PATTERN-DISCRIMINATION IN VERTEBRATES 325 



defect which impresses the writer as serious. An apparatus for 

 quantitative study should permit of change by insensible grada- 

 tions from a given st'mulus- value to any other stimulus- value 

 desired, within the limits of the instrument. The standard 

 apparatus permits such changes in the case of size-differences, it 

 being necessary only to prepare a sufficiently large number of 

 plates. This is not possible in the case of form-difference. 

 Suppose an animal learns to discriminate between a circle and 

 triangle, and maintains a perfect habit throughout a satisfactory 

 set of changes of size, of illumination, and even of surroundings 

 of the test-objects. We may safely conclude that he is reacting 

 to the constant difference of form of the test-fields. 'But we 

 cannot say, in terms of stimulus-factors, how much of this form- 

 difference is necessary to effect discrimination. It is indeed 

 possible to prepare a series of plates changing by imperceptible 

 degrees from the plane triangle through regular trilateral figures 

 (such as certain plane projections of spherical triangles) to the 

 circle; keeping the figures equivalent throughout the series. 

 No experimenter has yet attempted this plan. It involves con- 

 siderable expense and mechanical skill, and may for this reason 

 be impracticable. Meanwhile we are without an apparatus for 

 the study of form-discrimination which meets the fundamental 

 requirements of a quantitative study. 



To consider now a limitation of the problem of form-discrimina- 

 tion itself: it is questionable whether the data obtainable in 

 this study are indicative of an animal's ability to react to differ- 

 ences of visual detail. Perhaps a definite opinion is not warranted 

 by the evidence now at hand. But the circular and square 

 windows, for example, of the Yerkes- Watson apparatus merely 

 limit the perimeters of two light sources of equal brightness and 

 extent. If natural conditions never require an animal closely 

 to scrutinize the borders of objects so nearly alike in all other 

 respects, he may acquire such specialized discrimination with 

 great difficulty or not at all. Supplementary data are necessary 

 to justify one in referring the learning records to the character 

 of the animal's visual equipment, to the character of his 

 " attention " or to the limits of his ability to learn. It impressed 

 the writer some time ago that an improvement of the method 

 used by Casteel and the development of the problems suggested 

 by his work should throw light on the question at issue. 



