A fuel design being developed for the high-temperature gas-cooled reactor consists of microspheres (particles) of a very small kernel of dense, sintered, enriched 235UCO encapsulated by several layers of pyrolytic carbon and a layer of silicon carbide (SiC). The coated fuel particles are often called TRISO® particles. The SiC is derived via thermal decomposition of methyltrichlorosilane. This strong, dense layer is very important to the integrity of the particle and the retention of fission products. A fundamental understanding of failure mechanisms of unirradiated fuel particles is elucidated by measuring their failure rates when exposed to mechanical stresses. This was accomplished by compression testing of whole particles in two modes: (a) point loading and (b) dimple loading. Finite element stress modeling showed that point loading primarily exposed a small portion of the inner surface of the SiC layer to a maximum tensile stress. Stress analysis for the dimple loading showed that a significant area (inner and outer surface) of the SiC layer and a large volume of the SiC layer were stressed to near-maximum tensile levels. Various batches of archived particles were tested. Weibull methodology was used for analyses of failure statistics for groups of 500 particles. A scanning electron microscope was used for fractography, which identified critical flaws that were the likely fracture origins. The following is concluded from the strength tests. First, the dimple test yielded much lower strengths and different Weibull distribution curves than those resulting from the point-load test. This was attributed to a higher probability of finding flaws due to exposing more and different portions of the SiC layers to high stresses. Therefore, Weibull failure probabilities from the dimple test should give a more accurate prediction of in-service failures than point-load tests. The dimple test gave consistent and reproducible results. Second, fractography indicated that strengths were controlled by flaws, which were identified and categorized. Third, gold-colored spots were linked to large lenticular flaws oriented circumferentially in the SiC layer. These flaws were associated with diffuse iron impurities and silicon and carbon soot. The spots did not greatly affect the medium and high portions of the strength distribution because of their orientation to the tensile stresses. However, there was evidence that large gold spots were associated with a low-strength dog leg at low failure probabilities, and testing a minimum number of 1000 particle/batch is recommended to increase the confidence in fitting this portion of the probability plot. Fourth, compacting did not greatly affect the overall strength distribution of performance test fuel particles. Finally, burn-back and hydrofluoric acid etching procedures appear to accentuate the deleterious effect of some flaws.