When a stream of air is partially stopped by an inserted temperature probe, the temperature increase due to the conversion of kinetic energy affects the reading of the probe. The fraction of the total kinetic temperature rise which is registered by the probe, i.e., the so-called “recovery factor” of the probe, is a function of a number of variables. Tests dealing with the effect of probe shape and air velocity on this recovery factor, and with the influence of radiation on the accuracy of the measurements, are reported in this paper. Bare-wire probes gave recovery factors of approximately 0.65 in transverse flow and, in axial flow, approached 0.87 as the air velocity increased (in good agreement with theoretical predictions for flow over flat plates). With a spherical enlargement at the thermocouple junction, recovery approached 0.75. Recovery of twisted-wire couples varied from 0.72 to 0.83. A reduced-scale model of the Franz probe was found unsatisfactory after extensive study. Two simpler probes were developed, having high recovery (above 0.98 as velocity approaches sonic) and satisfactory insensitivity to yaw and radiation errors.