Tensile strength models for unidirectional composites require scaling of fiber strength distributions to small length scales in the range of the fiber radius to the ineffective length (5–200 μm). No test methods currently exist to measure and validate fiber strength at these length scales. In this paper, a novel continuous fiber bending test method is developed to measure the size and spatial distribution of defects in S-glass fibers. For this test method, a single glass fiber was placed between Kapton film and tape, and the sample is subjected to a sequence of controlled radii of curvatures ranging from 350 μm to 25 μm in a confocal microscope. From the bending tests, the number of defects and spacings between the defects over 780 mm length of fiber are obtained for each level of flexural stress. The Kapton film confines the fiber fragments and maintains fiber alignment allowing the same sample to be tested at lower radius of curvature (higher flexural strain/stress). The associated defect size is calculated using fracture mechanics. These results are used to map the size- and spatial-distribution of critical defects on the tension surface of a glass fiber. A methodology is presented to map additional defects around the circumference of the fiber, while maintaining the size and spatial distribution obtained from the bending tests. The results show that the lognormal distribution best fits the spatial distribution of defects along the fiber axis. The results also indicate that extrapolating the Weibull strength distribution (be it bimodal or unimodal) to gage lengths less than 200 μm leads to serious errors in fiber strength prediction.