Flow cytometry has become an important tool for research in cell biology (4, 15, 18, 21, 23). An important application of flow cytometry is to determine the distribution of DNA content per cell in a population of cells at a very rapid rate and with good statistics. However, the important parameters usually desired-the proportion of cells in each phase of the cell cycle-cannot be accurately determined without using a computer to model the distribution. Various mathematical models have been proposed to accomplish this, with varying degrees of success. These have been reviewed recently (12).These models may be categorized by two basic conceptual approaches. One approach is to develop a function that will accurately represent the GI, S and Gz + M components of the DNA histogram derived from flow cytometry, and thus obtain the relative proportion of cells in each phase (1,2, 5, 7, 8, 9). The other approach is more basic in the sense that it models the cell age distribution and then transforms that distribution into a DNA distribution (10, 11,17,24), usually by matrix manipulation. This approach requires knowledge of the rate of DNA synthesis during S-phase. It does allow the capability of determining additional information, such as the phase times, and in fact the percentage of cells in each phase is not usually the primary information desired from these models.' Supported by Grant CA2563601 awarded by the National Cancer Institute. DHEW. 71 The model presented here is based on the f i s t approach.Dean and Jett (5) presented the first serious attempt to model accurately the DNA distributions of asynchronous cell populations. With such populations, their model gives excellent results. The model fits normal (Gaussian) distributions to the GI and Gz + M peaks, and a broadened second-order polynomial to S-phase. The broadening accounts for dispersion from a variety of causes, including instrumental, staining and biologic factors. The limitation of the model is that it can only fit distributions where a second order polynomial can accurately represent the S-phase data. This leaves out nearly all synchronous populations. Fried (7-9) presented another model in which a series of normal curves is fitted to S-phase. The advantage of this model is that it can fit synchronous populations, with varying degrees of success. However, the model is very sensitive to the resolution of the data. Furthermore, it requires considerable operator monitoring to obtain proper parameters (12). The main difficulty is in choosing the spacing of the S-phase compartments. If this is not chosen carefully, the resulting fit can have large fluctuations in the estimated number of cells from one compartment to the next. The spacing is correlated with the coefficient of variation of the GI peak (8). This is a rather arbitrary factor because ideally there should be one compartment per channel since DNA synthesis is a continuous rather than discrete process.