It has been assumed by many bacteriologists that during the period of rapid growth, in a satisfactory culture medium, some bacteria will die in spite of good food and favorable environment. No doubt this aumption was derived from an analogy with populations of higher forms of life, of which a number of individuals are known to die before they reach the reproductive age even with good care.Wilson (1922) approached this question experimnentally, and found that there was a regular discrepancy between plate counts and direct microscopic counts which he explained by supposing that some bacteria would not grow on agar but could still be seen. Reichenbach (1911), in order to explain the logarithmic order of death, assumed that a certain proportion of the multiplying bacteria of each new generation ceased multiplying and became dormant, and that the resistance of these dormant forms increased with age; thus a logarithmic gradation of resistance might be established.These assumptions have been tested by us in the following manner: Bacteria, and in later experiments yeasts, from young liquid cultures were s.Pread on the surface of an agar plate; from this seeded plate, a square was cut out and used as a hanging block in a moist chamber, for direct observation of multiplication under the microscope. The technique observed was essentially that of Orskow (1922), but the same field was kept under the microscope during the entire observation, which usually covered four generations. The moist chamber was held in a stage incu-

In a previous paper, the author (1929) has calculated the order of death for organisms which are killed by the reaction of just one molecule in the cell, and also for organisms which require the reaction of 2, 3, 4 and more molecules before the cell is injured beyond recovery. It was shown that higher organisms follow the order of death as calculated for several reacting molecules, while bacteria, as a rule, follow the order computed for one reacting molecule per cell.This can be best illustrated graphically by plotting the logarithms of the survivors on a standard time scale where 100 units represent the time necessary to kill 99.9 percent of the initial number of organisms. Fig. 1 shows a straight line when the reaction of one certain molecule in the organism causes its death; all other cases show a different type of curve, bulging out above the straight line.It has been further shown that the death rate of organisms computed from the formula 0.434 K --_1 log _a t b(where a is the initial number of organisms, and b the number of survivors after the time t) is constant/or one reacting molecule, but increases steadily if more than one molecule reacts. While bacteria often show a constant death rate and a straight line survivor curve, it happens frequently that the death rate decreases and the survivor curve sags below the straight line (see dotted line in Fig. 1). Both these deviations indicate that this cannot be caused by more than one molecule reacting because the effect would have been just the opposite.

The general belief that uniform cells under uniform conditions will all multiply at the same moment implies that the smallest units of the chromosomes, i.e., either the genes or the molecules of which the genes are composed, all double at exactly the same moment in all cells. Since the doubling of chromosomes is a synthetic chemical process, it seems more probable that it would follow chemical laws. With the assumption that the corresponding molecules in a number of uniform cells obey the mass law in their process of doubling, a definite order in the multiplication of identical cells is established which can be formulated mathematically for the simplest case. This is the same assumption which the author has used to account for the differences in the order of death between bacteria and higher organisms. This theory demands a great variability of the growth rate of uniform cells, so great that it must be experimentally measurable even for cells with a million molecules to the chromosome. The theory demands further that the frequency curve of cell divisions plotted for successive time intervals, be skewed to the left, and that the relative range of variation become smaller as the number of genes or gene-type molecules increases. Experiments on the growth rate of Bacterium aerogenes and Saccharomyces ellipsoideus showed regularly a frequency curve skewed to the left. The yeast had a relatively narrower range of variability than the bacterium. Even with multicellular organisms, theoretical calculations show a range of variation of the growth rate from the egg cell which should still be measurable though it decreases relatively with the number of cells produced. An experiment on the size of bacteria colonies at different ages of development agreed with the theory.