Web decoration polymorphism in Argiope Audouin, 1826 (Araneidae) spiders: ontogenetic and interspecific variation

Spiders in the genus Argiope commonly include curious silk structures, termed web decorations or stabilimenta in their webs. Whilst interesting ontogenetic and interspecific variation in both the form and frequency of web decorations has been documented, to our knowledge this is the first study to compare this variation across a number of decorating species. Here we show that two sympatric species A. picta and A. aetherea construct different forms of web decorations as adults and that A. picta decorates at a higher frequency than A. aetherea. Furthermore, this difference in decoration frequency may be related to the different decoration forms (linear or cruciate) across this genus. We also show that native bees responded significantly more quickly to cruciate decorations than to linear decorations. Here we argue that consideration of the different decoration forms and the frequency at which spiders adorn their webs may help illuminate possible context‐dependent functions for these curious structures.


Introduction
Web decorations are conspicuous silk structures included in the webs of many species of diurnal orb-web spiders (Araneidae, Uloboridae, Tetragnathidae). The decorating silk is produced by the aciniform and piriform glands (Peters 1993;Foelix 1996) and arranged in various patterns on the web. Intriguingly, the expression of web decorations is characterized by very high levels of ontogenetic and interspecific variability. The ontogenetic variability has been noted in a number of species, especially in the genus Argiope. In this genus juveniles commonly include discoid decorations in their webs, whilst adults include cruciate or linear decorations (Herberstein et al. 2000a). Furthermore, there is significant interspecific variation with at least six forms of decorations found in adult female spiders (Herberstein et al. 2000a). Within just one genus, Argiope, adults are known to construct three different decoration forms; linear, cruciate, and discoid (Herberstein et al. 2000a;Bruce et al. 2005). The interspecific variation is not limited to the form of decorations but also the frequency of spiders including decorations in their webs. For A. aetherea and A. picta were conducted on three separate occasions, January to February 2002, July to August 2002, andMarch 2003. We searched extensively for individuals of all age classes, from Hervey Bay (25u129S, 152u489E) to Cape Tribulation (16u19S, 145u319E) on the coast of Queensland, Australia (Figure 1). When we located an individual we noted the presence or absence of web decorations. If decorations were present we classified them as either discoid, cruciate, or linear (sensu Herberstein et al. 2000a). We also measured the body length of the spider to the nearest 0.5 mm. Body length was used as an estimate of size because we could not accurately measure any other parameter in the field. As A. aetherea and A. picta are sympatric (sometimes within 1 m of each other) and small juveniles look similar, we collected individuals that could not be assigned to one or the other species and housed them in the laboratory in Sydney until they matured and we could identify them. Individuals that did not survive to this point were not included in the analysis. Adult female individuals were identified by the presence of a raised epigyne, with two openings, signifying a reproductively mature female. We included both female and male juveniles as males construct webs and decorations in both species. As the males of both species look similar, they were only included in the analysis if they were found in areas where only one species was present and therefore could be unambiguously assigned to that species.
We surveyed juvenile A. keyserlingi on three separate occasions in Sydney at Bicentennial Park, West Pymble in December 2003. We noted the features of the decorations and measured spiders as described above. Surveyed individuals were removed from the population after each survey to ensure they were not encountered on subsequent occasions. We obtained data on the frequency of decorating in adult A. keyserlingi from Herberstein (2000).
For the purposes of analysis, we assigned juvenile spiders to one of four size classes depending on their body length. These were: (A) below 4.0 mm; (B) 4.0-5.9 mm; (C) 6.0-7.9 mm; and (D) 8.0 mm and above. These size classes corresponded to roughly equal numbers of individuals for each species. We used chi-squared analyses to compare decoration frequency between these size classes for each species, between adults and juveniles of the same species, and between adults or juveniles across species. a was set to 0.05 in all cases.
Does satiation level affect frequency of decorating behaviour in A. aetherea and A. picta?
Juvenile and adult females of A. aetherea and A. picta were collected from Queensland and transported to the laboratory at Macquarie University, Sydney. We housed them in either upturned plastic cups (1166.5 cm) or PerspexH frames (15615650 cm) where they were fed a diet of Drosophila melanogaster Meigen, 1830 (Diptera), Lucillia cuprina (Wiedemann, 1830) (Diptera), and Acheta domestica (Linnaeus, 1758) (Orthoptera), and were regularly sprayed with water.
Adult female spiders from both species were randomly assigned to either a food-deprived or food-supplemented treatment 1 week prior to the initiation of the treatments. They were weighed and the length of the tibia-patella of the right front leg was measured in order to calculate initial body condition (see below). Each spider was then fed two juvenile crickets (A. domestica) and housed in an up-turned plastic cup. After 1 week, food-deprived spiders were each fed one cricket three times over 6 days and food-supplemented spiders were each fed three crickets three times over 6 days. Therefore, food-deprived spiders were fed three crickets each over 6 days and food-supplemented spiders nine crickets each over the same period. We sprayed all spiders with water every 2 days. After the final feed we transferred all spiders to PerspexH frames where they were allowed to construct a web. We measured the first web constructed by each spider, noting the size of the capture area (area covered by the sticky spiral) and distance between capture spirals (see Herberstein and Tso 2000 for calculations). We also noted the presence or absence of web decorations and measured their length. After an individual had constructed a web she was weighed in order to calculate her final body condition (see below).
To account for allometry we calculated the parameter ''body condition'', which is known to influence web-building behaviour (Sherman 1994). We performed a linear regression between ln-transformed tibia-patella length (independent variable) and ln-transformed initial or final weight (dependent variable) separately for each species. The residual portion of the variable ''body weight'' was then taken as an estimate of the body condition of an individual. All data were analysed for normality (Kolmogorov-Smirnov) and equality of variances (Levene's test). The results for A. aetherea and A. picta were analysed separately using either Student's t tests or the non-parametric Mann-Whitney U test. We used chisquared to compare decoration frequency between treatments. a50.05 for all tests. All data are presented as mean¡standard error.

Effect of decorating pattern (cruciate versus linear) on prey response
To investigate the influence of decoration orientation on prey response we conducted a paired Y-choice experiment. We used the webs of A. keyserlingi as this species was readily available and the reflectance spectrum of web decorations in A. keyserlingi is identical to that of A. aetherea and A. picta (Bruce et al. 2005). Webs with two decorative bands (half a cruciate decoration; one in the upper web half and one in the opposite lower web half) were harvested and offered to native Australian bees (Trigona carbonaria) in an opaque Y-maze ( Figure 2). Each web was used twice, always paired with an undecorated web. The decorations from each web were arranged in either a cruciate orientation (30u from vertical and identical to A. aetherea) or in a linear orientation (0u from vertical and identical to A. picta). The order (cruciate or linear first) and exit corresponding to the decorated web (left or right) were determined randomly.
Each trial consisted of the behavioural response of six bees entering the maze consecutively for each decoration pattern (cruciate or linear). Bees were allowed to Figure 2. Schematic of the Y-maze used to test the response of prey to cruciate and linear web decorations. acclimatize to the maze for 1 min in the isolation chamber ( Figure 2) and then released into the choice chamber. We conducted each trial with the webs presented against a background photograph of Lomandra longifolia Labill., a common plant in which decorating spiders are typically found. The photograph at the background of the right maze exit was the horizontal reverse of that at the left exit. Webs were positioned in front of the equivalent position in each photograph to ensure a consistent background for each web. The Y-mazes were covered with plastic food wrap at each exit to remove the influence of pheromones and air currents on prey choice. This foil does not influence wavelengths of light transmitted. A ''choice'' of web was recorded when the prey animal contacted the foil at one of the maze exits. Bees were removed from the maze after a ''choice'' was recorded (only one bee in the maze at any time) and the maze was cleaned with 70% ethanol (to remove olfactory cues) and allowed to completely dry. We conducted a total of 19 trials and thus recorded the choice of 228 bees (19 trials612 bees per trial) was recorded. We used a paired nonparametric Mann-Whitney test and Spearman correlation to analyse our data.

Decoration frequency meta-analysis
In order to determine whether the observed difference between decoration frequency of adult A. aetherea (cruciate) and A. picta (linear) was related to the difference in decoration form (cruciate or linear) we conducted a meta-analysis of published accounts of decoration frequency. We used the percentage of decorating spiders in a population reported by the authors to calculate a mean frequency of decoration for each form (cruciate or linear). These means were compared using the non-parametric Mann-Whitney test because the number of reported accounts of linear decorations was much lower than that of cruciate decorations.

Decoration frequency in wild populations
We were able locate a total of 83 adult female A. aetherea from Hervey Bay (25u129S, 152u489E) to Cape Tribulation (16u19S, 145u319E) ranging in length from 8.0 to 18.0 mm. Thirty adult female A. picta were located from Mackay (21u19S, 149u129E) to Cape Tribulation, ranging in length from 10.5 to 19.6 mm. We located 149 juvenile A. aetherea and 45 juvenile A. picta across the same ranges as adult spiders. Whilst the range of A. aetherea extends further south than that of A. picta, 70% of adult A. aetherea and 75% of juvenile A. aetherea were found in sympatry with A. picta. In a separate study, Herberstein (2000) surveyed the webs of 25 A. keyserlingi in West Pymble, NSW. To supplement these data we surveyed 80 juvenile A. keyserlingi from the same population.
Of the 83 A. aetherea adults 24 (28.9%) were on webs with one or more cruciate bands (Figure 3). Argiope keyserlingi adults also construct cruciate decorations, 58.7% of surveyed webs contained one or more cruciate bands (Figure 3; Herberstein 2000). By contrast A. picta are compulsive decorators with 93.3% of spiders being found on decorated webs ( Figure 3).
Furthermore, almost all adults of A. picta construct linear decorations (85.7%) with a small minority constructing cruciate decorations (14.3%). The proportion of spiders constructing decorations between these three species was significantly different (x 2 2 535.8, P,0.0001).
In all three species there was an ontogenetic shift in both decoration form and frequency during juvenile development and between juveniles and adults. In A. aetherea 40.9% of juveniles were on decorated webs, 77.8% of A. picta juveniles were on decorated webs, and in A. keyserlingi 36.3% of juveniles were on decorated webs ( Figure 3). As with adult spiders, there was an overall difference in the frequency of juvenile A. picta, A. aetherea, and A. keyserlingi adding decorations to their webs (x 2 2 522.91, P,0.0001). In both A. aetherea and A. keyserlingi juveniles up to 5.9 mm in length included discoid decorations, although this was much more common in A. keyserlingi (Figure 4a, b). In A. picta discoid decorations were constructed by individuals up to 7.9 mm (Figure 4c). In A. keyserlingi and A. aetherea there was a general decline in decoration frequency with size ( Figure 4a, b), although these frequency differences between size classes were not significant (A. keyserlingi: x 2 3 54.86, P50.18; A. aetherea: x 2 3 53.48, P50.32). Argiope picta juveniles also showed a decreased decoration frequency with increasing body size but only until they reach 8.0 mm, after which all juveniles were found with decorations (Figure 4c), again this difference in frequency between sizes was not significant (x 2 3 54.98, P50.17). Interestingly, the linear decorations characteristic of adults in this species only appear in the largest juveniles (8.0 mm and above) with smaller juveniles constructing cruciate or discoid decorations. Only one species, A. keyserlingi, showed a significant difference between juvenile and adult decorating frequency (x 2 1 54.41, P50.04), adults of this species construct decorations more frequently than juveniles.

Does satiation level affect frequency of decorating behaviour in A. aetherea and A. picta?
There was no significant difference in the initial body condition of food-deprived or foodsupplemented spiders in A. aetherea (Mann-Whitney U 33 5124.0, P50.67) or A. picta (Mann-Whitney U 25 560.0, P50.33). However, the body condition of food-supplemented spiders was significantly higher for both species after the experiment (A. aetherea: Mann-Whitney U 32 518.0, P,0.001; A. picta: U 25 513.0, P,0.001).
In both species food-deprived spiders constructed larger webs than food-supplemented spiders, although this difference was not significant in A. picta (Table I). There was no significant effect of the feeding treatments on the mesh height for either species (Table I).

Effect of decorating pattern (cruciate versus linear) on prey response
We found no difference between the frequencies of bees choosing cruciate or linear decorations from the same pair (Z 19 520.67, P50.50). Furthermore, there was no correlation between the frequency of bees choosing the decoration in the cruciate orientation and in the linear orientation (r520.24, n519, P50.33). We also analysed the time taken for bees to approach linear (39.6¡3.9 s) or cruciate decorations (28.4¡3.8 s). Of the bees that approached a decorated web, those approaching a cruciate decoration were significantly faster than those approaching a linear decoration (t 32 52.05, P50.05).

Decoration frequency meta-analysis
We found decorating frequency data for 12 Argiope species from 20 different populations (Table II). As there was considerable variation in the frequency of decorations in different populations of the same species, each population was included separately in the analysis. Of the 20 populations (18 from the literature and two from this study), 17 constructed cruciate decorations and three constructed linear decorations. One species, A. picta, has been recorded to construct both cruciate (Robinson et al. 1974) and linear (this study) decorations in different populations. Spiders that construct linear decorations are almost significantly more likely to decorate their webs (76.3¡5.0%) than spiders that build cruciate (52.9¡5.6%) decorations (Mann-Whitney U 19 510.0, P50.10).

Discussion
Our results revealed that there are pattern-specific frequencies in the web decorations of Argiope aetherea, A. keyserlingi, and A. picta. When spiders build linear patterns (A. picta) they decorate their webs far more frequently than when spiders build cruciate decorations (A. aetherea and A. keyserlingi). This was also apparent in juvenile spiders, with A. picta decorating more frequently than either A. aetherea or A. keyserlingi. Furthermore, patternspecific frequency is evident across the reported accounts of the decoration frequency in the Argiope. Moreover, A. aetherea responded to supplemental feeding in a similar way to other species (Blackledge 1998;Tso 1999;Herberstein et al. 2000b;Seah and Li 2002) by  Marples (1969) increasing decoration investment. However, A. picta did not alter its decorating behaviour. We also found evidence that a potential prey animal, Trigona carbonaria (Hymenoptera), showed a different response to linear than to cruciate decorations by approaching cruciate decorations significantly faster than linear decorations. Taken together, these lines of evidence suggest that either the different patterns perform a different function and/or there are different costs and benefits to building them. This difference is likely to be related to the angle of decorations (vertical or inclined at 30u) as there is no difference in the spectral properties of different patterns (Bruce et al. 2005). Indeed, both flies (Campbell and Strausfeld 2001) and bees (Lehrer et al. 1995) have the ability to discriminate between vertical lines and those at an angle. It is therefore possible that potential prey insects have the ability to discriminate between linear and cruciate decorations. This may then influence the attractiveness of decorations at different angles, depending on the patterns of other positive or negative cues in the environment. Whilst we did not explicitly test the function of decorations in this study, it is possible that the ontogenetic differences in decoration form indicate different functions at different life history stages. Indeed, Argiope versicolor shows decoration-specific anti-predator behaviours, with juveniles with discoid decorations more likely to shuttle to the other side of the web in the presence of a predator stimulus than those on undecorated webs (Li et al. 2003). Presumably the decoration creates a physical barrier between the spider and the predator. However, there is also evidence that discoid decorations are prey attractants in this species (Li et al. 2004). In adult spiders, there is a growing body of evidence that both cruciate (Herberstein 2000;Bruce et al. 2001) and linear decorations attract prey (Tso 1996(Tso , 1998Bruce et al. 2004), although in some species they may afford the spider protection against predators (Blackledge and Wenzel 2001). Clearly, more comparative experiments are vital to enhance our understanding of the relationship between pattern and function.
The other, non-mutually exclusive, possibility to explain interspecific differences in decoration patterns is that there is a difference in the costs and benefits of the patterns. In some species, such as A. keyserlingi, cruciate decorations attract prey and increase foraging success but they also attract predators (Herberstein 2000;Bruce et al. 2001), whilst in A. trifasciata (Forskål, 1775) linear decorations have been shown to both attract prey (Tso 1996(Tso , 1998 and provide protection against predators (Blackledge and Wenzel 2001). Therefore, cruciate decorations may be more costly in terms of predator attraction, prompting spiders with this pattern to decorate less frequently. Perhaps the payoff for cruciate decorators is that they are more attractive to prey. Whilst we found no overall difference in the number of native bees approaching linear and cruciate decorations, they approached cruciate decorations significantly faster, perhaps indicating that they have an innate preference for this pattern.
Habitat-specific factors therefore may play a role in determining the relative costs and benefits of different decoration patterns. Indeed, the low frequency of cruciate decorations in the webs of Argiope appensa on Guam compared to neighbouring islands was attributed to the lack of an avian predator on this island (Kerr 1993). Furthermore, in ''Araneus'' eburnus web decorations only influenced foraging success in undisturbed habitats, perhaps due to different prey assemblages (Bruce et al. 2004). It may be that the habitat occupied by the sympatric Argiope picta and A. aetherea favours the construction of linear decorations in adult spiders due to the presence of particular species of predators and/or prey. Further studies of the predator and prey assemblages of their habitat will assist in illuminating the costs and benefits of the different decoration forms. Interestingly, adult A. picta have also been recorded constructing mostly cruciate decorations (Robinson et al. 1974). Whilst this may simply be a case of mistaken identity, a more intriguing possibility is that this species is able to construct both forms and therefore uses the most profitable strategy in a particular environment. Indeed, in this study we recorded a low number of A. picta adults with cruciate decorations.
In this study, we have found intriguing, and to date unexplored ontogenetic and interspecific differences in the frequency of decorating behaviour in the genus Argiope. In adult spiders, these differences seem to be related to the pattern of web decorations (linear or cruciate) constructed by the different species. Specifically, consistent differences in linear versus cruciate decorations from convergent lines of evidence suggest that these patterns perform different functions and have different costs and benefits associated with them. This realization may help in resolving the controversy surrounding web decoration functions. Future functional studies should look at linear and cruciate decorations separately and consider the potential costs and benefits of these decoration forms.