The design and objective of a community study imply the selection of the appropriate ordination technique in terms of species response models and weighting options. In this paper, we start from the observation that existing two-table ordination techniques and related measures of niche breadth inevitably weight a sample in proportion to its abundance. We introduce a new multivariate method, which gives a more even weight to all sampling units, including those which are species poor or individual poor. We use this new method of analysis which we call OMI (for Outlying Mean Index) to address the question of niche separation and niche breadth. The Outlying Mean Index, or species marginality, measures the distance between the mean habitat conditions used by species (species centroid), and the mean habitat conditions of the sampling area (origin of the niche hyperspace), and OMI analysis places species along habitat conditions using a maximization of their mean OMI. Therefore, the position of the species depends on their niche deviation from a reference, which represents neither the mean nor the most abundant species, but a theoretical ubiquitous species that tolerates the most general habitat conditions (i.e., a hypothetical species uniformly distributed among habitat conditions). We demonstrate that OMI analysis is well suited for the investigation of multidimensional niche breadths in the case of strong limiting factors (e.g., meteorological conditions) or strong driving forces (e.g., longitudinal stream gradient). Furthermore, the analysis helps in finding which ecological factors are most important for community structure and organization and provides a separation of species based on their niche characteristics.