Abstract
The thermoelectric properties of a two-band model are discussed, and it is shown that use of the Nordheim-Gorter rule at high temperatures can give misleading values of impurity thermopower when the effects of electron relaxation time anisotropy are included. Experimental results for certain noble metal alloys are collected, and it is suggested that a critical comparison of high and low temperature experiments can yield information about anisotropy of electron diffusion thermopower.