16 BULLETIN 563, U. S. DEPARTMENT OF AGRICULTURE. 
same, or 25 per cent. As the total count becomes still higher, neces- 
sitating higher dilutions, the difference in bacteria per cubic centi- 
meter increases until at a dilution of from 1 to 1,000,000 the difference 
would be 75,000,000 per cubic centimeter. 
A study of this assumed case merely serves to show that even 
though there is a difference of 75,000,000 between two samples of 
ice cream, there is no greater percentage of variation between the 
samples than when the difference was only 750 per cubic centimeter. 
It also shows, however, that a difference between counts expressed 
in bacteria per cubic centimeter should never be considered by itself 
but should be interpreted in relation to the total number of bacteria 
per cubic centimeter in each sample. 
SUMMARY AND CONCLUSIONS. 
The method of collecting samples and making bacterial counts 
used in our experiments gives results which indicate that bacteria 
in commercial ice cream are distributed quite evenly and that an 
analysis of one sample from a gallon of ice cream gives results which 
will hold for any other similar sample from the same gallon. 
Storage of ice cream for 11 days in a commercial ice-cream cabinet 
or in a hardening room for a period of two months did not seem to 
cause an uneven distribution of bacteria. 
In a series of from 5 to 10 samples taken directly from a large 
commercial freezer the bacterial counts on each sample checked 
within the usual limits of error of bacterial analyses. 
No greater variation in bacterial counts between samples was 
observed when the plates were incubated at 37° C. (98.6° F.) for 48 
hours than when incubated at 30° C. (86° F.) for a period of five 
days. 
When dilutions were such that about 200 colonies were present 
on the plates a lower variation between counts of samples of ice 
cream was found than when there were 50 or fewer colonies per plate. 
The variation between a series of plates made from the same sample 
and dilution was found to range from 7 to 26.6 per cent. Among du- 
plicate plates a variation as high as 41 per cent was observed. This 
must be remembered in connection with the fact that the variation 
found in our experiments between average counts of different samples 
of ice cream from the same gallon lot ranged, generally speaking, 
between 20 and 30 per cent. To this variation between duplicate 
plates or a series of plates from the same dilution must be added the 
error introduced in removing 1 c. c. portions of ice cream from different 
samples. 
When interpreting bacterial counts, differences in the number of 
bacteria per cubic centimeter should never be considered except in 
relation to the total count of each sample. 
WASHINGTON \ GOVERNMENT PRINTING OFFICE : 1317 
