No. 509] 



NOTES AND LITERATURE 



whole of the theoretical curve in this case lies on the same side 

 of the observed line. The sum total of the plus deviations 

 equals only 1.15, whereas the sum total of the minus deviations 

 equals 199.05 ! 



Let us take still another example this time from near the end 

 of the first paper. In Table IX (p. 610) are presented Donald- 

 son's 7 data regarding the growth of the brain in the frog and 

 the fitted curve. In this table are 21 ordinates available for the 

 eomp-irison of theoretical curve and observational data. The 

 deviation between theory and observation is plus in 18 out of the 

 21 cases and minus in 3 cases only. Two ordinates (making with 

 the 21 the total of 23 tabled ) show exact agreement between 

 theory and observation. In spite of this extraordinarily uneven 

 and biased distribution of the errors this statement follows 

 Table IX: 



Surely a system ot errors i 



and only 14 per cent, are in defect and i 

 centage deviation per ordinate for the ] 

 cent, can not fairly be said to be "irregi 



Other examples showing the same thir 

 the papers. The tables which have been 

 of the point under discussion have been 

 others for two reasons; one that they wer 

 a fairly large number of ordinates. the 

 tional data in these tables were obt aim- 

 painstaking measuring and are absolutely 

 data, if anywhere, a theoretical curve i 

 to give good results. 



that the discr 

 great 



so biase< 

 ed that these ( 

 be held to afford evident 

 Robertson's ingenious, suf 

 hvDothesis. It is possibh 



