50 PART I. THE PROBLEM. 



of any matter, and from this it follows that in general scientific 

 work the Canons could play but a small part, whilst in cultural 

 investigations, for instance, there would be scarcely ever an 

 opportunity of applying them. No doubt, in the manipulation 

 of letters abode, abed, bed, ca,.dc we can readily detect what 

 purports to be the common or differentiating fact, but in nature 

 itself the component elements in a problem are unfortunately 

 not lettered, 1 and certainty is, therefore, an ideal to be respect- 

 fully approached rather than to be frequently attained in ex- 

 perience. A good illustration of this are the many obstinate 

 obscurities and difficulties encountered in consistently conceiving 

 and interpreting the Periodic Law in chemistry. What is hence 

 needed are additional Canons which shall deal with approximate 

 truths, for only such truths are the staple product of modern 

 science which has done once for all with the Noah's Ark world 

 postulated by the ancients. 



Scores of passages like the following could be quoted to en- 

 force the teaching of history that certainty is attained only 

 slowly and laboriously: 



"A good deal of evidence has been accumulated in favour of the view 

 that the meteorological conditions of our globe exhibit a periodicity of 

 thirty-five years in other words, that there is a tendency for a similarity 

 in the general run of the seasons to recur after the lapse of this interval 

 of time. Bruckner's study of the information available regarding the 

 variations of the water level in the Caspian Sea, first suggested this 

 period. Russian records also contain a good deal of information regarding 

 .floods or unusual shallowness of the rivers, and the dates of their opening 

 and closing to navigation, and a close examination of this material tended 

 to confirm the view. Subsequently, the investigation was extended to 

 the water levels of lakes in other parts of the world, having inland drain- 

 age, and the results were again in many instances broadly confirmatory 

 of Bruckner's cycle. Records of the advance and recession of Alpine 

 glaciers also supplied a certain amount of confirmation. The evidence 

 in favour of the existence of a periodicity of thirty-five years has had to 

 be culled, often with great labour, from historical documents in which 

 references to meteorological phenomena are only incidental. Only by 

 using such sources of information has it been possible to extend the 

 inquiry over the greater part of the last two centuries. Such indirect 

 evidence is not so satisfactory as we could wish, but the number of 

 meteorological records which are of sufficient length to be of service in 

 an inquiry of this sort is very small. Hann's examination of the rainfall 

 records from Padua, Milan, and Klagenfurt, which cover the years from 

 1726 to 1900, has shown some indications of the reality of a period of average 

 length of about thirty-five years. In the meteorology of the Southern 

 hemisphere, different authors have found indications of the existence of 

 a period of nineteen years. The records of Australia, South Africa, and 

 South America all show suggestions of such a period, but as yet the 

 evidence cannot be regarded as conclusive." (R. G. K. Lempfert, Weather 

 Science, pp. 76-77.) 



A second defect is revealed when we attempt to apply the 

 Canons. How many times must I determine agreement before 

 the Canon of Agreement, for example, is satisfied? Or what 



1 For an analogous criticism of Mill's alphabetic conception, see Whewell, 

 Philosophy of Discovery, pp. 263, 264. 



