[On the structure of visual similarity (author's transl)]

The purpose of the present study is to investigate the structure of visual similarity. Stimuli were two series of geometric figures, one of them was 16 rectangles (series A) and the other 18 "apparent" cubes (series B). Each stimulus series was given to eleven adults and eleven infants (five years old). Kruskal's nonmetric MDS was applied to observed (subjects mean average) similarity data. The main results were as follows: (a) The Euclidean model was more suited to the data than the city block model. (b) The number of necessary dimensions was two (series A), or three (series B) in both groups of the subjects.