Optical coherence tomography (OCT) is an emerging medical imaging technology able to detect tissue microstructure in vivo and in situ. However, many changes, associated with diseases such as cancer, result in cellular and sub-cellular variations which are very important for the diagnosis but are below the resolution limit of OCT. Since the spectrum of scattered light is structure-dependent, the spectral content of OCT signals depends on such features and can be used to extract information otherwise unavailable in standard OCT images. Spectral analysis of OCT has resulted in improved contrast which corresponds to scatterer size changes. In this paper, novel spectral analysis techniques, by means of autoregressive spectral estimation, discriminant analysis, clustering, and scatterer size calculations, are presented. These procedures result in quantitative and significantly more accurate measurements compared to previous techniques. They were initially tested on samples of microspheres, successfully identifying each with sensitivity and specificity of >90% and yielding diameter estimates within 16% of the actual size. The techniques were also applied to in vivo images of xenopus laevis tadpoles and ex vivo neurological samples and provided increased contrast between different tissues. Such methods could provide a very useful procedure for the identification of subtle changes in OCT images and could make available a tool for differentiating tissues based on scatterer size. Such a tool could prove extremely valuable in the evaluation of disease features which now remain below the resolution limit of OCT.