THE QUESTION OP FATIGUE FROM THE ECONOMIC STANDPOINT, 297 



tribution may be represented either by tables giving actual numbers for 

 each hour or part-period, or by diagrams measuring the time along their 

 base and the amount of output or accidents vertically therefrom. The 

 special advantage of diagrams is that sets of figures may be compared 

 that have differently delimited hours, by measuring the amounts at 

 slightly different points on the base. 



To enable the distribution of output over the working-day to be 

 treated statistically a unit of output must be found, in quantities of 

 which the statistics may deal. This unit must (1) be similar in quality 

 throughout the day, (2) be small enough to enable at least two such 

 units to be produced during the day. In short the output must be 

 ' repetition ' work. Processes thus suitable for statistical record are 

 found especially where a machine is used adapted to only one kind of 

 work and possibly provided with an automatic register of the amount of 

 work done. Where output does not lend itself to statistical compilation 

 owing to heterogeneity (absence of Condition 1), the piece wages earned 

 on the output might be recorded as affording a common denominator for 

 the different kinds of output. The accuracy of this denominator would, 

 as Weber points out, depend on how closely the piece rate for each 

 kind of output was estimated exactly proportionately to the comparative 

 effort required of the worker in each case (including under effort con- 

 centration rapidity and attention on the part of the worker). But where 

 scientific management is not yet adopted, piece rates are settled 

 empirically from results rather than by psycho-physical analysis, so that 

 the time distribution of piece-rate earnings is at best a somewhat rough 

 equivalent of output. 



In statistics of accident distribution, the statistical unit is the in- 

 dividual 'accident.' This may be either every case treated or given first 

 aid to as at Messrs. Cadbury's and Hans Eenolds' or only the cases 

 reported to Government ' certifying ' surgeons and inspectors. What- 

 ever definition of accident is adopted does not matter in statistics of the 

 daily distribution, so long as the definition is the same throughout the 

 day. 



Another chance of finding statistically the influence of the cause 

 ' previous duration of work ' on the output and less easily on accidents 

 occurs where the duration of the working day has been changed — where 

 there has been a ' change in hours ' either by way of a reduction through 

 reorganisation, strike, or trade depression, or of an increase through 

 overtime, &c. Here are compared the output or accidents per hour (or 

 other unit of time) or the state of health, before and after the change. 

 The importance of this comparison is that it makes it possible to seek 

 the causes of accumulated fatigue (see Section I.). On the other hand, 

 when such comparisons are made between different and sometimes dis- 

 tant periods, other factors besides duration of work are liable to have 

 altered, or may even have been purposely altered with the change of 

 hours of work. To attribute all differences in results to differences in 

 duration of work becomes, therefore, scientifically speaking, more 

 difficult. 



In the following table of double entry are grouped some of the 

 separate statistical investigations that have been made into Industrial 



