Abstract The Philippine Sea plate, located between the Pacific, Eurasian and Australian plates, is the world's largest marginal basin plate. The motion of the Philippine Sea plate through time is poorly understood as it is almost entirely surrounded by subduction zones and hence, previous studies have relied on palaeomagnetic analysis to constrain its rotation. We present a comprehensive analysis of geophysical data within the Parece Vela and Shikoku Basins—two Oligocene to Miocene back-arc basins—which provide independent constraints on the rotational history of the Philippine Sea plate by means of their seafloor spreading record. We have created a detailed plate model for the opening of the Parece Vela and Shikoku Basins based on an analysis of all available magnetic, gravity and bathymetric data in the region. Subduction along the Izu–Bonin–Mariana trench led to trench roll-back, arc rupture and back-arc rifting in the Parece Vela and Shikoku Basins at ∼30 Ma. Seafloor spreading in both basins developed by chron 9o (28 Ma), and possibly by chron 10o (29 Ma), as a northward and southward propagating rift, respectively. The spreading orientation in the Parece Vela Basin was E–W as opposed to ENE–WSW in the Shikoku Basin. The spreading ridges joined by chron 6By (23 Ma) and formed a R–R–R triple junction to accommodate the difference in spreading orientations in both basins. At chron 6No (∼20 Ma), the spreading direction in the Parece Vela Basin changed from E–W to NE–SW. At chron 5Ey (∼19 Ma), the spreading direction in the Shikoku Basin changed from ENE–WSW to NE–SW. This change was accompanied by a marked decrease in spreading rate. Cessation of back-arc opening occurred at ∼15 Ma, a time of regional plate reorganisation in SE Asia. We interpret the dramatic change in spreading rate and direction from E–W to NE–SW at 20±1.3 Ma as an expression of Philippine Sea plate rotation and is constrained by the spacing between our magnetic anomaly identifications and the curvature of the fracture zones. This rotation was previously thought to have begun at 25 Ma as a result of a global change in plate motions. Our results suggest that the Philippine Sea plate rotated clockwise by about 4° between 20 and 15 Ma about a pole located 35°N, 84°E. This implies that the majority of the 34° clockwise rotation inferred to have occurred between 25 and 5 Ma from paleomagnetic data may have in fact been confined to the period between 15 and 5 Ma.