REVISED CENSUS OF 189i. 



95 



referred to. Instead of tbe average correction of 75 per cent, one of 100 per cent 

 would have been more nearly correct. However, since in tbe original census of 1890 

 and in the census of 1897 Kitovi rookery has been taken as typical of the rookeries as 

 a whole, we may use its percentage of correction, 91, in the revision. After making 

 this correction and such alterations in the estimates of harems as experience shows to 

 be advisable, we have the following revised and completed census for the season of 

 1896: 



Eeviaed census, 1896. 

 ST. PAUL. 



ST. GEORGE. 



RECAPITULATION. 



1 A reduction of 30 harems from the original estimate. 



2 A reduction of 40 harems. 



3 The original count of 650 pups used in the census ^of 1896 was made from the cliffs above the rookery under 

 circumstances which make it certain that it is au underestimate It is therefore discarded. 



4 An increase of 42 harems. 

 6 An increase of 15 harems. 



6 A decrease of 25 harems. 



7 A decrease of 39 harems. 



THE VALUE OF THIS CENSUS. 



This elaborated and corrected census of 1896 gives our best possible judgment as 

 to the conditions on the rookeries of St. Paul and St. George during the season of 

 1890. The revision and alteration which have been found necessary make the estimate 

 more or less unsatisfactory. But no other result could be expected. The census was 

 a growth and an experiment from the start. The conditions under which it ought to 

 have been made were not understood until the time was passed. Notwithstanding 



