The signal-to-symbol problem is the task of converting raw sensor data into a set of symbols that Artificial Intelligence systems can reason about. We have developed a method for directly learning and combining algorithms that map signalsinto symbols. This new method is based on Genetic Programming (GP). Previous papers have focused on PADO, our learning architecture. We showed how PADO applies to the general signalto- symbol task and in particular the positive results it brings to natural image object recognition. Originally, PADO’s programs were written in a Lisp-like language formulated in (Teller 1994b). PADO’s programs are now written in a very different language. Using this new language, PADO’s performance has increased substantially on several domains including two vision domains this paper will mention. This paper will discuss these two language representations, the results they produced, and some analysis of the performance improvement. The higher level goals of this paper are to give some justification for PADO’s specific language progression, some explanation for the improved performance this progression generated, and to offer PADO’s new language representation as an advancement in GP.