The effects of plant structure on the spatial and microspatial distribution of a bromeliad-living jumping spider (Salticidae)

Authors

  • GUSTAVO Q. ROMERO,

    Corresponding author
    1. Departamento de Zoologia, Instituto de Biologia, Universidade Estadual de Campinas (Unicamp), CP 6109, CEP 13083–970, Campinas, SP, Brazil
      Correspondence: Departamento de Zoologia, Instituto de Biologia, Universidade Estadual de Campinas (Unicamp), CP 6109, CEP 13083–970, Campinas, SP, Brazil. E-mail: gqromero@unicamp.br
    Search for more papers by this author
  • JOÃO VASCONCELLOS-NETO

    1. Departamento de Zoologia, Instituto de Biologia, Universidade Estadual de Campinas (Unicamp), CP 6109, CEP 13083–970, Campinas, SP, Brazil
    Search for more papers by this author

Correspondence: Departamento de Zoologia, Instituto de Biologia, Universidade Estadual de Campinas (Unicamp), CP 6109, CEP 13083–970, Campinas, SP, Brazil. E-mail: gqromero@unicamp.br

Summary

  • 1In several regions of South America, the neotropical jumping spider Psecas chapoda inhabits and reproduces strictly on the bromeliad Bromelia balansae. Previous studies reported that this spider is more frequent on bromeliads in grasslands than on those growing in forests, and on larger plants, but only when the bromeliads are without inflorescence. Upon blooming, B. balansae fold their leaves back, thereby changing the plant architecture from a tri-dimensional to a bi-dimensional flattened shape, and our hypothesis is that this alteration affects the spider's habitat-selection decisions.
  • 2In the present study, we examined experimentally the effects of inflorescence, plant size and blockade of the axil of the leaves (spider shelters) of forest bromeliads on the colonization of a bromeliad by P. chapoda. By using sticky traps, we also compared prey availability in grassland and forest.
  • 3Plants with simulated inflorescence were colonized at a lower frequency than those without inflorescence simulation. Grassland bromeliads in which the rosettes were blocked with dry leaves were colonized less frequently than open bromeliads, whereas forest bromeliads from which dry leaves had been removed were not colonized. Spiders generally abandoned bromeliads in which three-quarters of the length of the leaves had been removed, although females with eggsacs remained on these plants. Prey availability (biomass and number) was up to 18 fold higher in the grassland than in the forest. These results suggest that microhabitat structure and prey availability shape the spatial distribution of P. chapoda populations.
  • 4Our findings suggest that Psecas chapoda can evaluate, in fine detail, the physical state of its microhabitat, and this unusual spider–plant association is readily destabilized by changes in the microhabitat (i.e., it is strictly dependent on the size and morphology of the host plant). This study is one of the few to report a strict association between a spider species and its host plant, and also one of the few to examine the effects of habitat and microhabitat structure on the spatial distribution of active hunters on plants.

Introduction

Structural components of the vegetation exert a strong influence on the density and diversity of many terrestrial arthropods (Lawton 1983; Morse et al. 1985; Scheuring 1991; Gunnarsson 1992; Gardner et al. 1995; Borges & Brown 2001), but spiders appear to be particularly strongly influenced by architectural variations in the vegetation (Riechert & Gillespie 1986; Gunnarsson 1996). Spiders do not eat plants, but plants are often important for them as sites for building webs (Lubin 1978; Rypstra 1983; Greenstone 1984; Figueira & Vasconcellos-Neto 1991; Herberstein 1997), for sheltering against desiccation (Riechert & Tracy 1975) or natural enemies (Gunnarsson 1990, 1996), for foraging (Morse & Fritz 1982; Morse 1990; Scheidler 1990; Schmalhofer 2001; Romero & Vasconcellos-Neto 2003, 2004a,b), and for mating and oviposition (Rossa-Feres, Romero & Gonçalves-de-Freitas 2000; Smith 2000; Romero & Vasconcellos-Neto 2003; Romero & Vasconcellos-Neto, in press a,b). That plant architecture affects the abundance and distribution of spiders has been shown for a variety of spiders (Colebourne 1974; Greenquist & Rovner 1976; Robinson 1981; Rypstra 1983; Gunnarsson 1990, 1992), and thomisid spiders have been shown actively to select plant-determined microhabitat (Morse & Fritz 1982; Morse 1990, 1993).

Even when there is an apparent relationship between changes in plant architecture and spider density, whether the causal relationship is direct or indirect is often uncertain because potentially important environmental factors, such as prey availability (Rypstra 1983; Greenstone 1984; Halaj, Ross & Moldenke 1998) can vary with the habitat structure. Any strong conclusions about plant architecture directly affecting spider distribution require experimental support (Wise 1993). For example, experiments based on artificial vegetation or direct manipulation of the structures of live plants have shown that spiders from different taxonomic groups and from different guilds have specific preferences for certain types of architecture (Robinson 1981). Other experimental studies have shown that the reduction in the number of foraging or shelter sites, and changes in the substrate abundance, negatively affected the density of particular spider species (Gunnarsson 1990; McNett & Rypstra 2000). However, all previous experimental studies have examined how the density web-building spiders is influenced by habitat structure. There have been no experimental studies on how plant architecture might influence density of hunting spiders.

The spider Psecas chapoda (Peckham & Peckham) (Salticidae) inhabits and reproduces strictly on Bromelia balansae Mez. (Bromeliaceae) (Rossa-Feres et al. 2000; Romero & Vasconcellos-Neto, in press a,b) in several regions of Brazil, Paraguay and Bolivia (Höfer & Brescovit 1994; Rossa-Feres et al. 2000; Romero & Vasconcellos-Neto, in press a,b; G. Q. Romero, unpublished data). This spider occurs at a higher density on B. balansae in grassland (open areas) than on bromeliads growing in forest. Romero & Vasconcellos-Neto (in press a) suggested that dry leaves falling from trees in the forest negatively affected the colonization of these bromeliads by blocking the base of bromeliad rosettes used as shelter by the spiders. Spider density was also lower in plants with inflorescence (or infrutescence), possibly because of the changes in plant architecture during blooming (Romero & Vasconcellos-Neto, in press a,b). Romero & Vasconcellos-Neto (in press a) found a positive relationship between length of bromeliad leaves and number of spiders on the plants, and suggested that the carrying capacity is positively related to bromeliad size.

Here we use this spider–plant system for investigating the effects of changes in microhabitat architecture on spider density. We consider the spider's selection of microhabitats for shelter, foraging, mating and oviposition. In our experiments, we examine how colonization and selection of microhabitats by P. chapoda is influenced by plant size, inflorescence and accessibility to rosettes of B. balansae. Specifically, we addressed three questions. (i) How many spiders colonize plants without inflorescence and without dry leaves in their rosette compared to how many colonize plants with inflorescence and dry leaves? (ii) Does spider age influence colonization pattern? (iii) Does leaf size affect the density of spiders on bromeliads? We also consider whether prey availability in forest and grassland differ.

Material and methods

study area and organisms

This work was carried out from October 2001 to February 2003 in a 250 × 60 m fragment of semi-deciduous forest and in an adjacent grassland area along the margin of a river. This site was near Dois Córregos city (22°21′ S, 48°22′ W), São Paulo state, in south-eastern Brazil. Local climate consists of a distinct dry/cold (May–September) and wet/warm (October–April) season. Mean annual rainfall is 1600 mm (Romero & Vasconcellos-Neto, in press b).

The entire life cycle of P. chapoda, including courtship behaviour, mating, oviposition, and spiderling recruitment occurs on the bromeliad B. balansae. Females lay 1–3 eggsacs on the concave side of the central region of the leaves. The eggsacs are wrapped in a plain silk cover (nest) that is spun at the edge of each leaf, and the females remain under the nest (Rossa-Feres et al. 2000). When we attempted to capture the spiders, they tended to run to the rosette base (G. Q. Romero, personal observation). This suggested that the bromeliad was especially suitable as a shelter against predation. In the study area, B. balansae was present in the grassland and forest at comparable densities (Romero & Vasconcellos-Neto, in press a). These bromeliads do not accumulate rain water, they bloom at the beginning of the rainy season (September–December) and up to 40% of the individuals release the inflorescence (Romero & Vasconcellos-Neto in press b).

experiment 1: simulation of the inflorescence

The objective in this experiment was to investigate whether the presence of inflorescence influenced the spider's inclination to take up residence in plants (bromeliad affinity). Different age classes of spiders were tested for evidence of whether bromeliad affinity changes during the life cycle. For this, 52 individuals of B. balansae were brought from other sites and planted in pairs (blocks) in the grassland of the study area. The bromeliads used were all of similar size (median leaf length c. 50–70 cm). A distance of 2 m separated plants of the same pair, and each pair was at least 4–6 m from its nearest neighbour. One bromeliad in each pair was randomly chosen for inflorescence simulation (experimental) while the other bromeliad was unaltered (control). To simulate inflorescence, an iron ring (16 cm in diameter) was placed on top of the central region of the rosette and then pressed down to force the leaves of the external and internal layers to the ground, where they remained parallel with each other and with the soil. The rings were kept in this position by three iron supports that were fixed perpendicular to each ring and anchored in the soil. The rings and supports were 5 mm in diameter. Only the leaves of the first one or two layers of the experimental plants were not bent by the ring. This was because of their small size. However, all other leaves were altered. In the control plants, the ring was positioned normally, but was not pressed down (i.e. plant architecture was not altered). There was no evidence that the rings changed the behaviour or abundance of spiders on the control plants.

Before the experiment began, all plants were inspected to ensure that they had not already been colonized by P. chapoda (individuals and eggsacs). The spiders and eggsacs on each plant were censused fortnightly after the beginning of the experiment (5 October 2001). Only new eggsacs (with plain silk cover, see Rossa-Feres et al. 2000) were censused in these samples. Age-specific patterns of spots and coloration were used to identify P. chapoda as spiderlings (3rd instar), young (4th and 5th instars), juvenile males (up to 1·1 cm in body length), juvenile females (6th instar), subadult male (7th instar) and adult males (8th instar) (Romero & Vasconcellos-Neto, in press a). Although sex-specific patterns of spots and coloration are useful for discriminating subadult and adult stages, subadult and adult females have the same spot and coloration patterns, and a similar size (up to 16 mm in body length), making it difficult to determine their instar in the field (Romero & Vasconcellos-Neto, in press a). Because the capture of these spiders on the bromeliads was very difficult, we included subadult and adult females (7th and 8th instars) in the same class for analysis.

experiment 2: addition/removal of dry leaves

The hypothesis here is that dry leaves falling from forest trees block the base of B. balansae rosettes (the shelter used by the spiders) and make them difficult for the spider to access. Dry leaves on the bromeliads therefore could be changing the patterns of spider distribution between habitats (grassland/forest). To test this hypothesis, 36 bromeliads of a similar size were brought from other sites and freed of spiders and eggsacs before planting in pairs (blocks) in the grassland of the study area, as in experiment 1. Both plants of each pair initially received dry leaves brought from the rosettes of forest bromeliads. However, for one individual chosen at random in each pair, all of the dry leaves were removed (control). In the forest area, dry leaves were removed from the interior of 10 bromeliads of similar size (experimental), 10–20 m apart from the grassland, while the 10 bromeliads of similar size closest to the experimental plants (within 0·5–1·5 m) were left with their natural accumulation of dry leaves (control). These forest bromeliads had no spiders or eggsacs. The spiders and eggsacs were censused once every 10–15 days after the start of the experiment (20 September 2002).

experiment 3: cutting of bromeliad leaves

Psecas chapoda uses B. balansae leaves as foraging sites, and a shortening of the leaves might decrease spider density by decreasing the carrying capacity of the microhabitat. Something similar has been reported by other studies for other spiders (Gunnarsson 1990). To test this hypothesis, 30 individuals of B. balansae of similar size (see experiment 1) in the natural population from grassland were selected along a 250-m transect and numbered. In the first plant found, three-quarters (3/4) of the total length of all of the leaves was removed (treatment 1), in the second plant, one-quarter (1/4) of the total length of all of the leaves was removed (treatment 2), and in the third plant no leaves were cut (control). This sequence of treatments was repeated until 30 plants had been included (10 plants per treatment). In the control plants, the leaves were slightly shaken to simulate the leaf cutting in the other treatments. The number of spiders and their eggsacs on each plant was censused before the start of the experiment (30 December 2002), and after 1, 2, 3, 4, 5 and 19 days.

prey availability

Even if rosettes blockage by dry leaves affects spider distribution, another influence might be prey availability. We consider this hypothesis by determining how prey number and prey biomass varied between forest and grassland. For this, we used 10 sticky-traps set up 40–60 cm above the ground at 10 m intervals in the vegetation amongst the bromeliads. These traps were put in place on two parallel transects (10–30 m apart), one in grassland and the other in forest, on 9 and 23 November 2002, 18 January 2003 and 12 February 2003. Each of these were sunny days with little or no wind. The traps remained in place between 9:00 a.m. and 6:00 p.m., which corresponded to the period of highest foraging activity of P. chapoda (G. Q. Romero, pers. obs.). Each trap consisted of a wooden frame (20 × 15 cm) covered with commercial transparent plastic (DAC Ltd, São Vicente, Brazil), as well as resin Tanglefoot (Tanglefoot Co., Grand Rapids, MI, USA) on one of its surfaces. This trap type was used because it captures flying insects that randomly colonize bromeliads, these being the insects that comprise the main part of P. chapoda's diet in nature (G. Q. Romero, pers. obs.). The insects captured were counted, measured (total length) and identified at least to order (and to family where possible). The general regression equations of Hódar (1996) were used to estimate the biomass. Because some of these insects may have come from an aquatic environment adjacent to the grassland, the specimens caught were designated as ‘aquatic’ if they were from an order or family known to develop in water during their larval period. All others were designated as ‘terrestrial’.

statistical analysis

A randomized-block experimental design (Hurlbert 1984) was used for experiments 1 and 2, with each plant of the pair (sample unit) receiving a treatment. The numbers of spiders and bromeliads were compared between the treatments using randomized block, repeated measures anova (Sokal & Rohlf 1995), in which the plant pairs were the blocks and time (samples) was the repeated factor. The blocks were the random effect and the treatments (inflorescence simulation and dry leaves) were fixed effects in the mixed-model anova. Experiment 3 was done using a systematic design (see Hurlbert 1984) and the number of spiders was compared using repeated measures ancova, with the initial number of spiders (pre-treatment: sample 0) as the covariate and time (samples) as the repeated factor. The number of spiders in the leaf-removal treatments (control, 1/4 and 3/4) over time were compared using the Dunnett's post hoc test (Sokal & Rohlf 1995), with α = 0·05. The number and biomass of the insects collected in grassland and forest were compared using repeated measures anova. The probabilities of the within subject factors for all the repeated measures analyses were corrected against sphericity using the Greenhouse–Geisser correction (G–G) (Sokal & Rohlf 1995). Prior to the tests, all the data were log or log(n + 1) transformed to homogenize the variances (Sokal & Rohlf 1995). The mean values (± 1 SE) presented in the figures and text were computed directly from untransformed data. Some instars of the spiders were not considered in some of the experiments because they were found only infrequently during the study period.

Results

Plants in which the architecture was changed by inflorescence simulation were less frequently colonized by P. chapoda than control plants (Fig. 1, Table 1). In general, individuals of several age classes, including adult males, more frequently colonized plants that had not been modified by inflorescence (Fig. 1, Table 1), except that juvenile males and females occurred with similar frequency on the control and the experimental plants (Fig. 1, Table 1). Adult + subadult females and eggsacs were more abundant on the control plants (Fig. 1, Table 1). During this experiment, the number of spiders per plant varied temporally, generally increasing at the start and decreasing at the end of the experiment (Fig. 1, Table 1).

Figure 1.

Number of spiders in each age class and the number of eggsacs of Psecas chapoda on individuals of Bromelia balansae with (filled circles) and without (open circles) inflorescence simulation (see Methods for details). Beginning of the experiment: 5 October 2001; samples: 1 = 20 October 2001, 2 = 2 November 2001, 3 = 21 November 2002, 4 = 8 December 2001. Error bars are ± 1 SE.

Table 1.  Randomised block, repeated measures anova examining the effects of Bromelia balansae inflorescence simulation (treatment) on colonization by Psecas chapoda and on the occurrence of eggsacs
ParametersSource of variationd.f.MSFPG–G
TotalTreatment 11·22012·13    0·002 
Error250·101   
Time 30·164 7·00< 0·001< 0·001
Time × treatment 30·016 0·70    0·558    0·551
Error750·023   
Females (adults + subadults)Treatment 10·44916·17< 0·001 
Error250·028   
Time 30·047 3·24    0·027    0·028
Time × treatment 30·023 1·56    0·206    0·208
Error750·015   
Males (adults)Treatment 10·098 7·63    0·011 
Error250·013   
Time 30·029 2·85    0·043    0·049
Time × treatment 30·010 0·94    0·425    0·418
Error750·010   
Males + females (juveniles)Treatment 10·011 1·20    0·284 
Error250·009   
Time 30·016 1·70    0·181    0·182
Time × treatment 30·005 0·55    0·652    0·650
Error750·009   
YoungTreatment 10·881 9·78    0·004 
Error250·090   
Time 30·049 1·93    0·132    0·136
Time × treatment 30·014 0·56    0·645    0·637
Error750·025   
EggsacsTreatment 10·214 7·86    0·010 
Error250·027   
Time 30·045 4·67    0·005    0·007
Time × treatment 30·018 1·86    0·143    0·152
Error750·010   

Grassland bromeliads with dry leaves in their rosettes were less frequently colonized by P. chapoda (Fig. 2, Table 2). These dry leaves specifically affected adult + subadult females and young (Fig. 2, Table 2). Although adult + subadult females occurred on plants blocked with dry leaves, they were found in this microhabitat only during the first two sampling periods and then abandoned it (Fig. 2). These females (adults) did not produce eggsacs in this microhabitat (Fig. 2). In the forest, only one individual of B. balansae from which the dry leaves had been removed (experimental groups) was colonized by a young P. chapoda (i.e. in only one sampling period); none of the bromeliads with dry leaves in their rosette (control groups) were colonized by this spider. The colonization of B. balansae by P. chapoda during experiments 1 and 2 was very rapid (within 10–15 days spiders of all age classes had already occupied the plants). In this short interval, the adult females had also produced eggsacs on these plants.

Figure 2.

The total number of spiders, and the number of adult + subadult females, young and eggsacs of Psecas chapoda on individuals of Bromelia balansae that received (filled circles) or did not received (open circles) dry leaves in their rosette (see Methods for details). Beginning of the experiment: 20 September 2002; samples: 1 = 30 September 2002, 2 = 15 October 2002, 3 = 26 October 2002, 4 = 9 November 2002. Error bars are ± 1 SE.

Table 2.  Randomised block, repeated measures anova examining the effects of the dry leaves added to the rosettes of Bromelia balansae (treatment) on the colonization by Psecas chapoda
ParametersSource of variationd.f.MSFPG–G
TotalTreatment 11·74330·76< 0·001 
Error170·057   
Time 30·025 0·94    0·4300·422
Time × treatment 30·021 0·78    0·5080·495
Error510·027   
Females (adults + subadults)Treatment 10·271 8·06    0·011 
Error170·034   
Time 30·038 4·12    0·0110·020
Time × treatment 30·008 0·885    0·4550·432
Error510·009   
YoungTreatment 10·80130·75< 0·001 
Error170·920   
Time 30·063 2·471    0·0720·073
Time × treatment 30·037 1·44    0·2420·243
Error510·026   

Leaf cutting negatively affected the density of P. chapoda on B. balansae (Fig. 3, Table 3). However, this effect was seen only for plants in which three-quarters of the leaf length had been removed relative to control plants, while there was no statistically significant difference between plants that lacked one-quarter of leaf length and control plants (Dunnett's post hoc test: control vs. 3/4 removed: P = 0·004; control vs. 1/4 removed: P = 0·296). When data for each instar were analysed, there was no evidence that the loss of surface area in their microhabitats affected adult + subadult females and young (Table 3). Most females on the plants that lost three-quarters of their leaf length (n = 9; 78%) were on their eggsacs.

Figure 3.

The total number of spiders, and the number of adult + subadult females and young Psecas chapoda on individuals of Bromelia balansae without (filled circles) and with partial leaf removal (filled square = 1/4 removal; open triangles = 3/4 removal) to reduce leaf length. Beginning of the experiment (sample 0): 30 December 2002; samples: 1–5 = 31 December 2002 to 4 January 2003; 6 = 18 January 2003. Error bars are ± 1 SE.

Table 3.  Repeated measures ancova examining the effects of leaf removal (1/4 and 3/4 of the total length) in Bromelia balansae (treatment) on the permanence of Psecas chapoda. The initial number of spiders was the covariate
ParametersSource of variationd.f.MSFPG–G
TotalTreatment  20·78311·64< 0·001 
Covariate  10·669 9·935    0·004 
Error 260·067   
Time  50·022 1·21    0·3060·309
Time × treatment 100·025 1·38    0·1950·214
Error1300·018   
Females (adults + subadults)Treatment  20·145 2·25    0·126 
Covariate  10·206 3·19    0·086 
Error 260·065   
Time  50·031 2·61    0·0270·049
Time × treatment 100·008 0·70    0·7250·669
Error1300·012   
YoungTreatment  20·119 2·07    0·146 
Covariate  10·012 0·21    0·650 
Error 260·058   
Time  50·005 0·24    0·9430·914
Time × treatment 100·036 1·60    0·1150·135
Error1300·023   

The biomass of terrestrial arthropods, and of aquatic arthropods that invaded the terrestrial environment as adults (especially chironomid dipterans), was 1·5-fold to 18-fold higher in grassland than in forest (Fig. 4, Table 4). Overall, 26% of the biomass of arthropods collected in grassland came from an aquatic environment, whereas only 2% of this biomass occurred in the forest. The number of terrestrial and aquatic arthropods was significantly higher in grassland than in forest (Fig. 4, Table 4).

Figure 4.

Biomass and number of terrestrial + aquatic (when in larval phase) arthropods collected with sticky traps in grassland (open circles) and adjacent forest (filled circles). Samples: 1 = 9 November 2002, 2 = 23 November 2002, 3 = 8 January 2003, 4 = 12 February 2003. Error bars are ± 1 SE.

Table 4.  Repeated measures anova examining the biomass (mg) and the number of terrestrial and aquatic arthropods (prey availability) in grassland and forest (environment)
ParametersSource of variationd.f.MSFPG–G
Biomass of arthropodsEnvironment 127·200 42·81< 0·001 
Error18 0·635   
Time 3 0·501  0·17    0·4140·411
Time × environment 3 0·175  0·34    0·7980·787
Error54 0·517   
Number of arthropodsEnvironment 1 4·929110·09< 0·001 
Error18 0·045   
Time 3 0·152  3·56    0·0200·024
Time × environment 3 0·137  3·20    0·0300·035
Error54 0·043   

Discussion

Changes in the rosette architecture of B. balansae during inflorescence affected the colonization of plants by P. chapoda. Upon blooming, the plants fold their leaves back and extend them parallel to the ground, and this changed the plant architecture from a conical tri-dimensional configuration to a flattened, almost bi-dimensional one. The functional significance of these changes for the bromeliads may be to give pollinators, such as humming birds, better access. However, at the same time, these changes appear to make the bromeliad less suitable for the spider through various mechanisms. First, when the plant alters its architecture, adult and immature spiders lose the shelter at the base of the rosette and most likely become more exposed to adverse climatic conditions and to predation. The funnel shape of B. balansae in vegetative phase and the presence of large spines on the leaves probably make these plants a very suitable structure of protection against vertebrate predators such as birds and small mammals. Second, architectural changes alter the spider's reproduction and oviposition sites. Females of P. chapoda place their eggsacs in the more internal layers of the plant. This placement protects their eggs against natural enemies and desiccation, and it may facilitate nursery localization (centre of the bromeliads) by the spiderlings (Romero & Vasconcellos-Neto, in press a). With the leaves inclined and the central inflorescence exposed, females lost their oviposition sites and the spiderlings lost their nursery. In addition, males court females in the upper region of the vertical leaves, with the females attending in the lower region of the same leaves (Rossa-Feres et al. 2000). This suggests that an architectural change of the leaves to an horizontal position disrupts mating. Finally, architectural changes can affect the foraging sites. The funnel-shape tri-dimensional architecture of Bromelia balansae may channel inwards the potential prey of P. chapoda that have fallen into the bromeliad. These potential prey may include other jumping spiders, web-building spiders, planthoppers (homopterans), flies (dipterans), dragonflies, moths and wasps, all of which we have seen P. chapoda feeding on in nature (G. Q. Romero, unpublished data).

Dry leaves in the rosettes of grassland B. balansae also affected the colonization of the microhabitats. Although adult females colonized bromeliads with dry leaves, they subsequently migrated and did not construct eggsacs in these plants. The presence of dry leaves probably reduces the availability of shelter for adult spiders and entirely eliminates the nursery (centre of the bromeliads) of the spiderlings.

Other studies have shown that web-building spiders (Figueira & Vasconcellos-Neto 1993) and sit-and-wait spiders that do not build webs (Fritz & Morse 1985; Morse 1990, 1992, 1993) evaluate oviposition sites and choose the more suitable sites. Evidently, females of P. chapoda make similar oviposition-site decisions.

Our results partly corroborated an earlier hypothesis (Romero & Vasconcellos-Neto, in press a) that P. chapoda does not occur in forest because dry leaves from trees block the plant base used as shelter. However, forest bromeliads from which dry leaves had been removed were not colonized, suggesting that other factors must affect the presence of spiders on these bromeliads. Because the biomass and the number of arthropods available in grassland were several fold higher than in forest, it is possible that habitat structure and prey availability may exert an additive effect on the spatial distribution of this spider population. Halaj et al. (1998) reported that spider abundance and species richness correlated positively with prey availability and habitat complexity. They suggested that these two factors played an important role in the community structure of spiders. However, Rypstra (1983) showed experimentally that prey availability was more important in determining the carrying capacity of a substrate for populations of the spider Achaearanea tepidariorum (Theridiidae), and McNett & Rypstra (2000) demonstrated experimentally that habitat complexity was a primary factor in determining habitat (plant species) selection by a web-building spider, Argiope trifasciata (Araneidae). These divergent findings (see also Greenstone 1984) illustrate the need for more experimental studies on spiders from different guilds before we can draw general conclusions about the relative importance of prey availability vs. habitat structure in determining spider distribution.

For the spiders, the base of the bromeliad rosettes is a useful shelter and the leaf surface provides a foraging sites. Changes in both of these structures affected how long the spiders stayed on plants. In grassland, there are more individuals of P. chapoda on plants with larger leaves (Romero & Vasconcellos-Neto, in press a). Prey may be more likely to land on the leaves of larger plants (i.e. large leaves may be better foraging sites for the spiders). However, analysis for each spider phenophase showed that cutting the leaves did not affect the number of adult + subadult females and young. Most of the adult females found on plants that had lost tree-quarters of their leaf length were on eggsacs and possibly remained on the plants to protect their offspring. Moreover, because the young (4th and 5th instars) are small, the loss of surface area did not affect them. In addition, to abandon the microhabitat in this phenophase might involve a high risk of mortality through predation or desiccation during migration.

In our experiments, Psecas chapoda colonized the bromeliads rapidly and individuals of almost all ages occupied the plants with unaltered architecture (without inflorescence or dry leaves), suggesting that they had a high efficiency in encounters with the host plant, and that they were capable of recognizing and evaluating the physical state of the microhabitat. This efficiency in locating the host plant may reflect an adaptation of these spiders to the plant and may have favoured the occupation of B. balansae by P. chapoda in different regions. Indeed, Psecas chapoda inhabits B. balansae in at least three South American countries (Höfer & Brescovit 1994; Rossa-Feres et al. 2000; Romero & Vasconcellos-Neto, in press a,b; G. Q. Romero, unpublished data), indicating a strong association between this spider and a bromeliad species. Juvenile males and females (antepenultimate instar) colonized with similar frequency the control and experimental plants in various experiments. Spiders of this age may have moved frequently among the bromeliads in search of suitable microhabitats, without older and larger spiders (adults or subadults) to prey on or expel them (G. Q. Romero, personal observation), with spiders in this instar remaining on any bromeliads until more suitable bromeliads are abandoned by older spiders. However, little is known about migration of spiders of different ages among bromeliads and levels of intraspecific competition among the different age groups.

In conclusion, the way bromeliad architecture changes during inflorescence appears to make the plants less suitable for P. chapoda. Dry leaves falling from trees also appear to affect the colonization of the bromeliads in the forest. Since prey availability was much higher in the grassland than in the forest, microhabitat structure and prey availability may be additional factors that shape the spatial distribution of P. chapoda populations. The several aspects that characterize the association of P. chapoda with B. balansae (microhabitats for shelter, foraging, mating and oviposition) appear to be strictly dependent on phytosociological parameters such as the size, morphology and habitat of the bromeliads. Alterations in these microhabitats can lead to instability in this spider–plant association.

Acknowledgements

The authors thank G. Machado, D. de C. Rossa-Feres, M. O. Gonzaga, A. J. Santos, T. J. Izzo, M. Menin and two anonymous referees for advice and for reviewing the manuscript, and A. E. B. Romero for logistic support in the field. M. Menin, T. J. Izzo, L. R. Medeiros, F. R. Torres and A. Cruz helped with data collection. The staff of Guedes farm kindly provided permission to work on their property. G. Q. Romero was supported by a research grant from Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP, grant no. 01/04610-0). J. Vasconcellos-Neto was supported by a grant from Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq, grant no. 300539/94-0). This paper is part of the BIOTA/FAPESP − The Biodiversity Virtual Institute Program (http://www.biotasp.org.br; grant no. 99/05446-8).

Ancillary