Abstract. We develop a structure theory for two classes of infinite dimensional modules over tame hereditary algebras: the Baer modules, and the Mittag-Leffler ones. A right R-module M is called Baer if Ext 1 R (M, T ) = 0 for all torsion modules T , and M is Mittag-Leffler in case the canonical mapWe show that a module M is Baer iff M is p-filtered where p is the preprojective component of the tame hereditary algebra R. We apply this to prove that the universal localization of a Baer module is projective in case we localize with respect to a complete tube. Using infinite dimensional tilting theory we then obtain a structure result showing that Baer modules are more complex then the (infinite dimensional) preprojective modules. In the final section, we give a complete classification of the Mittag-Leffler modules.Since the fundamental work of Ringel [29], the study of infinite dimensional modules has become one of the challenging tasks of the representation theory of finite dimensional hereditary algebras. In the present paper, we consider in detail two classes of infinite dimensional modules over tame hereditary algebras: the Baer modules, and the Mittag-Leffler ones.Besides proving structure results, we investigate further the surprising analogy with modules over Dedekind domains discovered in [29]. Indeed, the current progress relies heavily on applications of the recent powerful set-theoretic and homological methods developed originally for modules over domains, cf.[3] and [15].Baer modules can be defined in a rather general setting [26]: Let R be a ring and T be a torsion class in Mod−R. A module M ∈ Mod−R is a Baer module for T provided that Ext 1 R (M, T ) = 0 for all T ∈ T . In the particular case when R = Z, and more generally when R is a integral domain and T is the class of all torsion R-modules, we obtain thus the classical notions of a Baer group [8] and a Baer module [17]. It took quite a long time to prove that these notions actually coincide with the well-known notions of a free group and a projective module, respectively. Countable Baer groups were shown to be free already in 1936 by Baer [8], but the arbitrary ones only in 1969 by Griffith [16]. The projectivity of the classical Baer modules was also shown in two steps spreading over decades, however, in a different order. First, a reduction to countably presented modules was proved by set-theoretic methods by Eklof, Fuchs