Lisp-Stat was originally developed as a framework for experimenting with dynamic graphics in statistics. To support this use, it evolved into a platform for more general statistical computing. The choice of the Lisp language as the basis of the system was in part coincidence and in part a very deliberate decision. This paper describes the background behind the choice of Lisp, as well as the advantages and disadvantages of this choice. The paper then discusses some lessons that can be drawn from experience with Lisp-Stat and with the R language to guide future development of Lisp-Stat, R, and similar systems.