Despite the numerous experimental works on rolling contact fatigue, dealing with two-disk contacts, some phenomena still remain badly understood. Most of the test benches, used for that purpose, impose the rotational speeds to the disks: global slipping occurs and the tangential force is measured. Even if this configuration is found in some mechanical contacts, it does not reflect situations, where only microslipping occurs with high tangential loads. For these reasons, an original bench has been designed: a specimen disk rotates a braked stainless steel disk under a normal load . The tangential load , due to the braked disk, is set below the global slipping value; the specimen disks are transparent for the cracks observation and brittle to avoid any plasticity complication. A typical run consists in carrying out a succession of steps of increasing the number of cycles. Each step ends with several measurements on the cracks: their counting and their width and depth measurements. The results are divided in two categories: general observations and quantitative results. The most evident observation concerns the crack shape since it propagates along an ellipse on the contact path. Furthermore, the direction of propagation inside the disk is perpendicular to the surface. Lastly, a regular primary network of well-defined cracks is observed with cracks less marked. Concerning the effects of varying loads, the higher the , the faster the cracks initiate and propagate because of a higher tensile stress state. However, these effects can be partly overridden by beneath the contact path. As the disk material is brittle, the crack behavior is quite similar to the one observed on metallic specimens. Even if the results are obtained in an epoxy resin, a reasonable transposition is possible. The disk transparency makes it possible to quantify the cracks growth and to propose original 3D photographs of the cracks.