One hundred and twenty-five premature infants were divided into two groups, one being treated in incubators at an air temperature of 31.8°C (89°F) and the other in the same style of incubator, but with more careful regulation of body temperature by a radiant heat device actuated by a thermistor taped to the abdomen and set for 36°C (97°F). There was a trend toward lower mortality among the heat lamp babies which was present in males and females, whites and Negroes, and which was not related to an accidental distribution of high risk infants to the control group. The advantage was most pronounced among babies vaginally born in the vertex position between birth weights of 800 and 1,599 gm inclusive. The cause of the excess mortality was not revealed by autopsy data. The results confirm those of Silverman, but extend them in that his "normothermic" babies were in conditions similar to our cold or control babies. The trends shown in this paper, taken in combination with those of the previous studies of Silverman, of Jolly et al., and of Buetow and Klein14 in this issue, point to the desirability of care in the control of the temperature of premature infants.