Variation of Sentence-Constants in Literature 19 



the figures employed, but the magnitude of the probable error 

 in the 6"'s can only be appreciated by a glance at the figures from 

 which some of the averages are gotten. Thus we find Lyly's 17 

 to be the mean of the numbers 26, 14, 20, 15, 8/ these numbers 

 representing the simple-sentence-percentages for consecutive hun- 

 dreds. Similarly, De Quincey's 14 results from the numbers 

 10, 19, 15, 7, 21.^ If larger averagings had been made, essen- 

 tially different results would have been obtained in many cases 

 as the following example shows : 



VARIATION IN SIMPLE-SENTENCE-PERCENTAGES FOR CONSECUTIVE 

 HUNDREDS IN 



Spenser^ s Viezv of the Present State of Ireland:^ 



Average of 300 sentences yields 12.0 per cent simple sentences 



" 400 " " 11.5 " " " 



" " 500 " " 10.6 " " 



" 600 " " 10.5 



'• 700 " " 9.7 



" 800 " " 9.0 



" 900 " " 8.9 



" 1000 " " 8.0 



Even so great a stylist as Macaulay gives averages for 5* so 

 widely divergent as 41 and 27 for different parts of his History 

 of England, though each average is based upon 500 consecutive 

 periods.* 



Although disastrous for immediate progress, these facts did 

 not lessen my confidence in a more definite law than that which 

 was thus far apparent. The first problem was how to reduce 

 the probable errors to manageable proportions. Of course, the 

 obvious solution consists in making larger averagings, say, of 

 10,000 or more sentences from each work, but this involves an 

 enormous amount of unattractive labor. To avoid this, I de- 

 vised a seemingly crude experiment on Mr. Gerwig's figures for 

 English prose works. ^ I struck averages of both P and 5" for 



"^University Studies, vol. II, no. 1, p. 38. 



Ubid.,-p.^. 



»/bid., p. 42. 



*ldid., -pp. 25-26. The numbers represent the averages for the 27th to the 

 33cl, and for the 284th to the 289th hundreds respectively. 



^Ibid. , pp. 31-44 . These figures comprise averages of about 60,000 periods, 

 taken from 71 different authors. 



247 



