were encountered. Under tank condibions the ycast was subject t o variation or degeneration and yielded substrains which did not produce riboflavin. A larger inoculum was required since this organism did not grow as rapidly as Candida guillieimondiu, Therefore a new medium was developed to obtain vigorous growth in the inoculum phase. The composition of t,his medium was as fo1lon.s: Sucrose 40 grams per liter Defatted soya flour (trypsin 10 grams per liter Calcium carbonate 5 grams per liter digested 2 hours a t 37' C.) Shake flasks containing 100 nil. of this medium ~w r e st>er and inoculated from slants. Owing to the high iron level of this medium no more than 100 ml. could be added to a 23-liter f c~ mentation. This volume of inoculum did not supp cells to combat contaniinati0n; therefore, an intermed inoculum stage was introduced. Six liters of niedimn 4B \vert; sterilized in a bottle fitted with a porous candle aerator and ari air filter. The urea n-as sterilized separately to avoid decomposition. -4fter cooling, the urea was added to the sterile medium and 107' by volume of 24-hour soya medium shake culture was added as inoculum. During 20 to 24 hours' growth at) 30" C. under vigorous aeration, the pH fell from 6.2 to 5.0 or loircr. Five per cent of this culture TWS then used to inoculate 24 liters of nonsterile medium 4B. Using t,his technique, pot,encies as high as 3257 per nil. were obtained, but in some runs the above yields n-ere not obtained for unknown reasons. A summary of data from one experiment involving t,hree tanks using the above technique is given in Figure 2. The curve of riboflaviii production is closely related to the yeast growth curve.I t was felt that a possible csplanat,ion for the occasional failures in the tanks might be due to the sterilized medium used in the secpnd stage of inoculum culture. -4 new technique was developed which eliminated the usc of st>erilc medium. -4 sinal1 volume (1 to 3 liters) of nonsterile medium 413 mas inoculated with 100 ml. of soya medium culture and acratcd as 1x;fore. B s the growth increased, nonsterile medium 4B was added to maintain a constant but high ratio of cells to volume. When a volume of 6 liters was reached (pH 4.0 to 5.5), the entire stagc was used as inoculum for 18 liters of nonsterile medium 4B.Using this procedure, a series of 81 experimental 24-lit,er tanks were run in order to determine the average yield and reproducibility of this method. Of these 81 tanks, 03 ivere not contaniinat,ed, while 16 tanks viere low in potency due to coiitarnination and 2 were low due to the failure of the agitation system. The average riboflavin potency of all t,hc t,anlrs was 1617 per ml.The average potency of the 63 iioncontaminated tanks was 1917 per nil. The distribution of riboflavin yields among these 63 tanks is tabulated below: 01-er 300-, por ml. Between 200 and 3007 per inl. Between 100 and 2 0 0 y per ml.