ii SCIENTIFIC RECONSTRUCTION 255 



to the celestial, an extension at first tentative and then con- 

 firmed by the perfection of calculation and observation. 



All sound hypothesis, I would venture to say, falls back 

 on this method, though it may not begin with it. Thus, 

 Darwin made a new epoch in biology because he assumed 

 only such causes of variation as were known the selective 

 action of breeders, as he understood them and his argu- 

 ment went to show (a) that a partially though not wholly 

 similar selection was at work all through organic life, and 

 (b) that the cumulative action of such selection operating 

 through generations would explain the facts of the organic 

 order. Had Darwin been able to carry through his argu- 

 ment with the precision of Newton, he would equally have 

 proved his theory as a generalised extension to the whole 

 range of organic life of that which can be seen at work 

 in some phases of organic life. The true criticism of 

 Darwin is coming from those who are demonstrating (a) 

 the insufficiency of the kinds of variation and selection or 

 which he had knowledge, and (b) the existence, as a matter 

 of verifiable observation, of other kinds of variation. 

 None the less, Darwin's method was sound because it 

 rested on empirical generalisation. So also is the method 

 of those who rely on experimental breeding or on micro- 

 scopic examination of the structure and structural changes 

 of the germ cell. The rival method of assuming elements 

 of the germ cell and modes of action within it which cannot 

 be observed, and reasoning therefrom to get back to the 

 facts with as little awkwardness as possible, is one destined 

 to fill many volumes of controversy and to produce theories 

 which undergo transmutation after transmutation before 

 they settle down into accord with the verifiable facts. 



3. On the other hand, to limit the work of science to the 

 accurate and compendious description of observable data 

 is to be carried too far in reaction. A theory necessarily 

 aims at something more than the exact description of what 

 it finds. It aims at generalisation, that is, at inference 

 which will enable it to say not only what is found, but what 

 will be found, or would have been found by observation 

 in the past. Such generalisation is secure in proportion 



