Neural networks are being increasingly used in the visual pattern recognition field. They have already been applied with some success to such problems as the classification of fingerprints, the recognition of facial expressions and the identification of hand-written characters. This would suggest the possibility of applying a neural network system to the conceptually similar field of interferometric fringe analysis. Although the algorithmic techniques of phase stepping and Fourier fringe analysis have been successfully applied to many fringe analysis problems, there exist some areas, notably with complicated and noisy images, where a different approach may be required. Here, the backpropagation paradigm is applied to two simple fringe analysis problems. Firstly, to find the radius of a one dimensional curved surface from its simulated intensity distribution. Secondly, to identify four lens shaped objects of different radii of curvature from fringe patterns obtained under different conditions. Both cases were met with reasonable success, but the overall implications of the results were that, when applied to fringe analysis, neural networks would function more successfully at identification, rather than precise measurement.