The capacity to extract a polymorphous feature rule from several independent features is also likely to be influenced by the quality of the features themselves. By using photographs of natural scenes as stimuli, von Fersen and Lea (1990) were the first to show that pigeons can discriminate categories that have been synthetisized from natural categories in a polymorphous way. The features that defined category membership were themselves natural categories, with the usual polymorphous properties of such categories. Stimuli were photographs of two sites in Exeter, and the defining features were chosen to be of obvious importance to a free-flying pigeon. Features were 1) the site at which the photograph was taken (in front of Northcote House at the University of Exeter, and in front of the Crown and Sceptre, a pub in the center of the city), 2) the weather conditions (sunny or cloudy), 3) the distance of the building in the picture (near or far), 4) the orientation of the building (horizontal or oblique), 5) the camera height from which the pictures were taken (aerial or ground). One value of each feature was designated as positive, and pictures containing three or more positive feature values were members of the positive category. Although only half of the subjects came under the control of all five features spontaneously (the other half only after a special single-feature training), this experiment proved that pigeons successfully discriminate artificially defined categories that require to extract and combine as many as five 'natural' features. If natural categories are defined by a polymorphous combination of a large number of natural features that are themselves polymorphous in nature, the results can be taken with confidence as being in line with our conclusion that multiple feature learning is an adequate account of natural categorization.