I. AVOID FRtTlTLESS EXPERIMENTS 29 



Without tlii^, future workiM-s will not be ahle to use llie data in a 

 strictly (|uantitative sense. 



4. Use Appropriate Degree of Acciiraey 



In almost any experiment, some of the factors usually can be 

 measured with a high degree of precision, others with an entirely 

 satisfactory accuracy, while perhaps still others with only a com- 

 paratively crude subjective estimate. Since the various factors should 

 all be measured with al^out the same order of accuracy, it is often 

 somewhat ridiculous to go to great lengths to measure one item very 

 precisely while other items in the same experiment are little more 

 than guesses. If the same relative precision is to characterize all as- 

 pects of a particular investigation, an increase in the accuracy of the 

 least precise items in an experiment should be the first point of attack 

 in any effort to improve the over-all accuracy. 



Although quantitative measurements constitute the basis of ex- 

 perimental science, the necessity for accuracy may be overempha- 

 sized. The development of quantitative spectrographic analysis, for 

 example, probably was delayed many years because physicists 

 stressed the possible sources of inaccuracy in the method; however, 

 they did not recognize that it had other tremendous advantages over 

 existing methods and that the accuracy available was sufficient for 

 numerous applications. Today a useful technique of quantitative 

 analysis even employs a modified Bunsen burner as a source. 



Extreme accuracy can often be obtained if one is willing to pay an 

 extreme price. Unless the extra effort can be justified in terms of 

 relative usefulness of the data either in the same or other experiments, 

 one should not go to extremes. For example, one might embark on a 

 program to determine the melting points of all organic compounds to 

 a precision of a himdredth of a degree, but would this serve any 

 justifiable purpose either theoretically or experimentally in the fore- 

 seeable future? 



H. ANALYZE DATA OBJECTIVELY 



Experimental data should usually have the benefit of analysis by 

 statistical methods. By definition these methods are especially 

 adapted to the elucidation of quantitative data affected by a mul- 

 tiplicity of causes. Since much biological experimentation yields 

 data wherein the organisms or sj^stems studied are influenced by 



