Initial experiments were carried out as part of a project to train large-scale artificial neural networks (ANNs) using a reconfigurable optical processing system. A perception with binary inputs and bipolar outputs was trained with several two-input Boolean operations to analyze the soft learning properties of the system. It was possible to compensate for all of the system's time-invariant nonlinearities at the LCD panel. The remaining inherent nonlinearities introduced a level of noise and crosstalk which aided learning. The experiments addressed the constructional and operational difficulties in the final configuration, currently under development.