THEORIES AND MODELS 225 



Pursuing the chimera of "scientific induction," we are heartily dis- 

 satisfied with vague alkisions to "sagacity," "knack," "intuition," or 

 "feehng for order." But how else can one define the imaginative in- 

 sight required for the greatest conceptual advances in science— de- 

 manded, for example, first to see sameness in the fall of leaf and 

 apple, and then in fall of apple and moon? The multitude of such 

 insights achieved by certain individuals, and wholly denied to con- 

 temporaries amply skilled in logic and mathematics, forces us to 

 recognize how much more than those skills must enter into the 

 faculty most esteemed by scientists. To be sure, "imaginative in- 

 sight" is an unobservable, perhaps a fiction, and pronouncement of 

 these words solves no problem. But "logical induction" is also 

 strictly hypothetical, an unobservable, a fiction— and a thoroughly 

 dishonest fiction at that. When we say "imaginative insight"' we ad- 

 mit that the creation of new ideas poses an unsolved problem, and 

 we point to certain (psychological) areas in which at least a partial 

 solution may be sought. But when we say "logical induction" we 

 make solution impossible, either by denying the problem with a pre- 

 tension to having solved it, or by pointing in die wrong direction. 



Preparing the way for an intuition they themselves cannot supply, 

 logic and mathematics have some role even in the first stage of the 

 hypothetico-deductive operation; and they are of course centrally 

 involved in the second stage of appraising the fruits of intuition. 

 Some prove not peaches but lemons. The root of the progressivism 

 of science was earlier identified with the negative capacity for 

 reasonably prompt rejection of error reasonably convincingly ex- 

 posed: logic and mathematics constitute an essential part of the 

 machinery of rejection. No "Canons of Induction" can yield us a first 

 conception of a "true cause." But Mill himself stresses that such 

 Canons are primarily methods of elimination and, as part of the in- 

 formed common sense of scientists, Mill's Canons quite efiiciently 

 indicate the hypothetical causes we do well to reject. 



To the pure logician and the pure mathematician the choice of 

 postulates is in principle essentially free, limited only by the require- 

 ment of self-consistency. To the physical scientist that choice, though 

 underdetermined, is closely restricted: for him the only theoretic pos- 

 tulates worth considering are those from which he can derive the 

 known relations for which theoretic accommodation is sought. Just 

 so, Euclid selected his postulates, rather than others with which 



