1146 EARLE V. HARDENBURG 



from which to determine the most beneficial influences and practices. 

 The survey idea was first launched in New York by Dr. L. H. Bailey, 

 under whose direction horticultural studies were made thruout the State. 

 In 1903 Professor John Craig started an orchard survey campaign in west- 

 ern New York. Such of these surveys as were completed have been 

 published as Cornell bulletins (Warren, 1905, a and b; Cummings, 1909; 

 Martin, 1911). Under the direction of Dr. G. F. Warren, the survey idea 

 was extended to include the farm as a whole, with the result that whole 

 farming areas, with the farm as a unit, have been studied in what are 

 called farm-management surveys. The results of such studies have also 

 been published as Cornell bulletins (Warren and Livermore, 1911; 

 Thompson, 1915). Montgomery (1913), in discussing crop surveys, states 

 that their primary function is to determine how to grow the crop, while 

 farm-management surveys aim to determine when to grow the crop. 

 Warren (1914) attests the value of agricultural surveys by saying that there 

 are many kinds of agricultural information that can be found only by 

 survey methods, since the conditions in question exist only on the farms. 

 He states further that agricultural knowledge, to be of most value, 

 should be the result of both survey studies and experimental tests. 



The accuracy of survey methods depends very largely on such factors 

 as the personality of the man procuring the records, the manner in which 

 questions are asked, the number of records obtained for each region studied, 

 the unit used as a basis in the study of a factor, accuracy in tabulation, 

 and the final interpretation of results. The more extensive the record to 

 be obtained, the greater is the number of records necessary for final 

 accuracy. The principal faults in much of the survey work to date lie 

 in the attempt to include too much detail and in the use of too- few records. 

 Warren (1914) is of the opinion that ordinarily 1000 records should be 

 used, tho 500 may be enough in some cases. However, the necessity of 

 such large numbers depends somewhat upon the scope of the survey. 

 By the law of averages, large numbers tend to eliminate individual errors. 

 Spillman (1917) has said that the accuracy of any average depends on three 

 things: first, on freedom from bias; secondly, on the number of items from 

 which an average is obtained; and thirdly, on the accuracy of the individual, 

 items averaged. Considering the limitations of much of the experimental 

 evidence to date, large numbers of survey records are undoubtedly pro- 

 ductive of as nearly accurate results as are obtained by experimental 

 work. As . emphasized by Warren (1914), the region surveyed should be 

 an agricultural, not a political, unit. Furthermore, the records should be 

 taken only during a normal year unless records are to be obtained for con- 

 secutive years. Unfortunately for this study, the year 1912 was at first 

 drier than normal, but the abundant rain which fell late in the growing 

 season caused some blight rot; 1913, however, was a more nearly normal 

 year. 



