Most polymeric materials appear as complex mixtures of macromolecules characterized by distributions of specific properties that are essential to the quality of these products. Among such properties, the accurate determination of the glass transition temperature, and therefore, accurate representation of it, is a key issue. When analyzed using dynamic scanning calorimetry (DSC) techniques, many copolymers exhibit a wide range of temperature over which the glass transition takes place, and the width of the transition region is, therefore, not satisfactorily described by average T-g values, for example those computed from tangent curves drawn on thermograms. This article describes a method that allows us to characterize this spreading of the glass transition region by reconstructing weighted T-g distributions from DSC thermograms. As such an objective might appear as questionable from a strictly physical point of view, the significance of what is meant by "distribution" is specified in the text. A model is proposed that accounts for relaxation phenomena. The approach is validated by examining samples of BuA/Sty emulsion copolymers produced at different overall conversions and compositions, and examining the corresponding histograms of T-g were computed. The results show that accurate and consistent information on the glass transition behavior of the copolymer is obtained, and that the effective distribution is clearly connected with the composition drift in the polymer particles. The proposed algorithm allows one to obtain a maximum amount of information from DSC measurements, and provides a deeper insight into the "history" of complex polymer mixtures. (C) 2000 John Wiley & Sons, Inc.