A seminal result of Thomassen [40] says that there are only finitely many 6-color-critical graphs for the bounded genus graphs. This result is no longer true if we consider K 3,k -minor-free graphs. K 3,k -minorfree graphs are a significant generalization of boundedgenus graphs. They also contain infinitely many t-colorcritical graphs for all t with k + 2 ≥ t ≥ 4.Motivated by this fact, we first show that if G is 6-color-critical 4-connected K 3,k -minor-free graphs, then G has tree-width at most g(k) for some function g of k. This allows us to show the following algorithmic result, which is of independent interest.For a 4-connected graph G and any k, there is an O(n 3 ) algorithm to test whether or not G is 5-colorable.Note that testing the 3-colorability of K 3,kminor-free graphs is NP-complete, and testing the 4-colorability of them would require a significant generalization of the Four Color Theorem (because K 3,3 -minor-free graphs are essentially planar). Testing the 5-colorability of bounded genus graphs can be done in polynomial time, as shown by Thomassen [40]. Thus our result can be viewed as a strengthening of the bounded genus case.We then investigate minimal forbidden minors in non-5-colorable K 3,k -minor-free graphs. Such a graph may exist, as K 6 shows. However, we prove the following result.There is a computable constant f (k) such that every minimal forbidden minor in non-5-colorable K 3,k -minorfree graphs has at most f (k) vertices.Our proof of the above result implies the following algorithmic result, which is of independent interest.For a graph G and any k, there is an O(n 3 ) algorithm to output one of the following:1. a 5-coloring of G, or * National Institute of Informatics and JST ERATO Kawarabayashi Project, 2-1-2, Hitotsubashi, Chiyoda-ku, Tokyo, Japan. Email address: k keniti@nii.ac.jp † Research partly supported by Japan Society for the Promotion of Science, Grant-in-Aid for Scientific Research and by Mitsubishi Foundation 2. a K 3,k -minor of G, or 3. a minor R of G of order at most f (k) (f (k) comes from the above theorem) such that R does not have a K 3,k -minor nor is 5-colorable.Let us emphasis that the chromatic number in our main results does NOT depend on the graph K 3,k that we are going to exclude as a minor. This is a big contrast with the algorithmic result of Hadwiger's conjecture [21].