This paper describes a new shape matching algorithm using the correlation of the relational models. Relational descriptions of shapes are incorporated in a correlation matching method. From relational decriptions of a shape, a list of relational spectrums for each feature is constructed, which preserves the global information of the whole shape. Shape matching of a candidate shape against the prototype model is performed by correlating the relational spectrums of the prototype and candidate models. Although the correlation technique is adopted in matching process, the proposed matching algorithm is invariant under elementary transformations such as translation, rotation, and scaling since the items being correlated are global relations among shape features.