Description: Aim Biodiversity dynamics comprise evolutionary and ecological changes on multiple temporal scales from millions of years to decades, but they are often interpreted within a single time frame. Planktonic foraminifera communities offer a unique opportunity for analysing the dynamics of marine biodiversity over different temporal scales. Our study aims to provide a baseline for assessments of biodiversity patterns over multiple time-scales, which is urgently needed to interpret biodiversity responses to increasing anthropogenic pressure. Location Global (26 sites). Time period Five time-scales: multi-million-year (0-7 Myr), million-year (0-0.5 Myr), multi-millennial (0-15 thousand years), millennial (0-1,100 years) and decadal (0-32 years). Major taxa studied Planktonic foraminifera. Methods We analysed community composition of planktonic foraminifera at five time-scales, combining measures of standing diversity (richness and effective number of species, ENS) with measures of temporal community turnover (presence-absence-based, dominance-based). Observed biodiversity patterns were compared with the outcome of a neutral model to separate the effects of sampling resolution (the highest in the shortest time series) from biological responses. Results Richness and ENS decreased from multi-million-year to millennial time-scales, but higher standing diversity was observed on the decadal scale. As predicted by the neutral model, turnover in species identity and dominance was strongest at the multi-million-year time-scale and decreased towards the millennial scale. However, contrary to the model predictions, modern time series show rapid decadal variation in the dominance structure of foraminifera communities, which is of comparable magnitude as over much longer time periods. Community turnover was significantly correlated with global temperature change, but not on the shortest time-scale. Main conclusions Biodiversity patterns can be to some degree predicted from the scaling effects related to different durations of time series, but changes in the dominance structure observed over the last few decades reach higher magnitude, probably forced by anthropogenic effects, than those observed over much longer durations.