Numerical aspects of a method for rational approximation of analytic functions on regions in the complex plane are considered. The approximation method divides the problem of computing a rational approximant into three subproblems. First, one chooses a space of rational functions, then one selects a basis for this space, and finally one determines an element of the space by interpolation. For approximation on regions with bounded simply connected complement, we discuss the choice of space and basis from a numerical point of view. We illustrate with computed examples.1. Introduction. We describe a numerical method for rational approximation of analytic functions on regions in the complex plane with bounded simply connected complement. In this method one first chooses a space Qn of rational functions and a basis for this space. Then, a rational approximant is selected from Q, by interpolation at points on the boundary of the region. We discuss numerical aspects of the choices of space and basis and show that these choices should depend on the shape of the region. The distribution of interpolation points should in turn depend on the rational space, and we further consider the computation of such points.The questions of how to choose basis and interpolation points also arise in polynomial approximation and have, in this context, been discussed in Gautschi [2], Mason [4] and [12]. In this first section, we present some computed examples which illustrate the questions to be discussed in 2 (choice of space) and 3 (choice of basis). The examples do not show how to organize the computations most efficiently. We will return to that in 4, which contains rational approximations of conformal mappings.Let fl be the closed exterior of the ellipse E(a, b):={a cos (t)+ibsin (t), 0-<_ =< 27r}, a -> b. Denote by 012 the boundary of ll and let 12c be the co.mplement of 12 w.r.t, the extended complex plane C*. Let w ,(z) denote the conformal mapping from ll to Iwl-> 1 such that ,() = and ,(a)= 1. Thenwhere the branch of the square root is chosen so that I(a + b)-(z +,/z:-(a -b) )1 >--1 for z 12. In a neighborhood of infinity ,(z) has the representation (1.2) ,(z) c-z+ E dk z-k k=0 with c 1/2(a + b). We determine approximants of ,(z) by applying the approximation scheme to compute rational approximants of (1.3) f(z):=O(z)-c-z, which is regular in fl. We seek to approximate f(z) by functions of the form (1.4) r,(z):= a,q(z) k=l *