1. The primary aim of the experiment was to find out if certain combinations of container and gas tube give a greater proportion of positive results than others (with the same strength of inoculation) and if possible to elucidate the causes of any differences between combinations.

The earlier results indicated that the best combination was a 7 × 1 ½ in. test-tube with a 4 × ¾ in. gas tube. Later experiments failed to maintain this preference but did show a definite disadvantage in using a very small gas tube. Whether this is due to the absolute sinallness of the gas tube or the smallness of its volume relative to the total amount of liquid used cannot be decided, for the latter was the same in all combinations. This was the only shape or sizefactor that was definitely influential on the evidence of these experiments. Yet if size of gas tube is intrinsically important the absence of reactions with acid but no gas is puzzling.

Attempts to break down the points involved in changes in shape and size of gas tube and container into a number of simple factors showed that several considerations might be involved, too many to cover in one experimental lay-out; but by including extreme combinations it was hoped that some indication might be obtained of the more important factors. With the new and successful combination of 7 × 1½ in. test-tube with 4 × ¾ in. gas tube, the gas tube holds 22% of the total volume of medium plus inoculum, the same proportion as the standard 6 × ⅝ in. test-tifbe and in. gas tube. This seems to be in support of standard practice if it is relative volumes that are important.

2. The analysis of plate counts shows that the opacity method, with the strain employed, gives complete control of the population and the three methods of assessing the plate count were in good agreement with one another.

The χ2 values were as a rule subnormal. This result excludes the possibility of any lack of control of the population in any one set of plates, but the reason for the subnormality is of some interest. It may plausibly be attributed to the factor of competition among developing colonies for the nutrient medium available so that the chance of development of a colony is smaller when a large number of others are present than when there are only a few. A toxic emanation from a colony which excluded the formation of another colony within a certain distance of it would have the same effect.

3. In the main experiment the percentage of positive tubes was about half (20% as against 45%) that to be expected from the plate counts on the usual Poisson hypothesis; or looking at the matter the other way the mean number of organisms per tube (0·22) deduced from the percentage of positives was about half the 0·5 expected from the plate counts. We have found no reason for this and in two subsequent experiments the agreement was satisfactory.

4. No difference was found between 2 × ⅜ and If in. gas tubes with a 6 × ⅝ in. test-tube or between rimmed and rimless gas tubes.