An extensive geothermal research program within the German Continental Deep Drilling Program (KTB) covered an almost complete spectrum of experimental and theoretical aspects. The main results and conclusions are as follows: (1) Equilibrium temperature is 118.6 ¿C at 4000 m in the KTB pilot borehole (KTB-VB) and will be around 260 ¿C at 9100 m in the KTB main borehole (KTB-HB). Time required for thermal equilibration of the KTB-HB to within 1% of the initial perturbation will be about 13--16 years. (2) The failure to predict temperature correctly for the KTB was mainly due to an unaccounted perturbation by a transient ground surface temperature history. (3) Pleistocene surface temperature variations affect the present-day crustal temperature between 1.3 to 2.7 K up to a depth of 4000 m. Accordingly, present-day heat flow density is systematically too low down to approximately 1500 m. Model simulations indicate that groundwater flow does not eliminate paleoclimatic signals, even though it may translate them to a depth incompatible with both their diffusive age and their amplitude. These results emphasize the importance of an adequate consideration of paleoclimatic effects for the interpretation of thermal data. (4) Lateral heat transport is significant when steep inclination and folding of the rock formations coincide with contrasts in thermal conductivity. This is indicated by typical variations in the vertical components of temperature gradient and heat flow density, such as in the KTB-HB. In contrast, thermally relevant advection of heat is confined to the top 500--1000 m. In the vicinity of the KTB, this is about twice the maximum difference in elevation. (5) In the deeper crust, free convection systems require permeabilities greater than 10-17 m2 for large rock volumes, but simple numerical models indicate that the associated temperature regimes are incompatible with KTB borehole data. (6) Heat production rate shows no systematic variation with depth and is related to lithology at the KTB as in other deep boreholes in crystalline rock. Numerical models, using heat production rate derived from seismic velocities, yield temperatures compatible with KTB borehole data.¿ 1997 American Geophysical Union