162 A. T. Welford 



Four other methods of a very different kind are sometimes 

 useful for ensuring or checking the comparability of subjects 

 in different age ranges. 



(i) When we know, or have reason to believe, that the 

 observed age trends should progress smoothly over several 

 age groups we can assume, provided our experimental method 

 has been rigorous, that any marked deviation from a smooth 

 progression is due to an accident of sampling. What appears 

 clearly to be a result of this kind is contained in an experiment 

 by Weston (1949) on changes of visual acuity in middle age. 



(ii) Some investigators, notably Thorndike and co-workers 

 (1928) have given their subjects a preliminary task such as an in- 

 telligence test, and selected groups in different age ranges with 

 equal means and scatters of test score. This is undoubtedly a 

 powerful method of control, provided we know what the pre- 

 liminary task measures. The inference from the experimental re- 

 sults is, of course, affected in the sense that any trend is relative 

 to that which would be observed in results of the pre-test for an 

 unselected population, but this, far from being a disadvantage 

 opens the way to a rather precise method of assessing the relative 

 magnitude of age trends for different types of human capacity. 



(iii) Somewhat similar to this last method is one used by 

 Szafran (see Welford, 1951), Kay (1954), Clay (1956) and 

 others which consists of presenting the experimental task in 

 two or more different forms and noting the difference of age 

 trend between them. In all the cases cited one form has been 

 found to give much less age trend than another and we can 

 thus use it as a base line for a relative statement about the 

 effect with age of the factor by which the one form of the task 

 differs from the other. This method appears to be of wide and 

 rather simple application. Its value will depend, however, 

 upon the easiest form of the task not being so easy that every 

 subject can do it equally well. The task and the method of 

 scoring must be such that some individual variation can be 

 shown if the similarity between the older and younger sub- 

 jects' performances at the easiest form of the task is to be more 

 than a trivial result. In practice this means that in, say, a 



