296 REPORTS ON THE STATE OP SCIENCE. — 1915. ' 



The tests here used particularly are : (1) The output of work, (2) 

 the accidents occurring in the course of work. 



In addition to the main tests of output of work and accidents, 

 scientific use has been made of some other tests, which fall into three 

 groups : 



(1) Tests allied to the output of work test, such as the output of elec- 



trical or other power during working time, and the labour costs 

 of the factory. 



(2) Tests allied to the accident test, such as the proportion of mistakes 



or spoilt goods in the product. 



(3) Medical tests of the after-effects of work, such as the prevalence 



of sickness and particularly nervous disorders attributable to 

 overwork, and the comparison of blood-pressure and breathing 

 (cp. Eoth) and of cutaneous discrimination (cp. Griesbach) before 

 and after spells of work. 



This enumeration almost completes the non-laboratory tests of 

 fatigue capable of statistical treatment. They are non-laboratorial in 

 the sense that they are records (tests 1 and 2) of facts or of the results 

 of facts (test 3) occurring in ordinary life. In the laboratory, on the 

 other hand, either the fatigue itself is artificially produced or the tests 

 specially invented, and thus certain circumstances of ordinary life are 

 removed and the factors to be studied isolated ; but though the 

 ' analysis ' or explanation of results may thus be facilitated, yet different 

 factors besides fatigue are introduced such as a high average tension 

 and suggestibility in the ' subject,' and these may make the conclusions 

 inapplicable to industry. 



All the non-laboratorial tests enumerated are capable of statistical 

 treatment in that their discoveries can be stated numerically as certain 

 quantities of one statistical unit. On the other hand, opinions of 

 managers and foremen or inspectors or doctors in answer to * question- 

 naires ' as asked by Eisner, Bloch, Levenstein, and others, though they 

 may suggest what statistics to collect, can hardly themselves be treated 

 statistically. 



The collection of these non-laboratorial statistics must always wait 

 on the chance occurrence of events that permit comparisons to be made. 

 Of such ' chances ' there are for industrial output and accident statistics 

 and some of the ' medical tests ' two types. 



One chance of finding statistically the influence of the factor of 

 ' previous duration of work ' on ' output ' (or use of power) on acci- 

 dents (or proportion of mistakes) and on the workers' health, is by 

 dividing the working day into equal parts — the hour is generally the 

 most convenient part — and by recording the output and the accidents 

 occurring in each part, or the state of health after each part. Fatigue, 

 it is presumed, will show itself, if anywhere, in the latter part of each 

 working period (spell, day, or week) ; a comparison of the output, 

 accidents and health in and after the later with those in and aft^r the 

 earlier parts will then show the influence of such fatigue as occurs in 

 the course of the \\'Orking periods. 



It is this test of the ' distribution ' over the working period of output 

 and * accidents ' that we are here mainly using and discussing. This dis- 



