THE MICROSCOPE. 163 



ter, and hence will be separately discussed for reasons herein- 

 after stated. 



Six hundred and twenty-six check signatures of Mr. Alpheus 

 H. Snow, 133 of Mr. Parke Daniel, and several hundred checks 

 signatures of other parties were also examined. 



While making these measurments we also compared the 

 checks of the same persons respectively inter sese, with the view 

 of finding, if there were any, such as were exactly, or within 

 very narrow limits, identical in size and form, but found not a 

 single instance of such identity. We had not proceeded far in 

 the investigation before it became apparent that the length of 

 the signature depends to some extent upon the length of the 

 line upon which it is written ; and that even in cases where the 

 line is longer than is necessary for the signature, as ordinarily 

 written, it invariably affects the length of the signature ; so that 

 the average length of the signature diminishes with the length 

 of the line upon which it is written. It is obvious, therefore, 

 that therecan be no settled mean length of signatures unless they 

 are written upon lines of substantially the same length. By lines 

 in this connection is not meant lines running entirely across the 

 sheet of paper, but such lines of convenient length as are usually 

 found upon blank checks and other similar documents. Having 

 arrived at this conclusion, we endeavored to ascertain whether 

 those signatures written on similar blank checks under substan- 

 tially the same conditions were governed, as we suppose, by the 

 general law of probability, so that the platted curves herein- 

 after described, would form what is known as the " probability 



curve." 



The results arrived at were not conclusive upon this point, 

 though we are of the opinion that, given a sufficiently large 

 number of signatures written upon similar blanks and under 

 the same conditions, they would be governed by the law above 

 referred to. 



The method adopted was as follows : The lengths of the sep- 

 arate signatures of each person were added together and the 

 sum divided by the number of signatures, giving the mean 

 length. The length of each separate signature was then sub- 

 tracted from this mean length, giving a series of remainders 

 plus and minus. These residuals were then classified and curves 

 platted in which the ordinates represent the number of signa- 



