METHOD OF DISCOVERING CAUSAL LAWS 203 



accuracy ; and means of allowing for unavoidable deviations : 

 such are the main functions of what is nowadays really a distinct 

 science and art, known as Metrology} 



It would be beyond the purpose of a treatise on general logic 

 to enter on the interesting but intricate questions involved in 

 determining the various units of measurement. It will be suffici 

 ent to observe here that all actual measurement is relative, i.e. it 

 is the perception of some magnitude, not absolutely, but as com 

 pared with the unit magnitude; that the unit magnitude itself 

 cannot, as such, be measured ; that for its constancy on which 

 its accuracy as a unit depends our only guarantee is our own 

 sense-perception, aided by whatever devices mechanical science 

 can offer us ; that these devices, though improving our powers 

 of perception, do not make the latter perfect ; while the devices 

 themselves may contain errors, or sources of error, peculiar to 

 themselves. 



It is a problem of great practical importance, how to minim 

 ize the inaccuracies of measurement, and how best to make 

 allowance for the residual error which cannot be eliminated. The 

 needful degree of accuracy will, of course, depend on the nature 

 of the case under consideration. In building an iron bridge the 

 engineer must allow space for expansion due to heat ; else the 

 force of the expanding metal would burst the whole structure. 

 For this purpose, however, he does not need to determine the 

 coefficient of expansion of his materials with anything like the 

 precision and exactitude required by the maker of a naval 

 chronometer in regard to the compensation springs and balances 

 for this article. But, in all cases, it is the ideal of the scientist 

 to eliminate all error, as far as possible, from his calculations. 

 How, then, is he to deal with the various inaccuracies incidental 

 to measurement? In the first place, he must make allowance 

 for them as far as he is aware of their existence. In physical 

 experiments the possible sources of incorrect measurement are 

 manifold : variations of temperature, friction, atmospheric pres 

 sure, electrical or magnetic conditions ; defects in the measur 

 ing instruments, eg. through want of rigidity in the telescope, 

 through refraction in all optical instruments, through inaccurate 

 adjustment in chronometers, etc. Then, besides all these, there 

 are the sense-limitations of the individual observer, often involv 

 ing deviations l&amp;lt; in the same direction and to the same average 



1 POINCARE, Op. Clt., p. 20. 



