of determining Stream Lines and JEquipotentials. 24t3 
corner of the square the contours drawn at equal intervals of 
V must make equal angles with one another. One starts 
then by ruling out an accurate square, putting in the 
diagonals, joining the mid-points of its sides and setting off 
the equal angles with a protractor. It is convenient to 
divide the range of V into ten equal parts. Having thus 
prepared the paper, lines were sketched and amended until 
further improvement became very slow. The pencil-lines 
were then firmly fixed in ink. Coordinate lines were drawn 
in and the values of V at six points were read from the 
diagram and are given in parenthesis in the accompanying 
table. The whole work from the beginning of the drawing- 
took two or three hours. 
•5 
•5 
(•47) 
•466 
1-0 
•5 
(•40) 
•396 
(•365) 
•364 
1-0 
•5 
(•307) 
•300 
(•23) 
•223 
(•20) 
•202 
•5 




Not until this had been done did I look up the correct 
values which had been computed from the analytical solution 
V= n 2 (-1) 
m odd 
\=1 1 , mil , 
2 — seen ^r- cos ma . cosn mz. 
m 2 
These are given in the table beside the numbers read from 
the graph. From these we find the errors + *007, + *007, 
— -002 : +-004, +-001, +-004. Treating these as all of the 
same sign, their mean is "0042. 
The error of a graph may well be compared with the total 
range of V within which the determination was made 
freehand. In this case the range was 0*5 so that the mean 
