Ecology, 85(2), 2004, pp. 446±457 ᭧ 2004 by the Ecological Society of America

BENEFICIAL EFFECTS OF FLOWER-DWELLING PREDATORS ON THEIR HOST

GUSTAVO Q. ROMERO1 AND JOAÄ O VASCONCELLOS-NETO Departamento de Zoologia, Instituto de Biologia, Universidade Estadual de Campinas (UNICAMP), C.P. 6109, Campinas, SP, 13083-970

Abstract. We examined the effects of the sit-and-wait spider Misumenops argenteus (Thomisidae) on the herbivore assemblage and ®tness of the perennial woody shrub Tri- chogoniopsis adenantha (). Because crab spiders prey on both pollinators and phytophagous , they can have potentially negative and positive effects on . In a manipulative experiment using paired plants, spiders decreased the density of sucking and some endophagous herbivores on the leaves and capitula and reduced the number of damaged achenes produced by the plants. Damaged capitula had a higher proportion of fertilized achenes in plants with spiders than without spiders, but not undamaged capitula. These results indicate that M. argenteus exerted a double positive effect on seed production in T. adenantha. The effect of M. argenteus on herbivores may be taxon speci®c and vary among years with different herbivore abundances. Key words: Asteraceae; direct and indirect effects; ¯oral herbivory; phytophage assemblage; plant ®tness; pollination; seed predators; southeastern Brazil; thomisid spider; Trichogoniopsis ad- enantha; tri-trophic interactions.

INTRODUCTION Rypstra 1995, Snyder and Wise 2001, Denno et al. 2002, Moran and Scheidler 2002), and indirectly en- Species that comprise a community may be linked hance plant ®tness (Ruhren and Handel 1999, Snyder directly or indirectly through interactions between re- and Wise 2001). However, other studies have shown sources and consumption (Polis and Winemiller 1996). that spiders or arti®cial models of spiders can prey upon In systems involving three trophic levels, such as (e.g., Morse and Fritz 1982, Morse 1993, 1999) or repel plants, herbivores, and predators, a trophic cascade de- pollinators (Bristowe 1958, Dukas 2001) or change the scribes the positive top-down effects of the third tro- ¯oral architecture so as to impede their access (Ott et phic level on the biomass, richness, or composition of al. 1998). Based on ®eld observations, Louda (1982) the producer species (Hairston et al. 1960, Wootton showed a balance between cost and bene®t effects of 1994, Menge 1995, Abrams et al. 1996, Polis et al. the green lynx spider, Peucetia viridans (Oxyopidae), 2000). Predators frequently have an impact on the den- to the plant Haplopappus venetus (Asteraceae). P. vir- sity, spatial distribution, and diversity of herbivore as- idans reduced the proportion of ¯owers pollinated by semblages, thereby altering the patterns and levels of one-third. However, the release of viable, undamaged herbivory in plant communities (Heads and Lawton seeds was higher on in¯orescence branches with spi- 1984, 1985, Thomas 1989, Spiller and Schoener 1990, 1994, Dial and Roughgarden 1995, Schmitz and Suttle ders than on those without. Although spiders belonging 2001). Predators on ¯owers can also exert a positive to several families commonly forage on ¯owers, with effect by enhancing plant reproductive rates through the exception of the Louda's study, their overall effects the consumption of ¯ower and fruit predators (Schem- on plant ®tness through interactions with pollinators ske 1980, Stephenson 1982, Del-Claro et al. 1996, and seed predators or ¯ower herbivores is largely un- Oliveira et al. 1999). However, if the predators prey known. on or chase away pollinators (Wolf et al. 1975, Elliott In the present study, the ¯ower-dwelling crab spider and Elliott 1991, Lima 1991, Dukas 2001), their pres- Misumenops argenteus (Thomisidae) was experimen- ence becomes costly for the plant. tally manipulated to determine its effects on the as- As third trophic level organisms, spiders can dimin- semblage of capitula (¯owerhead) herbivores and on ish herbivore abundance in both natural and the ®tness of the native plant, Trichogoniopsis aden- agroecosystems (reviewed in Wise 1993). Spiders can antha (Asteraceae, see Plate 1). M. argenteus preys on also diminish herbivory or seed predation in plants pollinators and phytophagous insects (Romero and (Louda 1982, Riechert and Bishop 1990, Carter and Vasconcellos-Neto, in press, a) and hence may directly and/or indirectly alter the densities and occupation pat- Manuscript received 30 May 2002; revised 28 April 2003; terns of herbivores and consequently interfere with the accepted 3 June 2003. Corresponding Editor: D. A. Spiller. balance of bene®cial and harmful interactions between 1 E-mail: [email protected] the plant and second trophic level components. The 446 February 2004 EFFECTS OF FLOWER-DWELLING PREDATORS 447 main questions addressed were: (1) Does M. argenteus affect the abundance of ¯ower herbivores and the dam- age they cause to the ¯ower structures of T. adenantha, and (2) does M. argenteus affect the plant's reproduc- tive output?

METHODS Study site and organisms The study was done in the Ecological Reserve of the Serra do Japi (23Њ11Ј S, 46Њ52Ј W), a subtropical sem- ideciduous mesophilic forest in southeastern Brazil (el- evation ϳ1000 m), in mid to late summer (February and March) of 2000 and 2001. Trichogoniopsis adenantha (DC) is a small perennial PLATE 1. The crab spider Misumenops argenteus preying on a tephritid ¯y (Trupanea sp.) on the plant Trichogoniopsis shrub (up to 1.8 m high) with glandular trichomes adenantha (Asteraceae). Photo credit: JoaÄo Vasconcellos- (King and Robinson 1987) in which insects frequently Neto. become stuck (see Plate 1). This species is abundant along forest margins at the study site and blooms throughout the year, with a peak in the rainy season is always accompanied by much frass, which distin- (February to April). Each branch produces 2±7 asyn- guishes this damage from that other of phytophages. chronous capitula. Before fertilization, viable achenes The most common pollinators of T. adenantha at the are white in color, but turn black when fertilized. In a study site were the ithomiine butter¯ies Pseudoscada greenhouse, in the absence of ¯oral visitors, achenes erruca, Aeria olena, and Episcada carcinia (Nym- are not fertilized, indicating that plants require polli- phalidae), and several species of ctenuchine nators (Romero 2001). (Arctiidae), bees (Apoidea), and hover ¯ies (Syrphidae) T. adenantha capitula are visited by phytophagous (Romero 2001). insects and their parasitoids, as well as pollinators and The crab spider M. argenteus occurred frequently on pollen- and nectar-eating insects. Macrolophus ara- T. adenantha at the study site. It preferentially uses garcanus (Miridae, Heteroptera), aphids, and membra- ¯owering branches and capitula as foraging sites where cids (Homoptera) are the main sucking insects of the it preys on a variety of insects, including phytophagous plant (Romero 2001). The endophagous herbivores and pollinivorous/nectarivorous species (Appendix; were Trupanea sp. (Diptera, Tephritidae), Melanagro- Romero and Vasconcellos-Neto, in press, b). myza sp. (Diptera, Agromyzidae), Asphondylia sp. (co- rolla gall-maker) (Diptera, Cecidomyiidae) and Phal- Experimental design onidia unguifera (, ). Among The experiments described here were conducted endophagous species, Trupanea sp. and Melanagro- close to a natural population of T. adenantha. To de- myza sp. are most common (Romero 2001); adult fe- termine the effects of M. argenteus on the herbivore males oviposit in the ¯oral primordia and their larvae species and on T. adenantha, paired plants were set out feed within the capitula on immature achenes (Almeida along a forest margin, one of which received a single 1997). These Diptera are frequently parasitized by spe- spider and the other did not. The experiments were cies of Braconidae and Chalcidoidea (Hymenoptera). carried out in February±March of 2000 and 2001, and In addition, a common exophagous geometrid larvae used, respectively, 24 and 22 plant pairs, with a dif- (unidenti®ed species) consumes ¯oral tissue, including ferent set of plants each year. For each experiment, 20± achenes and bracts. 40 cm high plants with a single vegetative branch were Achenes in capitula attacked by Trupanea sp. have collected in the ®eld and transplanted to pots (25 cm a comma shape (Romero 2001) and are grouped in a diameter, 18 cm high) with soil from the same place palisade in the center of the capitulum. The puparia of and depth (of margins) to minimize ``bottom-up'' ef- this species are black and opaque (Almeida 1997). fects. When ¯oral primordia appeared, plants were set Achenes in capitula attacked by Melanagromyza sp. do out in pairs (blocks) in the experimental area (0.2 ha) not form a comma, although it also uses the central before the capitula could be colonized by their spe- achenes. Its puparia are light brown and translucent cialized herbivores. A distance of 0.5±1.0 m separated (Almeida 1997). Geometrid damage to achenes is rec- paired plants along the forest margin, and each pair ognized by the removal of the apical region, including was at least 2 m away from its neighbors. One member the pappus, which rarely occurs with endophages. of a pair was chosen randomly to receive a single ®eld- Young geometrid larvae feed on many stigmas and collected M. argenteus (spider-present treatment), and some corollas, but 4th and 5th instars generally infest the other plant served as the control (spider-absent buds and consume achenes. Damage by tortricid larvae treatment). 448 GUSTAVO Q. ROMERO AND JOAÄ O VASCONCELLOS-NETO Ecology, Vol. 85, No. 2

As new ¯owering branches were produced on spider- M. argenteus were compared by randomized-block present plants, one additional spider was added per ANOVA in which treatment was a ®xed effect and pairs branch. The spiders used were juveniles and adult fe- of plants (blocks) was a random effect. Three- and two- males (5th to 8th instar; cephalothorax±abdomen length factor ANCOVA was used to examine the effects of of 4.0±5.5 mm); adult males were not used because spiders, years (crossed, ®xed factors), and plant blocks they frequently move between plants. The spiders in- (random factor) nested within ``year,'' on the total num- troduced onto experimental plants built shelters by ber of endophages and on the number of Trupanea sp., folding the apical leaves and subsequently remained Melanagromyza sp., Phalonidia unguifera, and As- for many days (mean Ϯ 1 SE, 11.72 Ϯ 1.39 d, n ϭ 32, phondylia sp. per capitulum per plant. The factor range: 2±31 days). All plants were inspected daily and ``year'' was not used for the last two endophages be- spiders were added or removed as necessary, according cause they were absent in 2000. The total number of to the treatment. The spiders were kept on the plants capitula was used as the covariate. The data were until the capitula matured (pre-dispersion phenophase, log10(x ϩ 1) transformed so that the interaction terms after 25±30 d of development). (e.g., spider ϫ year) test for proportional differences Sucking phytophagous (exophagic) insects were between treatments. The same designs were applied to counted ®ve times, once every three days (at 1000 the total number of achenes damaged and those dam- hours, 7±22 March 2001), on the apex (up to 5 cm, aged by different herbivores in plants with and without including capitula, leaves, and stem) of each ¯owering spiders. For analysis of each herbivore separately, data branches of each plant. Sucking insects were not sam- for capitula with interspeci®c co-occurrence were not pled in 2000 because of low population sizes. Their considered because it was not possible to estimate the density is expressed as the number of insects counted exact quantity of damage attributable to each herbivore divided by the number of branches per plant. species. At maturity, the capitula (n ϭ 443 and n ϭ 428 in The proportion of open ¯owers with damaged stig- 2000 and 2001, respectively) were then collected sep- mas and with damaged corollas (effects of Geometridae arately at the end of February and continuing through- sp.) relative to the total number of open ¯owers was out March in each year. Capitula were taken to the similarly evaluated by three-factor ANCOVA, with the laboratory, where they were stored in plastic cups for number of capitula as the covariate. Randomized- a few days until the adult endophagous insects or their block, repeated-measures ANCOVA was used to test parasitoids had emerged. The capitula were then dis- the effects of spiders, year, and capitula condition sected and information was recorded on the number of (damaged vs. undamaged) on the proportion of intact, fertilized, unfertilized, damaged, and intact achenes, fertile achenes relative to the total number of intact damaged and intact corollas and stigmas, the type of (fertile ϩ unfertile) achenes. The total number of ca- damage and the number of aborted ¯owers. Counts pitula per plant was the covariate, the block was the were also made on the number of endophages (puparia random factor nested within year, and capitula condi- or adults) of each species and the number of hyme- tion was the repeated factor, nested within plants (sub- nopterans (parasitoids) emerging. These parasitoids jects). The effect of capitula damage was used to dis- emerged at the same time as the endophages. Some criminate the indirect effect of spiders on fertilized capitula were heavily damaged, making the counting achenes (on intact capitula), and the indirect effect of of achenes dif®cult. In these cases, the number of herbivory (by endophages and exophages) on the rate achene scars in the ¯oral receptacle was used as an of achene fertilization (on damaged capitula). In both estimate the number of achenes produced. years (2000 and 2001), some plants had no damaged The oviposition behaviors of each individual of Tru- or no intact capitula, respectively, resulting in missing panea sp. and Melanagromyza sp. found during the values for the analysis of pollination. To maintain a experiments were observed by using the focal balanced design, the blocks containing a plant with no sampling (Martin and Bateson 1986) in order to infer damaged or no intact capitula were removed, as rec- the vulnerability of each to M. argenteus. Measure- ommended by Underwood (1997). Prior to both tests, ments included the total time on the branch, the length the proportion values were arcsin square-root trans- of time spent static or in motion (including supination formed to homogenize the variances (Sokal and Rohlf in Trupanea sp.; Knio et al. 1996), and the maximum 1995). distance traveled on the branch. To assess whether the rate of parasitoid attack on endophages is affected by the presence of spiders and Statistical analysis by endophage species, and to determine whether the The general experimental design was a randomized parasitism rate varies with host density, ANCOVA was block (Hurlbert 1984) in which each plant (experi- used with proportion of endophages parasitized per mental unit) of a pair received a separate treatment plant as the dependent variable, treatment (spiders pres- (spiders present or absent). The densities of sucking ence vs. absence) as the main effect, and number of phytophages (number of individuals per number of each endophage per plant as the covariate. These pro- branches per plant per day) on plants with and without portions were arcsin square-root transformed to ho- February 2004 EFFECTS OF FLOWER-DWELLING PREDATORS 449

FIG. 1. Total number of endophages and number of Trupanea sp., Melanagromyza sp., Phalonidia unguifera, and As- phondylia sp. in capitula of Trichogoniopsis adenantha in the absence (open bars) and presence (solid bars) of the crab spider, in 2000 and 2001. P. unguifera and Asphondylia sp. did not occur in 2000. Error bars indicate ϩ1 SE. mogenize the variances. Parasitoids were counted only Melanagromyza sp. and Trupanea sp. were the only in 2000. endophages that fed upon capitula of T. adenantha dur- To infer the potential vulnerability of Trupanea sp. ing the experiment in 2000. These species, as well as and Melanagromyza sp. to M. argenteus, measures of the corolla gall-maker Asphondylia sp. and the endo- the time of permanence and the maximum distance phagous lepidopteran P. unguifera, occurred in 2001. moved of ovipositing females of each endophage spe- Geometrid larvae occurred during both years. Overall, cies on the branches of T. adenantha were compared the total number of endophagous insects (all species using Student's t test and the Mann-Whitney U test, combined) per plant decreased 1.2 fold and 1.9 fold in respectively. The proportion of the total time in which the presence of the spiders, in 2000 and 2001, respec- these endophage species remained moving (e.g., ovi- tively (Fig. 1). This decrease was signi®cant (spider position, displacements, etc.) relative to the total time effect, Table 1). The number of endophagous herbi- on the plant was compared using the t test. Prior to the vores increase from 2000 to 2001 (year effect, Table test, the proportions were arcsin square-root trans- 1), but there was no signi®cant interaction between formed to homogenize the variances. spiders and years. Individual species differed in their All analyses of variance were run using Type III magnitude of response to spiders (Fig. 1, Table 1). The sums of squares (SS). All statistical analyses were per- number of Trupanea sp. (the most common herbivore) formed using GLM of SYSTAT version 10.2 (SPSS was signi®cantly lower on plants with spiders. This 2002). The mean values (Ϯ1 SE) presented in the ®g- herbivore population grew from 2000 to 2001 (year ures, tables, and text were computed directly from un- effect), and the spider effect was proportionally about transformed data. the same in both years (spider ϫ year). In contrast, the numbers of Melanagromyza sp. and of other endopha- RESULTS ges did not differ on plants with and without spiders, Spider effects on the number of sucking phytophages although numbers of Melanagromyza sp. decreased and herbivores within capitula from 2000 to 2001. In 2001, the density of sucking phytophagous insects The analysis of covariance showed that the spiders was signi®cantly higher in the plants with spiders ab- did not affect the rate of parasitism by hymenopterans sent (0.86 Ϯ 0.04 individuals per capitula [mean Ϯ 1 of Trupanea sp. and Melanagromyza sp. (spider effect;

SE]) than with spiders present (0.23 Ϯ 0.09, F1,20 ϭ Table 2, Fig. 2), and that the proportion of endophages 12.31, P ϭ 0.002). Density of these insects was too parasitized was independent of the density of these low in 2000 for analysis. herbivores on the plants (Fig. 2). 450 GUSTAVO Q. ROMERO AND JOAÄ O VASCONCELLOS-NETO Ecology, Vol. 85, No. 2

TABLE 1. Randomized-block ANCOVA examining the effects of spiders (two levels) and year (two levels) on the variation in the total number of endophages and in the number of each endophage species on T. adenantha plants.

Source of variation df MS FP Total Spider 1 0.390 7.40 0.009 Year 1 1.686 31.95 Ͻ0.001 Spider ϫ year 1 0.082 1.55 0.221 No. capitula (covariate) 1 1.323 25.07 Ͻ0.001 Error 43 0.053 Trupanea sp. Spider 1 0.672 16.17 Ͻ0.001 Year 1 3.187 76.73 Ͻ0.001 Spider ϫ year 1 0.002 0.038 0.846 No. capitula (covariate) 1 0.169 4.07 0.050 Error 43 0.042 Melanagromyza sp. Spider 1 0.002 0.04 0.844 Year 1 1.294 22.44 Ͻ0.001 Spider ϫ year 1 0.063 1.10 0.301 No. capitula (covariate) 1 0.535 9.27 0.004 Error 43 0.058 Phalonidia unguifera Spider 1 0.019 0.26 0.617 No. capitula (covariate) 1 0.702 9.59 0.006 Error 20 0.073 Asphondylia sp. Spider 1 0.087 0.82 0.376 No. capitula (covariate) 1 0.189 1.78 0.197 Error 20 0.106 Notes: P. unguifera and Asphondylia sp. were absent in 2000. Blocks were nested within year.

Oviposition behavior of Trupanea sp. and Trupanea sp. females landed on a leaf close to or dis- Melanagromyza sp. tant from the area that was later chosen for oviposition During the experiments, Trupanea sp. oviposited lat- (capitula). In contrast, Melanagromyza sp. females ¯ew erally in very young (1±4 days old) buds (n ϭ 18) rapidly in circles over an oviposition site before landing when the ®rst apical ¯owers still covered the capitula. on the apical leaf and walked to the capitulum, or land- They sometimes perforated the younger leaves with the ed directly on it. To reach different branches of the ovipositor to reach the youngest capitula. Melanagro- plant, Trupanea sp. walked, whereas Melanagromyza myza sp. oviposited on the apical region of capitula 4±5 sp. ¯ew. days old that were already close to or above the level The maximum walking distance of Trupanea sp. of the ®rst apical ¯owers (n ϭ 21). Before laying eggs, (18.81 Ϯ 3.08 cm, n ϭ 10) was greater than for Me-

TABLE 2. Results for ANCOVA of the effects of spiders (two levels) on the proportion of the endophages Trupanea sp. and Melanagromyza sp. parasitized by hymenopterans per plant.

Source of variation df MS FP Total (Trupanea ϩ Melanagromyza) Spider 1 0.327 1.16 0.287 No. endophages/plant (covariate) 1 0.005 0.02 0.897 Error 43 0.281 Trupanea sp. Spider 1 0.066 0.26 0.612 No. endophages/plant (covariate) 1 1.122 4.41 0.042 Error 43 0.255 Melanagromyza sp. Spider 1 0.165 0.48 0.494 No. endophages/plant (covariate) 1 0.255 0.73 0.396 Error 43 0.347 February 2004 EFFECTS OF FLOWER-DWELLING PREDATORS 451

The total number of achenes damaged was lower in the presence of spiders (Table 3, Fig. 3). The harmful effects of the herbivores to the achenes varied between years (year effect; Table 3) and were stronger in 2001 since the populations of the principal herbivores were much larger in 2001 (see Fig. 1). However, the inter- action between spiders and year on the number of achenes damaged was not signi®cant, indicating that the bene®cial effect of spiders in protecting seeds is proportional to the herbivore density or impact. The number of achenes damaged by Trupanea sp. and by Geometridae sp. decreased in plants with spiders. The number of achenes damaged by Trupanea sp. was sig- ni®cantly higher in 2001 than in 2000, whereas those damaged by Geometridae sp. was about the same in both years. Again, no interactive effect of spiders and years on the number of achenes damaged by these her- bivores species could be detected. In contrast with the ®rst two herbivores, the number of achenes damaged by Melanagromyza sp. and by P. unguifera did not differ between plants with and without spiders. Me- lanagromyza sp. damaged signi®cantly less achenes in 2001 than in 2000. Spiders diminished the proportion of ¯owers with stigmas and corollas damaged by geometrids (Table 4, Fig. 4), but no signi®cant interactive effect occurred between spiders and years. Individual corolla gall-mak- ers (Asphondylia sp.) infested only one corolla and con- sequently damaged (aborted) only one achene (achenes became atrophic). Hence, the number of Asphondylia sp. (see Fig. 1) equals the number of damaged achenes and corollas. FIG. 2. Relationship between the percentage of endopha- gous Trupanea sp. and Melanagromyza sp. parasitized by Spider effects on achene fertilization hymenopterans (Braconidae and Pteromalidae) and the num- ber of endophages in plants with spiders (solid circles, solid Although spiders preyed on pollinators (Appendix), lines) and without spiders (open circles, dashed lines). they did not affect directly the proportion of fertilized achenes in 2000 or 2001 (spider, and spider ϫ year effects; Table 5, Fig. 5). The proportion of fertilized lanagromyza sp. (2.87 Ϯ 0.65 cm, Mann-Whitney U achenes differed among years, but not among capitula ϭ 36.0, P Ͻ 0.0001, n ϭ 8). Trupanea sp. spent more condition (damaged and undamaged; Table 5, Fig. 5), time on T. adenantha branches (29.8 Ϯ 3.22 min, n ϭ indicating that ¯oral herbivory do not affect plant re- 10) than Melanagromyza sp. (16.2 Ϯ 2.17 min, t ϭ production. Fertilized achenes did not differ between 3.20, P ϭ 0.005, n ϭ 8). Melanagromyza sp. remained plants with and without spiders in undamaged capitula, on the move for 31 Ϯ 5.5% (n ϭ 8) of the total time but increased in damaged capitula in plants with spiders on the plant, whereas Trupanea sp. moved for 73 Ϯ (Table 5), indicating that spiders indirectly enhance 3.3% of the time (t ϭ 6.55, P Ͻ 0.001, n ϭ 10). fruit set. Undamaged capitula tended to have less fer- Spider effects on ¯oral damage tilized achenes in plants with spiders in 2001 (Fig. 5), but the interaction terms among capitula, spider, and The number of capitula produced by plants without year of the ®rst analysis in Table 5 were not signi®cant. spiders (9.26 Ϯ 0.82 in 2000, 10.00 Ϯ 1.36 in 2001) and those with spiders (10.00 Ϯ 1.05 in 2000, 9.45 Ϯ DISCUSSION 0.96 in 2001) did not differ signi®cantly (paired t test: Spider effects on herbivore density in 2000, t ϭϪ0.84, P ϭ 0.410; in 2001, t ϭ 0.49, P ϭ 0.623). The same result was observed for achenes Crab spiders lowered the density of sucking phy- (spiders absent in 2000, 49.36 Ϯ 1.32; in 2001, 47.05 tophagous insects. This herbivore guild was frequent Ϯ 1.34; spiders present in 2000, 50.89 Ϯ 1.12; in 2001, in the spider's diet during the experiment (Appendix). 46.06 Ϯ 1.35; paired t test, in 2000, t ϭϪ1.07, P ϭ Young (1989) also observed that Misumenops spp. 0.298; in 2001, t ϭ 0.64, P ϭ 0.530). (Thomisidae) reduced the density of a sucking insect, 452 GUSTAVO Q. ROMERO AND JOAÄ O VASCONCELLOS-NETO Ecology, Vol. 85, No. 2

TABLE 3. Randomized-block ANCOVA examining the effects of spiders (two levels) and year (two levels) on the total number of damaged achenes and on the number of achenes damaged by different herbivores.

Source of variation df MS FP Total of achenes damaged Spider 1 2.272 9.42 0.004 Year 1 2.862 11.86 0.001 Spider ϫ year 1 0.199 0.83 0.368 No. capitula (covariate) 1 3.131 12.97 0.001 Error 43 0.241 Damaged by Trupanea sp. Spider 1 2.200 10.89 0.002 Year 1 7.559 37.42 Ͻ0.001 Spider ϫ year 1 0.183 0.90 0.347 No. capitula (covariate) 1 0.452 2.24 0.142 Error 43 0.202 Damaged by Melanagromyza sp. Spider 1 0.035 0.11 0.742 Year 1 11.82 36.51 Ͻ0.001 Spider ϫ year 1 0.568 1.75 0.192 No. capitula (covariate) 1 2.314 7.15 0.011 Error 43 0.324 Damaged by Geometridae sp. Spider 1 2.074 5.47 0.024 Year 1 0.276 0.73 0.398 Spider ϫ year 1 0.195 0.51 0.477 No. capitula (covariate) 1 0.192 0.51 0.480 Error 43 0.379 Damaged by Phalonidia unguifera Spider 1 0.453 0.62 0.440 No. capitula (covariate) 1 4.250 5.83 0.025 Error 20 0.729 Notes: P. unguifera did not occur in 2000. Blocks were nested within year.

Lygus lineolaris (Het., Miridae), on Aster pilosus (As- cornis (Knio et al. 1996). The lower density of Tru- teraceae). Hunting spiders and web-builders were con- panea sp. larvae on plants with spiders could also be sidered to be the principal predators of several sucking caused by predator avoidance during oviposition. One phytophages (e.g., planthoppers), and in some cases individual of Trupanea sp. was seen escaping from M. they can act as effective biological control agents (re- argenteus by ¯ying when this predator was preparing viewed in DoÈbel and Denno 1994). to attack (G. Q. Romero, personal observation). Some Our data indicate that the spiders also lowered the tephritid ¯ies avoid predation by salticid spiders by density of endophagous larvae, but this effect was tax- copying their courtship behavior (Whitman et al. 1988, on speci®c. Trupanea sp., the most common endopha- Hasson 1995). M. argenteus builds a shelter by folding gous herbivore, decreased in the presence of spiders, one or two apical leaves in the direction of the central whereas the number of Melanagromyza sp., Asphon- axis, which may serve as a cue to specialized herbi- dylia sp., and P. unguifera did not differ in response vores. Leaf- and ¯ower-folding behavior has also been to spiders. M. argenteus was seen preying on adult observed in other crab spiders (Morse 1992, Evans female Trupanea sp. many times, but was never ob- 1997, Ott et al. 1998). served to prey on Melanagromyza sp. (Appendix; Rom- The abundance of the gall-maker Asphondylia sp. ero and Vasconcellos-Neto, in press, a). The difference was not affected by the predator. This is perhaps due in susceptibility between Trupanea sp. and Melana- to the plant's architecture. Adult female Asphondylia gromyza sp. may re¯ect differences in prey agility sp. oviposit on the corolla after the bracts fall. The (agromyzid ¯ies are very agile) and oviposition be- capitula in these phenophases (pre-anthesis or anthesis) havior. Melanagromyza sp. oviposited quickly, moved are distant from each other. Since sit-and-wait crab little on the branches, landed on the apical leaves, and spiders limit themselves to foraging in only one of the then went to the very young buds to oviposit, whereas capitula, the remaining capitula are free to be visited Trupanea sp. moved long distances among the leaves, by this insect. Phalonidia unguifera was also not in- stems, and capitula over a longer time period looking ¯uenced by the spiders. This probably oviposits for oviposition sites. Behavior similar to Trupanea sp. at night when the spiders were probably not foraging. was observed for Trupanea bisetosa and for T. nigri- Our results suggest that the patterns of host-plant oc- February 2004 EFFECTS OF FLOWER-DWELLING PREDATORS 453

FIG. 3. Total number of T. adenantha achenes damaged and the number of achenes damaged by each herbivore species on plants with spiders absent (open bars) and present (solid bars), in 2000 and 2001. P. unguifera did not occur in 2000. Achenes of capitula with interspeci®c co-occurrence were considered only in the analysis of the total number of achenes damaged (see Methods). Error bars indicate ϩ1 SE. cupation by the herbivores studied depended on the sequence, intraguild predation was weak, and the co- speci®c behavior of the insect species and on the plant's occurrence of predators contributed to enhance herbi- architecture. Spiller and Schoener (1990, 1994) sug- vore (planthopper) suppression. gested that the vulnerability of herbivorous species de- Spider effect on ¯oral damage and pended on the mode of foraging by predators. Schmitz on achene fertilization and Suttle (2001) showed that spiders with different hunting behaviors can produce different responses in Our data indicated that a sit-and-wait spider bene- prey (grasshoppers) behavior or density. Hence, pred- ®ted its host plant by lowering the damage to corollas, ators within the same guild can have different trophic stigmas, and achenes. Previous studies have shown that effects on food webs. Schmitz (1998) stressed that be- stalking spiders (Oxyopidae, Salticidae) and ground havioral ecology plays a considerable role in under- runners (Lycosidae) can also decrease herbivore den- standing determining population- and community-level interactions. TABLE 4. Randomized-block ANCOVA examining the ef- Although the spiders also preyed on parasitoids (Ap- fects of spiders (two levels) and year (two levels) on the pendix), they did not affect the rate at which endo- proportion of T. adenantha ¯owers with stigmas and co- rollas damaged by Geometridae species per open ¯ower. phages were parasitized. Since parasitoid attacks on endophagous larvae occur in the pre-anthesis pheno- Source of variation df MS FP phase, when the capitula start to spread apart, the same Flowers with stigmas damaged explanation given for the effect of spiders on gall-mak- Spider 1 0.636 10.92 0.002 er ¯ies applies here. Our results suggest that intraguild Year 1 0.015 0.25 0.618 predation between natural enemies (spider predation on Spider ϫ year 1 0.070 1.20 0.279 parasitoids) is unimportant. This is perhaps due to the No. of capitula (covariate) 1 0.040 0.69 0.411 effects of in¯orescence architecture, forming a barrier Error 42 0.058 to the predator. Hence, spiders and parasitoids can act Corollas damaged together in suppressing the endophage, but in different Spider 1 0.292 9.28 0.004 Year 1 0.133 4.24 0.046 phenophases of the capitula. Finke and Denno (2002) Spider ϫ year 1 0.070 2.21 1.144 report a similar situation in which vegetation com- No. of capitula (covariate) 1 0.000 0.00 0.965 plexity by providing refuges modulates the interaction Error 42 0.031 between spiders and their hunter mirid prey. As a con- Note: Blocks were nested within year. 454 GUSTAVO Q. ROMERO AND JOAÄ O VASCONCELLOS-NETO Ecology, Vol. 85, No. 2

During the experiments, M. argenteus occupied all branches of the experimental plants. However, under natural conditions, M. argenteus occupied only 13% of the reproductive branches of T. adenantha in the sum- mer (February and March 2000) and 6% in the winter (July and August 1999) (Romero 2001; Romero and Vasconcellos-Neto, in press, a), but in nature their im- pact per branch may be greater than the experimental effects due to the following reasons: First, the spiders were sometimes found and removed from ``spiders- absent'' treatments and were missing on ``spiders-pre- sent'' treatments. Therefore, the experimentally deter- mined spider effect may be an underestimate of the full effect of spiders in nature. Second, spiders naturally inhabiting plants may on average have a greater impact than experimental spiders because they had already built their retreats and had time to become familiar with the plants. Third, spiders observed on one branch may affect herbivores on other branches of the same plant. This might happen if the herbivores can sense the pres- ence of the spider and subsequently avoid the entire plant or if the spider can detect herbivores on other branches of the same plants and have time to attack those herbivores. Thus, the effect of M. argenteus for a plant in the wild is not fully known, but will depend on the number of spiders and on the main herbivore population sizes, since M. argenteus bene®t to plants was proportional to herbivore abundance. The presence of crab spiders did not directly affect the proportion of fertile achenes in T. adenantha for FIG. 4. Number of ¯owers with (A) damaged stigmas and various reasons. First, during these experiments, the (B) damaged corollas per open ¯ower (OF), caused by Geo- most frequent ¯oral visitors were ithomiinae butter¯ies metridae sp., in T. adenantha with spiders absent (open bars) (n ϭ 116), and M. argenteus preyed less on Ithomiinae and present (solid bars), in 2000 and 2001. Error bars indicate ϩ1 SE. and more on Ctenuchinae (Arctiidae) (Romero and Vas- concellos-Neto, in press, a). During the study, only ®ve Ctenuchinae were observed. Second, thrips (Thysan- optera), which are potential pollinators of some As- sity and enhance plant reproductive output (Louda teraceae species (reviewed in Ananthakrishnan 1993) 1982, Ruhren and Handel 1999, Snyder and Wise occurred in great densities on T. adenantha ¯owers 2001). Since the herbivore suppression by M. argenteus (Romero 2001) and may have fertilized the achenes. was taxon speci®c, only achene damage by vulnerable Misumenops argenteus has not been observed preying herbivore species decreased in the presence of the spi- on these minute insects (Romero and Vasconcellos- ders. Geometridae sp. density was not measured, but Neto, in press, a). Third, the distance among the ca- damage attributable to this herbivore was lower on pitula during anthesis may hamper crab spider preying plants with spiders. TanhuanpaÈaÈ et al. (2001) demon- on pollinators. For instance, Louda (1982) showed that strated that crab spiders are very effective in reducing the lynx spider Peucetia viridans (Oxyopidae) reduces geometrid larvae on Betula birch. As the larvae of this both damaged achenes and fertilized achenes in Hap- moth are sessile and remain on branches for a long lopappus venetus. The ¯at-topped in¯orescence struc- time, they are easily caught by the spider. In the Ba- ture in H. venetus at anthesis may facilitate the lynx hama Islands, Spiller and Schoener (1990) suggested spider predation on pollinators, whereas T. adenantha that winged gall-makers were more vulnerable to orb- capitula are well separated at anthesis, thus restricting web spiders than to lizards, whereas more sessile her- the crab spiders to a single capitulum, leaving the re- bivores, such as homopterans (aphids and leafhoppers) maining capitula open to pollinators. Thus, in¯ores- and lepidopteran larvae, were more easily captured by cence architecture may in¯uence prey capture ef®cien- lizards than by spiders. Consequently, the leaves of cy of ¯ower-dwelling spiders. plants in the presence of more spiders had less galls, Crab spiders increased the proportion of fertilized and in the presence of more lizards had less scars and achenes only in damaged capitula, but not in undam- holes. aged capitula. This may have occurred because February 2004 EFFECTS OF FLOWER-DWELLING PREDATORS 455

TABLE 5. Randomized-block ANCOVA examining the effects of spiders (two levels), year (two levels), and capitula condition (damaged and undamaged) on the proportion of intact fertilized achenes of T. adenantha per total of intact (fertile ϩ unfertile) achenes.

Source of variation df MS FP Capitula undamaged and damaged Between subjects Spider 1 0.001 0.06 0.814 Year 1 0.461 24.76 Ͻ0.001 Spider ϫ year 1 0.012 0.66 0.422 No. of capitula (covariate) 1 0.155 8.34 0.007 Error 28 0.019 Within subjects Capitula 1 0.015 0.06 0.446 Capitula ϫ year 1 0.040 1.57 0.220 Capitula ϫ spider 1 0.050 1.98 0.170 Capitula ϫ year ϫ spider 1 0.078 3.09 0.089 Error 28 0.025 Capitula undamaged Spider 1 0.014 0.59 0.447 Year 1 0.365 15.74 Ͻ0.001 Spider ϫ year 1 0.010 0.42 0.522 No. of capitula (covariate) 1 0.017 0.72 0.402 Error 32 0.023 Capitula damaged Spider 1 0.127 5.93 0.020 Year 1 0.713 33.25 Ͻ0.001 Spider ϫ year 1 0.020 0.93 0.342 No. of capitula (covariate) 1 0.117 5.45 0.025 Error 35 0.021 Notes: Blocks were nested within year. The ®rst analyses were conducted using repeated measures. amounts of damage were higher on capitula with spi- 1997, Krupnick et al. 1999, Mothershead and Marquis ders absent than on those with spiders present, and only 2000). Capitula damaged by herbivores have fewer high amounts of damage reduce achene fertilization. open ¯owers than intact capitula and probably offer Although the present analysis did not detect differences less reward (nectar and/or pollen) to pollinators. among damaged and undamaged capitula in achene fer- Since M. argenteus did not directly affect the plant's tilization, other work on this same system found evi- reproduction, but suppressed herbivores, the indirect dence of an indirect, negative effect of herbivory by effect of this spider for T. adenantha was only bene- endophages and exophages on the proportion of achene ®cial. This bene®cial interaction is perhaps mediated fertilized (Romero 2001; A. T. SalomaÄo, L. F. Martins, by glandular trichomes that attract the predators. Mar- R. S. Ribeiro, G. Q. Romero, unpublished data). In- quis and Whelan (1996) suggest that plant morpholog- direct effects of herbivory on plant reproduction have ical traits, such as pubescence type, make the plants been reported for other systems (LehtilaÈ and Strauss more accessible to predators and can mediate tri-tro-

FIG. 5. The proportion of intact fertile achenes of T. adenantha relative to the total number of intact (fertile ϩ unfertile) achenes in damaged and undamaged capitula, in plants with spiders absent (open bars) and present (solid bars), during 2000 and 2001. Error bars are Ϯ1 SE. 456 GUSTAVO Q. ROMERO AND JOAÄ O VASCONCELLOS-NETO Ecology, Vol. 85, No. 2 phic interactions. A role for glandular trichomes is sug- DoÈbel, H. G., and R. F. Denno. 1994. Predator±planthopper gested from the fact that M. argenteus occurs at a higher interactions. Pages 325±399 in R. F.Denno and T.J. Perfect, editors. Planthoppers: their ecology and management. frequency on T. adenantha and on the sympatric Hyptis Chapman and Hall, New York, New York, USA. suaveolens (Lamiaceae) and is uncommon on other Dukas, R. 2001. Effects of perceived danger on ¯ower choice shrubs and herbs (Romero 2001). Both plants possess by bees. Ecology Letters 4:327±333. glandular trichomes to which insects frequently adhere, Elliott, N. B., and W. M. Elliott. 1991. Effect of an ambush predator, Phymata americana Melin, on behavior of insects and which might enhance capture by sit-and-wait spi- visiting Daucus carota. American Midland Naturalist 126: ders. Moreover, T. adenantha blooms and attracts prey 198±202. throughout the year. These two characteristics increase Evans, T. A. 1997. Distribution of social crab spiders in eu- the foraging ef®ciency of M. argenteus and can mediate calypt forests. Australian Journal of Ecology 22:107±111. the spider±plant interaction (Romero and Vasconcellos- Finke, D. L., and R. F. Denno. 2002. Intraguild predation diminished in complex-structured vegetation: implications Neto, in press, b). for prey suppression. Ecology 83:643±652. In conclusion, this study demonstrated that M. ar- Hairston, N. G., F. E. Smith, and L. B. Slobodkin. 1960. genteus reduced overall herbivore abundance and dam- Community structure, population control, and competition. age on T. adenantha capitula and decreased loss of American Naturalist 94:421±425. Hasson, O. 1995. A ¯y in spider's clothing: what size the achenes, but they did not directly reduce pollination spider? Proceedings of the Royal Society of London, B success. However, spiders indirectly increased achene 261:223±226. fertilized in damaged capitula. Hence, all detectable Heads, P. A., and J. H. Lawton. 1984. Bracken, ants and effects of spiders were bene®cial. extra¯oral nectaries. II. The effect of ants on the insect herbivores of bracken. Journal of Animal Ecology 53: ACKNOWLEDGMENTS 1015±1031. Heads, P. A., and J. H. Lawton. 1985. Bracken, ants and We are very grateful to R. J. Marquis, T. M. Lewinsohn, extra¯oral nectaries. III. How insect herbivores avoid ant W. W. Benson, S. E. Riechert, J. E. C. Figueira, M. A. Garcia, predation. Ecological Entomology 10:29±42. D. J. Futuyma, K. Del-Claro, A. V. Freitas, E. Ramires, D. Hurlbert, S. H. 1984. Pseudoreplication and the design of Spiller, and two anonymous reviewers for helpful suggestions ecological ®eld experiments. Ecological Monographs 54: and criticism on the manuscript; A. M. Almeida and T. M. 187±211. Lewinsohn for the discussions about the T. adenantha±insect King, R. M., and H. Robinson. 1987. The genera of the interactions; and the staff of the ``Base EcoloÂgica da Serra (Asteraceae). Monographs in Systematic Bot- do Japi'' for logistical support in the ®eld. G. Q. Romero was any 22:1±581. supported by a research grant from FundacËaÄo de Amparo aÁ Knio, K. M., D. R. Goeden, and D. H. Headrick. 1996. Com- Pesquisa do Estado de SaÄo Paulo (FAPESP, grant no. 98/ parative biologies of the cryptic, sympatric species, Tru- 15854-3). J. Vasconcellos-Neto was supported by a grant panea bisetosa and T. nigricornis (Diptera: Tephritidae) in from Conselho Nacional de Desenvolvimento Cientõ®co e Southern California. Annuals of the Entomological Society TecnoloÂgico (CNPq, grant no. 300539/94-0). This paper is of America 89:252±260. part of the BIOTA/FAPESP (The Biodiversity Virtual Insti- Krupnick, G. A., A. E. Weis, and D. R. Campbell. 1999. The tute Program [͗www.biotasp.org.br͘]) Proc. 99/05446-8. consequences of ¯oral herbivory for pollinator service to Isomeris arborea. Ecology 80:125±134. LITERATURE CITED LehtilaÈ, K., and S. Y. Strauss. 1997. Leaf damage by her- Abrams, P. A., B. A. Menge, G. G. Mittelbach, D. A. Spiller, bivores affects attractiveness to pollinators in wild radish, and P. Yodzis. 1996. The role of indirect effects in food Raphanus raphanistrum. Oecologia 111:396±403. webs. Pages 371±395 in G. A. Polis and K. O. Winemiller, Lima, S. L. 1991. Energy, predators and the behaviour of editors. Food webs. Integration of patterns and dynamics. feeding hummingbirds. Evolutionary Ecology 5:220±230. Chapman and Hall, New York, New York, USA. Louda, S. M. 1982. In¯orescence spiders: a cost/bene®t anal- Almeida, A. M. 1997. Patterns of occurrence in endophagous ysis for the host plant, Haplopappus venetus Blake (As- insects associated with ¯owerheads of Trichogoniopsis ad- teraceae). Oecologia 55:185±191. enantha (DC) (Asteraceae) [In Portuguese]. Dissertation, Marquis, R. J., and C. Whelan. 1996. Plant morphology and Department of Zoology, Universidade Estadual de Cam- recruitment of the third trophic level: subtle and little-rec- pinas, SaÄo Paulo, Brazil. ognized defenses? Oikos 75:330±334. Ananthakrishnan, T. N. 1993. The role of thrips in pollina- Martin, P., and P. Bateson. 1986. Measuring behaviour: an tion. Current Science 65:262±264. introductory guide. Cambridge University Press, Cam- Bristowe, W. S. 1958. The world of spiders. Collins, London, bridge, UK. UK. Menge, B. A. 1995. Indirect effects in marine rocky intertidal Carter, P. E., and A. L. Rypstra. 1995. Top-down effects in interaction webs: patterns and importance. Ecological soybean agroecosystems: spider density affects herbivore Monographs 65:21±74. damage. Oikos 72:433±439. Moran, M. D., and A. R. Scheidler. 2002. Effects of nutrients Del-Claro, K., V. Berto, and W. ReÂu. 1996. Effect of her- and predators on an old-®eld food chain: interactions of bivore deterrence by ants on the fruit set of an extra¯oral top-down and bottom-up processes. Oikos 98:116±124. nectary plant, Qualea multi¯ora (Vochysiaceae). Journal of Morse, D. H. 1992. Predation on dispersing Misumena vatia Tropical Ecology 12:887±892. spiderlings and its relationship to maternal foraging deci- Denno, R. F., C. Gratton, M. A. Peterson, G. A. Langellotto, sions. Ecology 73:1814±1819. D. L. Finke, and A. F. Huberty. 2002. Bottom-up forces Morse, D. H. 1993. Choosing hunting sites with little infor- mediate natural-enemy impact in a phytophagous insect mation: patch-choice responses of crab spiders to distant community. Ecology 83:1443±1458. cues. Behavioral Ecology 4:61±65. Dial, R., and J. Roughgarden. 1995. Experimental removal Morse, D. H. 1999. Choice of hunting site as a consequence of insectivores from rain forest canopy: direct and indirect of experience in late-instar crab spiders. Oecologia 120: effects. Ecology 76:1821±1834. 252±257. February 2004 EFFECTS OF FLOWER-DWELLING PREDATORS 457

Morse, D. H., and R. S. Fritz. 1982. Experimental and ob- Schmitz, O. J., and K. B. Suttle. 2001. Effects of top predator servational studies of patch choice at different scales by species on direct and indirect interactions in a food web. the crab spider Misumena vatia. Ecology 63:172±182. Ecology 82:2072±2081. Mothershead, K., and R. J. Marquis. 2000. Fitness impacts Snyder, W. E., and D. H. Wise. 2001. Contrasting trophic of herbivory through indirect effects on plant±pollinator cascades generated by a community of generalist predators. interactions in Oenothera macrocarpa. Ecology 81:30±40. Ecology 82:1571±1583. Oliveira, P. S., V. Rico-Gray, C. DõÂaz-Castelazo, and C. Cas- Sokal, R. R., and F. J. Rohlf. 1995. Biometry. Third edition. tillo-Guevara. 1999. Interaction between ants, extra¯oral W. H. Freeman, New York, New York, USA. nectaries and insect herbivores in neotropical coastal sand Spiller, D. A., and T. W. Schoener. 1990. A terrestrial ®eld dunes: herbivore deterrence by visiting ants increases fruit set in Opuntia stricta (Cactaceae). Functional Ecology 13: experiment showing the impact of eliminating top predators 623±631. on foliage damage. Nature 347:469±472. Ott, J. R., J. A. Nelson, and T. Caillouet. 1998. The effect Spiller, D. A., and T. W. Schoener. 1994. Effects of top and of spider-mediated ¯ower alteration on seed production in intermediate predators in a terrestrial food web. Ecology golden-eye phlox. Southwestern Naturalist 43:430±436. 75:182±196. Polis, G. A., A. L. W. Sears, G. R. Huxel, D. R. Strong, and SPSS. 2002. SYSTAT software. Version 10.2. SPSS Science, J. Maron. 2000. When is a trophic cascade a trophic cas- Chicago, Illinois, USA. cade? Trends in Ecology and Evolution 15:473±475. Stephenson, A. G. 1982. The role of extra¯oral nectaries of Polis, G. A., and K. O. Winemiller. 1996. Food webs. Inte- Catalpa speciosa in limiting herbivory and increasing fruit gration of patterns and dynamics. Chapman and Hall, New production. Ecology 63:663±669. York, New York, USA. TanhuanpaÈaÈ, M., K. RuohomaÈki, and E. Uusipaikka. 2001. Riechert, S. E., and L. Bishop. 1990. Prey control by an High larval predation rate in non-outbreaking populations assemblage of generalist predators: spiders in garden test of a geometrid moth. Ecology 82:281±289. systems. Ecology 71:1441±1450. Thomas, C. D. 1989. Predator±herbivore interactions and the Romero, G. Q. 2001. Experimental study of the association between Runcinioides argenteus (Araneae, Thomisidae) escape of isolated plants from phytophagous insects. Oikos and Trichogoniopsis adenantha (DC) (Asteraceae). [In Por- 55:291±298. tuguese.] Dissertation, State University of Campinas, SaÄo Underwood, A. J. 1997. Experiments in ecology. Cambridge Paulo, Brazil. University Press, Cambridge, UK. Romero, G. Q., and J. Vasconcellos-Neto. In press, a. Natural Whitman, D. W., L. Orsak, and E. Greene. 1988. Spider mim- history of Misumenops argenteus (Thomisidae): seasonality icry in fruit ¯ies (Diptera: Tephritidae): further experiments and diet on Trichogoniopsis adenantha (Asteraceae). Jour- on the deterrence of jumping spiders (Araneae: Salticidae) nal of Arachnology. by Zonosemata vittigera (Coquillett). Annals of the Ento- Romero, G. Q., and J. Vasconcellos-Neto. In press, b. For- mological Society of America 81:532±536. aging by the ¯ower-dwelling spider, Misumenops argenteus Wise, D. H. 1993. Spiders in ecological webs. Cambridge (Thomisidae), at high prey density sites. Journal of Natural University Press, Cambridge, UK. History. Wolf, L. L., F. R. Hainsworth, and F. G. Gill. 1975. Foraging Ruhren, S., and S. N. Handel. 1999. Jumping spiders (Sal- ef®ciencies and time budgets in nectar-feeding birds. Ecol- ticidae) enhance the seed production of a plant with ex- ogy 56:117±128. tra¯oral nectaries. Oecologia 119:227±230. Wootton, J. T. 1994. The nature and consequences of indirect Schemske, D. W. 1980. The evolutionary signi®cance of ex- tra¯oral nectar production by Costus woodsonii (Zingiber- effects in ecological communities. Annual Review of Ecol- aceae): an experimental analysis of ant protection. Journal ogy and Systematics 25:443±466. of Ecology 68:959±967. Young, O. P. 1989. Relationships between Aster pilosus Schmitz, O. J. 1998. Direct and indirect effects of predation (Compositae), Misumenops spp. (Araneae: Thomisidae), and predation risk in old-®eld interaction webs. American and Ligus lineolaris (Heteroptera: Miridae). Journal of En- Naturalist 151:327±342. tomological Sciences 24:252±257.

APPENDIX Observations of prey consumed by Misumenops argenteus on Trichogoniopsis adenantha plants during the experiments carried out in 2001 are available in ESA's Electronic Data Archive: Ecological Archives E085-009-A1.