Abstract The late Neoproterozoic–Cambrian interval is characterized by global-scale orogenesis, rapid continental growth and profound changes in Earth systems. Orogenic activity involved continental collisions spanning more than 100 myr, culminating in Gondwana amalgamation. Avalonia is an example of arc magmatism and accretionary tectonics as subduction zones re-located to Gondwana's periphery in the aftermath of those collisions, and its evolution provides significant constraints for global reconstructions. Comprising late Neoproterozoic ( c. 650–570 Ma) arc-related magmatic and metasedimentary rocks, Avalonia is defined as a composite terrane by its latest Ediacaran–Ordovician overstep sequence: a distinctive, siliciclastic-dominated cover bearing ‘Acado-Baltic’ fauna. This definition implies that Neoproterozoic Avalonia may consist of several terranes, and so precise palaeomagnetic or provenance determination in one locality need not apply to all. On the basis of detrital zircon and Nd isotopic data, Avalonia and other lithotectonically related terranes, such as Cadomia, have long been thought to have resided along the Amazonian–West African margin of Gondwana between c. 650 and 500 Ma – Avalonia connected to Amazonia, and Cadomia to West Africa. These interpretations have constrained Paleozoic reconstructions, many of which imply that the departure of several peri-Gondwanan terranes led to the Early Paleozoic development of the Rheic Ocean whose subsequent demise in the Late Paleozoic led to Pangaea's amalgamation. Since these ideas were proposed, several new lines of evidence have challenged the Amazonian affinity of Avalonia. First, there is evidence that some Avalonian terranes may have been ‘peri-Baltican’ during the Neoproterozoic. Baltica was originally excluded as a potential source for Avalonia because, unlike Amazonia, arc-related Neoproterozoic rocks were not documented. However, subsequent recognition of Ediacaran arc-related sequences in the Timanides of northeastern Baltica invalidates this assumption. Second, detailed palaeontological and lithostratigraphic studies have been interpreted to reflect an insular Avalonia, well removed from either Gondwana or Baltica during the Ediacaran and early Cambrian. Third, recent palaeomagnetic data have raised the possibility of an ocean (Clymene Ocean) between Amazonia and West Africa in the late Neoproterozoic, thereby challenging conventional reconstructions that show the ‘peri-Gondwanan’ terranes as a contiguous belt straddling the suture zone between these cratons. In this contribution, we critically re-evaluate the provenance of the so-called ‘peri-Gondwanan’ terranes, the contiguity of the so-called ‘Avalonian–Cadomian’ belt and the validity of the various plate tectonic models based on the traditional interpretation of these terranes. In addition, we draw attention to critical uncertainties and the challenges that lie ahead.