It appears, therefore, that all methods result in rings of approximately the correct size, and it is usual to check, and, if necessary, make an allowance for variation in dimensions. It is not possible to do this, even approximately, with soft rubbers, as the rubber gives under the pressure of the micrometer. No doubt a photographic method would give more accurate results, but would take too long. I have found that a very close approximation is obtainable by weighing the rings as the specific gravity of the standard rubber mix is known. It is not necessary to weigh each ring, but the whole five or ten taken for testing may be weighed together. The next point that arises is the choice of a formula for the test mix. Practically all the work to date has been carried out on mixtures of rubber with 7 to 10 per cent. of sulphur. For some purposes--_e.g._, detecting variation in rate of cure--this mixing is satisfactory, but for other purposes it is not. Nor is the behaviour of a rubber-sulphur mixing a sure guide to the behaviour of one containing other ingredients, such as litharge. Thus, two samples vulcanised satisfactorily when mixed with sulphur only, but one of them gave unsatisfactory results in the presence of litharge. It has long been recognised that mineral ingredients may modify the product when vulcanised, but the modification is not necessarily uniform. Consequently, tests should also be made, when practicable, with vulcanised rubber containing other ingredients in addition to sulphur. As regards physical tests on the vulcanised products, these usually involve determination of breaking load and elongation at rupture (usually recorded as final length--that is, including the original length reckoned either as unity or as 100 units). Simultaneously a load-stretch curve is recorded on an autographic attachment. The type of curve varies with (1) state of cure, or degree to which the rubber is vulcanised; (2) proportion of sulphur and/or other ingredients; (3) specific nature of the rubber used. The last factor is almost negligible compared with the two former--at any rate for average quality rubber. As (2) is kept constant for any batch of tests, or even for every test, it follows that the load-stretch curve is mainly dependent on the state of cure, and the degree of vulcanising may be measured by comparing either the elongation produced at a given load or the load produced at a given elongation. Either set of figures is readily determined by measuring up the load-stretch diagram.