Mantle convection drives planetary heat loss, as manifest at the surface via one of several potential tectonic modes: stagnant lid (hot or cold), sluggish or squishy lid, and mobile or active lid (<80% of surface behaving as a rigid plate) or plate tectonics (PT; >80% rigid plates). Additionally, an episodic (transient) mode reflects alternating states. The thermal state of the mantle after crystallization of a magma ocean sets the initial conditions for mantle convection. However, because systems with the same physical parameters can exhibit different evolution depending on the initial state, forward modelling of geodynamics is challenging. Thus, how and on what timescale PT emerged after the last magma ocean is unknown, with proposals ranging from the Hadean to the Neoproterozoic. An alternative is to take an inverse approach and test the null hypothesis that Earth has always had PT; accepting the possibility of preservation and survivorship biases, the inverse approach is constrained by geology. The characteristic features of PT include: large-scale rigidity of lithosphere plates; distinctive tectonic settings for sedimentation (e.g., passive margins, foreland basins); distinctive styles of plate boundary magmatism (mid-ocean ridge and calc-alkaline suite), metamorphism (ocean floor versus subduction- and collision-related), and deformation, with a characteristic spatial asymmetry from trench to hinterland and strong strain localization along lithospheric-scale thrusts formed during collisional orogenesis; and, differential movement between cratons and/or terranes, as evidenced by paleomagnetic data. Many of these features are widespread among cratons since the mid-Paleoproterozoic, and some are recorded in one or more cratons since the Paleoarchean, consistent with a transition from c. 3.2 Ga to c. 2.1 Ga. The reason why Earth's tectonic mode may have been different in the past relates to secular cooling of the mantle. Currently, Earth generates about half as much heat as it loses, but these processes have changed at different rates through time, such that Earth has been cooling by ~100°C/Gyr since 2–3 Ga, but before c. 3 Ga it is unclear whether heat production was in balance with heat loss or whether the mantle was warming to a peak. At present, the variation in potential temperature (TP) of the upper mantle is ~120°C. For an ambient mantle TP >200 °C warmer than at present with a similar variation, widespread stable subduction may not have been possible, but could have occurred where mantle TP was lower. At elevated mantle TP magmatism will be dominant, and a sluggish or squishy lid may have preceded a mobile or active lid. Subduction initiation is a poorly-understood process, and both endogenic (mantle plume) and exogenic (impact-driven) processes have been suggested as plausible triggers. It is likely that as Earth's mantle began to cool, during the interval 3–2 Ga, stable subduction was able to spread from one or more centers of plate-like behavior, leading to a connected plate boundary network by the early Paleoproterozoic. Mobility during the Neoarchean enabled the amalgamation of continental crust into several supercratons, but the supercontinent cycle, a probable demonstration of global PT, only began in the Proterozoic.