An experimental procedure for determining the activation energy distribution of isolated organic matter from source rocks is described. The method uses a nonisothermal pyrolysis technique coupled with computer analysis of the data. The computer analysis allows three options for the unknown distribution: a single activation energy for homogeneous organic matter, a log normal distribution, or a discrete distribution. In the majority of cases, the discrete distribution is superior to both the single-energy treatment and log normal distribution. Estimation of the discrete distribution is mathematically an ill-posed problem, and attention is given to an algorithm that combats the numerical difficulties. The method described assumes that there is a single preexponential factor for all possible activation energies associated with the discrete distribution. This assumption can lead to erroneous results, especially for organic matter associated with a very broad activation energy distribution, i.e., type III kerogens. In order to obtain a robust solution, the pyrolysis experiments should be performed at two or three widely differing heating rates. In the case of the discrete distribution, an energy spacing of 1 kcal/mol or less should be used to obtain a robust solution.