The strength of the relationship between the independent variables and the components of the PLS regression differed based on product type (bread crumb, bread crust, or cracker). Within bread crumb, good relationships were observed when both free and bound phenolic acids within the flour were included in the analysis (Q2cum = 0.545 and R2Ycum = 0.841). High Q2 and R2Ycum were also observed when both free and bound phenolic acids from the crumb of the bread were included in the regression analysis (Q2 cum = 0.401 and R2Ycum = 0.767). The bread crust, however, displayed low Q2 cum values, regardless of whether flour phenolics (Q2cum = 0.174) or crust phenolics (Q2cum = -0.029) were used in the regression. This indicates that the PLS model is not a good fit for the crust data and that the model has little (if any) predictive power. For this reason, the relations between phenolic acids and bread crust were not examined further. For crackers, the bound phenolic acids in both the flour and baked product produced the best model fit. It has been reported that phenolic acids can interact with Maillard reaction adducts, resulting in altered flavour and aromatic properties (Jiang and Peterson, 2010). Maillard reaction is limited within the crumb while the cracker and the bread crust undergo more Maillard browning during baking. This possibly explains why the free phenolic acids are correlated with the sensory properties in the crumb but not within the cracker or bread crust.