THE CONDUCTION OF RESEARCH 429 



Before beginning an experiment care should be taken that everything 

 is functioning properly, and that all disturbing factors have been taken 

 into account. Never take a chance, as every research man should real- 

 ize that not only his reputation, but that of the organization with which 

 he is connected, is endangered by his mistakes, and he should under no 

 conditions allow results to come from him, unless he is very sure of 

 their correctness. 



It might seem unnecessary to bring up the subject of honesty in re- 

 search, for there would seem to be very little reason to give results 

 deliberately in error. It has often happened, however, that an experi- 

 menter has shaded the values of his readings to make them come closer 

 to what he supposed was the true value, but he often finds later, that he 

 has gone in the wrong direction, and readings taken in this way will 

 give neither the true mean nor the probable error. Sometimes this 

 squeezing of results up or down is done quite unconsciously, and for 

 this reason when it is desired to make a check run, the results of the 

 first run should never be in sight, or it will not be a true check, even 

 though the recorder is not in any way intentionally dishonest. 



When it is desired to determine the accuracy of certain data, or to 

 be assured that it lies within the permissible limits of accuracy, it is 

 customary to make two or more runs under identical conditions, the 

 difference between the values obtained in each being an indication of 

 the accuracy that may be expected. This does not, however, tell the 

 whole story, as it does not take into account those errors in the design 

 or setting up of the apparatus, or the individualities of the experiment- 

 ers. For this reason, it is always well when making a check run to 

 reset the apparatus or, better, to use a different piece of apparatus and 

 different observers, in which case the results may be considered to give 

 a true indication of the probable error from all sources. It often 

 happens that a certain set of facts are not obtainable in a direct or 

 simple manner, nor is the best method that can be devised entirely satis- 

 factory. In such cases it is always best to obtain the results in several 

 different ways, and, although none of them may be satisfactory, yet, if 

 the several results show an agreement, it may be concluded with cer- 

 tainty that they are correct. Even in the more straightforward investi- 

 gations, wherever possible, the results should be checked up by an alter- 

 native method, as this is an excellent way to make others have confi- 

 dence in the data. 



Next to accuracy, the most important consideration in research is 

 efficiency, that is, the obtaining of the largest amount of results for the 

 least expenditure of time and money. Efficiency can only be attained 

 by the careful laying out of the work, the careful determination of 

 what is necessary to do to get the required accuracy, and, most im- 

 portant, to have a smooth running organization. The laying out of the 



