This study focuses on the specific problems of protein extraction from recalcitrant plant tissues and evaluates several methods to bypass them. Sample preparation is a critical step in a two-dimensional gel electrophoresis proteome approach and is absolutely essential for good results. We evaluated four methods: the classical trichloroacetic acid (TCA)/acetone precipitation, TCA/acetone precipitation and fractionation, an alternative based on fractionation and without precipitation, and phenol extraction methanol/ammonium acetate precipitation. We optimized the phenol extraction protocol for small amounts of tissue, which is essential when the study material is limited. The protocol was optimized for banana (Musa spp.) and was subsequently applied to two other plant species: apple (Malus domestica L.) and potato (Solanum tuberosum L.). Banana (Musa spp.) is a good representative of a "difficult" plant species since it contains many interfering metabolites. Only classical TCA/acetone precipitation and phenol extraction methods proved useful as standard methods. Both methods are associated with a minor but reproducible loss of proteins. Every extraction method and the subsequent analytical procedure have their physicochemical limitations; both methods should be investigated before selecting an appropriate protocol. The study, which is presented in this paper, is useful for guiding the experimental setup of many other nonmodel species, containing various interfering elements.