This study provides reliable, precisely defined and well-dated Early Permian (286 ± 6 Ma) palaeomagnetic poles for Australia from the Mount Leyshon Intrusive Complex (MLIC) and the Tuckers Igneous Complex (TIC). Both complexes are associated with prominent negative magnetic anomalies, indicating the presence of rocks carrying stable remanence of reverse polarity, with a Koenigsberger ratio greater than unity. The characteristic remanence carried by the intrusive phases and by locally remagnetized, contact-metamorphosed host rocks is always of reverse polarity, consistent with acquisition during the Permo-Carboniferous (Kiaman) Reverse Superchron. The corresponding palaeopoles confirm that Australia occupied high latitudes in the Early Permian. The pole positions are: MLIC: lat. = 43.2 °S, long. = 137.3 °E; dp=6.0°, dm=6.4°; Q =6; TIC: lat.=47.5 °S, long.=143.0 °E, dp=6.0°, dm=6.6°; Q = 6. Permian palaeomagnetic overprinting is detectable at considerable distances from the MLIC (2–3 km), well beyond the zone of visible alteration. The primary nature of the Early Permian palaeomagnetic signature is established by full baked contact/aureole tests at both localities. Other newdata from Australia are consistent with the poles reported here. Comparison of the Australian, African and South American Apparent Polar Wander Paths (APWP) suggests that mean Permian and Triassic poles fromWest Gondwana, particularly from South America, are biased by remagnetization in the Jurassic–Cretaceous and that the Late Palaeozoic–Mesozoic APWP for Gondwana is best defined by Australian data. The Australian APWP exhibits substantial movement through the Mesozoic. Provided only that the time-averaged palaeofield was zonal, the Early Triassic palaeomagnetic data from Australia provide an important palaeogeographic constraint that the south geographic pole was within, or very close to, SE Australia around 240 Ma. The new Early Permian poles are apparently more consistent with Pangaea B-type reconstructions of Gondwana and Laurussia than with the Pangaea A2 configuration. This may be partly an artefact of reconstruction problems within Gondwana, as systematic differences between approximately coeval, apparently reliable, Permo-Carboniferous poles from Africa, South America and Australia are evident in standard Gondwana reconstructions. These discordances require a tighter fit of the southern continents, suggesting that some attenuation of continental margins, not accounted for in the reconstructions, has occurred during breakup of Gondwana, or that the fit between East and West Gondwana needs to be substantially modified. If stretching of continental margins during breakup of supercontinents is a general phenomenon, it may help to ameliorate, but not solve, the long-standing controversy regarding Pangaea reconstructions. Although alternative Pangaea reconstructions, such as Pangaea B, may reconcile poles from Laurussia with Australian poles in the Late Carboniferous–Early Permian, no plausible reconstruction can bring the Early Triassic poles into agreement. This suggests that persistent departures from a pure dipole field may have been present in the Early Triassic. Lesser, but still significant, non-dipole effects may also have been present during the Late Carboniferous and Permian, and may help resolve the Pangaea A versus B controversy, without requiring substantial attenuation of continental margins or intracontinental deformation. We suggest that the most parsimonious interpretation of the palaeomagnetic and geological information is that Laurussia and Gondwana remained in a Pangaea A2-type configuration through the Permian and Triassic. Discordance between the APWPs for these two supercontinents is attributable mainly to persistent non-dipole components of the geomagnetic field, which were most important in the Early Triassic.