METHOD OF DISCO VERING CA USAL LA WS 1 93 



and misery of the overcrowded tenement lodgings in which they 

 have their &quot;homes&quot;: we may infer that such variation is not 

 merely accidental, that it is causal, that it reveals existence of some 

 law of causation ; but we may not safely infer that either pheno 

 menon is exclusively the ?#.$ whether partial or total and 

 the other exclusively the effect; for there may be interaction: 

 each may be a partial cause of the other, and each, or both, may 

 be partly due to a combination of many other causes, such as de 

 fective early training, religious indifference, thriftlessness, absence 

 of opportunity for healthy recreation, high rents, unemployment, 

 excessive inducements to drink, insufficient control of the drink 

 traffic, etc. Again, suppose &quot; that statistics show a close corre 

 spondence between a diminution in convictions for drunkenness 

 and an increase of money in saving-banks whether in one town 

 in successive years, or in different towns at the same time ; even 

 if other things do not remain the same, we should be justified in 

 concluding that both improvements, if not actually cause and 

 effect, depended on some common cause, improvement in wages 

 or education&quot;. 1 



244. METHOD OF RESIDUES. &quot;CONJUNCTION OF CAUSES,&quot; 

 AND &quot; INTERMIXTURE OF EFFECTS &quot;. From all that has been said, 

 so far, about perceptual and experimental analysis, it will be evident 

 that in all cases we endeavour to make use of whatever knowledge 

 we already possess about the phenomena under investigation, and 

 about the whole department of facts to which they belong. Such 

 previous knowledge may help us in various ways for instance, 

 by suggesting hypotheses to be tested, and the various grounds 

 of elimination to be applied successively in testing them. Now, 

 there is a special sort of consideration which will help further 

 analysis when we already know a great deal, comparatively speak 

 ing, about the nature and the exact measure of the causal agencies 

 and effects which make up the whole sphere of facts under ob 

 servation. It may, perhaps, for the sake of simplicity, be repre 

 sented in this way. If, in a whole complex process, abcdefgh, we 

 know already that the causal relations between abed and efgh, 

 between a and e, between b and /, between c and g, are all re 

 ciprocating causal relations ; then we can infer a similar relation 

 between d and h. Or, if h be absent from the whole complex 

 event that is, if there be nothing in the event to account for d: 

 which, in all such cases, is described as a RESIDUAL PHENOMENON 



1 Palaestra Logica, p. 113, 353. 

 VOL. II. 13 



