as 53%/4 hr), but the temporal pattern was similar in both 

 habitats (Fig. 3). Detectable losses to herbivores occurred only 

 during the day. 



Grazing on young versus old segments 



When branches of H^ incrassata with young (<24 hr old) 

 apical segments were paired with branches having older ( >48 hr 

 old) apical segments and transplanted onto the reef slope, 

 significantly more segments were consumed from branches with 

 young apical segments than from branches with older apical 

 segments (p<.05, Wilcoxon Paired-Sample Test, Fig. 8). On 

 average, branches with new apical segments lost 39% more segments 

 than branches with old apical segments. When similar assays were 

 conducted using H^ simulans in the shallow, mangrove-lined bay 

 near the Hydrolab shore base, grazing by small fishes was 

 concentrated on the newer segments. Sixteen percent of the new 

 apical segments (6 of 38) showed grazing scars typical of small 

 parrotf ishes; none of the older apical segments (0 of 38) showed 

 similar bite marks. The newer segments were therefore 

 significantly more prone to attack by these small herbivores 

 (p=.013, Fisher's Exact Test). Time lapse movies made in this 

 area showed that almost all grazing was done by small brown and 

 white striped parrotfishes which our field observations indicated 

 were juvenile Scarus species. 



Halimedatrial as an herbivore deterrent 



When halimedatrial was coated onto Thalassia blades as 1% of 

 their wet mass (our initial HPLC-derived estimates indicated that 

 halimedatrial comprised 2-6% of the blotted wet mass of 

 developing tips) and these blades were transplanted onto the 

 reef, halimedatrial significantly reduced losses by 36% (Fig. 9) 

 relative to ether-coated controls (p=.005, Wilcoxon Paired-Sample 

 Test). Field observations showed that the Thalassia transplants 

 at this site were fed on primarily by the red band parrotfish 

 Sparisoma aurof renatum . When similar assays were conducted using 

 the red alga Acanthophora spicif era , it was fed on only by 

 surgeonf ishes, and halimedatrial reduced grazing by 47% (Fig. 9, 

 p=.025, Wilcoxon Paired-Sample Test). Following these field 

 assays, remaining portions of the coated plants were extracted 

 and analyzed by TLC. Halimedatrial was still present on the 

 plants and had not degraded to some other compound. Since TLC is 

 not a quantitative method, the concentration of the remaining 

 compound could not be determined. 



Direct observation of parrotfish feeding during these 

 experiments indicated that coated blades did not differ 

 significantly (.10>p>.05, Contingency Table Analysis) from 

 control blades in their probability of being bitten at least once 

 during the course of our assays (Table 2). However, once bitten, 

 a parrotfish was significantly (.025>p>.01, Paired-Sample t-Test) 



385 



