220 ANNUAL. REPORT OF THE Off. Doc. 



Incorrect running. — Too low or too liigh speed or feed, a trembliug 

 bowl, a machine ill cleansed, out of repair, or out of balance may and 

 do alfect results. I have known positive flaws to exist in the me- 

 chanism which modified results. . 



While net exactly germane to the subject I may be permitted, 1 

 trust, to say a word in answer to the very common question at meet- 

 ings of this kind — What is the best separator? 



This is one of the most common questions asked of the station. 

 Our present feeling in the matter is that there is not, of necessity, 

 any one make that is "best" in all points; that machine of all the 

 more prominent makes are capable of doing a good grade of work 

 when properly handled; that, since good skimming is the rule, other 

 points, such as initial cost, durability, probable repair bills, ease of 

 operation, etc., are now more important; and, finally, since flaws may 

 occur in individual machines of anj^ make, that agents' claims and 

 records of other machines of the same make are of less value, 

 touching the quality of skimming, than is the analysis of the skim 

 milk of the individual machine offered. Many farmers in Vermont 

 have bought separators on the condition that the skim milk ehould 

 be submitted to the experiment station for analysis, purchase to fol- 

 loAV its favorable, and rejection its unfavorable report. The buyer 

 Thus has, free of expense, the advantage of the advice of disinterested 

 experts, which moreover, is given in ignorance of the kind of machine 

 under trial. 



We are now ready to consider a phase of the question which I want 

 to treat with the greatest care as to the words I use and the impres- 

 sion I leave. 



I believe that among the serious factors in this matter of milk 

 test variation are the errors of the testing operation. Let us dis- 

 cuss this possibility of error in the manipulation of the test under 

 the sundry subheads, sampling, apparatus, errors of ignorance and 

 errors of intent. 



SAMPLING. 



By no art of legerdemain can a milk analyst return a correct re- 

 sult from an incorrect sample. I am inclined to think that a con- 

 siderable part of the variation between tests is due to imperfect 

 methods of sampling. 



Three methods of sampling are more commonly in vogue, the dip- 

 per method, the core method and the automatic method. The latter 

 is applicable to milk sampling only, unless very large quantities of 

 cream are brought to the factory. The two former are used for 

 both milk and cream. The first na^.. d is the most widely used of 

 the three. From the mass of milk or cream more or less (almost al- 

 ways less) thoroughly stirred (and, indeed, often not stirred at all) 

 a gill or less is dipped for a sample. Such procedure may result in 



