Grime's C-S-R triangle theory has been discussed in plant ecology for two decades, but it has rarely been tested, and not often dispassionately evaluated. We consider the theory from a community viewpoint, and attempt to develop and test predictions for plant communities. C-S-R assumes that in high-disturbance (ruderal, R) patches or habitats, competition will be absent, or low in intensity. Testing this is problematic because of the difficulty of defining the intensity of competition, and we could find no rigorous evidence to support or refute the prediction. The theory also implies that in high-disturbance habitats there will be no difference in species composition between ‘competition’ and ‘stress’ sites, but from available evidence this does not seem to be true. C-S-R assumes that in stressful (S) habitats, competition will be low. This assumption is difficult to define or test, because of the overall difference in plant growth rate between habitats. A prediction from the theory is that in stressful habitats autosuccession should occur, i.e. that the climax species should regenerate directly, with no specialist secondary pioneer (R) species. There is some evidence that autosuccession occurs under the most extreme stresses of various types. Previous criticisms of C-S-R are evaluated. Only a few are considered valid, mainly those that emphasise that C-S-R theory is a considerable simplification of reality. Previous tests of C-S-R theory appear to be inconclusive. C-S-R theory is basically a combination of r/K theory and Leaf Amortisation theory. We conclude that there is limited support for the C-S-R theory. Whether the gain in generality that the theory offers justifies the loss via simplification that it involves is still an open question. As formulated, it has limited utility as a predictive model in community ecology. Yet, it is currently the most comprehensive and coherent theory for community ecology.