An equation which describes the effective tensile strength of fibers in a composite strand has been obtained from the consideration on a failure model assuming statistical accumulation of fiber fractures in a matrix with increasing load until a sufficient number of fractures occur at some crosssectional region of the strand. It is assumed that when a fiber in the composite strand breaks by increasing the load, a transmission of the stress along the fiber is to be intermitted by an ineffective length in the vicinity of the fracture point. The fibers in the composite strand are treated assuming a statistical strength distribution. It is also postulated that when a certain fiber in the composite strand breaks at a weak point of the fiber, some other fibers neighboring the fracture point break simultaneously and a total of k fibers break at the same time. The ineffective fiber length δ and the effective tensile strength σc, b of the fibers in the composite strand are given as and where α and σo are Weibull parameters of the tensile strength of single fibers, τb is the interface shear strength, Vf is the fiber volume fraction, Rf is the fiber radius, υm is the Poisson ratio of matrix, and Ef and Em are the tensile moduli of fiber and matrix, respectively. The value of k was evaluated from the experimental results for the carbon fiber composite strands comprising a series of matrix resins, the tensile modulus of which ranges from about 10MN/m2 to about 1, 000MN/m2. The value of k changed systematically and increased from k_??_1 to k_??_20 with increasing tensile modulus of matrix resin, Em. The change of the σc, b of carbon fibers with Em, which was observed experimentally, was rather small. The increase in Em decreases δ, and this should increase σc, b. By increasing Em, however, the carbon fibers in the composite strand become to be fractured into a micro-bundle, and this should decrease σc, b. It is considered that the change of the σc, b of carbon fibers with Em is virtually determined by these two, mutually opposed factors.