Amorphous polymers under their glass transition temperature (Tg) exhibit large inelastic deformations. Their mechanical behavior is highly dependent upon temperature, strain rate, pressure and loading mode (tension, compression, shear). They also exhibit small strain isotropic hardening, softening and large strain anisotropic rehardening. In addition, while in their glassy state, polymers are far from thermodynamic equilibrium so that their properties may change over time (physical aging). This complex behavior is reflected in the response of composites and affects the onset and propagation of damage therein. Therefore, in order to design polymer composite structures, it is fundamental to develop relevant tools and methodologies which aim at understanding, capturing and predicting the full thermomechanical response of glassy polymers. In this study, the thermomechanical behavior of a thermosetting polymer epoxy is characterized experimentally for temperatures below Tg. The intrinsic behavior of the polymer is obtained using a new methodology based on digital image correlation (DIC) in combination with video-monitored extensometry. In particular, inelastic flow localization patterns are discussed based on the full-field strain measurements and their connection to the stress-strain curves are highlighted. The Boyce-Parks-Argon polymer constitutive model, hereafter called the macromolecular model, has been enhanced to describe the thermomechanical behavior of epoxies. The identification of the material parameters involved in the model is described in a detailed procedure that builds on a limited set of experiments. The model is shown to represent adequately the thermomechanical behavior of the studied epoxy over a wide range of temperatures and strain-rates. Using additional high strain-rate data obtained from collaborators on Kolsky bars, the model capabilities are further discussed. Using finite-element implementations of the constitutive model in both quasi-static and dynamic codes, the processes of plastic flow localization are analyzed in tensile and compression specimens. Such analysis can form the basis of an alternative method for identifying the model parameters through inverse identification. Finally, a preliminary set of experiments were also conducted to investigate the effect of physical aging on the yield behavior and enhance the macromolecular model with the capability of modeling aging effects. Our interpretation of the aging experiments suggests that they are not conclusive and do not permit full determination of model parameters. Specific recommendations are tentatively formulated for conducting aging experiments in the future.