The laser flash method, as a means of measuring thermal diffusivity, is well established, and several manufacturers produce equipment for performing these types of experiments. Most analysis methods used for interpreting the data from these experiments assume one-dimensional transient conduction, with insulated surfaces during the time subsequent to the flash. More recently, models of greater sophistication employing nonlinear regression have been applied to flash diffusivity experiments. These models assume an instantaneous flash and are highly accurate for most samples of moderate diffusivity and sample thickness. As samples become thinner and more highly conductive, the duration of the experiments becomes very short. Since the duration of the flash is typically on the order of several milliseconds, the assumption that this period of time is instantaneous becomes less valid for very short experiments. A model accounting for the duration of the flash is applied to three samples of stainless steel of varying thicknesses and analyzed with two different mathematical models. One model accounts for the finite duration of the flash and the other does not. The model accounting for the flash duration generates results that are much more consistent between samples than the model assuming an instantaneous flash. Moreover, the conformance of the mathematical model accounting for flash duration is much closer to the measured data than the model which assumes an instantaneous flash. As part of the finite flash duration model, the length of the flash is estimated by nonlinear regression, optimizing the conformance of the model to the measured data. Additionally, the starting time of the flash is treated as a parameter and is determined simultaneously with flash duration, thermal diffusivity and flash intensity. Statistical methods are also used for showing the validity of the added level of sophistication of the more advanced mathematical model.