Two experiments test this hypothesis using data on temperature and the incidence of crime for summer months in two midwestern cities. The crime data were divided into violent and nonviolent crimes, and then correlated with maximum, minimum, and average temperatures, and several humidity measures. Violent crime correlated significantly with temperature; nonviolent crime did not. This relationship was linear in the first study, but curvilinear in the second. Inspection of the data suggests that the incidence of aggressive behavior (i.e., violent crime) does not drop off in the mid-eighties as laboratory findings and Baron and Ransberger's results would suggest, but continues to increase as temperatures rise into the nineties. The mean incidence of violent crime was higher for days in the 90's than for days in the eighties. A hypothesis for resolving this contradiction between real world and laboratory findings is discussed. 4 figures and 13 references. (Author abstract modified)