This paper is concerned with numerical methods for a class of two-dimensional quasilinear elliptic boundary value problems. A compact finite difference method with a nonisotropic mesh is proposed for the problems. The existence of a maximal and a minimal compact difference solution is proved by the method of upper and lower solutions, and two sufficient conditions for the uniqueness of the solution are also given. The optimal error estimate in the discrete L ∞ norm is obtained under certain conditions. The error estimate shows the fourth-order accuracy of the proposed method when two spatial mesh sizes are proportional. By using an upper solution or a lower solution as the initial iteration, an "almost optimal" Picard type of monotone iterative algorithm is presented for solving the resulting nonlinear discrete system efficiently. Applications using two model problems give numerical results that confirm our theoretical analysis.Key words. quasi-linear elliptic boundary value problem, compact finite difference method, error estimate, fourth-order accuracy, monotone iterative algorithm 1. Introduction. Finite difference methods have long been used to approximate the solution of ordinary or partial differential equations. Central differencing is one of the most popular finite difference methods. However, it has only second-order accuracy in general. A simple technique for improving the accuracy is to include additional mesh points into the difference approximations of the derivatives [6, 7, 30]. The higher-order methods derived in this manner are noncompact in the sense that the required stencils always utilize mesh points located beyond those directly adjacent to the node about which the differences are taken. The noncompact stencil not only complicates the formulations near the boundaries but also increases the bandwidth of the resulting coefficient matrix. Motivated by these problems, various higher-order compact finite difference discretization techniques for different equations have been developed (see, e.g., [1,14,18,19,20,21,36,39]).In recent years, there has been growing interest in developing fourth-order compact finite difference methods for elliptic differential equations. Such a method uses a similar compact stencil as the standard second-order central finite difference method, but it still achieves higher-order accuracy. The existing fourth-order compact finite difference methods are mostly derived for linear elliptic differential equations (see, e.g., [4,5,11,12,16,34,35,38,41]). The main idea of these methods is to increase the accuracy of the standard central finite difference approximation from the second-