696 



APPENDIX 



If we had taken all the intermediate integers from 499 with 500 to 440 

 with 559, we should have had ten times as many points which would 

 arrange themselves along the curve in Fig. 4. By increasing the number 

 of coins and decreasing the horizontal scale, we can get the points as 

 close together as we please. This curve in Fig. 4 is the so-called prob- 

 ability curve and it approaches very nearly the curve of error, or normal 

 frequency curve, which we are now prepared to discuss. 



SECTION VI NORMAL PROBABILITY CURVE 



It has been found that the frequency curves of a great many biological 

 measurements follow a curve variously known as the " probability curve," 

 "normal probability curve," "curve of error," or "normal frequency 

 curve." In particular it is known as the "curve of error," because if 

 errors which an observer makes in a refined set of direct measurements 

 on a given quantity be plotted as abscissas, the corresponding ordinates 

 of points on this curve represent the frequencies or probabilities of the 

 errors, 



r 



The general form of the curve is shown in Fig. 5. The origin is taken 

 at the mean. Then, if any mark of a class is above the mean, its devia- 

 tion is positive, and it would be plotted to the right from the origin O, while 

 if the mark of a class is less than the mean, its deviation is negative, and 

 it would be plotted to the left from O. For the benefit of those who are 



