after a 6-day fast weighed 1600 g. gen whatever was found in the liver. of the carbohydrates in the litchi nut. The animal was killed and no glyco-'l'hese results show in a most marked way the glycogenetic properties SEW HAVEN, CONN.

new process and the old process. The proportion tells us, however, that we must compare the small-scale run by the new process (n) with the small-scale run by the old process (o), to see how much difference is made by the change in process. Then we compare the small-scale run by the old process (o), with the factory run by the old process (0), to see how much difference is made by the change from smallscale to factory-scale. Here, we have only one factor varying in each comparison; hence, we can draw more reliable conclusions.A second common error is to attempt to design an entirely new commercial process with only small-scale runs as a basis, without any large-scale tests or "semiworks" runs. The dangers of doing this have often been described.2An attempt to set up the "Small Scale to Factory Proportion" brings us to the same conclusion. In developing an entirely new commercial process we have the smallscale runs but no factory-scale results whatever. We have two terms of the proportion, but not three, and we know that three terms are necessary for the solution of a proportion.Many people think that the small-scale products ought to be the same as the factory products and, as this is seldom the case, that little can be learned by small-scale runs.The proportion n:o = N:0 clearly shows us, however, that n need not equal N and that o need not equal 0. In fact, they seldom do so. 2Whiting, 8th Intern. Cong. Appl. Chem., Section Xa, p. 204. LimitationsNow for the limitations of this proportion. First, it must be emphasized that it gives only relative terms and that results must be expressed as "more" or "less." Results are not strictly quantitative and the proportion is not a rigid mathematical equation, though it has been written as such for the sake of brevity and simplicity. One side is only approximately equal to the other, so that, strictly speaking, our proportion should be written n:o = approximately N:0.Another precaution to be borne in mind is that the smallscale method must involve all the essential factors of the factory method. The more nearly the small-scale formula, times, temperatures, and methods of handling approximate the large-scale factory conditions, the more reliable will this proportion be. Experience in comparing small-scale results with factory results in any given series of experiments will indicate how closely this "Small Scale to Factory Proportion" can be followed. Wherever small-scale runs are made it is necessary to get an idea of this, just as it is necessary to form an idea of the size of the experimental error in any experimentation. If small-scale conditions differ so much from factory conditions that results on a small scale vary without relation to results on a large scale, this fact should be known as soon as possible. Such small-scale runs are evidently actually misleading and not worth making at all, being like work in which the experimental error is greater than the differences to be observed.

No abstract

than reaction 3 must account for the high percentage of carbon monoxide found in experiments having a relatively high concentration of foreign gas. A reaction such as CHOCHO' -> CO + HCHO would meet the requirements but why it would predominate only at high pressures is not obvious. However, it is evident from the large variation of gaseous products that the primary process for glyoxal is the formation of an activated molecule which does not spontaneously dissociate into formyl radicals.From the limited results obtained, it appears that the same chemical processes occur at X 3660, 3130 and 2537, the variation being only one of degree.

No abstract