In order to find an effective and convincing method to measure rock dynamic fracture toughness for mode I and mode II, cracked straight through flattened Brazilian disc specimens of marble, which were geometrically similar for three size, were diametrically impacted by split Hopkinson pressure bar on the flat end of the specimen with three load angle respectively. History of stress intensity factors (KI(t) for opening mode I, and KII(t) for sliding mode II), mode mixture ratio (KI(t)/KII(t)), as well as mode I and mode II dynamic fracture toughness at crack initiation (KId and KIId) were determined with the experimental–numerical method. It is found that there is a unique size effect for dynamic fracture test with the specimens, the mode mixture ratio is not solely determined by load angle (the angle between load direction and crack line) as in the static loading; the pure mode II load angle is 19° for the ϕ50mm specimen, however it is 10° for the ϕ130mm and ϕ200mm specimens; the mode II load angle decreases with increment of specimen size. Realization of pure mode II is justified by the mode mixture ratio approaching zero, it can be realized under certain load angle and loading rate for the specimen of specified size. KIId is generally greater than KId. Both KId and KIId increase with increment of specimen size, and this trend for KIId is more remarkable than that for KId.