A method of cross-section adjustment and evaluation is presented. Based on a finite element representation, the method is especially appropriate for cross sections that are well represented by continuous functions. It is also suitable for many problems outside the area of cross-section evaluation where one desires to fit curves that have complicated shapes not easily described by low-order polynomials. The algorithm is based on least-squares techniques that use complete covariance information and prior values. Nuclear model calculations, microscopic and integral data, and the results of prior multigroup adjustments can be combined in a single consistent evaluation. Results are presented for an illustrative example and for the evaluation of the 54Fe(n, p) dosimeter cross section.