Abstract. A main floristic coenocline is extracted from a Danish forest vegetation data set with Detrended Correspondence Analysis. Huisman-Olff-Fresco (HOF) modelling of 262 species occurrences relative to site scores along the coenocline is compared with Gaussian and Generalized Linear Modelling. HOF models allow for responses of variously skewed forms, which may result from interspecific interactions. Most species have symmetric (60%) or skewed (29%), and only few species have flat or monotone response curves. This provides evidence for the generality of unimodal plant species response curves, provided the gradient in question comprises a sufficient amount of compositional turnover. The Gaussian niche width is correlated with density of optima and probability of occurrence of the species. Detailed analysis of HOF estimates for 28 woody species indicates that early succession species such as Pinus sylvestris and Populus tremula have wider niches than middle to late successional canopy species such as Alnus glutinosa, Carpinus betulus, Fagus sylvatica, Fraxinus excelsior, Quercus robur and Tilia cordata. A marked overlap in niche range along the coenocline is interpreted in terms of species competition. Rubus idaeus intrudes into the niches of all other shrub species. Several canopy tree species, such as Fagus sylvatica, Fraxinus excelsior and Quercus robur overlap more than 60% of the ranges of all other species, while their internal overlaps are fairly equal, presenting evidence for their likely co-existence given more natural forest conditions. The combined analysis of coenoclines with HOF modelling of species behaviour is the most robust among the compared methods for estimating model parameters and niche characteristics of plant species. It is recommended for future ecological studies.