The ecological significance of differences in levels of toxicity as determined by laboratory toxicity bioassays was investigated for two species of marine sponges, the only sponges present in a survey of a coral reef lagoon off Yanutha Island, Fiji Islands. Endemic coral reef fishes that are not naturally sponge predators were maintained in aquaria and exposed to hand-squeezed (aqueous) sponge extracts. In the field freshly squeezed sponge extracts were released near lagoon coral heads occupied by reef fishes. Fish behavior was noted in both the laboratory and field bioassays. Extracts from the two sponges, Dysidea sp. and Carteriospongia sp., were significantly different statistically in the time of mortality of the laboratory test fishes. However, in field trials fishes simply avoided the sponge extract and suffered no visible adverse effects. The palatability of sponge tissue was compared to fish meat, bread, and bread soaked in hand-squeezed sponge extract in field feeding trials. Sponge tissue was never consumed; the bread and meat were always consumed; and, the soaked break was consumed to an intermediate degree. Bread soaked in Dysidea sp. aqueous extract was rejected approximately twice as often as bread soaked in Carteriospongia sp. aqueous extract. Individual fish appeared to learn from each other. A food item rejected by one fish was generally avoided by all other fishes. Ranking of organisms based on their toxicity may have little or at least ambiguous ecological meaning. Dysidea sp. had the following characteristics when compared to Carteriospongia sp.: 1) present at higher density; 2) greater coverage of substrate; 3) more toxic; and, 4) its aqueous extract made bread less palatable than bread soaked in the Carteriospongia extract. However, both sponges were completely unacceptable food items to fishes in field trials. We suggest that for prey with predators capable of relatively rapid learning, such as fish, palatability may be a more important defensive trait than toxicity. Further research utilizing sponge-feeding fishes is required to determine the relationship between toxicity, palatability, and susceptibility to predation for marine sponges.