Since the early 1970s, most geodynamic models proposed for the Eastern Mediterranean consider that the initiation of extension in the Aegean, during the late Middle Miocene or later, was a direct consequence of the westward extrusion of Anatolia away from the Arabia–Eurasia collision front. In the late 1990s, a few models have taken into account the increasing evidence for an earlier initiation of Aegean extension, arising mostly from the analysis of deformation within the various Alpine metamorphic complexes of the area. Based largely on our own investigations, we first present a review of the data that reliably constrain the minimum age for the initiation of extension in these complexes. It shows that regional-scale extension with a pattern of stretching orientations similar to that of the Pliocene–Pleistocene was already strongly active in the Aegean before the onset of Arabian indentation into Eurasia (Lower Miocene vs. Middle or Upper Miocene). This implies that the initiation of Aegean extension did not result from the lateral extrusion of Anatolia. Instead, extension may have started, owing to a process of gravity spreading of the continental lithosphere that had previously been thickened during Alpine collision. The main arguments favouring this interpretation are presented, and a preliminary scenario is proposed for the Neogene evolution of the Eastern Mediterranean, in which lateral extrusion of Anatolia occurred lately during southward spreading of the Aegean lithosphere. In order to test this hypothesis further, physical experiments simulating the horizontal spreading of a continental lithosphere toward a free boundary have been carried out. The analysis of various elements of the deformation field (displacements, strain, rotations, fault pattern and kinematics), both in the experiments and in the Aegean domain, reveals an excellent agreement between the two. Not only do early stages of Aegean extension appear to fit the gravity spreading hypothesis, but so does the recent to present internal deformation of the Aegean. Transverse (≈E–W) active shortening observed in the northern Aegean is commonly taken as one of the most striking pieces of evidence that Aegean extension results from the lateral extrusion of Anatolia. The experiments indicate that such transverse shortening may equally develop as a result of pure gravity spreading, associated with a pattern of faults and rotations that compares well with that of the northern Aegean. Pure dextral shearing along some strands of the North Aegean Trough fault system, and the related development of an asymmetric pattern of faults in the northern Aegean, are eventually the only features that evidently arise from the extrusion of Anatolia, since less than 3 Ma. Finally, we argue that during the early stages of Aegean extension, since at least 21 Ma and until ca. 11 Ma, the distribution of extensional strain on a regional scale has been controlled by the distribution of earlier thickening. Post-thickening thermal weakening, through partial melting of deep levels of the crust, probably accounts for the localization of extensional strain within the areas of thick crust. Restoring the Aegean in its most likely configuration at the onset of bulk crustal extension suggests that (1) the extensional strain rate remained roughly constant, or has eventually decreased, since the onset of Aegean extension, and (2) the amounts of extension before and after 16 Ma have been roughly equal.