Condition monitoring is a technique of sensing the health of electrical machines. Analysis of the monitoring data quantifies the condition of the machine, so that faults can be detected and diagnosed at the incipient stage itself. Machine vibrations are known to carry information on most of the faults. Vibration response measurements yield a great deal of information concerning faults within rotating machines. Studies show that bearings are the major contributors to an onset of a fault. The aim of this paper is to compare the variations of a time domain signal of machine vibrations when the bearing is (a) healthy (fault free), (b) having a fault in the outer race, and (c) having a fault in the inner race. The statistical approach is used to investigate and classify the faults. The well-known frequency domain analysis is carried out to verify results of the statistical analysis. The proposed analysis technique implies a clear indication of the changes that occur even when a fault is at an incipient stage.