Biodiversity studies in ecology and agriculture often begin with the analysis of sampling curves. To measure the completeness of sampling, a sampling yield-effort curve can be drawn that plots the number of taxa sampled against the sample size (Colwell and Coddington 1994, Zhang and Schoenly 1999a,b). This curve is a stepfunction with a slope that should decrease as sample size increases and as fewer taxa remain to be sampled. Many models or methods were developed to fit these functions. Most of these models, however, yielded fixed errors. Neural network methods, always with any desired accuracy, were widely used to fit functions in engineering and related research (Bian and Zhang 2000, Hagan et al 1996, Zhang and Qi 2002). Therefore, we expect a better goodness of fit for sampling yield-effort curves with neural networks. The back propagation (BP) neural network and the radial basis function (RBF) neural network algorithms are introduced and tested in this study to provide an effective tool to fit the sampling yield-effort curves and to document the sampling information, based on data sets of invertebrates sampled in tropical irrigated rice fields. Some conclusions on rice invertebrates are obtained from the fitted functions of neural networks.