In order for plate tectonics to operate, the convecting upper mantle (asthenosphere) must be weaker than the overlying plates. The viscosity change due to temperature alone would be gradual and moderate, so additional factors are likely to control asthenosphere viscosity. For several decades, a debate has been active over whether the asthenosphere's low viscosity is due to a small amount of partial melt or a high hydrogen content. We will examine new geophysical data that support the partial melt hypothesis. For instance, the seismic low velocity zone in the asthenosphere has been interpreted through both solid-state and partial melt mechanisms, but new experimental data show that it can easily be caused by a realistic amount of partial melt. Magnetotelluric (MT) data respond to the conductive anomalies caused by both hydrogen and partial melt and in some cases can be used to distinguish between them. While the interpretation of some datasets is ambiguous, we will discuss new MT data from close to a ridge in Svalbard and from old asthenosphere in the north-west Pacific that strongly support the presence of partial melt. In addition, a new review of sea-mount populations shows that significant amounts of partial melt are being extracted from the oceanic asthenosphere even after 70 Myr. Careful interpretation of partial melt from geophysical data can constrain viscosity models of the asthenosphere, leading to improved geodynamic and glacial isostatic adjustment models.