Levels of leaf herbivory (percentage area loss) in primary forest understory, a pioneer plot, and vegetation in forest gaps were compared at St. Elie, French Guiana. Leaf laminas were analyzed for nitrogen, total phenol, and condensed tannin concentrations, as bulked vegetation samples irrespective of species. Mature leaves of the shaded forest understory were more heavily (greater than threefold) grazed than those on pioneer trees and small trees which grow up from the saplings in lighted gaps, while young leaves did not differ significantly in levels of defoliation between vegetation types. Defoliation levels averaged 5.5 percent on mature leaves. Herbivory on mature leaves was significantly negatively correlated with total leaf phenols. In the context of the optimal defense hypothesis, these chemicals may be deterring herbivory on pioneer and gap leaves, but a more likely primary cause is that total phenols are higher in plants in direct light than in those in shade conditions. FEENY (1976) AND RHOADES AND CATES (1976) have put forward an optimal defense hypothesis which predicts that persistent, slow-growing, and apparent K-selected plants will defend their tissues against herbivory by investing in compounds of high molecular weight such as polyphenolics. These quantitative defenses will reduce the digestibility of the plant tissue by forming tannins which decrease the availability of nitrogen to the predator. Tannins would be expected to deter specialists, although generalists would tolerate this low-quality diet. In contrast, ephemeral, fast-growing, unapparent r-species such as pioneers, for which energy for competition and reproduction is supposedly at a premium, are expected to coevolve less metabolically costly defenses as toxins of low molecular weight (e.g., alkaloids). Since these latter, qualitative defenses require special detoxifying enzymes, they effectively deter all but the few herbivores specialized to circumvent the defenses of a particular species. To this must be added a temporal dimension: ephemeral plants are often difficult to locate by specialists, and the actual incidence of predation in nature may be kept low simply by this passive mechanism of escape in time (Fox 1981). A model for the coevolution between predators and plant defenses has been put forward by Rhoades (1979). In the tropical rain forest two distinct physionomic strategies have been observed (Richards 1952, Whitmore 1975). The first type consists of tree species which germinate in the shade and are shade tolerant in the early stages of their sapling growth. Apart from tree species which complete their life cyde totally in the undercanopy, these shade-tolerant species will respond to light later and grow up into the main canopy. The second type consists of species which germinate in the light and are light demanding in their growth. These conditions occur when one or more trees open a gap in the upper canopy, allowing, especially in the larger gaps, a pioneer flora to establish. Pioneers are relatively short lived compared with the main forest trees, are fast growing, vigorously competitive, and have a high reproductive output of widely dispersed seeds (Whitmore 1975). If natural selection has operated to produce two such different growth strategies, coadapted traits (Stearns 1977) which act as defenses to predation are expected in accordance with the optimal defense hypothesis. This paper reports a test of this combined hypothesis in the rain forest of French Guiana, South America, by comparing four different vegetation types: shaded primary understory saplings, saplings of the understory recently exposed to light in small and, separately, large gaps created by tree fall, and a wholly pioneer flora establishing in a large area cleared by man. Large and small gaps differ in that large gaps have a number of pioneer species present.