The determinants of patterns of plant species composition on small mountains are poorly known, especially in Central Africa. We aimed here to identify variation in tree species composition throughout the Ngovayang Massif (southern Cameroon) and determine the relative contributions of environmental factors and spatial autocorrelation in shaping tree species composition. Vegetation surveys were conducted in fifteen 1-ha (100 m × 100 m) permanent plots established along a transect from lowland (200 m) to submontane forests (900 m) in which all trees with a diameter (dbh) = 10 cm were inventoried. Data were investigated using ordination methods (Correspondence Analysis and Canonical Correspondence Analysis). At the local scale, the most important variable in determining tree species composition patterns was slope exposure, followed by distance from the ocean and altitude. Together, these environmental variables explained 28% of floristic variation among plots, and the spatial structure almost disappeared when the effects of these variables were removed. Spatial autocorrelation analysis showed that spatial variables (geographic coordinates of the plots) or geographic distance between plots explained only 1% of the total initial variance. Residual spatial variation not explained by the environmental variables probably reflects the history of vegetation and the effects of other climatic variables that were not included in this study. Floristic variation in the Ngovayang Massif is due to strong environmental heterogeneity. The sensitivity of floristic composition to environmental variables such as slope orientation and altitude suggests that tree species composition may shift with expected climate changes, such as changes in the movement of air masses, increase in mean annual temperatures or increasing severity of the dry season. Our study highlights the need for systematic on-the-ground measurements of climate variables in tropical montane areas in order to better understand the current climate regime and serve as a basis for modelling future changes.