This thesis challenges the existing composite design paradigm that prohibits the use of non-symmetric laminates. The restrictive nature of "symmetric-only" stacking sequences is identified, highlighting the potential benefit for structural optimisation on relaxing this constraint. The thermal response of such laminates during manufacture is investigated , noting layup induced variation of the coefficients of thermal expansion is often cited as a' major objection to their use. By extending known results for warp-free non-symmetric laminates, it is shown that flattening of (some) moderately warped plates induces nominal build strain . Allowing a tolerance to non-symmetry induced warping is presented in terms of increased design space for a specified level of induced strain. The direct benefits of non-symmetry, taking advantage of the coupling between in and out-of-plane responses, is also considered. The increased performance of non-symmetric configurations observed in this thesis counters the perceived wisdom that this coupling is always unfavourable. Specifically, moderately non-symmetric laminates indicate the potential to increase the load carrying capacity of stringer terminations. This was achieved by improving the resistance to debonding as investigated using both analytical and finite element techniques. To better understand the underlying mechanisms, and the effect of non-symmetry on stringer termination design, an optimisation framework is proposed. An established Ritz-Galerkin hybrid model is employed to predict structural response, in conjunction with analytical debonding and buckling constraints . A two-step process, making use of lamination parameters and genetic algorithms, is able to find optimised designs. Significant improvements over the quasi-isotropic configuration are observed including a preference towards non-symmetric configurations. Previously unknown limitations of the Ritz-Galerkin approach are identified during this study. The choice of basis functions a re discussed a long with the validity of implicit assumptions made regarding the geometric non-symmetry and energy formulation. A more general, non-dimensional, Ritz based model was proposed to mitigate these limitations. Although a general improvement is observed, there are further complexities in capturing localised behaviour due to stiffness discontinuities. A series of further refinements and investigations a re proposed as future directions of research. Finally, via the use of Legendre polynomials, it is possible to reformulate the Ritz method using the "triple-product" approach. Such an approach transforms the problem into a summation of standardised integrals containing three term products of the basis functions. Exploiting previously unused - in structural engineering - algebraic recursion relations a significant increase in computational performance can be realised . The number of numerical integrations that must be performed is shown to reduce by an order of magnitude. This result is significant for the analysis and optimisation of any variable stiffness plate/ shell-like structure where calculating these integrals often dominates the computational expense of analysis.