330 REPORTS ON THE STATE OF SCIENCE, ETC. 



equal appearing intervals does not yield an A magnitude because the rela- 

 tion of ' appearing equal ' is not symmetrical ; because the sensations are 

 tied to stimuli which cannot be interchanged. But let us suppose that the 

 experiment is conducted, in the customary manner, by two persons one of 

 whom, called the experimenter, alone knows the stimulus values, while the 

 other person, called the observer, alone judges equality of appearance. 

 In these circumstances the relation of ' appearing equal ' is purely intro- 

 spective. Is it not symmetrical ? 25.6,38 



C. By Mr. T. Smith. 



If by measurement we mean the association of numbers and properties 

 by a rational systematic procedure such as the physicist employs (and — to 

 say the least — this restriction appears necessary to avoid misunderstanding) , 

 I agree with Mr. Guild that sensations are not measurable. Apart from 

 measurement there may be personal associations of numbers as well as of 

 other concepts with sensations, but this in itself is not of great importance 

 since the associations are peculiar to a single individual, though the fact 

 of association may be of psychological interest. By training, the numbers 

 in these associations can often be modified and controlled so that they 

 correspond more or less to the numbers of some measurable property. In 

 the absence of special training the numbers assigned to the members of any 

 collection will vary notably from one observer to another. Ability to guess 

 fairly accurately what measurement will give is an accomplishment of 

 considerable utility, and this perhaps represents the nearest approach we 

 can get to measurement on a sensory basis. 



Simple tests I have made on a number of subjects suggest that the ' scales,' 

 if the word is permitted, of untrained observers show marked differences 

 from one another. Consistency only began to show when four objects 

 were presented, and substantial agreement was reached on the magnitude 

 which corresponds to the cross-ratio of four points on a straight line. In 

 these experiments the individual scales were therefore projections of a 

 common scale. While this is consistent with Mr. Guild's suggestion that 

 likeness in a relation is a recognisable quality, it also suggests that Mr. 

 Guild's interpretation of this relation in stimulus terms may be too narrow. 

 A constant ratio of the stimuli from similar-appearing pairs is only one of a 

 number of possible cases, and this or any other choice must be justified 

 experimentally, and not by an a priori argximent. 27.6.38 



D. By Dr. Wm. Brown. 



While appreciating the excellence of Mr. Guild's discussion of the 

 question ' Are Sensation Intensities Measurable ? ' and agreeing with most 

 of his arguments directed against Fechner's position, I am not satisfied 

 that he has demolished the case for the direct measurement of cont.rastes 

 sensihles (Delboeuf), commonly translated as ' sense distances.' ' Sense 

 distances ' can be bisected with some degree of accuracy provided that the 

 subject carries out the experiment a large number of times under ap- 

 preciably constant conditions. In other words, the result is a statistical 

 central tendency of statistical constancy as checked by its probable error. 

 I find nothing in Mr. Guild's argument that would move me to withdraw 

 anything that I have written on this matter in Chapter I of The Essentials of 

 Mental Measurement. On the other hand, I do realise that a very much 

 fuller discussion of the problem is needed, in the light of recent experi- 

 mental work, and I am glad that the Committee is asking for a further year, 

 during which it can deal with the question more fully. I-7.38 



