For a period of nearly one year temperature readings from small sensors (high quality platinum resistance thermometers) in conventional screens (Stevensson type) and emaller screens (Lambrecht, Young and Vaisala) have been compared to those from a sensor of the same type in a ventilated screen (Teledyne). The test was financially supported by the Swedish Civil Board of Aviation and CDS Mätteknik, Skara. The reading from the Teledyne screen was used as reference and considered the 'true' air temperature. The deviations from the reference were mainly due to two factors:- the thermal inertia of the screens- radiation errors.The thermal inertia is largest for the larger screens. With rapid air temperature changes and calm winds the larger (Stevenson) screens lag behind much more than the smaller ones. Also the radiation errors are largest for calm winds. The extreme errors then occur during calm winds and clear sky. The errors found are larger than given in the literature. The largest error noted, +3. 6°C for one of the Stevenson screens, occurred a calm, clear evening. Generally the extreme errors occurred at sunset and sunrise, not in the a~ternoon when the irradiation has its maximum, because it then usually is windy. Generally the smaller screens followed the reference better than the Stevenson ones. However, with calm winds, clear sky and snow cover the small screens may rapidly get overheated. The averages for longer periods (months)