392 PKOCEEDINGS OF THE AMERICAN ACADEMY. 



In reading the adjustment of the dividers, the distance was laid off 

 four times along a ruled line, and the result was divided by four. This 

 served to diminish errors in the use of the scale. The dividers were 

 always returned to the starting point as a test for change of adjustment, 

 but no change ever appeared. The error introduced in this process was 

 probably less than one jjer cent. 



The " Entfernungsmesser," as I have said, registers in centimetres. 

 To obtain an accurate reading, it was run over each line a number of 

 times (sometimes fifteen or twenty, if error were suspected), and the 

 reading for ten trips was taken direct, with the substitution of milli- 

 metres for the centimetres of the reading. Tiiis method was laborious, 

 but accurate. By actual test the average error was found to be less than 

 one per cent. 



III. Results. 



In the following tables, AD stands for Average Deviation of the 

 measurements from the mean of all ; CV for the Coefficient of Variation, 

 which is the average deviation divided by the mean. 



