Variation of Sentence-Constants in Literature 25 



These are problems which it seems to me need not remain 

 unsolved. 



The greatest immediate need is for more abundant data. In 

 the data thus far obtained it has been assumed that 300 periods 

 will furnish with comparative accuracy the constants sought, and 

 in but few instances have more than 500 periods from single 

 works been examined. Neither the probable nor the actual error 

 arising from this hypothesis has been determined. 



In order to get some light on this question I have computed 

 the following table from averages of 400 periods and 500 

 periods each of 50 authors:^ 



TABLE V 



This table shows that the constants L, P, and 5" obtained from 

 an average of 400 sentences, will differ by 2.35 per cent, 2.06 

 per cent, and 5.98 per cent respectively from the values of these 

 constants based upon averages of 500 sentences, and in some 

 cases variations amounting to 9.5 per cent, 9.7 per cent, and 28.8 

 per cent respectively may occur. We should, therefore, even by 

 examining 400 sentences from each of 50 authors, obtain less 

 than 2 per cent accuracy in the constants L aqd P, and about 6 

 per cent accuracy in 5". The numbers in the last column show 

 the probable deviation between 400 and 500 sentence-averages 

 for any other work whose constants have not yet been deter- 

 mined. It appears, therefore, that so long as averages are based 

 upon 500 or less periods the fractional parts in L and 5" may be 

 entirely neglected, and in P everything after the first decimal 

 place may be disregarded. 



^The table is based upon Miss Whiting's results for L and Mr. Gerwig's 

 tables for P and 6". 



253 



