THIRTY-FIFTH ANNUAL CONVENTION. 113 



loss in separation, loss in buttermilk, loss in shrinkage, or an 

 error in testing has caused us to pay for more butter fat than is 

 actually received, and this last point, namely, error in testing, 

 arising from the misuse of the Babcock test, is a subject which 

 has already been talked threadbare, but suggestions favorably 

 met with have too often been entirely forgotten. 



With no desire to cast reproach on the creameryman or 

 the methods he employs, facts go to show that more general dis- 

 satisfaction involving losses on either side, can be attributed to 

 the misuse of the Babcock test than to any other one factor in- 

 volved. One obvious reason for this is that testing does not 

 occupy the prominent place in creamery work that it should. 



The very simplicity of the test seems to have made fat deter- 

 minations a stumbling block to the creameryman and it is 

 divulging no trade secret to say that this whole operation, im- 

 portant as it is, is in too many instances occupying a "back seat." 

 If fat, as indicated by the Babcock test, is the basis upon which 

 we are going to establish our creamery pay roll, why not let 

 testing occupy' a legitimate portion of the day's work, and not 

 be a task to be done after the "day's work" is over? 



Suppose even that we think we are carrying on our testing 

 in the best possible manner and so far as we can tell are doing 

 our best to have this part of the work done well ; how many of 

 us can take a twenty-four bottle tester and obtain twenty-four 

 perfectly clear and uniform tests? Not a sure sign of accuracy, 

 to be sure, but an indication of good workmanship. Then how 

 many of us can take a dozen samples from the same source and 

 obtain results close enough to be within the limits of experimental 

 error, and what latitude of variation should we allow for such 

 error? These are all fair questions, and worthy of our honest 

 answer. 



It is a well known fact among you who are familiar with 

 cream and milk testing that from the very nature of the appara- 

 tus we use it is impossible to obtain results which are exact dupli- 

 cates. This is true, first, because of lack of sensitiveness in 

 cream scales ; second, because the best bottles on the market are 

 not graduated closer than a half per cent; then, finally, we are 



