The southwest Pacific is a frontier region for petroleum exploration. A complex series of subduction and back-arc basin forming episodes characterises the late Cretaceous to present day evolution of the region. Controversial aspects of the regional tectonic history include the presence or lack of subduction between 83 and 43 Ma, the polarity of subduction, the timing of back-arc basin formation, and whether or not Pacific plate motion can be tied to the motion of Australia via spreading in the Tasman Sea during the late Cretaceous-early Cenozoic. A combination of tectonic and geodynamic models has previously been used to propose that there was no subduction to the east of Australia between 83 and 43 Ma, with the Lord Howe Rise being part of the Pacific plate during this time period, contrary to alternative plate models that include a plate boundary to the east of the Lord Howe Rise. Determining which plate circuit to use for Pacific motion is critical for producing regional reconstructions for the southwest Pacific, and addressing specific problems on the chronology of tectonic and basin-forming events. To help resolve these long-standing disputes we test a recently published plate reconstruction in global mantle flow models with imposed plate motions. We use the 3D spherical mantle-convection code CitcomS coupled to the plate reconstruction software GPlates, with plate motions since 200 Ma and evolving plate boundaries imposed. We use seismic mantle tomography models to test the forward-modelled subduction history in the region. The reconstruction that we test incorporates east-dipping subduction from 85-45 Ma along the western margin of the Loyalty-Three Kings Ridge to close the South Loyalty Basin. Following collision of the Loyalty Ridge with New Caledonia, west-dipping Tonga-Kermadec subduction initiates along the eastern margin of the Loyalty Ridge and opens the North Loyalty, South Fiji, Norfolk and Lau basins. Contemporaneous with westdipping Tonga-Kermadec subduction, there is short-lived eastdipping subduction between 36-18 Ma along the western margin of the Loyalty-Three Kings ridge. We find that subduction to the east of Australia during the period 85-45 Ma is necessary to account for the distribution of lower mantle slab material that is imaged by seismic tomography beneath New Zealand. Pacific plate motion therefore cannot be directly tied to Australia via spreading in the Tasman Sea as a convergent margin must have existed to the east of the Lord Howe Rise at this time. We suggest that adopting a plate circuit that ties Pacific plate motion directly to the Lord Howe Rise should be avoided for this period. An unexpected result is that the regional lower mantle structure provides strong evidence for a long-lived intra-oceanic subduction zone, located to the northeast of Australia at about 10-25°S and 170°E-170°W, that was active during at least the late Cretaceous. We propose that this subduction zone was located outboard of the plate boundary that separated the palaeo-Pacific ocean from the Tethys, and we speculate that arc remnants may be preserved in southeast Asia.