In space, the radiative heat transfer is the dominant phenomenon. Therefore, the spacecraft thermal analysis pays particular attention to the computation of the radiative terms of the heat transfer equation (radiative couplings and external heat fluxes). In the European space industry, widely automated softwares such as THERMICA (distributed by MMS1), based on a Monte-Carlo Ray-Tracing method simulating the behaviour of photons, have permitted to reach a satisfactory level of accuracy while keeping an acceptable cpu-time consumption.However, even if results converge when the number of photons increases, the adequate number of photons required to achieve a given accuracy is not available. This lack of information may produce “hidden errors” in the radiative model.To solve this problem, a statistical method, based on the central limit theorem, was developed in 1992  and improved in 1993 , in order to compute confidence intervals from the variance of the results. The optimum number of photons, corresponding to the accuracy required by the thermal engineer is then deduced from this error estimation on radiative couplings and on fluxes.A full scale implementation of the method in THERMICA started in October 1994. All test results gave excellent conclusions (see  and ).In the future, a global accuracy control on temperature fields may be considered.