Appendix A. Supplemental materials and methods for assessing seed viability following germination trials and measuring seedling growth rates as well as supplemental results for community and specieslevel analyses.
Supplemental Materials and Methods
Seed Viability following Germination. Following germination trials, seeds that did not germinate were opened to determine the status of the embryo, which was characterized as viable, dead, or with damage clearly attributable to insects or fungi. Because many of the seeds at the end of the germination trials were rotten, there were many dead seeds with no known cause of death that could have been killed by either insects or fungi. Predation clearly attributable to insects was recorded if larvae, frass, or emergence holes were observed, and damage clearly attributable to fungi was recorded if fungal hyphae were observed either internally or externally. Both internal and external fungal damage were included because observed fungus outside of the seed may include fungus that colonized seeds during the period seeds were germinating in the greenhouse, and may therefore not reflect fungal infection at the predispersal stage.
Seedling Growth Rates. For Anacardium excelsum and Castilla elastica, the relative growth rates of seedlings were measured to determine whether vertebrates, invertebrates, and pathogens at the predispersal stage affect the seedling stage. For the seedling study, germinating seeds were randomly chosen within each treatment. Sample sizes differed per species and depended on the number of viable seeds per treatment. For C. elastica, 38–70 individuals were randomly chosen per treatment, and for A. excelsum, 13–29 individuals were chosen. Initial measurements of seedling height, leaf number, leaf length and width were taken when first leaves were fully expanded for C. elastica and one week after leaf expansion for A. excelsum. Thirty seeds from nontreated branches were grown and harvested at the same time to determine linear relationships between initial measurements and initial biomass and leaf areas.
Seedlings from treatments were transplanted into 4" x 6" pots filled with soil collected locally from Gamboa and maintained in the growing house. Soil was collected from the first 30 cm, sieved to remove large rocks, roots, and clay clumps, and autoclaved for 1 hour. Seedlings were watered to field capacity and rotated three times a week to ensure similar light conditions. After six weeks, seedlings were harvested and seedling height, leaf area, and biomass were measured. At time of harvest, roots of seedlings had not filled the pots. Height was measured from the base of the stem to the apical meristem. To measure leaf area, leaves were placed in plastic bags with moist paper towels, transferred to Barro Colorado Island in a cooler, and measured with a LICOR Leaf Area Meter (ModelLI3100). Leaves, stem, roots, and remaining cotyledons were dried at 60°C until constant weight and final biomass was determined.
Statistical Analyses. We use a general linear mixed model (GLMM) with binomial errors to analyze variation in the proportion of seeds (of the total seeds planted) with damage clearly attributable to insects or fungi in response to vertebrate, invertebrate, and pathogen removal treatments. Analyses follow statistical analyses of variation in germination (see text for further details). We used Pearson’s productmoment correlation test to analyze whether the proportion of seeds with internal and external fungal damage were correlated and aggregated seeds by branch. For some species, damage clearly attributable to insects or fungi was not detectable either because the seeds were too small or because they were too decayed at the end of the experiment; these are noted in the specieslevel results below.
Relative growth rate of seedlings was calculated as RGR = (ln W_{2} – ln W_{1})/(t_{2} – t_{1}), with separate calculations based on height (RGR_{height}), biomass (RGR_{biomass}), and leaf area (RGR_{leaf area}). We estimated seedling biomass from height at the first time step using a linear regression of log_{10} (initial biomass) versus log_{10} (height) from initial harvests (Castilla elastica: a (intercept) = 2.9153, b (slope) = 0.8997,r^{2} = 0.53; Ancardium excelsum: a = 2.1714, b = 0.9395,r^{2} = 0.46). Leaf area at the first time step was estimated from leaf length and width based on a linear regression of leaf area and the product of leaf length and width from initial harvests (C. elastica: a = 0.1789, b = 0.0057,r^{2} =0.88; A. excelsum: a = 0.9979, b = 0.0063,r^{2} =0.98). To determine differences in relative growth rate among treatments, we used an ANOVA, including treatment, a covariate to control for initial size, and their interactions as explanatory variables. In C. elastica, the covariate was diaspore mass, and for A. excelsum, we used initial seedling height. For C. elastica, we randomly selected 38 individuals from each treatment to achieve equal sample sizes; this gave similar results to the full dataset, and therefore the full dataset was used.
Results
Below we detail results of analyses on fruit development, seed germination, seedling growth rates, insect seed predation, and fungal damage for each species (Table A6). For each species, standardized diaspore mass did not differ among treatments.
Cecropia peltata
Vertebrate exclosures significantly increased final fruit length relative to the controls (t = 2.55, p < 0.05) while insecticide or fungicide treatments had no significant effect (Fig. 2, Table A6). There was a significant positive effect of initial fruit length on final length (t = 3.39, p < 0.05). Vertebrate exclosures reduced germination success (z = 1.74, p = 0.0816), while the insecticide and fungicide treatments had no effect (Fig. 3, Table A6).
Insect seed predation was not detectable. Vertebrate exclosures marginally increased external fungal damage (z = 1.79, p = 0.0734). There were no significant treatment effects on internal fungal damage of C. peltata seeds. The proportion of total seeds with external vs. internal fungus was significantly correlated (t_{17} = 3.56, p < 0.01, r = 0.65, 95% CI = [0.28,0.85]).
Luehea seemannii
Vertebrate exclosures significantly increased the probability of fruit maturation (z = 1.98, p < 0.05), insecticide marginally increased fruit maturation (z = 1.65, p = 0.0987), and fungicide did not affect development (Fig. 2, Table A6). Insecticide (z = 2.06, p < 0.05) and fungicide (z = 1.72, p = 0.0861) reduced the proportion of seeds that germinated before the breaking of dormancy. Adjusting for diaspore mass, fungicide had a marginally negative effect on germination (z = 1.82, p = 0.0695), while germination increased with standardized diaspore mass (SDiaM; z = 2.1, p < 0.05, Table A6). Following the hot water treatment, treatments did not affect germination (Fig. 3), however germination significantly increased with SDiaM (z = 4.84, p < .001; Table A6). The difference in treatment effects before and after the hot water treatment suggests that pesticides may inhibit germination of L. seemannii seeds, potentially due to some chemical properties of the pesticides or because natural enemies may break mechanical dormancy. SDiaM did not vary among treatments (mean of diaspore mass + 1 SD: 2.3 mg + 0.4).
L. seemannii externalfungal attack was very low and not analyzed (Table A1). Insect seed predation and internal fungal damage were not detectable.
Antirhea trichantha
In 2008, the fungicide treatment significantly reduced the probability of immature fruits reaching maturity (z = 2.61, p < 0.01; Fig. 2), while the other treatments did not affect development. Insecticide significantly increased germination (z = 2.14, p < 0.05) and the probability of germinating was significantly higher in 2008 compared to 2007 (z = 4.17, p < 0.001; Table A6).
In 2008, diaspore mass was also measured for A. trichantha. The main effects of treatments and SDiaM did not significantly affect germination, but there was a marginally significant negative interaction between vertebrate exclosures and SDiaM (z = 1.96, p = 0.0503; Table A6). SDiaM did not differ among treatments (mean of diaspore mass + 1 SD: 16.4 mg + 5.6). In 2008, 8.4% of A. trichantha seeds were observed germinating from the second locule after germinating once. The probability of germinating a second time significantly increased in the insecticide treatment (z = 2.38, p < 0.05), while SDiaM, fungicide, and vertebrate exclosures had no effect (Table A6).
Natural enemy removal treatments and SDiaM had no affect on external fungal damage (Table A6). Insect seed predation and internal fungal damage were not detectable.
Stigmaphyllon hypargyreum
The fungicide treatment reduced the probability of fruit development (z = 2.05, p < 0.05), while the vertebrate exclosure and insecticide treatments did not affect the probability of fruit maturing during development (Fig. 2). The insecticide treatment significantly increased germination (z = 4.34, p < 0.0001), while fungicide and vertebrate exclosures had no effect (Fig. 3).
The insecticide treatment tended to reduce the number of seeds with damage clearly attributable to insects (z = 1.79, p = 0.073) and external fungal damage compared to controls (z = 1.67, p = 0.0957; Table A6). There was no effect of treatments on internal pathogen damage. The proportion of seeds with internal damage and external fungal growth were not correlated (t6 = 1.30, p > 0.05, r = 0.47, 95% CI = [0.35, 0.88]).
Bonamia trichantha
Treatments did not affect the probability of fruit maturation or germination (Fig. 2, 3); however, the vertebrate exclosures reduced the proportion of healthy versus aborted diaspores per fruit (z = 2.67, p < 0.01; Table A6).
Overall, damage clearly attributable to insects of germinating seeds was low (Table A1), and thefungicide treatment marginally increased insect seed predation (z = 1.71, p = 0.087; Table A6). Treatments had no effect on external fungal damage (Table A6), and internal fungal damage was not detectable.
Castilla elastica
The main effects of natural enemy removal treatments did not affect probability of development or germination, but germination significantly decreased with SDiaM (z = 4.14, p < 0.001; Table A6). In germination analyses, there was a significant positive interaction between SDiaM and the fungicide (z = 5.23, p < 0.001) and a marginally significant positive interaction between SDiaM and vertebrate exclosures (z = 1.76, p = 0.0782). SDiaM did not differ among treatments (mean of diaspore mass + 1 SD: 376.5 mg + 103.7).
Of the seeds that recruited into seedlings, diaspores were an average 12% heavier in the fungicide treatment compared to controls (t = 2.64, p < 0.01), whereas the other treatments did not differ from the control. RGR_{height} tended to increase with the fungicide treatment (t = 1.74, p = 0.0844) and significantly so with diaspore mass (t = 2.04, p < 0.05). Seed mass significantly increased RGR_{biomass} (t = 6.12, p < 0.001) with no effect of treatment. There were no significant effects of treatments on RGR_{leaf area} (Table A6)
Insecticide significantly reduced detectable insect seed predation (z = 1.96, p < 0.05). When including SDiaM in the model available for half of the seeds studied, insecticide marginally reduced insect seed predation (z = 1.65, p = 0.0996; Table A6).
Treatments did not affect damage clearly attributable to internal fungi, while insecticide (z = 1.90, p = 0.0573) and vertebrate exclosures (z = 2.14, p < 0.05) reduced external fungal damage. When including SDiaM in the analyses, it significantly increased with external (z = 1.99, p < 0.05) and internal (z = 3.55, p < 0.001) fungal damage. In this model, vertebrate exclosures tended to reduce both fungal external (z = 2.49, p < 0.05) and internal (z = 1.88, p = 0.0607) damage, and insecticide tended to reduce external fungal damage (z = 1.90, p < 0.0581). The proportion of seeds with external fungus and internal damage was significantly correlated ( t_{19} = 2.92, p < 0.01, r = 0.56, 95% CI = [0.16, 0.80]).
Anacardium excelsum.
All enemy exclusion treatments significantly increased the probability of fruit developing to maturity (fungicide: z = 2.07, p < 0.05, insecticide: z = 2.74, p < 0.01, vertebrate exclosure: z = 2.69, p < 0.01; Fig. 2). The fungicide significantly increased germination (z = 2.14, p < 0.05), and the vertebrate exclosures marginally increased germination (z = 1.82, p = 0.0685), with no effect of the insecticide treatment (Fig. 3).
Initial seedling height was marginally lower in the insecticide treatment than controls (t = 1.69, p = 0.0952). RGR_{height} and RGR_{mass} was higher in the vertebrate exclosures (p < 0.05), and there was a significant negative interaction between the exclosures and initial height (p < 0.05). RGR_{leafarea} was logtransformed to meet assumptions of normality. Logtransformed RGR_{leafarea} was marginally higher in vertebrate exclosures (t = 1.79, p = 0.0774), and there was a marginally significant negative interaction between vertebrate exclosures and initial height (t= 1.69, p = 0.0953).
Insect seed predation was low overall (Table A1), and no insect seed predation was observed in the insecticide treatment. Fungicide and insecticide significantly reduced the proportion of seeds with damage clearly attributable to internal fungi, while vertebrate exclosures had a marginally significant reduction (fungicide: z = 2.05, p < 0.05, insecticide: z = 2.09, p < 0.05; vertebrate exclosures: z = 1.71, p = 0.0875). Treatments did not affect external fungal damage. The proportion of seeds with external fungus and internal damage was significantly correlated across treatments (t_{26} = 3.66, p < 0.01, r = 0.58, 95% CI = [0.27, 0.79]).
TABLE A1. Summary data for each study species.
Species 
Nu. trees 
Nu. branches 
Nu. fruit 
Total maturation (%) 
Length of fruit development experiment (days) 
Nu. seeds 
Total germination (%) 
Length of germination trials (days) 
Total insect predation (%) 
Total external fungal damage (%) 
Total internal fungal damage (%) 

Cecropia peltata 
3 
20 
225 
54 
116 
1621 
77 
90 
_ 
23 
14 

Luehea seemannii 
3 
28 
1668 
89 
102 
1637 
86 
77 
_ 
0.4 
_ 

Antirhea trichantha 
(2007) 
3 
24 
NA 
NA 
64 
1094 
34 
110 
_ 
_ 
_ 
(2008) 
2 
15 
1957 
52 
184 
925 
61 
299 
_ 
6 
_ 

Stigmaphyllon hypargyreum 
1 
8 
442 
60 
91 
773 
13 
251 
11 
13 
13 

Bonamia trichantha 
1 
14 
995 
67 
84 
955 
85 
86 
7 
12 
_ 

Castilla elastica 
3 
24 
1060 
76 
69 
1015 
57 
75 
6 
16 
20 

Anacardium excelsum 
3 
29 
1688 
18 
103 
321 
73 
216 
3 
22 
24 
TABLE A2. Correlations between morphological fruit traits (^ P < .1, * P < .05, ** P < 0.01, *** P < 0.001)

Capsule: fruit ratio 
Protective structure: diaspore ratio 
Log (fruit dry mass) 
Log (fruit length) 
Log (fruit width) 
Log (seed reserve dry mass) 
Log (no. seeds per fruit) 
Pulp: fruit ratio 
0.427 
0.481 
0.097 
0.131 
0.325 
0.701^ 
0.508 
Capsule: fruit ratio 

0.028 
0.187 
0.394 
0.199 
0.050 
0.170 
Protective structure: diaspore ratio 


0.706^ 
0.444 
0.951*** 
0.663 
0.071 
Log (fruit dry mass) 



0.779* 
0.643 
0.148 
0.658 
Log (fruit length) 




0.451 
0.202 
0.806* 
Log (fruit width) 





0.609 
0.047 
Log (seed reserve dry mass) 






0.632 
TABLE A3. Summary of generalized linear mixed model for fruit development probability in response to natural removal treatment. The intercept is the mean (log of odds ratio) of the control group when each principal component is zero. Coefficients of the treatments are differences from the control group with all PC’s at their mean, zero. The coefficient of each principal component covariate describes the change in the response with one unit change in the PC for the control group. Coefficients of interactions between treatments and PC’s describe the differences in the slopes between treatments relative to the control in response to each PC. In bold are p values significant at the 0.1 level.
Variable 
Estimate 
Std. Error 
z value 
p value 
Intercept 
1.2210 
0.3095 
3.945 
<0.0001 
Fungicide 
0.2562 
0.2485 
1.031 
0.3025 
Insecticide 
0.2978 
0.2446 
1.218 
0.2234 
Vertebrate exclosure 
0.0086 
0.2407 
0.036 
0.9715 
PC1 
0.0067 
0.1325 
0.051 
0.9597 
PC2 
2.0542 
0.4259 
4.823 
<0.0001 
PC3 
0.1097 
0.2234 
0.491 
0.6234 
Fungicide: PC1 
0.3217 
0.1075 
2.993 
0.0028 
Insecticide: PC1 
0.1062 
0.1030 
1.031 
0.3024 
Vertebrate Exclosure: PC1 
0.1502 
0.1010 
1.486 
0.1372 
Fungicide: PC2 
0.1801 
0.3255 
0.553 
0.5801 
Insecticide: PC2 
0.2825 
0.3229 
0.875 
0.3817 
Vertebrate Exclosure: PC2 
0.3862 
0.3213 
1.202 
0.2293 
Fungicide: PC3 
0.2146 
0.1699 
1.263 
0.2065 
Insecticide: PC3 
0.1018 
0.1674 
0.608 
0.5432 
Vertebrate Exclosure: PC3 
0.3097 
0.1644 
1.884 
0.0596 
TABLE A4. Summary of generalized linear mixed model for seed germination probability in response to natural removal treatment; interpretation as for Table A3. In bold are p values significant at the 0.1 level.
Variable 
Estimate 
Std. Error 
z value 
p value 
Intercept 
0.7880 
0.2917 
2.702 
0.0069 
Fungicide 
0.1898 
0.2848 
0.666 
0.5052 
Insecticide 
0.4191 
0.2865 
1.463 
0.1435 
Vertebrate exclosure 
0.1792 
0.2802 
0.639 
0.5226 
PC1 
0.1120 
0.1548 
0.723 
0.4695 
PC2 
0.3598 
0.2123 
1.695 
0.0900 
PC3 
0.6346 
0.2406 
2.638 
0.0083 
Fungicide: PC1 
0.3539 
0.1514 
2.338 
0.0194 
Insecticide: PC1 
0.0936 
0.1595 
0.586 
0.5576 
Vertebrate Exclosure: PC1 
0.2189 
0.1542 
1.420 
0.1557 
Fungicide: PC2 
0.0829 
0.2207 
0.376 
0.7071 
Insecticide: PC2 
0.2556 
0.2088 
1.224 
0.2208 
Vertebrate Exclosure: PC2 
0.3527 
0.2069 
1.705 
0.0882 
Fungicide: PC3 
0.2140 
0.2264 
0.945 
0.3447 
Insecticide: PC3 
0.2256 
0.2331 
0.968 
0.3332 
Vertebrate Exclosure: PC3 
0.1538 
0.2264 
0.679 
0.4969 
TABLE A5. Summary of linear mixed model for seed survival probability integrated over fruit development and seed germination in response to natural removal treatment; interpretation as for Table A3. In bold are p values at the 0.05 significance level.
Variable 
Estimate 
Std. Error 
t value 
p value 
Intercept 
0.6781 
0.0630 
10.772 
< 0.05 
Fungicide 
0.0084 
0.0513 
0.164 
> 0.05 
Insecticide 
0.0523 
0.0512 
1.023 
> 0.05 
Vertebrate exclosure 
0.0441 
0.0494 
0.892 
> 0.05 
PC1 
0.0407 
0.0337 
1.209 
> 0.05 
PC2 
0.1515 
0.0454 
3.334 
< 0.05 
PC3 
0.1438 
0.0528 
2.725 
< 0.05 
Fungicide: PC1 
0.0988 
0.0286 
3.457 
< 0.05 
Insecticide: PC1 
0.0455 
0.0290 
1.566 
> 0.05 
Vertebrate Exclosure: PC1 
0.0608 
0.0277 
2.194 
< 0.05 
Fungicide: PC2 
0.0501 
0.0392 
1.279 
> 0.05 
Insecticide: PC2 
0.0759 
0.0373 
2.035 
> 0.05 
Vertebrate Exclosure: PC2 
0.0160 
0.0371 
0.433 
> 0.05 
Fungicide: PC3 
0.0024 
0.0421 
0.057 
> 0.05 
Insecticide: PC3 
0.0085 
0.0430 
0.198 
> 0.05 
Vertebrate Exclosure: PC3 
0.0060 
0.0407 
0.147 
> 0.05 
TABLE A6. Coefficient estimates (standard errors) for specieslevel analyses. Species are ordered in increasing seed size. Coefficients in bold are significant at the 0.1 level. Abbreviations for types of fungal damage: in = internal, ex = external.