Roasted coffee samples of the two major trade species (Coffea arabica and C. canephora) were studied to identify sensory descriptors that might be used to determine blends production and evaluation, following the expectations of consumers. Coffee beans were roasted at 220 + 10 °C, for 7, 9, and 11 min, and the sensory profiles of the beverages were assessed. From descriptive analysis the eigenvalues allowed the identification of two principal components (PCs), being the variance between samples 68.9% and 21.1%. In the first PC the characteristic odor, astringency, body, bitter flavor, burned aroma, and residual, typical, and burned tastes prevailed. The correlation coefficient between the second PC and citric acid flavor and aroma reached 0.96 and 0.78, respectively. It was concluded that in beverages of these species, the descriptors of both components can be separated according to bean roasting time. Considering roasting time, the overall quality was also rated.