Abstract The general weakness of axles is within the region of the wheel fit near the inside hub face where axle fatigue fractures occur in service. This paper presents data from laboratory fatigue tests of full-size 11½-in. diameter driving axles to investigate such failures. A description and discussion of the suitability of the present design of axle-testing machine is submitted to indicate the reliability of the data obtained from this type of accelerated axle test for use in the design of railroad axles. These axles show fatigue cracks developing at such low nominally calculated bending stresses as 10,500 lb per sq in. The propagation of fatigue cracks is considerably retarded by rolling the axle wheel seat. Without rolling, the axle breaks off in 5¾ million revolutions (20,500 equivalent miles) at 19,000 lb per sq in., whereas, tested at the same stress, a similar axle, except rolled at the wheel seat, operated 25 times as long (144 million revolutions or 510,000 equivalent miles) with development of a crack 7/16 in. deep. Rolled axles compared with unrolled axles show little improved resistance to the development of incipient fatigue cracks as produced by a large number of repetitions of low stresses obtained under ordinary railroad operating conditions of speed and load. Under prevailing conditions of high speeds and load, however, the occasional stresses are considerably increased and cause these incipient fatigue cracks to propagate to breaking off of the axle. In this respect, rolling provides immense increase in resistance to these high occasional stresses to retard the crack propagation, thus giving greater safety. A correlation of the 11½-in. and the 2-in. axle tests is given.