The theory of critical slowing down, i.e. the increasing recovery times of complex systems close to tipping points, has been proposed as an early warning signal for collapse. Empirical evidence for the reality of such warning signals is still rare in ecology. We studied this on Zostera noltii intertidal seagrass meadows at their southern range limit, the Banc d'Arguin, Mauritania. We analyse the environmental covariates of recovery rates using structural equation modelling (SEM), based on an experiment in which we assessed whether recovery after disturbances (i.e. seagrass & infauna removal) depends on stress intensity (increasing with elevation) and disturbance patch size (1 m2 vs. 9 m2). The SEM analyses revealed that higher biofilm density and sediment accretion best explained seagrass recovery rates. Experimental disturbances were followed by slow rates of recovery, regrowth occurring mainly in the coolest months of the year. Macrofauna recolonisation lagged behind seagrass recovery. Overall, the recovery rate was six times slower in the high intertidal zone than in the low zone. The large disturbances in the low zone recovered faster than the small ones in the high zone. This provides empirical evidence for critical slowing down with increasing desiccation stress in an intertidal seagrass system.