METHOD OF DISCO VERING CA USAL LA WS 171 



This outline of the experimental discovery and proof of a cause, contem 

 plates only a comparatively simple case. As a rule, the agencies operative in 

 nature are so complex, so interwoven, and so dependent on one another, that 

 the actual process of analysis cannot be adequately represented by any such 

 simple symbolism as we have been employing. The outline given will, how 

 ever, help us to realize that in every successful discovery and proof of a causal 

 relation &quot; there is a comparison of the phenomenon \J&amp;gt;~\ both in the presence 

 [5* = x + A 2 + p~\ and in the absence \S = A 52 ] of that particular condition \x\ 

 we are investigating &quot;- 1 The observation of positive instances will help us to 

 include in the field of investigation, 6&quot;, all the really relevant operative influences 

 in regard to the phenomenon under investigation,// while the observation of 

 negative instances will help us to determine what part of 6&quot; is indispensable to 

 the occurrence of the phenomenon, p. But in both sorts of instances there 

 are difficulties to be surmounted. Practically speaking, we can never com 

 pletely eliminate the &quot; residue &quot; from our positive instances ; we can never get 

 a positive instance of p without a residue A*, A*, 1 or A 12 , accompanying the sup 

 posed m, Im, or x. And hence we have to make sure (i) that this residue 

 &quot; includes nothing which is not known to be present, and whose influence, if it 

 existed, could be determined and allowed for &quot; ; 2 and (2) that this residue, if 

 it cannot be totally eliminated, is at all events really irrelevant lop. We try 

 to make sure of these points and so to convince ourselves that there is nothing 

 really operative unknown to us in the positive instances, besides the supposed 

 cause by directing our attention to the residue, and analysing it in a series of 

 negative observations or experiments, i.e. instances in which the phenomenon, 

 p, does not occur. We remove, if possible, the supposed cause, leaving the 

 residue, in order to see if the phenomenon will disappear. Or, finding a case 

 in which both the latter and the supposed cause are absent, and the residue 

 alone present, we introduce, if possible, the supposed cause, to see whether the 

 phenomenon also will appear. This process of comparing positive with nega 

 tive instances is likely to bring to light operative influences of which we were 

 previously unaware, if there were really any such present in the unanalysed resi 

 due in the positive instances. It is by the negative instances that we prove 

 our supposed cause to be indispensable, and everything else irrelevant, to the 

 effect. 



But this proof will be cogent only in so far as we are sure that in passing 

 from the positive to the negative instance, or vice versa, we have eliminated, 

 or introduced, nothing else but the supposed cause. And it is not easy to be 

 sure of this, on account of the complexity and interdependence of the causal 

 agencies that are operative in nature. In experimenting, our working prin 

 ciple should be, as far as possible, never to introduce or eliminate more than 

 one element at a time. If we vary more than one element at a time, we shall 

 not know to which of the elements, so varied, any resulting change is to be 

 attributed. 



It will now be sufficiently clear that the conduct of analysis, by observation 

 and experiment, cannot be reduced to any set of mechanical rules or formulae. 

 The history of the inductive sciences, of the ways in which great scientific truths 

 have been de facto discovered and established, makes this fact still more evi 

 dent. A number of interesting and instructive examples are given by Professor 



1 WELTON, op. cit., p. 118. a ibid., p. 121. 



