Directional web strikes are performed by ray spiders in response to airborne prey vibrations

Spider orb webs intercept and adhere to flying insect prey and consist of spirals of sticky capture silk overlaid on a frame of strong radial silk. Most orb webs are static structures, constructed and fixed in one place, with the spider waiting for prey to fly into the web. When an insect hits the web, the radii stretch and stop its flight while the sticky capture silk holds onto the insect. Vibrations are then transmitted to the spider via the silk threads (Barrows, 1915), triggering the spider to attack. The spider must then move across the web to attack and subdue the insect before the insect frees itself (Watanabe, 2000; Zschokke et al., 2006). Thus, orb webs typically play a passive role in prey capture, while the spider actively moves to attack the insect only after impact occurs.

Some spiders actively actuate their webs, so that both the spider and its web move during prey capture. The triangle weaver (Hyptiotes cavatus) uses its entire web as a spring-loaded net by first tensioning the web, then releasing the web once an insect contacts the silk. The triangular web then rapidly moves sideways over the insect accelerating up to 772 m s⁻², wrapping multiple strands of sticky silk around the prey in as little as 4 ms (Han et al., 2019). While Hyptiotes’ web release is triggered by insects physically touching the web, other spiders react to prey in the air. The bolas spider throws a line of silk ending in a sticky ball of glue at passing moths, likely detecting the prey through auditory cues (Eberhard, 1977). Similarly, the ogre-faced spider (Deinopis sp.) can detect airborne prey sounds behind itself and performs backwards strikes with a small silk net held in the spider's front legs (Coddington and Sobrevila, 1987; Stafstrom et al., 2020). In both instances, the spiders gain access to prey that might otherwise avoid their webs.

Ray spiders (Theridiosomatidae) pull their orb webs into tensed cones that can be released during prey capture (Alexander and Bhamla, 2020; Coddington, 1986; McCook, 1881) (Fig. 1). After building a typical looking orb web, the spider adds a tension line that connects the center of the web to a nearby twig or rock. The spider will then face away from the web, hold onto the center of the orb with its back legs and pull itself forwards on the tension line, thus tensing the planar web into a cone. The web strike is triggered when the spider releases the tension line and the web snaps back into its planar form (McCook, 1881; Alexander and Bhamla, 2020).

The function of the slingshot-like release of Theridiosoma gemmosum webs is unclear. The web of T. gemmosum moves from a cone into a flat circle so that insects in T. gemmosum's web will not be hit by multiple strands of silk moving across its body in the same fashion as in the web of Hyptiotes spiders (Han et al., 2019). Instead, web release may allow ray spiders to strike at insects before the prey contact silk. In the field, T. gemmosum releases webs in response to nearby finger snapping (Challita et al., 2021). These spiders often prey upon small nematoceran flies (Coddington, 1986), such as mosquitoes (Coddington, 1986). These flies typically hold their front legs outstretched and can avoid webs by rapidly reversing their flight after touching a silk thread (Chacon and Eberhard, 1980). Therefore, we propose the hypothesis that T. gemmosum uses airborne vibrations to detect flying insects and releases its web to capture prey that would otherwise sense the silk and flee. We test the predictions that airborne vibratory cues are sufficient to trigger web release and that T. gemmosum will preferentially release webs when insects approach the front, rather than the back, of the web cone. We also examine the kinematics of web release to determine how quickly the web can move.

We collected Theridiosoma gemmosum (L. Koch 1877) from several locations in the Summit Metro Parks and from a residence in Richfield, Ohio (both Summit County, OH, USA). All spiders were housed in the laboratory under individual 25.5×15×19 cm inverted terrariums, which were placed on trays of shallow water to mimic the spider's natural moist environment. Twigs were placed in Styrofoam bases to provide the spiders with ∼10 cm×10 cm frames to build their orb webs. Additional small twigs were provided near the frames for the spiders to fasten their tension lines (Fig. 1). This setup allowed the terrarium to be lifted gently so that the spider and web were fully accessible and minimally disturbed during trials. Spiders were maintained in the laboratory for ∼4 weeks and fed several wingless Drosophila melanogaster once each week.

We used 19 spiders that built multiple webs (n=26) over our testing period. We used two stimuli to test for web release: a 256 Hz weighted tuning fork (American Diagnostic Corporation, NY, USA) (n=29) that provides high amplitude vibrations in the mid-range of common wingbeat frequencies for insects (Dudley, 2002), and live tethered mosquitoes (N=68) to provide natural prey stimuli. Tuning forks served as a control lacking both visual resemblance to an actual insect and the bulk air flow from moving wings. The tuning fork frequency is also in the mid-range for many North American mosquito species (300–600 Hz; Kim et al., 2021). Mosquitoes were tethered to very thin strips of black construction paper attached to the abdomen or hind legs using a small dab of gel SuperGlue (cyanoacrylate). This allowed the mosquitoes to beat their wings freely, similar to free-flying mosquitoes, while being maneuvered around the webs (Fig. 2; Movies 1 and 2). Mosquitoes occasionally stopped moving their wings and had to be prodded gently to re-stimulate flight (Movie 2). We observed 97 total web approaches, 68 from tethered mosquitoes and 29 from tuning forks (two videos were subsequently excluded because of insufficient clarity). Of the 68 mosquito approaches, 33 were presented in front of the cone and 35 were presented behind the cone (Fig. 1). Of the 29 tuning fork approaches, 16 were from the front and 13 from the back of the cone. Most webs were used more than once because mosquitoes could often be carefully pulled from capture threads without tearing the silk. But eventually the webs would be torn by hitting the mosquito or tuning fork. Most webs were used four times or less, but one exceptional web persisted through eight mosquito approaches and a second web persisted through seven tuning fork approaches. Regardless, the number of front versus back approaches was kept approximately equal for any individual web. Spiders were also allowed to rest after each approach for at least 5 min after the spiders re-tensioned their webs. We included web as a covariate in the analysis to control for these repeated tests.

To test for directionality of strikes we define the capture cone as the volume of space that the web would move through when released. Stimuli were then moved toward the spider from either in front of or behind the capture cone (Fig. 2). Approaches were done by hand, starting ∼10 cm away from the spider, and the stimuli were moved steadily toward the spider at 1 cm s⁻¹ (Movie 1). If the initial approach did not result in a release, the active mosquito was held within a few centimeters of the spider’s body for an additional 1–2 s to see if the spider responded. Including versus excluding these trials had no statistical impact on our conclusions.

Two S-900D LED light panels lighted the web for recording. All tests were recorded at 500–1000 frames s⁻¹ with a Photron SA4 Fastcam Camera using the Photron Fastcam Viewer (PFV) program (Photron, Tokyo, Japan). Recordings were done with the camera lens parallel to the spider's movement to maximize in-focus viewing of the movement of the spider.

The videos were analyzed using ProAnalyst (Xcitex, Woburn, MA, USA) to quantify the position of the spider with respect to time. The position data were then smoothed using a fourth order polynomial. This smoothed position data were then used to calculate the velocity and acceleration of the spider. To determine if the spider's release was directional, we compared the proportion of times that webs were released for frontal versus back approaches, using logit regression to control for the repeated use of some webs using the GLZ function in Statistica v.10 (StatSoft, OK, USA). We also measured strike distance to check whether frontal releases occurred while the stimulus was within the capture cone of the web such that sticky silk would contact the stimulus after release, or if webs were also released when stimuli were too far away to be hit. To measure strike distance, we measured the distance from the center of the spider's body to the stimulus. For mosquitoes, we measured from the center of the spider's body to the center of the mosquito's body. For the tuning fork, we measured from the center of the spider's body to the center of the weighted end of the tuning fork. Distances from stimuli were compared for front versus back approaches using t-tests, separately for mosquitoes and tuning forks. A t-test also compared the strike distance for frontal approaches by mosquitoes versus tuning forks. Finally, we compared results where the mosquito stimulus was held in one place after the initial approach with the subset of tests where there was a smooth approach to see if we got the same probability of web release with all the combined data.

We tested if Theridiosoma gemmosum spiders released webs directionally in response to airborne cues, before insects contacted silk. The spider’s body accelerated up to 504.47 m s⁻² (mean±s.d.:187.75±122.34 m s⁻²) during web release, reaching velocities of up to 0.97 m s⁻¹ (mean±sd: 0.53±0.16 m s⁻¹). Acceleration peaked immediately after release, usually within the first 1–2 ms, and velocity increased until around 20–30 ms. Both acceleration and velocity then declined as the stored energy in the silk was expended (Fig. 3). Logistic regression showed that spiders released webs more often when live mosquitoes were presented toward the front of the capture cone, where the mosquito would be hit by silk, compared to the back of the cone (76% vs 29%, P<0.0005), with no effect of web identity (P=0.6). Mosquitoes triggered web strikes at an average distance of 1.45±0.97 cm during frontal approaches compared to 1.03±0.35 cm for back approaches (Table S1). In 100% of the frontal releases the spider released the web only after the mosquito was within the cone of capture, such that the insect was hit by moving silk in a fraction of a second (Fig. 2). Webs were released much more often in response to high amplitude tuning forks with no influence of direction of approach (P=0.19). Tuning forks triggered web strikes at an average distance of 2.88±1.02 cm during frontal approaches compared to 2.23±0.97 cm for back approaches (Table S1). Spiders responded to tuning forks at significantly longer distances compared to mosquitoes for both front (p<0.001) and back (p<0.004) approaches. We therefore suspect that the higher amplitude vibrations of the tuning fork may have acted similar to a supernormal stimulus or that they mimicked a larger parasitoid or predator instead of prey so that webs were released defensively to evade an attack. Regardless, the tuning fork serves as an important control suggesting that the decision to release a web is made in response to sound rather than bulk flow of air or visual outline of an insect. More importantly, release always occurred without insects or tuning forks touching the webs. These data support our hypothesis that ray spiders can detect flying prey solely through airborne vibrations, and that one of the functions of the ray spider’s web release may be to attack flying insects that would otherwise sense and avoid a static orb web.

How does the spider determine when the prey is in front of the web and within range of attack? Visual cues are unlikely because: (1) the spider’s body is at the vertex of the center of the orb and the anchor line, facing away from the web cone; (2) T. gemmosum, like many orb weavers does not have the developed eyes of many visual hunting spiders (Foelix, 2010); and (3) we observed one trial where a spider did not respond to a motionless mosquito held within the capture cone, but then immediately released its web immediately after the mosquito began to beat its wings (Movie 2). The decision to release a web is therefore likely based upon vibrational information. Spiders have long setae called trichobothria, which are used to detect both sound and air currents. Theridiosomatids have numerous tibial trichobothria, with especially long trichobothria on the third and fourth tibia (Coddington, 1986). Theridiosoma gemmosum faces away from its web, so that its back legs are placed closest to the cone. It is possible that these sensitive setae on the spider's back legs detect the air currents or acoustic vibrations from the wing beats flying mosquitoes. Spiders would also sometimes raise a foreleg and swivel the leg around when a mosquito was close, possibly altering the spiders' sensitivity to vibrations. Vibrations can also be transmitted to the spider through silk. In static orb webs spiders can rapidly locate prey by comparing vibrational information from different legs that are spread out among different radial threads (Landolfa and Barth, 1996). Static orb webs can also pick up airborne vibrations that are sensed by the spider (Zhou et al., 2022). These observations suggest the hypothesis that T. gemmosum may use some comparison of sound perceived through its body versus through the web to determine whether or not an insect is inside the strike cone.

In summary, T. gemmosum discerns distance and direction of flying prey though airborne sound and releases the stored energy of their webs only once an insect is within catching range of the sticky cone. This research opens up questions about how ray spiders assess vibration transmission, and whether or not the tension or shape of a silk web can affect that transmission. Speedy response times are important to prevent escape of prey from static webs (Blackledge and Zevenbergen, 2006; Rypstra, 1982). And many orb spiders employ taxon-specific attack behaviors that increase chances of capturing insects with diverse escape behaviors (Olive, 1980; Robinson, 1969). Given that static webs can pick up airborne sounds (Zhou et al., 2022), it is plausible that spiders hunting in these webs might also discern useful information about the approach, size and/or behaviors of flying insects before they impact webs. If this hypothesis is correct, such information could significantly improve the odds of spiders successfully capturing prey.

We thank the Summit Metro Parks for allowing us to collect spiders within their park system. We also thank The University of Akron and Bath Nature Preserve for providing additional field sites for us to collect spiders and mosquitoes. Lastly, we thank the spiders and mosquitoes that were used during this study.