Random field regression models provide an extremely flexible way to investigate nonlinearity in economic data. This paper introduces a new approach to interpreting such models, which may allow for improved inference abour the possible parametric specification of nonlinearity.