This dissertation develops theory and methodology for the evaluation and assessment of finite mixture models. New methods for simulating finite mixture models satisfying a pre-specified level of complexity defined through the notion of pairwise overlap, are developed. Corresponding software is publicly available at CRAN. This dissertation also develops methodology for assessing significance in finite mixture models with applications to model-based unsupervised and semi-supervised clustering frameworks. The dissertation concludes with an application of finite mixture models to two-dimensional gel electrophoresis.