Object recognition methods are usually suited to objects which contain specific kinds of features. However, a collection of real world objects seldom contains only one kind of feature. The premise of the paper is that it would be possible to recognize more objects if the results from differing methods were combined. The combination is not only to widen the domain of discourse of objects to be recognized but also to help improve the quality of the recognition. The authors propose a scheme in which the results from more than one method can be combined. The authors illustrate their scheme with an implementation which combines three methods for object recognition: one method uses geometric features; the second is appropriate for jagged structures in objects; the third is for objects which can be well described with a skeleton. The authors then present some results.>