Turbulent flows involving chemical reactions are a basic feature of many combustion and propulsion systems. The development of calculation procedures for turbulent reacting flows requires the understanding and modeling of the coupling between turbulence and combustion. Second-order closure modeling of turbulent flows provides a convenient framework for studying these interactions between turbulence and chemical reactions. Models for the scalar probability density function (pdf) have to be developed to achieve closure of turbulent transport equations for mixing and reacting flows. A delta function 'typical eddy' model has been developed for the joint pdf of the scalar variables. It has been demonstrated that delta functions are a necessary part of pdf's in order to attain the extremums of the statistical constraints on the moments. The statistical bounds on a number of moments of interest have been derived. It has been proven that a rational pdf composed of a set of delta functions alone can always be constructed at any point within the statistically valid moment space. The model provides a good representation of actual pdf's in two-species, variable-density mixing flows. The model has been directly compared to experimental pdf measurements and good agreement for higher-order moments has been demonstrated. It can be shown that the delta function pdf model is significantly simpler than other proposed pdf models and is more than adequate for the closure of the transport equations. (Author).