166 EMPIRICAL TOOLS AND EMPIRICISM 



word is creative. The historv of science furnishes us a multitude of 

 examples familiar to all [e.g., universal gravitation, energy, field]. 



How materialize "the soul of the fact"? The latent meaning of experi- 

 mental data appears only with the application of a complex concep- 

 tual developer, the formulation of which is irreducible to Method. 



THE POWER OF STATISTICAL METHODS 



The elaboration in recent years of refined techniques of statistical 

 analysis may be thought to rehabilitate Method. Teaching us a su- 

 perior design for empirical studies, and showing us how to get more 

 out of empirical data, the statistical techniques generally simplify the 

 empiricist's tasks. Consider, for example, the use of experimental 

 controls. To investigate one \'ariable we ordinarily seek a pair of 

 systems diflFering in that variable but matched in all others possibly 

 relevant. When many such variables come in question, we are then 

 committed to the huge e£Fort of preparing great numbers of paired 

 systems matched in all but one variable. Furthermore, hewing to 

 this pattern, we must inevitably fail ever to observe such possible 

 effects as those "synergisms" that materialize only from simultaneous 

 variation of at least two variables. Finally, and most important, we 

 are entirely debarred from certain investigations in which no match- 

 ing of all the potentially interesting variables is possible: in agricul- 

 tural studies, for example, often we simply cannot find two sizable 

 plots of land exactly matched in soil, exposure, drainage, and every 

 other significant detail. 



The pioneering studies of Fisher and others have taught us how 

 to meet these problems; taught us how, by statistical analysis, to ex- 

 tract meaningful results from the examination of multi-v SLriant sys- 

 tems. With the aid of such analysis the requisite number of trials is 

 sharply reduced, and conclusions otherwise uncertain or wholly in- 

 accessible are brought firmly widiin our grasp. Recall the co-ordi- 

 nated simplification and complication we earlier found to accompany 

 the shift from observation to experiment. To win simplification of our 

 conceptual problems we accept the empirical complications of con- 

 triving experimental systems tliat match or control die relevant vari- 

 ables. Now Fisher and his colleagues have forged for us a new con- 

 ceptual tool of unprecedented power. When strictly controlled ex- 

 periment is impossible or inconvenient, tliis tool permits us to extract 



