Oil and gas exploration uses accurate interpretations of the sedimentary basins that populate the continental crust in rift zones and on continental margins, in order to map source and reservoir rocks and hydrocarbon traps, and to interpret their evolutionary histories. The acquisition of much new high-resolution seismic data, essential for these tasks, has been driven by the industry’s move into deep sedimentary basins at extended continental margins. Complemented by studies of obducted analogues, for example in the Alps, and regional compilations of potential field data, these data have also been applied to the interpretation and understanding of the processes that complete the breakup of continents. Among the industry’s other requirements are estimates of paleogeography and paleostress, which need to be calculated or interpreted from plate tectonic reconstructions. Usually, one of the industry’s principal target times for this is the margin development phase, which follows the development of an intracontinental rift zone and immediately precedes seafloor spreading. As well as their use in interpretations or models of paleogeography or paleostress, reconstructions of this phase have been seen as a context in which to interpret the processes of extended continental margin development. Blaich et al. (2011) for example, interpreted a complex history of margin generation in the context of Torsvik et al.’s (2009) South Atlantic reconstruction. Dunbar and Sawyer (1989) and Williams et al. (2011) attempt to undo the effects of continental stretching in passive margins using the divergence direction of the surrounding plates, taken from near-fit plate reconstructions, as constraints. Intuitively, most plate reconstructions for the margin development phase are based on fitting extended continental margin features, the principal products of breakup, together. Interpretations of continental margin processes made in the context of plate reconstructions, therefore, whilst self-consistent, are based on circular reasoning. Errors or inaccuracies in the margin interpretations will be reflected in the plate reconstruction, which can only serve to validate the original interpretation. This may not be a problem if the inaccuracies and hence the possible errors related to reconstructions of interpreted margin features are small. But how accurate might we expect reconstructions of the margin development phase to be? Any plate reconstruction assumes that at any time Earth’s surface hosts a few large rigid plates edged by narrow deformation zones and moving on circular paths that can be described using Euler rotations. Reconstruction requires identifying and reuniting linear conjugate isochrons (LCIs); the mirror images of ancient divergent margins on two plates. In early margin-development phase reconstructions, the continental shelf was used as a proxy LCI in spite of knowledge that their shapes had altered by post-breakup processes such as sedimentation and volcanism (Bullard et al., 1966). Later work moved on to use the concept of continent-ocean boundaries (COBs) interpreted from seismic reflection, refraction, or potential field data. Characteristically, COBs determined in these different ways may not agree at much better than 50 km resolution, suggesting there is no universal physical signature of the moment at which oceanic crust begins to be produced. COBs are usually only dateable at the stratigraphic resolution of associated ‘break-up’ unconformities, which are unlikely to be instantaneous time markers over long distances. The blurry nature of the COB concept is due to the action of processes, such as hyperextension and excess volcanism that can alter the physical properties of pre-existing continental crust to such an extent that they lie within the natural variability of oceanic crust. These and other processes define the concept of a continent ocean transition zone (COTZ), which is therefore non-linear, non-rigid, and diachronous. Typical COTZ widths are 50-200 km, which may serve as a pessimistic indication of the spatial resolution we might expect of COB-related plate models. More importantly, if such models are used predictively, errors within this range can propagate to much larger proportions by rotation, leaving the COB concept as damaging as well as nebulous. Despite this, and despite the lack of a process-based definition of the term ‘continent-ocean boundary’, COBs remain widely used in the plate reconstruction literature for the lack of alternative near-fit LCIs. Post- margin development phase reconstructions use the simpler features of oceanic crust as LCIs for modelling that is independent of interpreted COTZ processes. Such features include fracture zones, which can be mapped from orbit at the resolution of a satellite radar altimeter’s footprint, and marine magnetic anomalies that pinpoint the ancient ridge crest with an accuracy equivalent to the distance travelled by an average pillow lava flow before it solidifies. In both cases, this means we can locate the products of rigid plate tectonics to around 15 km or better, and date them to within a few tens of thousands of years, both an order of magnitude more precisely than for an undefined line within an average COTZ. Even after error propagation, this guarantees plate reconstruction resolution in many locations that is still sharper than the width of an average COTZ. By extrapolating models like this backwards from modelled epochs that are close as possible to the time of continental margin formation, we may be able to constrain plate-linked continental extension and COTZ formation independently of anything interpreted, or modelled, from the passive margins or COTZs themselves. We use the South Atlantic ocean to illustrate this. We generate a high-resolution set of Euler rotations by minimizing the misfits of small circle segments about finite rotation poles to picks taken along fracture zones, and of target and rotated isochron picks to great-circle segments about the same poles. This approach is more flexible than the well-known Hellinger (1981) technique, which is confined to use conjugate isochron fits only, allowing the use of fracture zone constraints all the way to where they merge with COTZs. The technique yields smooth plate motion estimates with relatively few magnetic isochron data. Misfit distributions suggest that no more than two plates diverged in opening of the South Atlantic. At 95% confidence level, the rotations are about poles significantly different than those derived with COBs. At 100 Ma, as the ocean’s northern continental margin segments formed, the differences can be shown to be preferable at the 95% confidence level for paleomagnetic data (Font et al, 2010). At earlier times, although the conjugate lengths of COTZ segments are similar, the differences in plate motion azimuth are visually appreciable in terms of the orientations of real and synthetic fracture zones (Eagles, 2007). The model is clearly more accurate in describing plate divergence than those based on COBs can be. What does this new accuracy allow us to interpret about the development of the continental margins? Synthetic flowlines overlap the northern continental margins by 550 km. This illustrates the well-known fact of breakup in the north being much later than in the south. This diachronous breakup must have been accommodated by northwards propagation of the mid-ocean ridge, but what did the remaining boundaries of the plates look like during the propagation? There is fragmentary evidence of these boundaries in the form of Cretaceous faults, basins, and volcanism in Africa and South America. This is interpreted usually as evidence for multiple plate involvement in South Atlantic opening (e.g. Burke et al., 1971), a concept that has been used in COB-centric paleogeographic reconstruction studies. Nürnberg and Müller (1991) Moulin et al., (2010) and Torsvik et al., (2009) used ‘permissible’ amounts of motion on these features to improve COB fits. Heine et al. (2014) tried to directly constrain and apply plate motions from the fragmentary evidence in them. Our new model, in contrast, reveals no evidence for the divergence of more than two plates. The Cretaceous intracontinental deformation of Africa and South America can therefore be most parsimoniously interpreted as evidence for successive locations of the continental portion of an evolving two-plate boundary, rather than a set of aulacogens or other third- or nth plate features. It can be used to complete margin development phase paleogeographies for any time slice during that migration, and to describe the closed boundaries of two plates that can form the context for paleostress interpretations or calculations. Both are upcoming aims of the Continental Margin Research Group at Royal Holloway. The two-plate model can also be used to quantify plate-linked strain in continental Africa and South America during the northward ridge propagation. It shows that the continental interiors need to have deformed by an amount equivalent to around 1100 km of margin normal extension. A well-defined set of Cretaceous intracontinental basins (e.g. Colorado and Salado Basins, Benue Trough, Iullemedin Basin) can account for 47-60% of this deformation if they developed from an original crustal thickness of 35 km. The remainder might be attributed to a mixture of variable extension of COTZs (Vink, 1982), transfer motions, and modest extension on more enigmatic features associated with Cretaceous alkaline volcanism and crustal deformation such as the Ponta Grossa Arch, Transbrasiliano Lineament or Lucapa graben.