An algorithm for learning a subclass of erasing regular pattern languages is presented. On extended regular pattern languages generated by patterns π of the form x0α1x1 . . . αmxm , where x0, . . . , xm are variables and α1, . . . , αm strings of terminals of length c each, it runs with arbitrarily high probability of success using a number of examples polynomial in m (and exponential in c ). It is assumed that m is unknown, but c is known and that samples are randomly drawn according to some distribution, for which we only require that it has certain natural and plausible properties. Aiming to improve this algorithm further we also explore computer simulations of a heuristic. 1 Supported in part by NSF grant number CCR-0208616 and USDA IFAFS grant number 01-04145. 2 Supported in part by NUS grant number R252-000-127-112. 3 Supported in part by NUS grant number R252-000-212-112. Most work was done while F. Stephan stayed with the National ICT Australia which is funded by the Aust...