Statistical modelling can be used to relate biological survey data to environmental factors, thereby providing a basis for predictive mapping of species or communities. However, there has been little discussion about the effect of scale on the predictive power of the variables used for species prediction. In this study, we analysed if the relative importance of environmental factors for the distribution of aquatic species was scale dependent, using data on the cover of five common benthic species (four macrophytes and one animal), from 1731 sites along the Swedish Baltic Sea coast. We modelled the cover and distribution of the five species in relation to salinity, depth, slope, wave exposure and substrate in scale steps from 25 to 1500 km, and analysed the relative contribution of the environmental variables to each species model. The average total deviance explained by the models was generally quite high, and decreased with increasing scale for all macrophyte species, while it increased for the animal, the Baltic Sea blue mussel, Mytilus edulis. The average contribution of salinity increased for all species when moving from local to Baltic Sea scale, and for the Baltic Sea blue mussel it was the single most important factor at the Baltic Sea scale. The average contribution of depth decreased with increasing scale for all species. However, regardless of scale, depth was the most important environmental factor to explain the distribution of all but one of the investigated macrophyte species. The relative contribution of different environmental variables changed with scale, and responses also differed between species. Factors measured on a fine scale, and thus describing local conditions were more influential at the local scale, whereas the large scale salinity gradient increased in importance with scale.