NOTES ON THE HISTORY OF CORRELATION. 
Being a paper read to the Society of Biometricians and 
Mathematical Statisticians, June 14, 1920. 
By KARL PEARSON, F.R.S. 
(1) As I have often stated, Laplace anticipated Gauss by some 40 years. In 
his memoir of 1783, Histuire de V Academie, pp. 423 — 467, he gives the expression 
for the probability integral 
1 
V2 
and suggests (p. 433) its tabulation as a useful task. It is clear that to do this is 
to recognise the existence of the probability-curve 
y 
\l27r 
or in its doubly pi'ojected form 
N - ^ 
y ~ / - - ^ • 
V 27rcr 
Laplace's investigation while not pioceeding from the very simple axioms of 
Gauss, which lead directly to the above ecjuation, is more satisfiictory than Gauss' 
because we see better the nature of the approximations by which the curve is 
reached and get hints of how to generalise it. Many years ago I called the Laplace- 
Gaussian curve the normal curve, which name, while it avoids an international 
question of priority, has the disadvantage of leading people to believe that all other 
distributions of frequency are in one sense or another ' abnormal.' That belief is, of 
course, not justifiable. It has led many writers to try and force all frecpiency by aid 
of one or another process of distortion into a ' normal ' curve. 
Gauss starting with a normal curve as the law of distribution of errors reached 
at once the method of least squares. To understand the origin of the correlational 
calculus we must really go back to Gauss' fundamental memoirs on least squares, 
namely the Tlieofia covibinationis ohservationum erruribus iiiiniiiiis obnoxiae of 1823 
and the Supplementum of 1826. 
