The present paper reviews the principal aspects of some work carried out between 1958 and 1961 to obtain a thermocouple capable of operating in the range of 1200-2000°C in the combustion chamber and reheat tailpipes of advanced turbojet aero-engines with an inherent accuracy of about ± 1 per cent. The work of previous experimenters is reviewed and a consideration of existing thermocouple combinations indicates that oxidation-resistant refractories might be well suited to the applications envisaged. A number of refractories and high-temperature metals and alloys were selected for evaluation on an oxidation-resistance basis, and their thermoelectric output against EY4 graphite (Morgan Crucible nomenclature) is presented. Thermocouple pairs were selected on the basis of the above tests and complete assemblies were evaluated in the laboratory and on test-bed engines, a total of over 1000 hours being accumulated for the final design. Such factors as repeatability, stability and response time are presented. It is concluded that in order to obtain adequate response times only refractories having a high thermal conductivity are suitable. Greater immersions are therefore required in order to avoid excessive cooling of the hot junction by conduction. This in turn requires greater strength in the thermocouple in view of increased stresses due to gas loading. As a result a refractory thermocouple will be dimensionally larger than conventional thermocouples, but the thermoelectric output will be one magnitude greater. By stringent control during sintering or hot moulding and subsequent selection, adequate accuracy may be maintained and repeatability achieved. Some limitations exist as regards shape and size in view of the difficulty in obtaining sintered or hot-moulded bodies of high length/diameter ratio. In view of the fact that the thermocouple impedance could be less than 10 Ω in conjunction with thermoelectric outputs of 200 mV or more, the possibilities of power generation are considered.