234 ANNUAL REPORT SMITHSONIAN INSTITUTION, 1954 



Now, of course, even in our present state of ignorance, a fair number 

 of models might be designed to meet these conditions, although an 

 increasing number are being ruled out. So the next step must be to 

 study in detail how such a model would work, and how it would go 

 wrong ; then to go back to the brain to discover whether it does likewise. 

 The discrepancies we find can, if we are lucky, suggest the lines along 

 which we should refine or redefine our model for the next comparison. 



Let me give you an example. No one knows how the visual system 

 enables us to recognize the shape of an object in^espective of its size. 

 One hypothesis suggested a model in which the visual map in the cortex 

 was rhythmically magnified and diminished in size about ten times per 

 second, in step with the electrical rhythm of the same frequency which 

 can sometimes be detected at the back of the head. 



There are, of course, anatomical and other tests of this type of model, 

 and I do not think that it meets them very satisfactorily ; but one way 

 of testing such a model would be to ask : Is there any way in which the 

 mechanism could be tricked ? If so, can we trick the brain in the same 

 way? 



One obvious trick might be to present the model with a pattern whose 

 size was fluctuating at the same rate as its internal magnifier. In the 

 simplest form of model this could either prevent the pattern from being 

 recognized, or at least give rise to perceptual anomalies. It is not diffi- 

 cult to try this experiment on a human subject. A television tube can 

 draw a square or triangle whose size fluctuates at the same rate as the 

 electrical potential on the subject's scalp. But in several such tests I 

 have never found any evidence of anomalous perception (MacKay, 

 1953). 



A null result of this sort does not, of course, prove that rhythmic 

 "scanning" does not occur. It proves only that the hypothetical model 

 of the process was wrong in at least one respect. I think there are 

 other grounds for believing that scanning, in the engineer's sense, is 

 unlikely to occur, but that is another story. 



PROGRESS AND PROSPECTS 



From the example just given, you may get the impression that our 

 model making is still at a very rudimentary stage ; you will be correct. 

 There are still far too many possible models that cannot yet be ruled 

 out on the evidence at hand.^ Some, like the digital-computer model, 

 will fall gradually by the wayside as other models are evolved which 

 beg fewer questions. The choice of a model is much more a matter of 

 intuitive judgment than of strict deduction. 



* Readers may consult the eight pages of references given by F. L. Stumpers in 

 "A Bibliography of Information Theory," Massachusetts Inst. Techn. Rep., Feb. 

 2, 1953. 



