Classic island biogeography theory predicts that very small islands, near the extreme lower end of the species–area relationship, should support very few species. At times no species may be present, however, due to randomness in the immigration–extinction dynamics. Alternatively, a lack of vegetation on very small islands may indicate that such islands do not contain the appropriate habitat for the establishment or long-term survival of plants, or that disturbances are too frequent or intense. These potential mechanisms were evaluated in the central Exumas, Bahamas, where surveys of 117 small islands revealed that over a third of the islands supported no terrestrial plant life. Area and exposure were significant predictors of whether a small island was vegetated or not in multiple logistic regressions. No islands naturally devoid of vegetation were colonized over a 17-yr period, and only two naturally vegetated islands lost all vegetation. Experimental introductions of two species –Sesuvium portulacastrum and Borrichia arborescens– revealed that a number of islands naturally lacking vegetation were able to sustain introduced populations over the long term (up to 15 yr). Drought and hurricanes appeared to have reduced the establishment success and possibly long-term survival of the introductions, although some populations survived four major hurricanes. Turnover rates of both introduced species were often an order of magnitude higher on the experimental introduction islands than on other islands in the archipelago. It appears many of the islands in this system that naturally lack vegetation may be physically capable of supporting terrestrial plant life, yet have no plants primarily due to barriers to colonization.