Current micromechanical models of composite longitudinal compression strength are contradictory about the role of matrix strength and matrix modulus. A resolution among these models has been difficult because past experiments had difficulty controlling other potentially critical variables, such as fiber misalignment, fiber distribution, and so on, while changing the intended experimental variables. These difficulties were overcome in the present work by measuring thermoplastic composite compression strength as a function of temperature. Matrix properties are therefore changed while microstructural variables stay constant. Rate of change of matrix modulus and matrix strength with temperature were different, thereby making a resolution between the influence of these properties possible. Polycarbonate matrix composites with XAS and AS-4, sized and unsized, carbon fibers were tested. Both bending and direct compression (IITRI) test geometries were employed. Failure was by microbuckling. Composite longitudinal compression strength was found to depend proportionally on the matrix shear modulus. Matrix strength did not directly correlate to compression strength. The range of fiber/matrix interfacial strengths tested also did not influence the compression strength, suggesting that above an interfacial strength threshold this is not a critical variable in composite compression strength. Finally, using a recently developed technique for measuring small-angle fiber misalignment distributions, the data suggested that small random fiber misalignments on the order of 1 to 2 degrees do have a significant affect on compression strength.