Caribbean reefs have experienced unprecedented changes in the past four decades. Of great concern is the perceived widespread shift from coral to macroalgal dominance and the question of whether it represents a new, stable equilibrium for coral-reef communities. The primary causes of the shift—grazing pressure (top-down), nutrient loading (bottom-up) or direct coral mortality (side-in)—still remain somewhat controversial in the coral-reef literature. We have attempted to tease out the relative importance of each of these causes. Four insights emerge from our analysis of an early regional dataset of information on the benthic composition of Caribbean reefs spanning the years 1977–2001. First, although three-quarters of reef sites have experienced coral declines concomitant with macroalgal increases, fewer than 10% of the more than 200 sites studied were dominated by macroalgae in 2001, by even the most conservative definition of dominance. Using relative dominance as the threshold, a total of 49 coral-to-macroalgae shifts were detected. This total represents ~ 35% of all sites that were dominated by coral at the start of their monitoring periods. Four shifts (8.2%) occurred because of coral loss with no change in macroalgal cover, 15 (30.6%) occurred because of macroalgal gain without coral loss, and 30 (61.2%) occurred owing to concomitant coral decline and macroalgal increase. Second, the timing of shifts at the regional scale is most consistent with the side-in model of reef degradation, which invokes coral mortality as a precursor to macroalgal takeover, because more shifts occurred after regional coral-mortality events than expected by chance. Third, instantaneous observations taken at the start and end of the time-series for individual sites showed these reefs existed along a continuum of coral and macroalgal cover. The continuous, broadly negative relationship between coral and macroalgal cover suggests that in some cases coral-to-macroalgae phase shifts may be reversed by removing sources of perturbation or restoring critical components such as the herbivorous sea urchin Diadema antillarum to the system. The five instances in which macroalgal dominance was reversed corroborate the conclusion that macroalgal dominance is not a stable, alternative community state as has been commonly assumed. Fourth, the fact that the loss in regional coral cover and concomitant changes to the benthic community are related to punctuated, discrete events with known causes (i.e. coral disease and bleaching), lends credence to the hypothesis that coral reefs of the Caribbean have been under assault from climate-change-related maladies since the 1970s.