This paper assesses the performance of a new evolutionary algorithm on test data generation for path testing. The Generalized Extremal Optimization (GEO) was evaluated on benchmark programs, with or without loops, and the results were very encouraging.