The classification of the character of cigarette brands was attempted using their smoke-component data. The samples were seven commercial cigarettes. Their semivolatile compounds were analyzed by GC. The peak area ratio to the internal standard peak were analyzed by the linear discriminant analysis program of the BMCP software program. The discriminant functions with ten semivolatile compounds completely classified the seven brands into their groups. The canonical discriminant analysis generated two equations to discriminate the seven brands using ten selected semivolatile compounds. The first and second canonical values of each sample were plotted on an X-Y graph. Three brands formed a triangle on the graph and the other four brands were distributed into the triangle. The peak area ratios of ten semivolatile compounds of thirty-six brands were analyzed by GC and the canonical values of each brand were calculated with the generated canonical functions. The canonical values were plotted on the graph. The samples were distributed on the graph with their leaf blend types.