The success of a prediction therefore only shows that the theory on which it is founded has had practical value so far as a working hypothesis. As working hypotheses, and as long as they are kept down to brief steps _which can be verified_, the scientific theories are very valuable--indeed we could not do without them; but when they are treated as objective facts--when, for instance, the "law of gravitation"--derived as it is from a brief study of the heavenly bodies--has a universal truth ascribed to it, and is made to apply to phenomena extending over millions of years, and to warrant unverifiable prophecies about the planetary orbits, or statements about the age of the earth and the duration of the solar system--all one can say is that those who argue so are flying off at a tangent from actual facts. For as the tangent represents the direction of a curve over a small arc, so these theories represent the bearing of facts well enough over a small region of observation; but as following the tangent we soon lose the curve, so following these theories for any distance beyond the region of actual observation we speedily part company with facts.[22] To proceed with a few more words about the general method of Science. Science passes from phenomena to laws, from individual details which can be seen and felt to large generalisations of an intangible and phantom-like character. That is to say, that for convenience of thought we classify objects. How is this classification effected? It is effected through the perception of identity amid difference. Among a lot of objects I perceive certain attributes in common; this group of common attributes serves, so to speak, as a band to tie these objects together with--into a bundle convenient for thought. I give a name to the band, and that serves to denote any unit of the bundle by. Thus perceiving common attributes among a lot of dogs--as in an example already given--I give the name foxhound to this group of attributes, and thenceforth use the name foxhound to connect these objects by in my mind; again perceiving other common attributes among other similar objects, I invent the word greyhound to denote these latter by. The concept foxhound differs from the objects which it denotes, in this respect that these latter are (as we say) _real_ dogs with thousands and thousands of attributes each: one of them has a broken tooth, another is nearly all white, another answers to the name "Sally," and so on; while the concept is only an imaginary form in my mind, with only a few attributes and no individual peculiarities--a kind of tiny G.C.M. arising from the contemplation of a long row of big figures.