EMPIRICAL TOOLS AND EMPIRICISM 163 



this major complication by setting up our paired experimental sys- 

 tems with pairs of experimental subjects exactly alike— thus (hope- 

 fully, and often actually) arranging to cancel the effects of variables 

 not even known to us as such. Powerful though it is, however, use of 

 an experimental control is no panacea. Literally identical specimens, 

 in systems literally identical in all save one respect, we do not have 

 and cannot secure. We have at most specimens and systems matched 

 in those (relevant) characteristics that, on conceptual grounds, we 

 regard as assuring effective identity. The use of experimental con- 

 trols makes the problem of relevant variables somewhat less oppres- 

 sive, but ultimately the problem remains just as real here as when we 

 seek to control the variables explicitly and directly. 



Whether in a controlled experiment or in one with experimental 

 controls, we sometimes encounter erratic fluctuations in our results. 

 When these fluctuations substantially exceed our estimate of experi- 

 mental error, they teach us that we have overlooked some relevant 

 variable(s). But however our data may assist us, neither they nor 

 Method then proclaim the identity of the relevant but uncontrolled 

 variable(s). Always it is ive who must go on to guess that. Such 

 guessing is not always easy. Those shrewd investigators Scheele and 

 Priestley long failed to grasp what shortly after was realized by 

 Ingen-Housz: intensity of illumination is a highly relevant variable 

 in the chemical interaction of plants with the atmosphere. 



The choice of observables. I cannot control, or even match, all the 

 variables possibly relevant to experimental production of a given 

 phenomenon. But also, among myriad possibilities presented to view, 

 I cannot possibly note all the observables. As early as the 16th cen- 

 tury Tycho Br ahe— himself neither theoretician nor even experi- 

 menter but simply observer— recognized the utter impossibility of 

 "pure" obsers^ation. One must have a lead indicating where to look 

 for something worth observing. This essential role of premonitory 

 hypotheses is generalized by Cohen in the following emphatic state- 

 ment. 



Accidental discoveries of which popular histories of science make 

 mention never happen except to those who have previously devoted 

 a great deal of thought to the matter. Observation unillumined by 

 theoretic reason is sterile. . . . Wisdom does not come to those who 

 gape at nature with an empty head. Fruitful observation depends not 

 as Bacon thought upon the absence of bias or anticipatory ideas, but 



