The D-DNA double helix model of poly(dA-dT).poly(dA-dT) proposed in the literature is not in accordance with some notable experimental facts and physicochemical conditions to which it is related. Thus, the fibre X-ray diffraction pattern of D-DNA obtained at a relative humidity lower than that giving the A-DNA form is singularly not taken into account when one assumes that there is only one D structure of B-DNA type. We rather suggest that there are actually two different forms of D-DNA, namely DA which partakes in the D-A-B transitions and DB associated with the D-B change of conformation. Although these two DNA structures have the same helical parameters (pitch and number of residues per turn), in agreement with X-ray data, their detailed conformations are considerably different. Whereas DB is indeed the structure generally defined as D-DNA, a critical analysis based on a comparison between different possible DNA double helices leads us to propose dihedral angles, a set of atomic coordinates and a stereo view of another new form of D-DNA, the DA structural model. It is a right-handed double helix with a dinucleotide as the repeat unit. The furanose rings are of the A-DNA type (C3′ endo) and the bases are hydrogen bonded according to the reversed Hoogsteen pairing. Such a disposition renders the DA model unsuitable for poly(dI-dC).poly(dI-dC), the other alternating polynucleotide observed in the DB structure. The consistency of these two different D-DNA structures of poly(dA-dT).poly(dA-dT) with the general aspects of hydration and helix-helix transitions of DNA, as well as with the conformational variability of AT base sequences, is discussed.