In spite of numerous works, the relaxation phenomena observed at high temperature (between room temperature and the melting temperature TM) are still under discussion. Because relaxation peaks were observed in single crystals, it is generally considered that the basis of the relaxation mechanism is linked to the dislocation network. The main difficulty for high temperature damping measurements is the great sensitivity of internal friction with several experimental parameters: the heating/cooling rate, maximal applied strain amplitude, sample purity, thermomechanical history of the sample, microstructure, etc. This sensitivity can explain the large scatter in experimental results published by various authors. Moreover, internal friction (IF) measurements performed during continuous heating or cooling and using an apparatus working at a quasi-static frequency, do not allow to completely describe the relaxation phenomena. On the contrary, isothermal mechanical spectroscopy (measurements of internal friction in a large frequency, temperature and maximal strain amplitude ranges) improves the experiments or evidences new relaxation effects. This is illustrated in this paper for various examples: slightly cold worked single crystals, polycrystals after a large cold work and recrystallization, non thermally activated peaks observed in metallic alloys, and relaxation peaks at very high temperature (above 0.9 TM).