In this paper, a neurofuzzy model is proposed for the development of an explanatory model of an industrial gold leach circuit based on historic plant data. A regression tree was used to partition the data in the predictor variable space and the structure of the tree was used as a basis for the derivation of fuzzy rules that could be integrated into a fuzzy expert system useful for operator decision support or process control. The position and shape of the membership functions were optimised with a backpropagation algorithm in a neural network framework. The approach yielded a significantly better explanatory model of the gold losses in the plant than comparable linear models. Although a regression tree could perform somewhat better in accounting for the variance of the dissolved gold losses, the tree was less robust than the fuzzy model.