THE ABYSS OF CENTURIES. 293 



the lake is wearing away a bluff at a rate which has been 

 ascertained. The submerged stump of the bluff permits a 

 measurement of the whole cubic contents removed. From 

 this and the rate of erosion, the time required appears to be 

 2,720 years. The material of the bluff was deposited in a ter- 

 race at the south end of the lake, and this, therefore, has the 

 same age as the erosive process. Next, he finds above the 

 bluff, at a higher level, two sand beaches which must have 

 been deposited when the lakes were at flood. How long a 

 time was required for their deposition? By comparing their 

 volume with that of the sand beach at the south end of the 

 lake, whose age is known as 2,720 years, he finds the high 

 beaches required about 2,570 years. Therefore, from the be- 

 ginning of high water to the present, the interval is expressed 

 by the sum of the age of the upper beaches and the age of 

 the erosion, that is 5,290 years. Dr. Andrews thinks the real 

 interval is somewhere between 5,300 and 7,500 years. 



As the upper beaches must have worn away to some ex* 

 tent during their prolonged exposure ; their age is somewhat 

 greater than Dr. Andrews has estimated. Making allowance 

 for this, the result is not far from that based on the recession 

 of the Falls of St. Anthony. 



We have then several results based on rates of erosion, 

 which stand satisfactorily accordant. For myself, these meth- 

 ods appear more reliable than those depending on mathemat- 

 ical calculations. The results of mathematics are so precise 

 and demonstrative when the data are adequate, that we are 

 tempted to trust them. But very slight errors in the con- 

 stants assumed often lead to enormous errors in the result; 

 and it must be admitted that we discover too many chances 

 of error in the data involved in mathematical calculations 

 respecting the age of the world, to feel that the results are 

 trustworthy, or equally trustworthy with the results based 

 on observed data which bear a larger ratio to the unknown 

 quantity. 



