Signalling conflict between prey and predator attraction


  • M. J. Bruce,

    1. Department of Zoology, University of Melbourne, Melbourne, Vic. 3010, Australia
    Search for more papers by this author
    • *Present address: Matthew Bruce, Department of Biological Sciences, Macquarie University, NSW 2109, Australia. Tel.: +61-29850-8190; fax: +61-29850-8245; e-mail:

  • M. E. Herberstein,

    1. Department of Zoology, University of Melbourne, Melbourne, Vic. 3010, Australia
    Search for more papers by this author
    • **Present address: Marie E. Herberstein, Department of Biological Sciences, Macquarie University, NSW 2109, Australia.

  • M. A. Elgar

    1. Department of Zoology, University of Melbourne, Melbourne, Vic. 3010, Australia
    Search for more papers by this author

Matthew Bruce, Department of Biological Sciences, Macquarie University, NSW 2109, Australia. Tel.: +61-29850-8190; fax: +61-29850-8245; e-mail:


Predators may utilize signals to exploit the sensory biases of their prey or their predators. The inclusion of conspicuous silk structures called decorations or stabilimenta in the webs of some orb-web spiders (Araneae: Araneidae, Tetragnathidae, Uloboridae) appears to be an example of a sensory exploitation system. The function of these structures is controversial but they may signal to attract prey and/or deter predators. Here, we test these predictions, using a combination of field manipulations and laboratory experiments. In the field, decorations influenced the foraging success of adult female St. Andrew’s Cross spiders, Argiope keyserlingi: inclusion of decorations increased prey capture rates as the available prey also increased. In contrast, when decorations were removed, prey capture rates were low and unrelated to the amount of available prey. Laboratory choice experiments showed that significantly more flies (Chrysomya varipes; Diptera: Calliphoridae) were attracted to decorated webs. However, decorations also attracted predators (adult and juvenile praying mantids, Archimantis latistylus; Mantodea: Mantidae) to the web. St. Andrew’s Cross spiders apparently resolve the conflicting nature of a prey- and predator-attracting signal by varying their decorating behaviour according to the risk of predation: spiders spun fewer decorations if their webs were located in dense vegetation where predators had greater access, than if the webs were located in sparse vegetation.


Animal signals comprise of a wide variety of forms and may incorporate tactual elements, vision, hearing, olfaction, or a combination of these. Signals may mimic a cue to which the sensory system of a receiver is pre-evolved to respond (Johnstone, 1997; Manning & Dawkins, 1998). For example, flowering plants have evolved colours and patterns that exploit the inherent sensory biases of pollinating insects (Chittka et al., 1994). Although signalling systems are often co-operative, where the interests of the signaller and the intended receiver are closely matched (Guilford & Dawkins, 1991), there is considerable opportunity for a conflict of interest between signaller and receiver (e.g. Johnstone, 1997). For example, predators can produce signals that exploit the sensory biases of their prey. Bolas spiders, Mastophora spp., capture prey using a viscous silk ball, which the spider swings from a silk thread. The spiders increase the effectiveness of this hunting method by producing a chemical that mimics the female sex pheromone of their lepidopteran prey, thereby attracting male moths to within the range of the swinging bolas (Stowe et al., 1987; Yeargan, 1994). Alternatively, predators may ‘eavesdrop’ on the signal communication of their prey. The yellowjacket wasp, Vespula germanica (Hymenoptera) locates its dipteran prey by targeting the pheromones released by lekking and mating males (Hendrichs et al., 1994; Hendrichs & Hendrichs, 1998). Similarly, the cusorial spider, Habronestes bradleyi, locates its prey, workers of the meat-ant, Iridomyrmex purpureus engaged in territorial disputes, by detecting the ants’ alarm pheromone (Allan et al., 1996). Thus, any attempt to examine the evolutionary significance of a particular signal should take into account both intended and unintended receivers.

The silk decorations (or stabilimenta) constructed by numerous species of orb-web spiders represent an interesting example of a signalling system whose function has remained unresolved for over a century (Herberstein et al., 2000a). These decorations typically consist of densely woven bands of silk, arranged in linear, cruciate or spiral patterns at the hub of the web, where the spider resides (Fig. 1). Proposed functional explanations for web decorations include web stabilization (Robinson & Robinson, 1973) and thermoregulation (Humphries, 1992). However, a visual signalling function is considered more likely than alternative mechanical or physiological functions because these decorations are only found in diurnal species that rest at the hub (Eberhard, 1990; Scharff & Coddington, 1997; Herberstein et al., 2000a).

Figure 1.

 A schematic representation of an orb-web constructed by A. keyserlingi, adorned with four decorative bands of a diagonal cross (from Herberstein, 2000).

Three basic visual signal functions of web decorations have been suggested. First, the signals may exploit the sensory biases of insects by either mimicking the colour patterns on flowers (Craig & Bernard, 1990) or by reflecting UV-light that may resemble gaps in the vegetation, which are used by insects as escape routes (Craig & Bernard, 1990; Goldsmith, 1990; Watanabe, 1999). Second, silk decorations may deter predators by either concealing the position of the spider at the hub (Hingston, 1927; Bristowe, 1941; Ewer, 1972; Eberhard, 1973, 1990; Tolbert, 1975; Edmunds & Edmunds, 1986; Schoener & Spiller, 1992), obscuring its shape (Edmunds, 1986; Blackledge & Wenzel, 2001), or making the spider appear larger (Schoener & Spiller, 1992). Finally, decorations may advertise the web as noxious to predator and nonpredator species, especially birds, which may cause web damage (Horton, 1980; Eisner & Nowicki, 1983; Blackledge & Wenzel, 1999). These functions are not mutually exclusive and web decorations may serve more than one function (Herberstein et al., 2000a). Nevertheless, unintended receivers may compromise the benefits of producing silk decorations. For example, the web decorations constructed by Argiope aurantia appear to reduce web damage caused by birds. However these decorations also reduce prey capture success because the prey apparently perceive the decorations and thus avoid the web (Blackledge, 1998a; Blackledge & Wenzel, 1999).

The signalling function of web decorations remains controversial because several studies provide conflicting results. For example, the reduction in prey capture reported by Blackledge (1998a) and Blackledge & Wenzel (1999) contrasts with the increase in foraging success reported in other studies (Craig & Bernard, 1990; Tso, 1996, 1998; Hauber, 1998; Herberstein, 2000). This controversy may be best resolved by using direct experiments to test several functions simultaneously and by considering both intended and unintended receivers (e.g. Blackledge & Wenzel, 1999). Here, we use this approach to examine the signalling effects of web decorations in Argiope keyserlingi on predators and prey through a combination of field and laboratory experiments. If web decorations exploit the sensory biases of their prey to attract them to their webs, then spiders that build decorations are expected to intercept more prey than those without decorations. In contrast, if spiders use web decorations to signal their unsuitability as prey to predators, we predict that predators will target spiders on undecorated webs rather than decorated webs.


Study animal and study site

We conducted field and laboratory experiments with the St. Andrew’s Cross spider, Argiope keyserlingi Karsch 1878, which is common along the eastern seaboard of Australia (Levi, 1983; Bradley, 1993; Elgar et al., 1996). Adult female spiders construct webs with between zero and four decorative bands in a cruciate configuration (Fig. 1). We made field observations of adults during January 2000 at Bicentennial Park, West Pymble, Sydney, Australia, and collected adult females for further laboratory experiments between February and August 2000.

The effect of web decorations on prey capture

The influence of web decorations on prey capture was examined by manipulating the presence of web decorations in the field. Between two and six webs that contained two bands of decorations were selected each day for 10 days, and no spider was selected more than once. The webs were measured (web area, decoration length, web height) and randomly assigned to an experimental or a control treatment. The spiders were removed from the web and the tibia-patella length of the first leg and the abdominal width of each spider were measured. The decorations in the webs of the experimental treatment were completely removed by cutting the radial threads attached to the decorations. The control treatment had equivalent sections of web removed without disturbing the decorations. The spiders were returned to the hub of their webs after the manipulations. A mesh insect trap (27.5 cm × 18.5 cm) was erected within 1 m, and at the same height and aspect of the web to provide a comparative estimate of the prey available to the spiders. The traps were covered on both sides with a clear sticky paste (Tanglefoot® Pest Barrier, Tanglefoot Co., Grand Rapids, MI, USA).

The webs were censused each half-hour for 7 h to monitor prey capture events. It is unlikely that capture events were missed because small items are rarely attacked and large items take longer than 30 min to ingest (personal observations). At each census, the length of each new prey item was measured and assigned to two size classes, small prey (<2 mm) and large prey (≥2 mm). Prey items were not removed from the web, thereby minimizing web disturbance and damage and maintaining natural prey capture conditions. The distance and position with respect to the hub for each prey item was measured to prevent double counting. The traps were also surveyed and the number and length of insects captured over the 7-h period were recorded.

The effect of web site on decorating behaviour

The relationship between the presence of web decorations and the position of the web within the vegetation stratum was assessed by surveying 30 adult female spiders every day for 9 days. Each spider was marked with a bee tag and the number, orientation and length of decorations were recorded. We assigned the position of the web according to one of three categories of vegetation density surrounding the web: high vegetation density (more than 75% of the vertical plane of the web obscured by vegetation), moderate density (between 75 and 25% of the web obscured by vegetation), and low density (less than 25% of the web obscured by vegetation). None of the webs was shaded by overhead vegetation. Only spiders that constructed webs at the same web site for at least 6 of the 9 days were included in the analysis. The average decorating behaviour for each individual was calculated to provide a single representative data point for each spider.

The effect of web decorations prey attraction

We examined the effect of web decorations on the flight paths of potential prey by allowing the fly, C. varipes (Diptera: Calliphoridae) to choose between two exits of a darkened Y-maze (Fig. 2). The distribution (Pont, 1973) and size (Pont, 1973; Kurahashi, 1981) of this fly makes it a potential prey species of A. keyserlingi. Furthermore, this fly was available in culture and therefore could be obtained in sufficient numbers for experimentation. We observed the preference of the fly to use an exit from the maze that did or did not have decorations. At each of the two maze exits one 14.5 cm diameter wire ring was suspended at a distance of 8 cm from the end of the maze. The rings, which contained either a decorated or an undecorated web (but with the spider removed), were randomly placed at the left or the right exit. The web hubs were aligned with the centre of the exits. The exits were covered with a plastic film (cling wrap) to eliminate the influence of olfactory cues or air currents. A black cardboard screen was placed behind the webs to provide maximum contrast between the webs and the background. The webs were illuminated from the front by 100 W natural light globes. The maze was washed with 70% ethanol and the cling wrap was replaced between each trial. Multiple webs were collected from 68 spiders but no web or combination of webs from two spiders was used more than once. Webs from the same spider were never used in the same experiment.

Figure 2.

 Schematic diagram of the Y-maze used in prey and juvenile mantid choice experiments.

A fly was selected for the trial and anaesthetized with carbon dioxide and immediately transferred to the recovery chamber of the maze. It was allowed to recover for several minutes, after which the recovery chamber was opened and the trial commenced. The exits of the maze were observed continuously from a concealed position, and the first choice was scored when the insect flew into the cling wrap at either of the maze exits. The time was recorded from the opening of the slide until a choice was recorded. The fly was destroyed at the end of the trial.

The effect of web decorations on the risk of predation

We examined whether decorations affect potential predators by providing praying mantids, Archimantis latistylus (Mantodea: Mantidae) with a choice of decorated and undecorated webs. We have observed this mantid capturing and consuming A. keyserlingi at Bicentennial Park on five occasions (no other predator has been identified at this point). In these experiments, we used adults and juveniles that had been raised in the laboratory and had had no experience of spiders or web decorations.

Juvenile mantids were placed into the Y-maze apparatus (Fig. 2) by hand and we stimulated their movement by allowing a constant air current to pass into the maze from directly behind the animal (Boyan & Ball, 1986). The trials commenced as soon as the mantid was released and were terminated after 1 h. Mantids that did not move were subjected to a subsequent trial no sooner than 24 h later.

Experiments using adult mantids were conducted using an open Y-maze apparatus because they were too large for the confined Y-maze. This open maze comprised of a perspex Y-junction (length: 16 cm, maximum width: 27 cm) that was covered with strapping tape to provide a purchase for the mantids. A large perspex frame (57.3 cm × 56.5 cm × 14.5 cm) containing a web with A. keyserlingi at the hub was placed eight centimetres from the end of each of the two arms. The abdomen of each spider was positioned at the same height relative to the end of the maze. One web contained no decorations and one web contained three or four decoration bands randomly assigned to the left or the right side of the maze. The Y-maze was inclined at an angle of 19° to encourage the mantid to walk towards the webs. The apparatus was fully enclosed within black cardboard with a black base and the only light came from overhead fluorescent tubes. The mantid was placed by hand at the beginning of the maze and its behaviour was viewed from a concealed position.

We included data only from trials in which the mantid moved its head from side to side, looking in both directions, proceeded along one of the maze arms, and finally reached for one of the spiders. A trial was terminated if the mantid did not approach either spider after 45 min. At the conclusion of each trial the web area, spider size (weight, tibia-patella length) and mantid weight was measured. The strapping tape was changed after each trial to eliminate any olfactorial cues from previous mantids.

Statistical analysis

Statistical analysis was performed using SYSTAT 9.01 (Wilkinson, 1992) with transformations (ln + 1) being performed on skewed data. All values are mean ± SE unless otherwise indicated. The number of prey items captured by each web in the field was too low to calculate a mean number of prey items in each size class per web. Therefore, prey size was analysed by randomly selecting one prey item from each web or trap. This process was repeated 25 times and a mean value for each web and trap was obtained. This process reduced the influence of any single web or trap on the analysis.


The effect of web decorations on prey capture

A total of 102 prey capture events were recorded over the experimental period. Most of the prey captured (65%) were <2 mm long although large Diptera and Hymenoptera (up to 7.7 mm) were also captured. Spiders on decorated webs (control treatment) captured an average of 7.0 ± 1.4 prey items per 14 h foraging period, whereas spiders in the experimental treatment captured an average of 5.1 ± 0.6 items per foraging period. An analysis of covariance revealed that this difference was not significant, however, there was a significant interaction between the treatments and the available prey as estimated by the traps (Table 1). When decorations were retained, prey capture rate increased with the available prey (R2=0.37, < 0.01, n=19). However, there was no relationship between prey capture rate and available prey when decorations were removed (R2=0.01, P=0.65, n=19; Fig. 3). There was no difference in the proportion of small (<2 mm) or large (≥2 mm) prey captured by webs in both treatments and the prey available to each treatment (χ23=4.05, P=0.26; Fig. 4).

Table 1.   Results of ANCOVA on field capture rates of webs with decorations and webs where decorations have been removed. Thumbnail image of
Figure 3.

 Scatterplot of available prey capture rate vs. actual prey capture rate for decorated (control) and undecorated (experimental) webs. Actual prey capture rate was log-transformed to normalize the data. The regression values are: (○) control treatment R2=0.37, < 0.01 (indicated by dashed line) and (+) experimental treatment R2=0.01, P=0.65 (indicated by solid line). The equation for the control treatment regression line is ln(y + 1)=0.17 + 0.88x. n=19 in each treatment.

Figure 4.

 Comparison of the number of small (<2 mm) and large (≥2 mm) prey items between treatments for both webs and traps. Values were calculated by averaging the percentage of small and large prey items over 19 webs or traps in each treatment.

The effect of web site on decorating behaviour

The decorating behaviour of spiders was associated with the location of their web. Spiders in vegetation with the lowest density constructed more bands (1.9 ± 0.2) than those in moderate (0.8 ± 0.3) and highest density of vegetation (0.2 ± 0.1; Kruskal–Wallis K2=14.23, < 0.001). The mean variance in the number of bands was similar between the groups (Kruskal–Wallis K2=3.88, P=0.14).

The effect of web decorations on prey attraction

Adult C. varipes typically weigh between 0.01 and 0.02 g and are between 4.0 and 6.0 mm long (Kurahashi, 1981). C. varipes showed a significant preference for decorated over undecorated webs (15 of 21; binomial P=0.026). The mean time taken to reach the exit of the maze was 44.3 ± 8.1 min.

The effect of web decorations on the risk of predation

Adult praying mantids are large, sexually dimorphic predators: females typically weigh 1.5–2.2 g and males 0.6–1.1 g. The juveniles used in the experiments weighed 0.49–0.55 g. Sex cannot be determined in juveniles. Both adult and juvenile mantids approached decorated webs more frequently than undecorated webs. Nine of 10 adult mantids approached decorated webs (binomial P=0.001) and 17 of 26 juvenile mantids approached decorated webs (binomial P=0.047). These preferences were not influenced by the size of the spider or the web: juvenile mantids were only exposed to webs of standardized dimensions and there was no significant difference between web size (F1,18=0.072, P=0.8) or spider size (weight: F1,18=3.39, P=0.08; tibia-patella: F1,14, P=0.84) in the adult mantid experiment.


The cruciate decorations incorporated in the webs of A. keyserlingi attract both prey and predators to the web. In the field, A. keyserlingi with decorations exploit increases in prey abundance. Furthermore, direct choice experiments reveal that the fly, C. varipes is also attracted to decorated over undecorated webs. Similar results were obtained from choice experiments using Drosophilla melanogaster as prey (Craig & Bernard, 1990; Watanabe, 1999) and from field manipulations using web decorations placed in artificial traps (Tso, 1998). These results contrast with those of Blackledge & Wenzel (1999), who found that the presence of web decorations reduced the prey capture of A. aurantia. There are several possible explanations for these differences. First, web decorations may have different effects in different habitats, as a result of differences in prey populations and light levels. Second, Blackledge & Wenzel (1999) did not control for web characteristics such as web area and mesh height, which influence prey capture (e.g. Charcon & Eberhard, 1980). Finally, Blackledge & Wenzel (1999) placed their experimental webs randomly in the field at unknown heights and orientations. In our study (see also Tso, 1998), web characteristics that may affect capture success were controlled for and the field experiments were carried out in the same microhabitat as the spider. Exposure of webs and decorations outside these microhabitats may affect the visibility of the web and thus prey capture success (Craig, 1988).

The reflectance spectrum of web decorations is crucial if they are to provoke a wavelength dependent response among insect prey. The visual system of insects consists of several photoreceptors, each sensitive within only a specific range of wavelengths (Goldsmith & Bernard, 1974; Goldsmith, 1990; Briscoe & Chittka, 2001). The number and sensitivity of these receptors may vary between different families of insects as well as between different species (see Briscoe & Chittka, 2001 for a summary). Insects detect and respond to visual signals if one or two classes of photoreceptors are excited and there is sufficient background contrast (Lunau, 1996). If the reflectance spectrum of web decorations is flat, thus exciting all receptors equally, they will not provoke a response and may be cryptic against a background of soil or green foliage (Blackledge, 1998a). Indeed, the decorations of the uloborid spider Octonoba sybotides are spectrally flat between 300 and 550 nm (Watanabe, 1999). However, the web decorations of A. argentata, a close relative of A. keyserlingi, have a reflectance peak in the near-UV (350 nm; Craig & Bernard, 1990). Therefore, it is likely that some insects perceive colour reflected by these web decorations.

Trichromatic (UV, blue, and green) colour vision, such as occurs in bees, is optimal for identifying flower colour (Chittka, 1997). However, the evolution of these colour receptors in arthropods predates the evolution of flower colour by approximately 400 million years (Chittka, 1996). Thus, flower colour may be adapted to exploit insect vision, rather than vice versa. Similarly, web decorations are unlikely to mimic flowers rather they exploit insect colour vision independent of the evolution of flowers.

Earlier studies suggest that web decorations reduce the risk of predation by hiding the location of the spider on the web (Hingston, 1927; Bristowe, 1941; Ewer, 1972; Eberhard, 1973, 1990; Tolbert, 1975; Edmunds & Edmunds, 1986; Schoener & Spiller, 1992) or making the spider appear larger (Edmunds, 1986; Schoener & Spiller, 1992). Experimental evidence in support of this function comes from a study on birds (Horton, 1980). Furthermore, A. trifasciata on undecorated webs appear to have been attacked more frequently by mud-dauber wasps (Blackledge & Wenzel, 2001), but the generality of these results is limited by the small (n=4) sample size. Other, indirect evidence for this function has been inferred from natural patterns of web damage (Eisner & Nowicki, 1983; Blackledge & Wenzel, 2000).

In contrast, Seah & Li (2001) demonstrate experimentally using similar choice experiments as we present here, that the predatory and sympatric jumping spider, Portia labiata, is attracted to the web decorations of A. versicolor. Similarly, we also found that the web decorations of A. keyserlingi did not deter mantid predators, but rather they attracted these predators to within striking distance of the spider. Knowledge of mantid vision is limited. However, mantids have a primary visual pigment absorbing maximally at 515 nm and a secondary maximum at 370 nm (Goldsmith & Bernard, 1974), which corresponds to the UV spectrum. Our results suggest that they are able to distinguish web decorations from a dark background, however, different background colours may affect the ability of mantids to detect decorations. Clearly, the effect of decorations on various potential predators depends on the visual acuity of the individual predator and the background colour. It follows that different selective pressures on decorating behaviour will apply where there are predators that respond differently to web decorations, for example predatory wasps (Blackledge & Wenzel, 2001).

Web decorating behaviour in A. keyserlingi creates a conflict between increasing prey capture success and increasing the risk of predation. One solution to this conflict is for individuals to adjust facultatively their decorating behaviour according to the risk of predation. For A. keyserlingi, this may be reflected in the density of the surrounding vegetation. In dense vegetation mantids may have greater access to spiders because they use foliage to approach the web. Once within reach, the mantid typically remains on the vegetation and plucks the spider from the web with its raptorial forelegs (personal observations). Thus, A. keyserlingi may construct fewer decorations in dense vegetation in order to remain cryptic to praying mantids. However, spiders that capture more prey also construct more decorations (Blackledge, 1998b; Herberstein et al., 2000b) and open web sites may lead to higher prey capture. Additionally, the facultative change in decorating behaviour with different vegetation density may also provide protection against flying predators such as predatory wasps and nonpredator bird species that can cause extensive web damage (Marson, 1947; Eberhard, 1973; Horton, 1980; Eisner & Nowicki, 1983; Kerr, 1993; Blackledge, 1998a; Blackledge & Wenzel, 2001). In the field, the presence of web decorations does not influence the short-term mortality of A. keyserlingi (Herberstein, 2000), perhaps because spiders are able to alter their decorating behaviour according to the threat posed by praying mantids. By contrast, in a different predator–prey system, individuals of A. argentata that frequently vary the number and size of web decorations benefit from a lower mortality than those that rarely decorate or those that often decorate (Craig et al., 2001). These results also suggest that web decorations attract predators. However, the facultative change in decorating behaviour may be the result of genetic variation between individual spiders. Indeed, there is a heritable component to the frequency of decorating behaviour but its expression is correlated with environmental factors (Craig et al., 2001).

The effectiveness of signals varies in different habitats and is influenced by factors such as light intensity and background contrast (Goldsmith & Bernard, 1974; Endler, 1992). The signal created by web decorations must be sufficiently general to target a number of receivers and must be perceptible over long distances. As some wavelengths are scattered or absorbed more than others over long distances, a signal incorporating a variety of wavelengths (appearing as white to humans) is best for long distance signalling (Manning & Dawkins, 1998). The distance over which a signal can be perceived is also a factor of the visual acuity of the individual receiver and the density, length and thickness of the decoration silk. This may explain why these spiders invest in decorations despite the potential of greater exposure to predators, during and after construction. Visual signals are more effective if they are not obstructed, so therefore constructing decorations in less dense vegetation would enhance the efficacy of the signal (see also Elgar et al., 1996), increasing the probability that it reaches its intended target (Guilford & Dawkins, 1991).

Why have not insects vulnerable to orb-web spider predation developed a counter strategy to avoid decorations? To understand how such a visual signalling system can remain stable, all stimuli that elicit the same reaction in insects must be considered. Approaching flowers or vegetation gaps provides benefits for insects. If the signals created by web decorations only make up a small sub-set of the signals insects respond to, then the selective pressure on insects to avoid decorations is negligible (see also Grafen, 1990; Hasson, 1994). Indeed, it is in the interests of a manipulative signaller to remain at a low density compared with the positive stimuli to avoid the evolution of avoidance mechanisms.

Although our study suggests that decorating behaviour in A. keyserlingi is maintained by increased fitness through foraging success, extrapolation to the evolutionary origins of web decorations is not possible without comparative data. These signals have evolved nine times independently among orb-web spiders and were lost once (Scharff & Coddington, 1997). Conflicting evidence for the function of web decorations suggests that different selective pressures are responsible for their maintenance in different species. Furthermore, the responses of a number of different classes of receivers must be taken into account when investigating a signalling system. There are a large number of potential receivers for visual signals, such as web decorations, so considering any one of them in isolation could result in misleading conclusions. Visual signals can be cryptic to one group of receivers and visible to others, as well as being exploited by nonintended receivers.


We thank Sue, Kevin, Kristy Burton and Trent Rivett for logistic support, Terry Beattie for constructing the Y-maze, and Melanie Archer and Aescha Hill for supplying flies. Cay Craig provided helpful discussion and Mark Hauber, I-Min Tso, Todd Blackledge, Daiqin Li, and two anonymous reviewers provided helpful comments on the manuscript. Our research was supported by the Australian Research Council (grant no.: 19930103 to MAE).


  1. *Present address: Matthew Bruce, Department of Biological Sciences, Macquarie University, NSW 2109, Australia. Tel.: +61-29850-8190; fax: +61-29850-8245; e-mail:

  2. **Present address: Marie E. Herberstein, Department of Biological Sciences, Macquarie University, NSW 2109, Australia.