106 Report of the Department of Bacteriology of the 



weighing of a few samples from different pipettes demonstrates, 

 however, that the error is ordinarily due more to faulty graduation 

 and form of tip than to the inability to measure successively quan- 

 tities which weigh approximately the same. This is shown in 

 Table XII which gives the weights of several samples of milk 

 obtained from different pipettes, together with the percentage error. 

 The computation of this error is based upon the specific gravity 

 of milk as 1.032. The percentage error of the same pipettes as 

 determined by calibration with mercury is also given. 



Pipettes A, B and C were straight-bored with square tips (See 

 Fig. 2) . Pipettes D and E were straight-bored with beveled or rounded 

 tips (See Fig. 1). The tip of pipette D was, however, so slightly 

 rounded that no difficulty was experienced in depositing the whole drop 

 of milk. Pipette F was the one shown in Fig. 3. Evaporation of 

 milk during weighing may have caused some of the irregularities in 

 the weights given, but this error must have been practically the same 

 in all cases. One thing shown by these figures is that it is unwise 

 to trust to the calibration of such pipettes by commercial firms. 

 The tests indicate that capillary pipettes such as these should be so 

 calibrated with mercury as to have an error of approximately plus 

 5 per ct. in order to deliver the correct amount of milk (.0103 grams). 

 Pipette C was the one chosen for use in the present investigation. 



Small samples not representative. — Objection has been raised to 

 the studying of so small a sample of milk because of the possibility 

 that it will not be representative of the whole. Table XIII shows 

 counts made on three samples of milk, from each of which four dupli- 

 cate smears were prepared and counted, which indicate that this is 

 not a serious error. These three samples are typical of a large series 

 of duplicate microscopic counts which have been made. The irregu- 

 larities are no greater than those found in duplicate counts made by 

 the plate method. 



Few or many bacteria. — The error of count is greater where there 

 are only a few organisms present, because it is easily possible to over- 

 look some. The finding of one organism or the failure to find one 

 organism means a difference of several thousand in the final count, 

 the exact amount depending upon the number of fields counted. 

 If an accurate count must be made on such samples then it is neces- 

 sary to count a large number of fields. Under ordinary circumstances 

 it is not necessary to do this, for it is soon seen that the sample con- 



