Adult age-at-death assessment is one of the most difficult problem encountered in paleoanthropology. Many procedures have been proposed using either skeletal remains or dental records, but most show systematic bias. Data processing of current methods are a source of error because they neglect that process of biological ageing is very variable between individuals and populations. The aim of this study is to test the potentiality of artificial neural networks (ANN) as a prediction tool. ANN have been used for a wide variety of applications where statistical methods are traditionally employed. But it performs better to solve linearly non separable patterns. We applied this technique after observation of several features' aging changes of the pubic symphysis and the auricular surface of the ilium. Although we failed to reduce the size of the intermediate class (30-59 years), the neural network identifies, with better reliability than previous works, the youngest (20-29 years) and the oldest (above 60 years) individuals.