TEACHING AND LEARNING 13 



be so planned as to call for a simple answer, and of course should 

 be as little complicated and as little damaging to the plant as 

 possible. It properly follows upon careful observation, and 

 usually is a testing of hypotheses suggested by reasoning thereon. 

 Most commonly an experiment is undertaken to find out the 

 relation existing between the processes of the plant and some 

 particular external condition; and practically the first and 

 most natural step is to observe the effect upon the plant when 

 that condition is removed or neutralized. The ideal experi- 

 ments are those in which only this single condition is altered; 

 but, partly on account of the closeness with which different 

 conditions are yoked together, and partly because of the rela- 

 tive crudeness of even our finest methods of experimenting, 

 this is very rarely possible. Hence in order to make sure that 

 the result obtained is really connected with the condition changed, 

 and not with some secondary influence introduced by the manip- 

 ulation in the experiment, it is usually necessary, and always 

 best, to try at the same time a parallel experiment in which a 

 similar plant is placed under precisely the same external and 

 experimental conditions as the first plant except that the given 

 single condition is not changed. Here, in both experiments, 

 all the secondary conditions are the same; the difference is only 

 in the given primary condition; and hence it is a fair inference 

 that an observed effect is connected with the change in the pri- 

 mary condition. Such a parallel experiment is called a control, 

 and an impulse to control experimenting is an essential part 

 of the experimental habit. It is through lack of it that our 

 elementary botanical text-books are disfigured by descriptions 

 of some experiments which are scientifically illogical, and which 

 only by accident give correct results, a subject on which I shall 

 comment at the proper places in this book.* 



Control experimenting tends to neutralize some of the grossest 

 of the sources of error which beset all scientific investigation; 

 but there remain many others whose detection and elimination 



*I have also discussed it, somewhat fully, in School Science, 6, 1906, 297. 



