on strategies to selectively stimulate the production, survival, and activity of this set of lymphocytes that is instrumental in preventing immune pathology. We have argued that the culprit of this inability of basic research to solve major clinical problems has been the self-righteousness of recessive tolerance champions, from Ehrlich to some of our contemporaries. It is ironical, however, that none of us--including the heretic opponents of horror autotoxicus--had understood that self-tolerance, or its robustness at least, is in part determined by the frequency and intensity of the responses to non-self. In the evolution of ideas on immunological tolerance, the time might be ripe for some kinds of synthesis. First, conventional theory reduced self-tolerance to negative selection and microbial defense to positive selection, while the MM96 solution was the precise opposite: positive selection of autoreactivities for self-tolerance (Treg) and negative selection (of Treg) for ridding responses. In contrast, it would now appear that positive and negative selection of autoreactive T cells are both necessary to establish either self-tolerance or competence to eliminate microbes, two processes that actually reinforce each other in the maintenance of self-integrity. Second, V-region recognition has generally been held responsible for specific discrimination between what should be either tolerated or eliminated from the organism. In contrast again, it would now seem that both processes of self-tolerance and microbial defense (self/non-self discrimination) also operate on the basis of evolutionarily ancient, germ-line-encoded innate, nonspecific receptors (Medzhitov and Janeway 2000) capable of a coarse level of self/non-self discrimination (Coutinho 1975). It could thus be interesting to revisit notions of cooperativity between V-regions and such mitogen receptors, both in single cell functions (Coutinho et al. 1974) and in the system's evolution (Coutinho 1975, 1980) as well. After all, major transitions in evolution were cooperative (Maynard-Smith and Szathmary 1995).