2000. The Journal of 28:346±350

DOES THE PRESENCE OF POTENTIAL PREY AFFECT WEB DESIGN IN KEYSERLINGI (ARANEAE, ARANEIDAE)?

Marie E. Herberstein, Anne C. Gaskett, Deborah Glencross, Simon Hart, Sue Jaensch and Mark A. Elgar: Department of Zoology, University of Melbourne, 3010, Victoria,

ABSTRACT. Orb-web may anticipate their future prey environment by detecting the presence of prey and adjusting their web building behavior accordingly. Here we investigate the effect of different prey sizes and density on the web size and mesh height of the orb webs constructed by Argiope keyserlingi. The experimental design allowed the transmission of prey vibrations but prevented any capture. We found that A. keyserlingi constructed webs more frequently in the presence of prey, but did not alter the web size or mesh height of their webs. Keywords: Orb web, mesh height, foraging, behavior

Orb-web spiders (Araneae, Araneidae) em- mites emerged, potentially using rainfall and ploy remarkable ¯exibility in their foraging humidity as cues (Sandoval 1994). Experi- behavior. For example, following periods of mental evidence also suggests that spiders starvation, orb-web spiders increase the size vary mesh height due to the presence of dif- of their webs and attack prey less selectively ferently sized prey (Schneider & Vollrath while sated spiders reduce web size and reject 1998). In a similar case, Zygiella x-notata less pro®table prey (e.g., Sherman 1994; Her- (Pasquet et al. 1994) anticipated prey density berstein et al. 1998; Herberstein et al. 2000). before web construction. More abundant po- Web construction is energetically the most ex- tential prey induced the construction of small- pensive component of a 's foraging ef- er webs earlier in the evening. Presumably, fort (Opell 1998), and webs cannot be modi- smaller webs were ®nished more quickly, al- ®ed following completion. Decisions made lowing prey capture to commence earlier. during web construction in¯uence subsequent Here, we examine the effect of the size and capture success until a new web is built. Thus, number of potential prey on the web building it may be advantageous to design a web in behavior of Argiope keyserlingi Karsch 1878. anticipation of the future prey environment, We predict that larger potential prey will in- rather than simply relying on past events. duce increased mesh height, and that higher Web-building spiders may make some pre- prey density will decrease web area. emptive foraging decisions in response to the Experiments were conducted in March and density or size of potential prey. Sandoval April 1998 and January 1999, using adult Ar- (1994) concluded that the orb-web spider, giope keyserlingi collected in and bistriata is able to exploit swarms , Australia. In the laboratory, spiders of unusually large prey. Parawixia were housed in upturned plastic cups (13.5 ϫ bistriata typically constructed small, ®nely 9 ϫ 9 cm) where they were watered and fed meshed webs at night that trapped tiny dip- blow ¯ies (Lucilia cuprina, Diptera) ad libi- teran prey (Sandoval 1994). At the onset of tum. The spiders were unable to construct a the rainy season, the spiders dramatically functional web in the upturned cups apart changed their activity patterns and web de- from a few supporting threads. Thus, prey sign. At this time, they built additional webs capture did not involve a web. Instead, the during the day with increased web area and spiders generally grasped the ¯ies buzzing mesh height (the average distance between around in the cup. capture spirals). Interestingly, the spiders The spiders were starved for four days prior seemed to anticipate the timing of the swarms: to experimentation. This ensured that the spi- they changed their web design before the ter- ders' energetic status was uniform. Addition-

346 HERBERSTEIN ET AL.ÐPREY AND WEB DESIGN 347 ally, by depriving spiders of web-building space we minimized the in¯uence of previ- ously built webs on the foraging decisions made during experimentation (see Herberstein et al. 2000). The spiders were weighed and transferred to three-dimensional frames (40 ϫ 50 ϫ 8.5 cm) and allowed to construct com- plete webs in the presence of different sizes and densities of potential prey. Frames either contained 30 Drosophila (Diptera), one blow ¯y, or 30 blow ¯ies. Prey were held in iden- tical plastic jars (diameter: 4.7 cm, height: 6.8 cm), covered by ®ne mesh. This setup allowed Figure 1.ÐThe mean (Ϯ SE) area of webs con- prey movement and the transmission of air- structed in the presence of one large prey, 30 small borne vibrations created by the buzzing of the prey and 30 large prey in 1998 (●) and 1999 (⅙). ¯ies, but prevented capture. We selected the two prey types because was no interaction effect of year and treat- they differ in body length (blow ¯ies: 7.8 Ϯ ment (F2, 43 ϭ 0.61, P Ͼ 0.05). Web area (Fig. 0.12 mm, n ϭ 20; Drosophila: 2.5 Ϯ 0.06 1) did differ between the two years (F1, 43 ϭ mm, n ϭ 20). To control for differences in 30.79, P Ͻ 0.01): in 1999 spiders constructed weight and therefore energy return, treatment larger webs compared to the previous year. one consisted of 30 Drosophila per jar. This This is probably because spiders were main- approximated the weight of one blow ¯y per tained in the laboratory for approximately two jar as used in treatment two (one blow ¯y: months before use in 1998, whereas the ex- 0.022 Ϯ 0.0006 g, n ϭ 39; 30 Drosophila: periment was commenced within two weeks 0.021 Ϯ 0.0004 g, n ϭ 21). The third treat- of collection in 1999. Varying the size and ment, 30 blow ¯ies, allowed comparison of density of potential prey did not affect web the webs built for different prey densities, and size (F2, 43 ϭ 0.007, P Ͼ 0.05), nor was there for different prey types. Only the ®rst web an interaction effect between year and treat- spun by each individual was measured and ment (F2, 43 ϭ 0.79, P Ͼ 0.05). The size of the used to evaluate the effects of the prey treat- frame, and thus the available web building ments. This minimized the in¯uence of pre- space may have limited the foraging decision vious foraging history on web design. We es- of the spiders. However, the maximum web timated the web area and the mesh height size observed (approx. 850 cm2) was less than using various formulae that only require a few half of that available (2000 cm2). measurements (Herberstein & Tso 2000). Mesh height (Fig. 2) was similar in both

Statistical analyses were conducted using years (F1, 43 ϭ 1.40, P Ͼ 0.05) and was un- Systat 5.2 (Wilkinson 1992) and G●Power affected by prey treatment (F2, 43 ϭ 0.34, P Ͼ (Buchner et al. 1997). Data were log trans- 0.05). Contrary to prediction, the presence of formed if they were not normally distributed large prey did not result in larger mesh height (Kolmogorov-Smirnov). Web area, mesh and compared to small prey. Power analysis re- spider weight were compared using ANOVA vealed that our sample size was suf®cient to with treatment and year as factors. All values detect a treatment effect (1 Ϫ␤ϭ0.68). are mean Ϯ SE unless stated otherwise. Again, the interaction between year and treat-

Data from 49 spiders were included in the ment was not signi®cant (F2, 43 ϭ 1.63, P Ͼ analyses. There was no signi®cant difference 0.05). in body weight between the spiders used in These results are contrary to both of our 1998 and 1999 (for 1998/1999: 30 blow ¯ies predictions, and the results of previous studies 0.255 Ϯ 0.028 g / 0.266 Ϯ 0.023 g, one blow (Schneider & Vollrath 1998; Pasquet et al. ¯y 0.269 Ϯ 0.020 g / 0.293 Ϯ 0.029 g, 30 1994) that found a relationship between the Drosophila 0.253 Ϯ 0.019 g / 0.219 Ϯ 0.038 size and density of prey and web design. g; F1, 43 ϭ 0.00001, P Ͼ 0.05). The weight of However, these previous experiments released spiders allocated to the three treatments was prey into the web-building frames with the similar (F2, 43 ϭ 1.37, P Ͼ 0.05), and there spiders. In the laboratory, we frequently ob- 348 THE JOURNAL OF ARACHNOLOGY

Table 1.ÐThe mean (ϮSE) for the web area and mesh height of spiders constructing webs when there are no ¯ies, one ¯y or 30 ¯ies enclosed with the spider.

Sam- Treat- ple Mesh height ment size Web area (cm2) (cm) No ¯ies 11 1053.0 Ϯ 112.6 0.517 Ϯ 0.02 1¯y 12 1015.3 Ϯ 108.4 0.503 Ϯ 0.04 30 ¯ies 14 1086.8 Ϯ 100.3 0.517 Ϯ 0.02

Figure 2.ÐThe mean (Ϯ SE) mesh height for webs constructed in the presence of one large prey,

30 small prey and 30 large prey in 1998 (●) and area (F2, 34 ϭ 0.29, P Ͼ 0.05; Table 1) be- 1999 (⅙). tween the treatments. However, fewer spiders constructed a web when no ¯ies were present serve orb-web spiders housed in both frames (16 out of 24 spiders). In contrast, almost all and cups grasping and consuming prey with- spiders (21 out of 22) presented with a jar of out webs. As the spiders in these previous ex- 30 blow ¯ies, and 19 of 26 spiders presented periments (e.g., Schneider & Vollrath 1998) with only one blow ¯y, built a web. We com- had the opportunity to capture prey during pared these frequencies using a contingency web building, it is unclear whether their webs table, which revealed that the likelihood to represented an anticipatory prey assessment, build a web was signi®cantly different be- 2 or past experience. In the present study, en- tween the three treatments (␹ ϭ 6.3, P ϭ closing the prey in mesh-covered jars pre- 0.044). These results indicate that our exper- vented such confounding effects. However, imental design allowed the spiders to detect the absence of any signi®cant difference in the presence of potential prey, and they ad- web design between the prey treatments sug- justed the frequency of web construction ac- gests two explanations; either our experimen- cordingly (see also Pasquet et al. 1994), but tal design did not allow the spiders to detect not web size or design. It may be that spiders the prey, or A. keyserlingi does not make pre- are unable to detect differences between the emptive adjustments to web mesh size and airborne vibrations created by different sizes area to suit varying sizes and numbers of po- and densities of prey. Alternatively, spiders tential prey. may be able distinguish between prey densi- To distinguish between these two explana- ties and sizes, but do not alter the web design tions, we repeated the experiments in January in response. Behavioral tests, such as those and May 2000 using identical methods but in- presented here, cannot distinguish between cluding a control treatment (no ¯ies), where these two alternatives. we measured web area and mesh height in a Adjusting web building frequency in re- sub-sample and the frequency of web con- sponse to the presence of prey may re¯ect risk struction in a larger sample of individuals. We sensitivity, where foragers react to variation in predicted that, if these spiders can detect air- prey encounter rates by changing web sites or borne vibrations created by the enclosed ¯ies, web size (e.g., Herberstein et al. 2000; Gilles- we should ®nd differences between treatments pie & Caraco 1987). Web building spiders in- that included no blow ¯ies (empty container), vest a substantial amount of energy into silk one blow ¯y and 30 blow ¯ies. Any difference production and web construction (e.g., Peakal in the web-building behavior between the no- & Witt 1976; Higgins & Buskirk 1992), and ¯y treatment versus the ¯y treatments would rely on prey coming into contact with the web. indicate that the spiders were able to detect As such, prey encounter can be highly unpre- the presence or absence of prey in the con- dictable and spiders may conserve energy by tainers. not building a web when there is little indi- We found no signi®cant differences in cation of abundant prey. In contrast, when mesh size (F2, 34 ϭ 1.28, P Ͼ 0.05) or web prey is in close proximity and in high density, HERBERSTEIN ET AL.ÐPREY AND WEB DESIGN 349 increased web building activity may allow power analyses for the Macintosh (Version 2.1.2; these spiders to exploit abundant prey. Computer program). University of Trier: Ger- Numerous ®eld studies have also failed to many. ®nd a consistent relationship between mesh ChacoÂn, P. & W.G Eberhard. 1980. Factors affect- height and prey size (McReynolds & Polis ing numbers and kinds of prey caught in arti®cial 1987; Herberstein & Elgar 1994; Herberstein spider webs with considerations of how orb webs trap prey. Bulletin of the British Arachnological & Heiling 1998). Simulations (Eberhard Society 5:29±38. 1986) and laboratory manipulations (Nentwig Craig, C.L. 1986. Orb-web visibility: The in¯u- 1983) further con®rm that orb-webs do not ence of ¯ight behaviour and visual phys- function as ``sieves.'' Mesh height may ful®ll iology on the evolution of web designs within alternative functions. A narrow mesh may fa- the Araneoidea. Behaviour 34:54±68. cilitate the retention of larger prey, as more Craig, C.L. & C.R. Freeman. 1991. Effects of threads are in contact with the item (Eberhard predator visibility on prey encounter: A case 1990). However, more spiral turns also re¯ect study on aerial web weaving spiders. Behavioral more light thus increasing the visibility of the & Sociobiology 29:249±254. web to prey (Craig 1986; Craig & Freeman Eberhard, W.G. 1986. Effects of orb-web geometry 1991). Mesh height may therefore indicate a on prey interception and retention. Pp. 70±100, compromise between prey retention and web In SpidersÐWebs, Behavior, and Evolution. (W.A. Shear, ed.). Stanford Univ. Press. Stanford, visibility. A larger capture area results in a California. higher prey interception rate (ChacoÂn & Eber- Eberhard, W.G. 1990. Function and phylogeny of hard 1980) and by increasing the distance be- spider webs. Annual Review of Ecological Sys- tween sticky spirals spiders may enlarge over- tems 21:341±372. all capture area without greater energy Gillespie, R.G. & T. Caraco. 1987. Risk-sensitive expenditure. Accordingly, food deprived spi- foraging strategies of two spider populations. ders commonly increase web area to enhance Ecology 68:887±899. prey encounter (Sherman 1994; Herberstein et Herberstein, M.E., K.E. Abernethy, K. Backhouse, al. 2000). Finally, it seems unlikely that spi- H. Bradford, F.E. de Crespigny, P.R. Luckock & ders would tailor their webs for small and pos- M.A. Elgar. 1998. The effect of feeding history sibly unpro®table prey. Spiders often ignore on prey capture behaviour in the orb-web spider small prey entangled in the web (Uetz & Hart- Argiope keyserlingi (Araneae: Araneidae). Ethol- sock 1986; Herberstein et al. 1998) which ogy 104:565±571. may subsequently escape. Logically, any web Herberstein, M.E., C.L. Craig & M.A. Elgar. 2000. should target pro®table prey items worthy of Foraging strategies and feeding regimes: Web and decoration investment in Argiope keyserlingi attack and more permanent retention through Karsch (Araneae: Araneidae). Evolutionary silk wrapping. Ecology Research 2:69±80. Web design re¯ects several trade-offs be- Herberstein, M.E. & M.A. Elgar. 1994. Foraging tween the different functions of various web strategies of Eriophora transmarina and Nephila elements and is in¯uenced by internal physi- plumipes (Araneae: Araneoidea): Nocturnal and ological states and previous experience. Inter- diurnal orb-weaving spiders. Australian Journal preting orb-webs as size ®lters is likely to of Ecology 19:451±457. oversimplify this complex foraging invest- Herberstein, M.E. & A.M. Heiling. 1998. Does ment. mesh height in¯uence prey length in orb-web We thank Fleur de Crespigny and Sharada spiders? European Journal of Entomology 95: Ramamurthy for their technical help and sup- 367±371. port; Simon Blomberg, Diana Fisher and Mat- Herberstein, M.E. & I.M. Tso. 2000. Evaluation of thias Herberstein for logistic support; John formulae to estimate the capture area and mesh Mackenzie and Janet Yen for providing the height of orb webs (Araneoidea: Araneae). Jour- ¯ies, the Austrian Science Foundation for ®- nal of Arachnology 28(2):180±184. nancial support to MEH (J1318-BIO and Higgins, L.E. & R.E. Buskirk. 1992. A trap-build- J1500-BIO) and the Australian Research ing predator exhibits different tactics for differ- ent aspects of foraging behaviour. Animal Be- Council to MAE (ARC 19930103). haviour 44:485±499. LITERATURE CITED McReynolds, C.N. & G.A. Polis. 1987. Ecomor- Buchner, A., F. Faul & E. Erdfelder. 1997. phological factors in¯uencing prey use by two G●Power: a priori, post-hoc, andcompromise sympatric of orb-web spiders, Argiope 350 THE JOURNAL OF ARACHNOLOGY

aurantia and Argiope trifasciata (Araneidae). variable prey type. Functional Ecology 8:701± Journal of Arachnology 15:371±383. 707. Nentwig, W. 1983. The non-®lter function of orb Schneider, J.M. & F. Vollrath. 1998. The effect of webs in spiders. Oecologia 58:418±420. prey type on the geometry of the capture web of Opell, B.D. 1998. Economics of spider orb-webs: Araneus diadematus. Naturwissenschaften 85: The bene®ts of producing adhesive capture 391±394. Sherman, P. 1994. The orb web: An energetic and thread and of recycling silk. Functional Ecology behavioural estimator of a spider's dynamic for- 12:613±624. aging and reproductive strategies. Animal Be- Pasquet, A., A. Ridwan & R. Leborgne. 1994. haviour 48:19±34. Presence of potential prey affects web-building Uetz, G.W. & S.P. Hartsock. 1987. Prey selection in an orb-weaving spider Zygiella x-notata. An- in an orb-weaving spider: Micrathena gracilis imal Behaviour 47:477±480. (Araneae: Araneidae). Psyche 94:103±116. Peakal, D.B. & P.N. Witt. 1976. The energy budget Wilkinson, L. 1992. `Systat: Statistics.' Version of an orb web-building spider. Comparative Bio- 5.2, Systat: Evanston. chemistry & Physiology A 54:187±190. Sandoval, C.P. 1994. Plasticity in web design in Manuscript received 20 August 1999, revised 30 the spider Parawixia bistriata: A response to June 2000.