Define a finitary combinatorial principle to be a first-order sentence which is valid in the finite but falsifiable in the infinite. We aim to compare the strength of such principles over bounded arithmetics. We distinguish “weak” and “strong” principles based on their behaviour with respect to finite structures that are only partially defined. We show that over relativized T^1_2 “weak” principles do not imply “strong” ones. The proof applies a general forcing method to produce models of relativized T^1_2.