The evolution of size and excess temperature of starting plumes is calculated as a function of (1) Newtonian and power law (stress-dependent) mantle rheology, and (2) buoyancy flux into the plume through a feeding conduit in mantle of temperature- and pressure-dependent viscosity. The initial size and excess temperature that allow plumes starting at the core-mantle boundary to reach the base of the lithosphere 200 K or more warmer than the surroundings are determined for a range of conditions. The results suggest that starting plumes can be divided into slow and fast plumes. Slow plumes become large and cool through entrainment of the surroundings during ascent, so that they are likely to have little impact on the evolution of the lithosphere. Fast plumes, on the other hand, rise so rapidly that cooling and entrainment become unimportant. They reach the lithosphere while they are still hot (often with an excess temperature similar to that at the source) and are likely to strongly perturb the lithosphere. The change from slow to fast plumes occurs over a small increase in plume buoyancy at the source, particularly in power law mantle. Small buoyancy flux through the conduit affects only slow plumes, whereas high flux causes plumes of different initial radius to reach the lithosphere with similar radius and excess temperature. The results suggest that the study of important starting plumes (those that will have considerable geological consequences) can be simplified to that of (1) nonentraining (constant radius) thermals (plume heads isolated from the source) for small buoyancy flux and (2) nonentraining plumes for high flux. In all cases, most entrainment occurs in the lower mantle. Comparison between power law and Newtonian mantle viscosities suggests that the upper mantle will respond as a power law fluid because it yields lower effective viscosity when submitted to the buoyancy stress of plumes than does a corresponding Newtonian mantle. The power law response of the upper mantle is unlikely to considerably influence ascent time of plumes, but it will considerably change the way plumes interact with the lithosphere.¿ 1997 American Geophysical Union