Background: The classification systems for fractures of the proximal part of the humerus provide low interobserver and intraobserver reliability when radiographs or two-dimensional computed tomography scans are used. The purpose of this investigation was to determine whether the use of three-dimensional computed tomography scans could improve interobserver and intraobserver reliability of AO/OTA (Arbeitsgemeinschaft fur Osteosynthesefragen/Orthopaedic Trauma Association) and Neer classifications and treatment recommendations. Methods: Two trauma surgeons, one shoulder surgeon, two senior orthopaedic residents, and two junior orthopaedic residents reviewed the radiographs and two and three-dimensional computed tomography scans of forty fractures of the proximal part of the humerus. Each imaging modality was reviewed in isolation, and fractures were classified according to the Neer and AO/OTA classifications and treatment recommendations were provided. This process was repeated for intraobserver analysis. Interobserver agreement was calculated within and between levels of training for each classification and treatment recommendation with respect to radiographs and two and three-dimensional computed tomography scans. Results: Among attending orthopaedic surgeons and senior residents, the use of three-dimensional computed tomography did not improve agreement compared with the use of two-dimensional computed tomography for the Neer classification based on planes, the AO/OTA classification, or the treatment recommendation, but it did improve agreement among junior residents. Comparing between levels of training, three-dimensional computed tomography increased agreement only between junior residents and more experienced reviewers for the Neer classification based on planes and for the AO/OTA classification but not for the treatment recommendation. Intraobserver agreement for each reviewer for classification and treatment ranged from slight to fair and was not improved through the use of three-dimensional computed tomography. Conclusions: In this investigation, the use of three-dimensional computed tomography imaging did not offer improved interobserver and intraobserver agreement compared with the use of two-dimensional computed tomography imaging with regard to classification and treatment of fractures of the proximal part of the humerus, except among reviewers with limited clinical experience.