An earlier exploratory study on the effects of restraint of thermal expansion on the fire resistance of prestressed concrete showed that for similar specimens, made of normal weight concrete, maximum thermal thrust was a function of the allowed linear expansion. In the present study, data were obtained for three additional groups of specimens: lightweight prestressed, lightweight reinforced, and normal weight reinforced. A method is presented for estimating the maximum thermal thrust that occurs during a fire test of a concrete floor, roof, or beam. Results of 12 fire tests of restrained concrete flexural members indicate that the method predicts the maximum thrust within about 15 percent. The method is applicable to specimens restrained both longitudinally and laterally as well as those restrained in only one direction. The method underestimates thrust for specimens with embedded underfloor ducts and overestimates thrust for specimens partly insulated from direct contact with the fire. In all 43 fire tests the specimens supported their loads considerably longer than would have been anticipated for either simply supported or fully restrained conditions.