Modifications to a grammatical inference scheme by Feldman et al. are presented. A comparison of the relative performance of the original and modified schemes is made using the complexity measures of Feldman and Wharton. The case where a complex model is used to generate the sample set is then analyzed. A set of 104 samples was found that trained the program to infer the grammar that corresponded to the original model. The results of a study of the performance of this algorithm when there is a large number of samples is then presented. The major conclusion of this study is that the modified scheme has a superior performance on small sample sets but is highly unsuitable for large ones.