Our objectives were to compare the Couchman-Karasz-predicted glass transition temperature, TCK, to the measured glass transition temperature of a mixture, Tgm, of model confectionary systems and develop an empirical correction to improve the accuracy of prediction. Results showed that the original Couchman-Karasz equation fit the data better than the modified equation; although both generally overestimated Tgm. Blends containing sorbitol had the largest ΔTgmCK, where ΔTgmCK=TCK−Tgm. While Tgm varied with composition, the increase in Tgm with decreasing moisture content was linear (R2¯=0.984) and consistent across blends (4.5 ± 0.9°C/1% moisture, wb). The increase in Tgm with increasing cook temperature was best described by a polynomial model (R2¯=0.998), but adequately described by a more generalizable linear model (R2¯=0.979). Application of an empirical correction based on moisture content or cook temperature and TCK of dry ingredients reduced the average ΔTgmCK from 20.1 °C and 11.3 °C for modified and original equations, respectively, to <5.6 °C. •Original and modified Couchman-Karasz equations are applied to model confections.•Modified Couchman-Karasz equation overestimates measured Tg of samples.•Original Couchman-Karasz equation provides better fit to measured Tg data.•Tg decreases linearly with moisture content and increases with cook temperature.•Empirical corrections improve Couchman-Karasz accuracy to within 5 °C on average.