This work presents a new approach to analyze a series of m n  p tables X(1); : : : ;X(m) of symbolic interval variables. In this new approach, we firstly define a space of intervals with laws of composition ; 1; . This allows extending this reasoning to the matrices of intervals. Then, we define a n  p compromise matrix X = 􀀀Xij
i=1;:::;n; j=1;:::;p; of type intervals, a measure of covariance between interval variables, a new measure of correlation  between interval variables and the product operator 2 between a matrix n  p of intervals and one p vector u. This way, we achieve a symbolic PCA of compromise. To express the variability of tables X(1); : : : ;X(m), they are projected on the principal axes of PCA of intervals of compromise. For the interpretation of factorial map, a new measure of correlation  will be used.